University of Southern California Dissertations and Theses

Thin-film impedimetric sensors for chronic in vivo use: design and application to hydrocephalus treatment

(USC Thesis Other)

Thin-film impedimetric sensors for chronic in vivo use: design and application to hydrocephalus treatment

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
THIN-FILM IMPEDIMETRIC SENSORS FOR CHRONIC IN VIVO USE:
DESIGN AND APPLICATION TO HYDROCEPHALUS TREATMENT

by

Alexander Barnes Baldwin

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)

August 2018

Copyright 2018 Alex Baldwin

i

To Matt and Kevin

ii

Acknowledgements

My time as a Ph.D. student has been long and difficult, and it’s with a huge sense of relief
that I write the last portion of this thesis. There are many people who have helped me, in both
research and life, throughout this process. Without them I would not be here today. I’d first like to
thank all the older PhD students in Ellis’s lab that mentored me when I first joined. Roya Sheybani,
Seth Hara, Curtis Lee, and Brian Kim, you taught be what it meant not only to be a good researcher
and student but what it means to be a good person and a good friend. You four were all amazing
individually, but together you became a perfect mentor and example for a kid who just got out of
college and had no idea what to do in life. Lawrence Yu, thanks for teaching me all about lab safety
and properly documenting my research (might be a little sarcasm here…). In all seriousness, you
were a great mentor and an amazing friend, and a really solid partner to explore Shanghai and
Vietnam with. Angelica Cobo, you have taught me a ton about how to be a good friend (and about
where to find good KBBQ!). Whether you end up in LA, Miami, or somewhere else, I know you’ll
continue to do good things in life. I also have to thank all the current and (hopefully) future PhD
students in the Meng Lab. Eugene Yoon, working shoulder to shoulder with you for the last several
years has inspired me to up my game when it comes to lab organization, building great testing
setups, persistence in the face of research setbacks, and just generally living my life with intensity.
Trevor Hudson, you’ve been a great friend and partner over the last three years, and I couldn’t be
more confident in handing my research projects off to you. James Yoo, I know you pretend to be
all cynical but your love for life, your family, and your friends always shines through. Chris
Larson, you rock (literally) and you’ve been a great friend to chill with both in and out of lab. Jess
Ortigoza, you’ve been through what seems like one of the hardest PhD processes I’ve ever seen,
and you’re coming out the other end stronger for it; I know you’ll go on to do great things. Kee
Scholten, I remember your first presentation for the lab, when I thought “wow, he’s way too smart
to come to USC and be our postdoc.” I’m glad you proved me wrong! Seriously, thank you for all
your mentorship, encouragement, trivia, support, and friendship. Finally, Ahuva Weltman, you’re
one of the most positive, driven, badass, thoughtful, and kind people I’ve ever met; thanks for all
the words of encouragement over the last four years, and go kick ass in med school!
The other group of people I am indebted to is the undergrads and students who I have had
the pleasure of mentoring during my time as a grad student. Madelina Pratt, you were an amazing
assistant and a constant inspiration with your energy and positivity. Your love for God and for
others is apparent in everything you do; go do great things in this life, and congratulations to you
and Pablo on the upcoming wedding! Grace Yakutis, in the short amount of time we worked
together you proved that you’re a quick learner, a hard worker, and a good friend. Whoever you
end up working for in lab will be lucky to have you! Janice Park, thanks for all your help with
research and I’m sorry I got your backpack stolen out of my car. Duke Alloh and Henri Etame,
you taught me so much about how people think who have never experienced research before, and
you showed humility as a teacher in spending a summer as a student learning something new again.
iii

Henri, thank you for the long conversations on leadership and management, and good luck with
your own studies on education.
Finally (in terms of research), I owe a huge amount to Dr. Meng for taking me in to her
lab, mentoring me over the years, and showing me how to be an effective leader. Ellis, you
managed to combine kindness and understanding with a strict requirement for perfection in a way
that made all of us better researchers and better people. Thank you for mentoring all of us and for
creating a lab where I could grow into the person I wanted to be.
Outside of lab, there have been a huge amount of people at USC who have helped and
supported me. Mischal Diasanta, your presence in the BME department has made this school a
better place, and you’ve been a huge help in my life whenever I sent you a frantic email, along
with being a good friend to all of us grad students. You’ll be missed around here. George
Tolomiczenko, thanks for teaching me all about customer discovery and what it takes to be a
medical entrepreneur. Brian Horwich, Martin Shapiro, Rupan Bose, Joe Juliano, Patrick Heindel,
I wish I had more room to thank you all personally but go out and be awesome doctors! Alexa
Hudnut, Gene Yu, Adam Mergenthal, Ivan Trujillo, Sahar Elyahoodayan, Nethika Ariyasinghe,
without you guys I never would have succeeded in grad school or in life. Nick Thompson, Kristen
Cotter, Tyler Ahlf, you guys were good party buddies and even better friends. Finally, shoutout to
my bro Collin Canella, who is killing it at hospital administration and in life, and Rob Curry, who
is going to teach industrial engineering to future Navy officers at Annapolis!
Last but most important, I want to thank my family for loving me and supporting me
through this long and sometimes very painful journey. There isn’t room on this or any page to
describe my feelings for them, so I’ll be brief. Mom, Dad, Matt, Kevin, you guys give meaning to
my life and I love you more than anything. Thank you.

iv

Table of Contents
Acknowledgements … ………………………………………………………………........ii
Table of Contents …… ……………………………………………………………..........iv
List of Figures ………………………………………………………………….............viii
List of Tables ………… ………………………………………………………................xx

Introduction …………………………………………………………………....................1

Chapter 1 Hydrocephalus and the Need for Chronically Implantable Sensors.......... 5
What is Hydrocephalus? .........................................................................................6
1.1.1 Pathology of hydrocephalus ................................................................................6
1.1.2 Causes of hydrocephalus.....................................................................................7
History of Hydrocephalus Treatment ....................................................................8
Current Shunt Technology....................................................................................11
1.3.1 Proximal and Distal Catheters ..........................................................................11
1.3.2 Valves ...............................................................................................................13
1.3.3 Valve Design Improvements.............................................................................17
Shunt Failure ..........................................................................................................20
1.4.1 Treating shunt failure ........................................................................................21
1.4.2 Diagnosing shunt failure ...................................................................................21
Smart Shunts ..........................................................................................................24
1.5.1 Prospective smart shunt technology..................................................................26
Conclusion ..............................................................................................................29
References ...............................................................................................................31

Chapter 2 Impedimetric Temperature and Flow Sensors for Physiological Fluids . 38
Background ............................................................................................................39
2.1.1 Measuring Flow in the Body.............................................................................39
2.1.2 Impedimetric Thermal Flow Sensing................................................................42
2.1.3 Previous work: a bubble-based flow sensor......................................................43
Sensor Fabrication .................................................................................................45
2.2.1 Parylene C as a Substrate Material ...................................................................45
v

2.2.2 Platinum as a Sensing Material .........................................................................46
2.2.3 Fabrication Details ............................................................................................47
2.2.4 Electron-Beam Deposited Platinum vs. Sputtered Platinum ............................48
Temperature Measurement using Electrochemical Impedance ........................50
2.3.1 Background: Measuring Temperature in Physiological Fluids ........................50
2.3.2 Theory: Temperature Sensitivity of Electrolyte Conductivity ..........................51
2.3.3 Experimental Design .........................................................................................56
2.3.4 Results ...............................................................................................................57
A Time-of-Flight Flow Sensor Using Electrochemical Impedance ...................61
2.4.1 Theory: Flow Transduction via Rate of Change of Impedance ........................61
2.4.2 Sensor Design ...................................................................................................63
2.4.3 Testing Methods................................................................................................65
2.4.4 Heater Characterization ....................................................................................66
2.4.5 Flow Transduction ............................................................................................67
2.4.6 Fluid Temperature and Ionic Concentration Experiments ................................69
2.4.7 Increasing Overheat Temperature .....................................................................71
A Calorimetric Flow Sensor Using Electrochemical Impedance ......................73
2.5.1 Theory: Flow Transduction via Difference in Impedance Dip .........................73
2.5.2 Sensor Design ...................................................................................................74
2.5.3 Experimental Methods ......................................................................................75
2.5.4 Calorimetric Results..........................................................................................76
Additional Experiments.........................................................................................79
2.6.1 Platinum-Iridium Coatings on Flow Sensors ....................................................79
2.6.2 Delamination .....................................................................................................81
2.6.3 Human CSF Tests .............................................................................................85
Discussion and Comparison to State of the Art ..................................................88
Conclusion ..............................................................................................................92
References ...............................................................................................................93

Chapter 3 Passive, Wireless Impedimetric Sensors using Reflected Impedance….100
Background ..........................................................................................................101
3.1.1 Impedance measurement in the body ..............................................................101
3.1.2 Techniques for wireless signal transmission...................................................101
3.1.3 Reflected Impedance .......................................................................................102
Reflected Impedance Theory ..............................................................................104
3.2.1 Circuit models .................................................................................................104
3.2.2 MATLAB simulations ....................................................................................106
Test Coils...............................................................................................................107
3.3.1 Design .............................................................................................................107
3.3.2 Fabrication ......................................................................................................108
3.3.3 Testing and characterization ...........................................................................109
vi

Test Coil Results ...................................................................................................110
3.4.1 Basic coil properties ........................................................................................110
3.4.2 Conductivity measurement .............................................................................111
3.4.3 Coil offset and misalignment ..........................................................................112
3.4.4 Temperature measurement ..............................................................................114
Reflected Impedance Patency Sensor.................................................................115
3.5.1 Design and testing of first-generation sensor ..................................................115
3.5.2 Experimental results........................................................................................116
3.5.3 Second-generation patency sensor design ......................................................116
3.5.4 Preliminary second-generation results ............................................................117
Reflected Impedance Glucose Sensor.................................................................122
3.6.1 Diabetes...........................................................................................................122
3.6.2 Diabetes diagnosis and treatment....................................................................123
3.6.3 State of the art in glucose sensing ...................................................................124
3.6.4 Test coil functionalization experiments ..........................................................126
3.6.5 Test coil results ...............................................................................................127
3.6.6 Second-generation glucose sensor ..................................................................127
3.6.7 Future biomarker sensing work ......................................................................129
The Reflected Impedance Method: Discussion and Comparison to State of the Art
......................................................................................................................................131
Conclusion ............................................................................................................123
References .............................................................................................................133

Chapter 4 Using Kirigami Principles to Develop Stretchable Parylene C Devices.142
Kirigami ................................................................................................................143
Parylene C Kirigami Test Devices ......................................................................145
4.2.1 Devices design ................................................................................................145
4.2.2 COMSOL simulations ....................................................................................146
4.2.3 Fabrication ......................................................................................................149
4.2.4 Testing methods ..............................................................................................151
Characterization ..................................................................................................151
4.3.1 Mechanical characterization ...........................................................................151
4.3.2 DC electrical characterization .........................................................................153
Strain Transduction .............................................................................................154
4.4.1 DC strain sensitivity ........................................................................................154
4.4.2 High-frequency trace impedance ....................................................................156
4.4.3 Inter-trace capacitance ....................................................................................156
Discussion..............................................................................................................157
Potential Applications ..........................................................................................158
Conclusions and Future Work ............................................................................159
vii

References .............................................................................................................160

Chapter 5 Clinical Tests of a Multi-Sensor Module in Pediatric EVD Patients ..... 163
A Multi-Sensor Module for Hydrocephalus Shunts .........................................164
5.1.1 Impedimetric sensors for hydrocephalus shunts .............................................164
5.1.2 A multi-sensor module ....................................................................................166
Designing a Sensor System for External Ventricular Drains ..........................168
5.2.1 Background on EVDs .....................................................................................168
5.2.2 Integrating the multi-sensor module into EVDs .............................................170
5.2.3 Sterilization .....................................................................................................171
5.2.4 Full system packaging.....................................................................................172
Clinical Study Round 1 ........................................................................................173
5.3.1 First-generation electronics design .................................................................173
5.3.2 Benchtop testing – sensors and first-generation electronics ...........................176
5.3.3 Clinical issues .................................................................................................180
5.3.4 Clinical results ................................................................................................181
5.3.5 Conclusion to clinical study round 1 ..............................................................186
Clinical Study Round 2 ........................................................................................186
5.4.1 Sensor fabrication issues .................................................................................186
5.4.2 Second-generation electronics design .............................................................187
5.4.3 Benchtop testing..............................................................................................189
5.4.4 Preliminary clinical results .............................................................................194
Discussion and Future Work ..............................................................................197
References .............................................................................................................199

Chapter 6 Conclusion………………….…………………………………....................202
Appendix A: Fabrication Flow and Photoresist Recipes ………………....................206
Appendix B: Device Release and Packaging ……....……………………...................208
Appendix C: Fabrication Masks ……….……....…………………….........................213

viii

List of Figures
Figure 1.1 (Left) A child with hydrocephalus
2
. (Right) Computerized tomography (CT) scan of a hydrocephalus
patient showing cranial tissue compression and ventricular enlargement
3
........................................................... 6
Figure 1.2 Diagram showing the ventricles of the brain. After being produced in the choroid plexus, CSF flows first
into the lateral ventricles, then through the foramina of Monro to the third ventricle, then through the aqueduct
of Sylvius to the fourth ventricle. The fourth ventricle contains openings which allow CSF to drain into the
subarachnoid space, where it is reabsorbed. A blockage in any of these passageways, whether congenital or
acquired, can lead to hydrocephalus. ................................................................................................................... 7
Figure 1.3 (Left) Diagram of the brain’s ventricles by Leonardo da Vinci, 1510. (Right) Medical pamphlet
describing hydrocephalus, Nuremberg, 1556.
5
.................................................................................................... 8
Figure 1.4 An early VP shunt made using Silastic. This shunt contains a “flushing chamber” roughly equivalent to
pumping reservoirs in current valves and features a slit valve at the distal end
17
. © 1967 AANS .................... 10
Figure 1.5 Modern hydrocephalus shunts consist of a proximal catheter implanted in the ventricle, a valve which
opens in response to a pressure gradient between the brain and the rest of the body, and a distal catheter which
drains fluid into the peritoneal, pleural, or atrial cavities. .................................................................................. 11
Figure 1.6 A proximal catheter for hydrocephalus shunts, containing 16 holes of 500 µm diameter. Proximal
catheters are made of silicone rubber and are infused with barium as a contrast agent for x-ray and CT
imaging. ............................................................................................................................................................. 11
Figure 1.7 Flow through the proximal catheter occurs primarily the top hole or hole pair
26
. © 2016 Springer Nature
........................................................................................................................................................................... 12
Figure 1.8 A flanged proximal catheter, designed by Portnoy in 1971
29
. © 1976 Springer Nature ............................ 13
Figure 1.9 Diagram of a distal slit valve
30
. Distal slit valves were the first type of valves put in use, but problems
with clogging led to them being quickly phased out. ......................................................................................... 13
Figure 1.10 The operation of a ball and cone valve. Cracking pressure is defined by the spring’s stiffness
30
. .......... 14
Figure 1.11 Diagram of a flow-regulating (or anti-siphoning) valve, which uses the difference between atmospheric
pressure and the negative pressure due to flow through the distal catheter to progressively increase fluidic
resistance and close the valve when the flow rate is too high
47
.......................................................................... 18
Figure 1.12 The ProGAV gravitational valve from Miethke contains both a standard ball-in-cone valve and a
gravitational assist device
41
. When the patient is lying down, the fluid’s flow is not impeded, but when the
patient is standing vertically steel balls fall and block flow through the channel, preventing overdrainage due
to self-siphoning. ................................................................................................................................................ 19
Figure 1.13 A CT image of the brain of a shunted hydrocephalus patient. The ventricles are somewhat enlarged and
the shunt’s proximal catheter can be seen inserted into the left ventricle
64
. ....................................................... 22
Figure 1.14 A conceptual diagram of a smart shunt system, which would relay physiological and diagnostic
information wirelessly to a physician. A more advanced smart shunt could also include closed-loop feedback
which adjusted the valve position in response to changes in pressure in the brain. ........................................... 24
ix

Figure 1.15 A diaphragm-based silicone pressure sensor can fail if used chronically in vivo from multiple factors,
including corrosion of the silicon substrate, biofouling, fatigue of the flexible sensing diaphragm, and
compromised integrity of the vacuum cavity used as a reference pressure
76
. © 2016 IEEE .............................. 26
Figure 1.16 Similar to the pressure sensor, a traditional MEMS thermal flow sensor can undergo failure due to
corrosion or biofouling. In addition, to operate in a saline environment requires polymer encapsulation of the
heater and temperature sensitive elements to protect from short circuit and corrosion. This encapsulation
increases thermal resistance between the sensor and the fluid under consideration and can increase thermal
leakage through the substrate. ............................................................................................................................ 26
Figure 1.17 The ShuntCheck system involves (A) placing a temperature sensor on the skin above the shunt’s distal
catheter and (B) placing an ice pack on the neck upstream of the sensor. If the sensor registeres a decrease in
temperature, it is assumed that fluid is flowing through the shunt
104
. ................................................................ 28
Figure 1.18 (Left) The Triggerfish contact lens device measures changes in IOP via changes in the eye’s radius and
transmits pressure measurements to an external reading coil
109
. (Right) The Miethke Sensor-Reservoir allows
doctors to place a traditional MEMS pressure sensor in line with a hydrocephalus shunt. Pressure data can be
wirelessly transmitted on demand to an external receiver
110
. ............................................................................. 29
Figure 2.1 Thermal flow sensors operate by measuring the effects of convective heat transfer in a flowing fluid.
Several methods of thermal flow sensing exist; the most popular methods are time-of-flight sensing, where a
heater generates a heat pulse in the fluid and the pulse travels downstream past one or more temperature
sensors, and calorimetric sensing, where the difference in temperature upstream and downstream of a heater is
used to transduce flow. ....................................................................................................................................... 40
Figure 2.2 (a) Micrograph of sensor as fabricated and released from wafer and (b) attached to flat flexible cable for
testing via zero insertion force connector. Additional electrodes on die were utilized for other electrochemical
impedance measurements. © 2015 IEEE ........................................................................................................... 43
Figure 2.3 Left: Bubble flow sensor operates by (a) generating a bubble upstream via electrolysis. The bubble then
flows downstream past the second (b) and third (c) electrode pairs, generating an impedance signal (d) which
can be used to measure flow rate. Right: the time of flight (TOF) of an electrolyzed bubble was found to be
inversely proportional to flow rate. © 2015 IEEE ............................................................................................. 44
Figure 2.4 Parylene C is a monochlorinated poly(para-xylylene) chain which possesses properties such as
flexibility, biocompatibility, inertness in vivo, and micromachinability. These properties have led to its use in a
number of biomedical and MEMS devices. ....................................................................................................... 45
Figure 2.5 The sensor fabrication process involved (a) deposition of 12 µm Parylene C on a silicon carrier wafer,
(b) electron-beam deposition of 2000 Å platinum and patterning using AZ5214 liftoff resist, (c) exposure of
contact pads and electrodes with a switch-chemistry deep RIE in oxygen plasma, and (d) cutout etch and
removal from the carrier wafer by peeling while immersed in DI water. © 2016 IEEE .................................... 47
Figure 2.6 Electron-beam deposited platinum contained cracks which rendered devices useless. This most likely
occurred due to thermal expansion of the Parylene C substrate during platinum deposition. ............................ 48
Figure 2.7 Temperature sensitivity of conductivity, which is comparable to TCR, of an aqueous solution of sodium
ions at infinite dilution. © 2016 IEEE ................................................................................................................ 54
Figure 2.8 The temperature sensitivity of a Na
+
ion in relation to ionic concentration according to the Debye-
Huckel-Onsager equation, at 25°C, where 0 is infinite dilution. The blue arrow indicates the ionic
concentration of cerebrospinal fluid. .................................................................................................................. 55
Figure 2.9 The electrochemical impedance between two electrodes consists of the solution resistance R S, as well as
a charge transfer resistance R ct and double layer capacitance C dl at each electrode interface
104
. ....................... 56
x

Figure 2.10 Impedimetric temperature sensing was tested using a pair of platinum electrodes on a Parylene C
substrate. A microfabricated platinum RTD, fabricated on the same substrate, was used for benchmarking. ©
2016 IEEE .......................................................................................................................................................... 57
Figure 2.11 Electrochemical impedance spectroscopy of a pair of platinum electrodes at temperatures between 30°C
and 50°C. Within the resistive range, as the temperature increases the impedance magnitude decreases. ........ 58
Figure 2.12 Left: The real part of impedance at 100 kHz was roughly linear with temperature, with a temperature
coefficient of -58.29 Ω/°C and resolution of 0.02°C. This compared favorably to the RTD, which showed a
temperature coefficient of 1.21 Ω/°C and a resolution of 0.08°C. Right: Minimal hysteresis was observed after
multiple cycles between 15°C and 40°C. ........................................................................................................... 58
Figure 2.13 EIS showing the magnitude and phase of temperature sensing electrodes in 1× PBS, 10× PBS, and DI
water. Higher solution resistivity resulted in a shift of the resistive range towards lower frequencies. ............. 59
Figure 2.14 Temperature coefficients of real impedance in the resistive range stayed relatively constant between 1×
PBS, 10× PBS, and DI water. ............................................................................................................................ 59
Figure 2.15 Electrode impedance was measured in 1× PBS for 14 hours and compared to the temperature as
measured by the RTD......................................................................................................................................... 60
Figure 2.16 The sensor transduces flow through the transfer of heat from a resistive heater to flowing fluid and
from the fluid to a pair of electrodes. The electrodes sense changes in temperature via changes in
electrochemical impedance (Z sense). Orange represents the temperature distribution being distorted by flow,
blue represents the flowing liquid and light blue represents the polymer substrate. © 2016 IEEE.................... 61
Figure 2.17 (A) Simulations show that the temperature profile 1 mm away from the heater is dependent on flow
and (B) that the maximum rate of change of temperature is related to flow velocity. © 2016 IEEE ................. 63
Figure 2.18 Fabricated sensor die showing the resistive heater and two sets of impedance electrodes. The die also
contains microfabricated pressure and patency sensors. © 2016 IEEE .............................................................. 64
Figure 2.19 A sensor die (A) just after being released from its silicon carrier wafer and (B) packaged in a luer lock
connector for fluidic testing. © 2016 IEEE ........................................................................................................ 64
Figure 2.20 To transduce flow rate, the impedance during heating was normalized to a baseline value and the
minimum rate of change in impedance, which occurred 1-2 s after heater activation, was recorded. © 2016
IEEE ................................................................................................................................................................... 66
Figure 2.21 (A) The overheat temperatures at the heater and (B) the response of an electrode pair 1 mm away for
various current levels at no flow. © 2016 IEEE ................................................................................................. 67
Figure 2.22 (A) The percent change in impedance at electrodes 1 mm away from heater during constant 1°C
heating and (B) the instantaneous rate of change. The minimum (peak) of the rate of change is used to
transduce flow velocity. © 2016 IEEE ............................................................................................................... 67
Figure 2.23 The minimum rate of change of impedance 1 mm away at flow velocities from -800 to 800 µm/s. ©
2016 IEEE .......................................................................................................................................................... 68
Figure 2.24 The sensor response of electrodes spaced 0.5 mm, 1 mm, and 2 mm away from the heater. The
response of electrodes at 4 mm was negligible. © 2016 IEEE ........................................................................... 69
Figure 2.25 A comparison of sensor response in 1x PBS and aCSF, which shows no significant differences. © 2016
IEEE ................................................................................................................................................................... 70
xi

Figure 2.26 (A) The minimum rate of change versus flow velocity at various dilutions of PBS. 0.25, 0.5, 1, and 2×
PBS have ionic concentrations of 70, 140, 280, and 560 mM respectively. (B) The sensor response versus the
baseline impedance of the temperature sensing electrodes. © 2016 IEEE ......................................................... 70
Figure 2.27 The sensor response using 1°C overheat temperature at ambient temperatures between 19°C and 30°C.
The minimum rate of change becomes slightly less sensitive at higher temperatures. © 2016 IEEE ................ 71
Figure 2.28 Comparison of sensor response between (A) 1°C and 2°C and (B) 1°C and 10°C. Higher overheat
temperatures are somewhat noisier but possess a significantly higher dynamic range, resulting in a net
improvement in measurement resolution. © 2016 IEEE .................................................................................... 72
Figure 2.29 The calorimetric/impedimetric thermal flow sensor consisted of a central heater with pairs of
impedance electrodes upstream and downstream. The uneven diffusion of heat due to flow resulted in higher
temperatures, and larger impedance dips, downstream of the heater. © 2018 IEEE ......................................... 73
Figure 2.30 To measure flow rate, the impedance across an electrode pair is measured in response to an applied heat
pulse. The impedance dip, defined as the percent change in impedance at the end of heater activation, is used
to transduce flow velocity. © 2018 IEEE ........................................................................................................... 74
Figure 2.31 Image of microfabricated flow sensor, which consists of a platinum resistive heater and two pairs of
impedance-sensing electrodes on a Parylene C substrate. © 2018 IEEE ........................................................... 75
Figure 2.32 Calorimetric flow sensor packaged in a luer lock spacer and sealed with EpoTek 353NDT
biocompatible epoxy. ......................................................................................................................................... 75
Figure 2.33 A custom multiplexer (MUX) setup was designed to simultaneously measure impedance from two
electrode pairs with a sampling rate >5 Hz each. Phosphate-buffered saline (PBS) was flowed using a syringe
pump, electrode impedance at 100 kHz was measured with an LCR meter, and a Keithley SourceMeter
delivered 3.3 V to the heater. © 2018 IEEE ....................................................................................................... 76
Figure 2.34 Data from a single pair of electrodes downstream of the heater revealed that sensitivity increased with
heat pulse length up to 10 s, but there was no significant improvement between 10 and 20 s. © 2018 IEEE ... 76
Figure 2.35 Left: Using a custom multiplexer setup, both upstream and downstream electrode response to heat was
simultaneously measured. Heat was asymmetrically distributed upstream and downstream of the heater as a
function of flow velocity. Right: Both upstream and downstream electrodes can be used to simultaneously
measure flow rate. Using multiple, independent flow measurements increased accuracy and added redundancy.
These are important for chronic in vivo applications in which electrodes may experience biofouling or
degradation. © 2018 IEEE ................................................................................................................................. 77
Figure 2.36 The difference between upstream and downstream impedance dip response allows highly sensitive flow
measurement, with a 2σ resolution of 19.1 µm/s for ultra-low flow velocities (-200 to 200 µm/s). © 2018
IEEE ................................................................................................................................................................... 78
Figure 2.37 Microscope image of a flow sensor electrode (A) before the coating process and (B) after being coated
with Pt-Ir. The rough, fractal nature of electroplated Pt-Ir causes it to appear dark under the microscope. ...... 79
Figure 2.38 Impedance magnitude and phase of a flow electrode before and after coating with platinum-iridium.
The lowest frequency available for sensor transduction dropped from 100 kHz to less than 100 Hz, and
impedance magnitude decreased by 81× at low frequencies. ............................................................................. 80
Figure 2.39 Left: Flow measurements taken with a sensor coated in platinum-iridium at 100 kHz, 10 kHz, and 1
kHz, compared to measurements taken with a bare platinum sensor at 100 kHz. Right: Drift in flow sensor
measurements over a 12-hour period decreased from 2.1%/hr before Pt-Ir coating to negligible (0.16%/hr)
after Pt-Ir coating. .............................................................................................................................................. 80
xii

Figure 2.40 Delamination occurs when adhesion is lost between adjacent Parylene C layers. (A) This can occur due
to movement causing stress at a mechanical interface, such as around the point that a Parylene device is in
contact with epoxy. (B) This delamination can propagate throughout the device and cause complete separation
of layers. ............................................................................................................................................................. 81
Figure 2.41 Delamination also occurs during soaking due to water penetration disrupting the Van der Waals forces
keeping the Parylene and platinum layers together. Here you can see platinum traces becoming detached and
moving around between Parylene layers in a severely delaminated device. ...................................................... 82
Figure 2.42 EIS across flow sensor electrodes with early signs of delamination, a low-frequency peak in the phase,
which results from signal transmission through liquid between traces. ............................................................. 82
Figure 2.43 Circuit model of the electrical response of a delaminated electrode. The solution and electrode-
electrolyte interface are modeled as a Randles circuit, with a constant phase element (Y dl & a dl) used for the
double layer capacitance. Delamination is modeled as an alternate conducting path to the solution resistance,
with a delamination resistance (R delam) modeling the resistive path and another constant phase element (Y cross &
a cross) modeling the capacitance across the insulation. A wire capacitance (C wire) is also included to model
parasitic capacitance within connection cables, which is responsible for the phase rolloff observed at high
frequencies. ........................................................................................................................................................ 83
Figure 2.44 The results of modeling EIS spectra between a thin-film platinum electrode insulated between 10 µm
Parylene C layers and a large platinum counter electrode during a 14-day soak in PBS at 37°C. Both R S and
R delam drop after the first day, while the magnitudes of both the double layer capacitance and the cross-
insulation capacitance steadily increase over the course of the test. .................................................................. 84
Figure 2.45 Flow electrodes responding to heater activation on (left) a partially delaminated device and (right) a
completely delaminated device. ......................................................................................................................... 85
Figure 2.46 Flow response curve of an R-type time-of-flight flow sensor (electrodes spaced 0.5 mm away from the
heater) in both human CSF and 1x PBS. There was no significant difference in sensor response. ................... 86
Figure 2.47 Response from a flow sensor tested in human CSF at room temperature for 80 days. (A) the slope
response of electrodes 0.5 mm away from the heater showed no significant drift, while (B) the impedance dip
response only showed drift during the first week of testing. (C, D) Electrodes 2 mm away from the heater
showed much higher relative variability but no net drift. .................................................................................. 87
Figure 2.48 The impedance magnitude and phase at 100 kHz of flow sensor electrodes in human CSF during an 80
day soak test at room temperature. ..................................................................................................................... 88
Figure 2.49 A comparison between simulated results and experimental results, both with approx. 2°C heating. The
results show a similar pattern over the majority of the tested flow rate range, though due to heat loss through
the walls of the flow channel the experimental results have a lower dynamic range. © 2016 IEEE ................. 89
Figure 3.1 Transmission distance versus antenna diameter for the reflected impedance test coils, compared to
passive, wireless sensors reported in literature that employ inductive coupling, RFID, ultrasound, or RF
coupling
1,14,22-25,40-42,51-54,57,60,62
. There is a roughly linear correlation between size and transmission distance
which is maintained across techniques. Reflected impedance devices perform worse than average, but further
developments in miniaturization and high-frequency operation promise future improvements. ..................... 103

xiii

Figure 3.2 The circuit model for the wireless sensing system: a primary coil is magnetically coupled to a secondary
coil which terminates in a pair of sensor electrodes. On the primary side (left), L P is the primary coil
inductance, C P is the parallel capacitance between windings, and R P is primary coil winding resistance. On the
secondary side (right), L S is the secondary coil inductance, R S is the trace resistance of the secondary coil
windings, C dl is the double layer capacitance of each electrode, and R Sol is the solution resistance between
electrodes. M represents the mutual inductance between primary and secondary coils. © 2017 Springer Nature
......................................................................................................................................................................... 104
Figure 3.3 The linearized circuit model of Figure 3.2, showing the leakage inductances (L P-M and L S-M). Z
represents the complex impedance of the circuit as seen from the primary coil terminals. © 2017 Springer
Nature ............................................................................................................................................................... 105
Figure 3.4 Results of simulating the effects of changing solution resistance on the reflected impedance as seen from
the primary coil. The (i) impedance at resonance and (ii) resonant frequency were both found to change with
solution resistance under order-of-magnitude estimates of circuit element values (R P = 1Ω, L P = 30 µH, C P = 2
nF, L S = 20 µH, R S = 100Ω, C dl = 5 nF, k = 0.4). In another simulation, increasing the coupling coefficient k
was found to increase the sensitivity of both (iii) impedance magnitude and (iv) resonant frequency to changes
in solution resistance (R P = 1Ω, L P = 24 µH, C P = 2 nF, L S = 4 µH, R S = 1Ω, C dl = 5 nF), and by varying the
secondary inductance L S it was found that (v) the sensitivity of impedance magnitude is maximized and (vi)
the sensitivity of resonant frequency has an inflection point where the primary and secondary resonant
frequencies are the same. © 2017 Springer Nature .......................................................................................... 107
Figure 3.5 (i) Thin-film gold coils were fabricated on a flexible Parylene C substrate. (ii) Each coil terminated in
two exposed electrodes of area 300 x 300 µm
2
. © 2017 Springer Nature ........................................................ 108
Figure 3.6 Three variations of the secondary coil, with 1, 5, or 16 turns each, were fabricated and tested. All coils
had an outer radius of 30 mm. © 2017 Springer Nature .................................................................................. 108
Figure 3.7 (i) Microfabricated coils were placed in a petri dish with varying concentrations of PBS and centered
over a commercial primary coil connected to an LCR meter (ii) A 22-turn primary coil with a rated resonant
frequency of 5 MHz (Wurth Electronics, 24 µH, Q = 180@125kHz, 50mm diameter). © 2017 Springer Nature
......................................................................................................................................................................... 109
Figure 3.8 (i) The impedance of the primary coil – secondary coil system between 3 and 10 MHz shows a clear
resonance peak close to 5.4 MHz. (ii) Both the resonant frequency and the impedance magnitude at resonance
were used to transduce solution resistance across the measurement electrodes on the secondary coil. © 2017
Springer Nature ................................................................................................................................................ 111
Figure 3.9 The system impedance magnitude at resonance is related to the conductivity of solution across the
electrodes, which is inversely proportional to measured impedance at the sensing electrodes. Error bars are
present but not visible in most cases; the average standard deviation is 13.0 Ω. © 2017 Springer Nature ...... 111
Figure 3.10 The resonant frequency is also related to conductivity across the electrodes of the 16-turn and 5-turn
devices, but the effect is only substantial for the 16-turn device. Error bars are present; the average standard
deviation is 659 Hz. © 2017 Springer Nature .................................................................................................. 112
Figure 3.11 Reflected impedance of a 16-turn secondary coil was measured as the secondary coil was separated
vertically from the primary coil using 1 mm thick glass slides. (i) When 1× PBS was applied to the secondary
coil, the impedance at resonance increased linearly as the coils were separated, while the resonant frequency
saw a slight decrease and then relative stability. However, the sensitivity of both (ii) the impedance at
resonance and (iii) the resonant frequency to changes in solution conductivity decreased when separated from
the primary coil. © 2017 Springer Nature ........................................................................................................ 113
xiv

Figure 3.12 When the coil was offset from center, both the (i) impedance magnitude and the (ii) resonant frequency
increased. This increase began to level off once the secondary coil was offset 15 mm, which is equal to its
radius. © 2017 Springer Nature ....................................................................................................................... 114
Figure 3.13 The (i) impedance magnitude at resonance decreased and the (ii) resonant frequency increased when the
solution temperature increased, which is consistent with a decrease in solution resistance due to changes in
ionic mobility when an ionic solution is heated. © 2017 Springer Nature ....................................................... 114
Figure 3.14 (i) A secondary coil for measuring catheter patency in hydrocephalus shunts, made by wrapping 34
AWG copper wire around a 3D-printed module compatible with current shunt technology. L = 4.31µH, Q =
23.4@1MHz, R DC = 1.2Ω. (ii) A mock proximal catheter, consisting of a silicone tube 1 mm wide with 12
open holes of 500 µm diameter each. © 2017 Springer Nature ....................................................................... 115
Figure 3.15 The impedance at resonance increased with decreasing number of holes, indicating that reflected
impedance could be used to predict proximal catheter occlusion .................................................................... 116
Figure 3.16 Diagram of a patency sensing reservoir design. .................................................................................... 116
Figure 3.17 SolidWorks model (left) and 3D-printed prototype (right) of a second generation reflected impedance
patency sensor. The second-generation prototype consisted of a 30-loop coil of 34 AWG copper wire, which
terminated in platinum wire electrodes (not shown). ....................................................................................... 117
Figure 3.18 (A) The second-generation sensor showed a large shift in the reflected impedance signal when
electrodes were placed in saline. (B) However, when electrodes were placed in the mock proximal catheters
there was no significant difference in signal between catheters with different numbers of open holes. .......... 118
Figure 3.19 To avoid the resonance produced by the parasitic inter-trace capacitance within the coil, secondary coil
tuning was attempted using a discrete tuning capacitor in series with the coil and electrodes. ....................... 119
Figure 3.20 Tests with the second-generation prototype + tuning capacitor showed that the reflected impedance
signal was highly sensitive to the presence or absence of fluid in the mock catheters, but there was still no
significant difference between catheters with different numbers of holes. ...................................................... 119
Figure 3.21 To test the performance limit of using a tuning capacitor, the 5.4 MHz commercial primary coil was
paired with a 25 MHz secondary coil, which was then tuned down to 5.4 MHz using a discrete capacitor in
series. Load resistors of various values were used to simulate solution resistances. ....................................... 120
Figure 3.22 Testing the “ideal” commercial coil system with resistive loads showed that the impedance at resonance
was highly sensitive to load resistance until around 6 kΩ, at which point the signal saturated. Strangely, the
resonant frequency did not correlate with load resistance at all, suggesting that the shifts in resonant frequency
seen with the thin-film test coils may have been due to variations in the double-layer capacitance instead of in
solution resistance. ........................................................................................................................................... 121
Figure 3.23 Glucose oxidase catalyzes the reaction of D-glucose with oxygen to produce D-gluconolactone and
hydrogen peroxide. Amperometric glucose sensors use an electrode held at a 0.7V bias to break down
hydrogen peroxide into water and oxygen, a process which results in a measurable current through the bias
electrode. Even without a voltage bias, however, increasing the concentration of hydrogen peroxide in solution
lowers solution resistance due to increasing the local ionic concentration. ..................................................... 124
Figure 3.24 MATLAB simulations show that both the secondary (double-layer) capacitance and the solution
resistance are related to the impedance at resonance and the resonant frequency. Therefore, an electrode
coating which changes either the double-layer capacitance or the local solution resistance should allow
biomarker detection via reflected impedance. .................................................................................................. 126
xv

Figure 3.25 A 16-turn coil coated with a GOx/titanium isopropoxide sol-gel was alternatingly exposed to PBS with
and without 100 mM d-glucose. When impedance was measured at the primary coil’s resonant frequency, (i)
the impedance magnitude increased and (ii) the phase slightly decreased on exposure to the glucose-containing
solution. This effect was still present up to 4 days after coating (data shown). © 2017 Springer Nature ........ 127
Figure 3.26 Fabrication of coils with two metal layers involves (a) deposition of metal on a Parylene C substrate
and patterning via liftoff, (b) deposition of a second, thin (1-5 µm) Parylene layer, (c) patterning and liftoff of
the second metal layer, deposition of a final Parylene insulation layer, and etching in oxygen plasma, and (d)
release from the silicon carrier wafer by peeling while immersed in DI water. .............................................. 128
Figure 3.27 Prototype second generation reflected impedance designs, constructed of two layers of gold within three
Parylene C layers. ............................................................................................................................................ 129
Figure 3.28 Proposed continuous glucose monitor using reflected impedance. Placing the electrodes on a Parylene
tab allows a three-dimensional device to be constructed out of a two-dimensional substrate.......................... 130
Figure 3.29 The reflected impedance glucose monitor would consist of an electrode tab, which would sit in the
interstitial fluid, and a coil adhered to the surface of the skin. (A) The electrode tab would be attached to an
insertion tool for placement in the skin. (B) The insertion tool would be used to punch the electrodes through
the skin into the interstitial fluid and then (C) removed, leaving a highly flexible, low-profile implant. (D) An
external reader could be used to measure glucose levels on demand. .............................................................. 130
Figure 4.1 Kirigami techniques use slits or cuts in a two-dimensional substrate (such as paper) to generate complex
three-dimensional shapes
4
. Credit: McEwen Lab, Nature, DOI: 10.1038/nature14588 .................................. 143
Figure 4.2 Left: Solar cells printed on polyimide with kirigami slits can change their angle with linear motion.
Credit: A. Lamoureux, et.al., 2015
12
. Right: Kirigami slits in graphene allow the substrate to stretch by 267%,
compared to 4% for the substrate without slits. Credit: T.C. Shyu, et.al., 2015
20
. ........................................... 144
Figure 4.3 A kirigami slit pattern etched into a 10 µm thick Parylene C substrate. The devices has not yet been
released from its silicon carrier wafer. ............................................................................................................. 145
Figure 4.4 Parylene strain sensors consist of an actuation section 5 mm long and 4 mm wide containing gold traces
wound through an etched kirigami slit array, with two contact pad regions at either end. .............................. 145
Figure 4.5 (Top) Type A devices contained slits spaced 400 µm apart in offset rows, with 200 µm separation
between rows. (Bottom) Type B devices contained slits spaced 150 µm apart in offset rows with 100 µm
separation. Traces either wind around slits consistently (right two traces) or alternated sides (left trace). ..... 146
Figure 4.6 Top view of the A-type and B-type models used for COMSOL simulations .......................................... 147
Figure 4.7 The von Mises stress within the A-type kirigami device when stretched (i) by 1 mm and (ii) by 3 mm.
Stress was concentrated in lateral bands between the rows of offset slits. Strain is shown at a 1:1 scale, with an
outline showing the geometry of the unstretched device. ................................................................................ 147
Figure 4.8 The von Mises stress inside B-type devices when stretched by (i) 2 mm and (ii) 14 mm. Stress levels are
much lower in B-type devices than in A-type devices for the same stretch distance. Strain is shown at a 1:1
scale, with an outline showing the geometry of the unstretched device. .......... 148Error! Bookmark not defined.
Figure 4.9 A close-up of the stress inside of a B-type device stretched by 6 mm. Stress in these devices is much
more uniform than in A-type devices, and is concentrated around the ends of the slits. ................................. 148
Figure 4.10 The maximum stress in A and B type kirigami devices versus strain, taken from COMSOL simulations.
Stress exceeded the yield strength of Parylene C after 1.696 mm for A-type devices and after 11.84 mm for B-
type devices. ..................................................................................................................................................... 149
xvi

Figure 4.11 Sensor fabrication involved (a) deposition of Parylene C on a silicon carrier wafer, (b) electron beam
deposition and liftoff patterning of 2000 Å Au with 200 Å Ti adhesion layer, (c) etching of slits via a
switched-chemistry deep reactive ion etch in oxygen plasma and (d) release from the carrier wafer by peeling
while immersed in DI water. © 2017 IEEE ..................................................................................................... 150
Figure 4.12 Devices were attached to PEEK backing and inserted into ZIF connectors during electrical testing. A
stretched device is shown between the ZIF connectors. © 2017 IEEE ............................................................ 150
Figure 4.13 Representative force vs displacement curves for device with A-type and B-type kirigami slit patterns,
compared with an identical Parylene device without kirigami slits. A-type devices begin plastically deforming
at around 6 mm displacement, while B-type devices stay in the elastic deformation range until around 10-12
mm displacement, though this transition is not well defined. Parylene devices without slits deform and break
almost immediately. ......................................................................................................................................... 152
Figure 4.14 Break force and break distance for devices with A-type kirigami slits, B-type kirigami slits, and no slits
(bare Parylene). A-type devices break at around 101% strain, while B-type devices break at around 350%
strain. Devices without slits break almost immediately (14% strain). Break force was inversely proportional to
break distance. .................................................................................................................................................. 152
Figure 4.15 Microscope image of a B-type device stretched to 9 mm (180% strain). Stress is concentrated at two
inflection points near the end of the slit. © 2017 IEEE .................................................................................... 153
Figure 4.16 The DC resistance of a trace on A- and B-type devices stretched at 0.2 mm/s until failure. Tighter slit
spacing increased the stretch distance to failure from 78% to 180% strain and showed that wider spacing
between slits did not improve the stability of embedded traces. A-type traces have a lower DC resistance than
B-type traces due to shorter path length; the baseline resistance of A-type traces is approximately 50 Ω versus
108 Ω for B-type traces. ................................................................................................................................... 154
Figure 4.17 The DC resistance of two different traces on a B-type device over three displacement cycles. The top
trace shows a strain sensitivity of 0.131 Ω/mm with 3σ resolution of 1.07 mm (21.4% strain), while the bottom
traces has a strain sensitivity 0.100 Ω/mm with 3σ resolution of 0.81 mm (16.2% strain). Both traces are of the
same design; differences in baseline resistance are likely due to process variations. ...................................... 155
Figure 4.18 The average DC resistance change versus strain for a trace over three displacement cycles between 0
and 4 mm. This trace showed a sensitivity of 0.156 Ω/mm with a resolution of 0.21 mm. A small amount of
hysteresis is observed, most likely due to stress relaxation when stretched. The baseline resistance for this
trace was 118.7 Ω. ............................................................................................................................................ 155
Figure 4.19 The high-frequency impedance magnitude shows a non-linear relationship with displacement, with
(left) a curved relationship for alternating traces and (right) a sigmoidal relationship for consistent traces. ... 156
Figure 4.20 The capacitance between two adjacent traces measured at 100 kHz. At around 2 mm displacement, the
capacitance begins to increase linearly with strain. .......................................................................................... 157
Figure 4.21 Kirigami slits could be used to mechanically isolate neural probes from relative motion between the
skull and brain, increasing effective compliance and leading to less inflammation and glial encapsulation. .. 159
Figure 5.1 The pressure sensor operates by creating a confined microbubble. The microbubble is contained in a
Parylene C fluidic chamber, and its size is measured using electrochemical impedance. The initial size and
dissolution rate of the microbubble can be used to transduce absolute pressure, and high-frequency pressure
changes can be transduced by measuring the compression and expansion of a stable bubble. ........................ 165

xvii

Figure 5.2 The pressure sensor consists of two pairs of platinum electrodes placed inside a microfabricated fluidic
chamber (filled with red photoresist in this image), with open ports at each end. Electrolysis produces a bubble
in the nucleation core, which then fills the bubble chamber, where its size is measured by electrodes at either
end. ................................................................................................................................................................... 165
Figure 5.3 The patency sensor operates by sensing the electrochemical impedance between an electrode inside and
an electrode outside of the proximal catheter. Any occlusion of the catheter end will cause a corresponding
increase in impedance. ..................................................................................................................................... 166
Figure 5.4 The full sensor system combines pressure, flow, and patency sensors on a single die. Two sensor designs
were fabricated; Z type devices contain a flow sensor with electrodes spaced 1 mm and 3 mm downstream of
the heater and a pressure sensor oriented parallel to the direction of flow, and R type devices contain flow
sensor electrodes spaced 0.5 mm and 2 mm downstream of the heater and a pressure sensor normal to the
direction of flow. .............................................................................................................................................. 167
Figure 5.5 Fabrication of our sensors involves (a) CVD deposition of 10 µm Parylene C on a silicon carrier wafer,
(b) electron-beam depositing platinum and patterning via liftoff, (c) depositing a second 10 µm Parylene C
layer and exposing electrodes via DRIE, (d) defining the fluidic chamber using sacrificial photoresist and a 5
µm Parylene C layer, (e) DRIE etching of fluidic ports, electrodes, and contact pads, and (f) release from the
carrier wafer, followed by dissolving the sacrificial photoresist in an acetone bath. ....................................... 168
Figure 5.6 The Becker External Drainage and Monitoring System from Medtronic, which allows pressure
measurement, draining pressure setting, and CSF sampling ............................................................................ 169
Figure 5.7 The multi-sensor module was integrated into hospital EVDs to test sensor performance under real-world,
clinical conditions. Sensors were packaged into luer lock modules which can interface directly with EVD
hardware. These sensor modules were attached to a custom datalogger box which measures flow, pressure,
and patency at set intervals and saves data to a microSD card for later analysis. ............................................ 170
Figure 5.8 H 2O 2 plasma sterilization caused discoloration on the surface of the EpoTek epoxy used to protect
electrical connections, but did not affect the integrity of the device. ............................................................... 171
Figure 5.9 The electrochemical properties of an electrode were tested via (a) electrochemical impedance
spectroscopy and (b) cyclic voltammetry before and after sterilization. No significant differences were found.
© 2016 Springer Nature ................................................................................................................................... 172
Figure 5.10 Patency was transduced before and after sterilization, revealing no significant differences in sensor
performance. © 2016 Springer Nature ............................................................................................................. 172
Figure 5.11 For hospital use the datalogger electronics were packaged in a white case with a power-indicating LED
and an on-off switch. A custom flexible cable connected the electronics to the luer-lock packaged sensor
module.............................................................................................................................................................. 173
Figure 5.12 The datalogger electronics consisted of a Teensy 3.1 microcontroller board which was integrated with
an SD card shield. For impedance measurement, a 10 kHz wave was sent through voltage and current limiting
components and into a voltage divider, with the output going directly to a 16-bit ADC. Electrolysis was
accomplished by using an LM334 current source, and flow sensing involved directly stimulating the heater
with a 3.3 V square wave. ................................................................................................................................ 174
Figure 5.13 The microcontroller and other electronic components were integrated into a custom PCB .................. 174
Figure 5.14 Using calibrated resistors, a good correlation was found between resistance and ADC input. ............. 175

xviii

Figure 5.15 Patency transduction was tested by drawing PBS through silicone catheters with varying numbers of
open holes and into the luer-lock module. Two methods of patency transduction were tested: measuring
between the two patency electrodes on the Parylene die, or measuring between one of the electrodes and an
external platinum counter electrode (not shown) placed in the beaker of PBS. ............................................... 176
Figure 5.16 When using an external catheter, a clear relationship was observed between the number of open holes
and the impedance measured by the datalogger. .............................................................................................. 177
Figure 5.17 (Left) Measurements with an LCR meter revealed that the number of open holes is directly proportional
to the impedance between the two patency electrodes on the sensor die. (Right) When using the datalogger
electronics to measure impedance, accuracy decreases and noise increases, and no significant relationship
between the impedance between patency electrodes and the number of open holes was found. ..................... 178
Figure 5.18 Raw impedance data from the datalogger electronics (blue) and a 10-point moving average (black).
Voltage was delivered to the heater between 10 and 20 seconds. Two methods for transducing flow were
tested with the electronics; the first, described in chapter 2, involves measuring the slope of impedance as the
heater is activated, while the second involves finding the difference between the heated average and the
baseline average. .............................................................................................................................................. 179
Figure 5.19 (Left) Measurements using the LCR meter revealed a direct relationship between flow rate and the rate
of change of impedance, but this was not reflected in measurements by the datalogger electronics. (Right) A
relationship between flow rate and the impedance difference before and after heater activation was seen when
both the LCR meter and the datalogger box were used for measurement. ....................................................... 179
Figure 5.20 (Left) When immersed in PBS, the fluidic chamber of the pressure sensor filled between 13:05 and
13:49 after immersion. However, even after the sensor was filled, no bubble nucleation events were observed.
(Right) After 24 hours of soaking, EIS was performed on the nucleation core electrodes. The EIS results were
purely capacitive, indicating that fluid had not filled the nucleation core. ....................................................... 180
Figure 5.21 The impedance between the two test electrodes in four patients. For three of these patients, impedance
began to dramatically rise between 10-12 hours after device activation. ......................................................... 182
Figure 5.22 The impedance data from patient 5 showed an electrical short circuit during ~7 hours of testing. This
may have been caused by water or CSF shorting a connector. ........................................................................ 182
Figure 5.23 Patient 3 had a device active for the longest time (123 hours). Test electrode impedance was stable for
the first 120 hours except for a region of higher impedance between 25 and 40 hours. This could have been
caused by a bubble partially occluding the sensor or a piece of tissue sticking to the surface. ........................ 183
Figure 5.24 Photograph of electrodes on devices recovered from (A) patient 5 and (B) patient 6, showing severe
delamination of the exposed platinum from the Parylene C substrate. ............................................................ 183
Figure 5.25 (Left) The minimum rate of change of impedance versus time, from the first pair of flow electrodes in
Patient 5. Minimum rate of change stays at zero until 8 hours after device activation, at which point it begins
to trend downward, indicating progressive delamination. (Right) The difference between heated and non-
heated impedance of the first pair of flow electrodes is positive, indicating that impedance rises when the
heater is activated. This suggests that some delamination has occurred before the first flow measurement,
though it becomes much worse after the 12 hour mark. ................................................................................... 184
Figure 5.26 The impedance data from a single heater activation / flow rate measurement attempt reveals that
impedance jumps sharply upon heater activation, with large spikes when the heater is turned on or off. This
could indicate capacitive coupling through the Parylene due to delamination at the heater. ........................... 184

xix

Figure 5.27 SEM images of a flow electrode from a device which has not been tested on the benchtop or in vivo.
The platinum surface is smooth and free from debris. ..................................................................................... 185
Figure 5.28 The surface of a flow sensor electrode from patient 9, active for 12 hours in vivo. Delamination and
cracking of the platinum surface is clearly visible, as well as contamination by what could be salt debris or
bacteria. Higher-magnification images reveal spindle-like structures on the platinum surface, which could
indicate bacterial or fungal growth adhering to the electrode surface. ............................................................. 185
Figure 5.29 Images of another flow electrode from patient 9 reveal omnipresent delamination and minor cracking,
as well as scum covering a large portion of the electrode surface. .................................................................. 185
Figure 5.30 Block diagram of the second-generation impedance measurement system, showing the AD5933
network analyzer IC connected to the Teensy 3.1 microcontroller via an I
2
C port and to electrodes on the
impedimetric sensors through a multiplexer and an analog front-end. ............................................................ 187
Figure 5.31 Picture of the second-generation datalogger board, with the network analyzer and multiplexer chips
highlighted. The second-generation electronics were layed out on a 3-layer PCB, which was designed to fit
into the same white plastic package used for the first-generation electronics. ................................................. 188
Figure 5.32 Difference in baseline noise when measuring the impedance of a resistor using (top) wall power,
delivered via a store-bought USB power converter connected directly to the Teensy board, and (bottom) the
lithium-ion battery to power the electronics board. Using wall power adds a significant amount of noise to
measurements, most likely due to switching noise within the power converter. Figure credit: Trevor Hudson,
Biomedical Microsystems Lab ......................................................................................................................... 189
Figure 5.33 EIS between 100 Hz and 100 kHz across test electrodes on a device tested in PBS for 12 days. Despite
some drift in impedance at lower frequencies, there is no significant change in magnitude or phase of
impedance at the target measurement frequency (50 kHz). ............................................................................. 190
Figure 5.34 EIS across test electrodes tested in human CSF also revealed little to no impedance drift in the
frequencies of interest. ..................................................................................................................................... 190
Figure 5.35 Flow electrode from device 2 under SEM. The electrode, which was tested in human CSF for 14 days,
shows no wrinkling or delamination, but cells can be observed adhered to the electrode surface. .................. 191
Figure 5.36 The baseline impedance magnitude at 50 kHz of flow sensing electrodes on three devices tested in
human CSF for >12 days. Two devices showed some baseline drift within the first two days, but all three
devices showed stable baseline impedances between 2 and 14 days of testing. .............................................. 191
Figure 5.37 During benchtop testing of a sensor/board in PBS, three bubbles entered the device and adhered to the
Parylene sensor. This is visible as a jump in impedance magnitude and a drop in the phase on the flow sensor
electrodes. Each time a bubble was detected, the electronics correctly identified the bubble and did not activate
the flow sensor’s heater. ................................................................................................................................... 192
Figure 5.38 The impedance across the pressure measurement electrodes was monitored by a second-generation
electronics box (wall power) while the sensor was pressurized to 400 mmHg in PBS. (A) When the jig was
first filled with PBS, a sharp drop in impedance resulted, due to PBS contacting the electrodes on each end of
the fluidic port. (B) After almost an hour of pressurization the air pocket membrane inside the pressure
channel “burst”, creating a fluidic pathway between the electrodes and causing a second drop in impedance
magnitude. ........................................................................................................................................................ 193

xx

Figure 5.39 Several flow measurements at 37°C recorded by the second-generation electronics board on battery
power, with 30 seconds of cool-down time between each measurement. Without the switching noise from the
wall power converter, the electronics are able to measure the heating signal with high precision and to
accurately transduce flow using either the initial slope or the impedance dip method. Figure credit: Trevor
Hudson, Biomedical Microsystems Lab .......................................................................................................... 193
Figure 5.40 The magnitude (left) and phase (right) between the two patency electrodes during the first six hours of
ex vivo use. The impedance magnitude decreased during this time period, stabilizing after 6 hours. ............. 194
Figure 5.41 The impedance magnitude (left) and phase (right) between the two patency electrodes during 72 hours
of recording. Data is shown for every 12 hours. Impedance magnitude and phase was stable for the first 40
hours. ................................................................................................................................................................ 195
Figure 5.42 The impedance between the patency electrodes went from stable to open circuit between hours 40 and
41. This was most likely caused by damage to the Hirose-FFC connection. ................................................... 195
Figure 5.43 The baseline impedance of the flow sensor electrodes was relatively stable for the first 88 hours of use.
Within this time period, the impedance magnitude drifted by an average of -0.1%/hr. ................................... 196
Figure 5.44 Both the initial slope and the impedance dip signals were stable for the first 72 hours, then jumped to
zero. .................................................................................................................................................................. 196
Figure 5.45 The impedance magnitude across the pressure sensor’s measurement electrodes. This data indicates that
the sensor’s microchannel filled with fluid after 5 hours of use. ..................................................................... 197
Figure 6.1 Integrating multiple sensing modalities into a hydrocephalus shunt would enable the detection and
location of multiple shunt failure modes. For example, a patency sensor would be most useful in identifying
proximal catheter occlusion, a pressure sensor could accurately identify valve or distal catheter occlusion, and
a flow sensor would be useful for correcting for overdrainage. ....................................................................... 202
Figure 6.2 A block diagram of a proposed implantable electronics system, which could record data from multiple
sensors and output it digitally via antenna admittance modulation. ................................................................. 203
Figure 6.3 A proposed power and data telemetry system for the implantable electronics. Power would be delivered
via an inductive coupled coil, and back-telemetry would be achieved by selectively de-tuning the power coil
via a transistor, producing a phase-shift keyed signal in the coupled impedance response. ............................ 204
Figure 6.4 A prototype of the implantable electronics was implemented using a PCB connected to a secondary coil
composed of 34 gauge copper wire wrapped around a 3D printed mock implantable module. Wireless power
delivery via inductive coupling was tested and achieved. ................................................................................ 204
Figure 6.5 SolidWorks model of a candidate package for implantation in line with hydrocephalus shunts. ............ 204

xxi

List of Tables
Table 1.1 Select valves which are currently available in the US market. Many valves combine a ball-and-cone
differential pressure system with either anti-siphon or gravitational device to prevent overdrainage. Most
valves also contain a pumping reservoir, which can be used for shunt taps or to clear valve occlusion by
manually pumping (though this has been shown to have little value). . ............................................................. 16
Table 2.1 A comparison of several thermal flow sensors, both commercial and reported in literature. Work
described in this thesis is bolded. (Acronyms: PSG = polysilicon glass, SiN = silicon nitride). ....................... 41
Table 2.2 Temperature coefficient of resistance (TCR) of metals and semiconductors which have been used in
resistive temperature sensors, as well as several ionic solutions including human cerebrospinal fluid.33,87,90
© 2016 IEEE ...................................................................................................................................................... 51
Table 2.3 The calorimetric flow sensor compared with other EI flow sensors reported in literature2,58,113,114. The
sensor exhibits a 4× higher resolution at ultra-low flow velocities when operating in the calorimetric mode.
Resolution has been reported as twice the standard deviation (2σ) for all sensors. (ToF = Time of Flight) ©
2018 IEEE .......................................................................................................................................................... 90
Table 3.1 A comparison of several methods of data transmission within the body. Work described in this thesis is
bolded. .............................................................................................................................................................. 102
Table 3.2 The DC resistance, and the inductance and quality factor at 5 MHz, of 1, 5, and 16 turn microfabricated
thin-film coils. © 2017 Springer Nature. ......................................................................................................... 110
Table 4.1 This work compared to other proposed strain sensors. CNT = carbon nanotubes, EGaIn = eutectic
gallium-indium (liquid metal).. ........................................................................................................................ 158
- 1 -

HE ability to record physiological signals from inside the body without invasive surgery or
dangerous testing was a revolutionary development in medicine that has accelerated our
understanding of the human body and the ability to diagnose, manage, and cure diseases.
However, many physiological signals cannot be accurately measured using non-invasive imaging
techniques. These signals include hormones and biomarkers in blood, the neural signals which
make up cognition, reason, and perception, and cerebrospinal fluid dynamics. Even for diseases
which can be non-invasively monitored, several barriers exist that limit the usefulness of medical
imaging. Imaging studies can only be performed during doctor’s visits, which generally occur once
every six months or less. This limits a doctor’s ability to understand and manage chronic
conditions. Patients can experience very different physiological responses in a doctor’s office
compared to normal activity, a phenomenon colloquially known as “white coat syndrome”.
Measurements made in a doctor’s office also preclude disease monitoring during exercise or
strenuous activities. To improve medical care and advance our understanding of chronic diseases,
techniques are needed to remotely monitor patients outside of the clinic, to acquire data on a
disease state during periods of physical activity, to take measurements at time scales of minutes or
hours instead of months, and to record biosignals that are not currently detectable using non-
invasive medical imaging. This would give us unprecedented access to a patient’s disease state,
enabling the detection or prediction of adverse events before the patient has any symptoms and
leading to rapid, targeted treatment. It would also advance our scientific understanding of chronic
diseases, allow high-precision trialing of therapies, and may lead to cures for diseases which are
now considered life-long conditions. Continuous glucose monitoring in diabetes patients acts as
early validation of continuous biosignal monitoring outside the clinic; numerous studies have
I NTR O DUC T ION
T
- 2 -

shown that even self-monitoring of blood glucose by patients results in improved long-term
outcomes.
Since the invention of the transistor in 1947, the microelectronics field has exploded in size
and scope, enabling the extreme miniaturization of complex devices and systems. The self-
fulfilling prophecy of Moore’s Law has resulted in transistors with 7 nm or smaller features,
enabling the creating of millimeter-sized microcontrollers and solid-state sensors. Piggybacking
off silicon manufacturing techniques developed for integrated circuits, the development of
microelectromechanical systems (MEMS) has brought physical actuation and sensing to the solid-
state electronics world. Commercial MEMS devices include pressure sensors, accelerometers, and
gyroscopes; these MEMS sensors are made by etching mechanical structures out of a monolithic
silicon substrate. The combination of microelectronics and MEMS devices has enabled orientation
detection, tactile feedback on cell phones, and tire pressure measurement in cars. Miniaturized
MEMS sensors and ultra-low power microelectronics hold enormous potential for the medical
field as well. Potential devices would allow doctors to remotely monitor a patient’s condition,
catch disease progression before it becomes symptomatic, and intervene before complications
occur. These remote medical sensors could be as simple as a fall-detector for elderly patients or as
complicated as a flexible skin-based platform for monitoring multiple biomarkers in sweat. Recent
manufacturing advances have led to the development of polymer MEMS sensors, which can be
placed inside the body for long periods of time without corrosion or sensor drift. Polymer MEMS
technology could enable novel sensors for monitoring blood or CSF dynamics, biomarkers in the
blood, or neural signals.
One group of patients which stand to benefit from continuous remote monitoring of their
disease state is those with hydrocephalus. Hydrocephalus is an incurable, chronic disease
characterized by an accumulation of excess cerebrospinal fluid in the brain’s ventricles. The
disease affects 1 in every 500 children born in the US, and 1 million Americans live with
hydrocephalus today. If left untreated, patients will experience brain damage, coma, and eventual
death. Luckily, effective treatment involving drainage of excess fluid to the abdomen using an
implanted hydrocephalus shunt has existed since the 1960s. These permanently implanted shunts
have saved millions of lives since their introduction, but they fail at an alarming rate due to
occlusion, infection, mechanical breakage, or improper drainage rate. The tools used by physicians
to diagnose shunt failure are non-specific and inaccurate. The standard of care for shunt failure
diagnosis involves one or more CT scans and a shunt tap, in which a doctor inserts a needle into
the patient’s shunt to measure pressure. CT scans must be compared against past images to identify
fluid build-up, and shunt taps often give incorrect pressure measurements when the shunt is
clogged; often, physicians cannot truly diagnose shunt failure until the patients undergoes a shunt
resection surgery. Shunt failure is a progressive process, and patients will experience non-specific
symptoms such as headaches, irritability, and dizziness before progressing to more serious
symptoms such as coma and brain damage. The ability to remotely monitor shunt status, and to
diagnose or predict shunt failure before serious symptoms occur, would result in a huge
improvement in the treatment of hydrocephalus and in the quality of life of patients.
- 3 -

The following thesis describes work toward developing a wireless multi-sensor module for
remote monitoring of hydrocephalus shunts status. This work focuses on the development of novel
flow sensors and wireless catheter patency sensors for hydrocephalus shunts, as well as an ongoing
clinical study of a multi-sensor module in external ventricular drains. Work toward improving
hydrocephalus treatment has also inspired related projects in implantable sensors and biomedical
monitoring. These projects include the development of a passive, wireless glucose sensor, a novel
method of temperature measurement, the use of electrodeposited platinum-iridium coatings for
improving sensors, and polymer strain sensors based on kirigami patterns.
Chapter 1 describes hydrocephalus, including its history and the current standards of care.
Hydrocephalus has been identified since ancient times, but it is only in the past century that patients
were able to survive the disease. Patient survival is thanks to the hydrocephalus shunt, which
revolutionized hydrocephalus treatment after its introduction in the 1960s. This chapter describes
the pathophysiology of hydrocephalus and the early development of the hydrocephalus shunt. It
then presents an overview of modern advances in hydrocephalus shunt technology, including
gravimetric devices and magnetically adjustable valves. Shunt failure is introduced, and the causes
and current state of shunt failure diagnosis is discussed. Finally, the chapter describes physicians’
desire for a “smart shunt” involving sensors and self-cleaning technology, and the technological
issues which have thus far prevented a smart shunt from entering the market.
Chapter 2 describes a novel flow sensor designed for chronic use in cerebrospinal fluid.
Commercial MEMS flow sensors are composed of a microfabricated heater and multiple
temperature sensors, and measure flow velocity via heat transfer through a fluid. However, these
sensors utilize high overheat temperatures and are fabricated on silicon or other semiconductor
materials, preventing their use in the body. This chapter first reports on a novel method of
temperature measurement using the high-frequency electrochemical impedance between
electrodes exposed to physiological fluid. The conductivity of a fluid is highly sensitive to
temperature, so temperature can be transduced using an electrical signal through the fluid, with
sensitivities an order of magnitude higher than the current state of the art. This temperature sensing
technique was used to design a thermal flow sensor for physiological fluids that is highly sensitive
to flow velocity in the range expected in hydrocephalus shunts while only using a 1°C overheat
temperature. A second-generation flow sensor is also described, which uses a slightly different
measurement technique to improve sensitivity at ultra-low flow rates. This chapter closes with
three projects investigating the practicality of flow sensor operation in vivo: sensor improvements
after coating with electrodeposited platinum-iridium, an investigation of the major expected failure
mode (delamination) during in vivo operation, and a long-term drift study using human
cerebrospinal fluid.
Chapter 3 describes a novel method for passive wireless transduction of electrochemical
impedance. A high-frequency electric signal applied to a coil of wire generates an oscillating
magnetic field, allowing energy transfer to a nearby coil. If a load is placed on the second coil,
energy will be drawn from the shared magnetic field, changing the impedance of the primary,
powered coil. If the load on the secondary coil is composed of a pair of electrodes immersed in
- 4 -

physiological fluid, the solution resistance between those electrodes can be transduced by
measuring the reflected impedance of the primary coil. This chapter describes the theory and
simulation of this transduction method, as well as the fabrication and testing of thin-film
biocompatible coils terminating in exposed gold electrodes. Both theory and experiment revealed
that the solution resistance between the electrodes could be wirelessly transduced using reflected
impedance. The effects of changing coil orientation and separation were also investigated. To
apply the reflected impedance technique to hydrocephalus treatment, a patency sensor was
developed which wirelessly measured the resistance between an electrode inside and an electrode
outside a shunt, enabling direct transduction of shunt blockage. In addition to hydrocephalus
treatment, the use of reflected impedance for biomarker detection is explored, and preliminary data
on glucose sensing is presented.
Chapter 4 describes a kirigami strain sensor that was inspired by colleague’s work on
Parylene C microfabrication. Parylene C is a flexible, biocompatible polymer which can be used
to create chronically implantable sensors. However, Parylene C does not stretch significantly when
strain is applied. Kirigami refers to the ancient art of cutting paper to produce complex three-
dimensional patterns. One kirigami technique which has recently been explored in the MEMS
community is the use of offset rows of slits to allow normally stiff planar thin films to stretch. This
chapter describes the development of a stretchable Parylene C device using a kirigami slit pattern
created via oxygen plasma etching. Kirigami slits enabled devices to stretch to over four times
their original size before mechanical failure occurred. Electrical traces on a Parylene C kirigami
device were tested during stretching, and multiple electrical parameters were found to change with
strain, suggesting the development of a novel biocompatible strain sensor.
Chapter 5 describes a clinical study of a multi-sensor module in pediatric patients with
external ventricular drains, or EVDs. EVDs are used in acute cases of hydrocephalus to drain
excess fluid to an external collection bag. Integrating sensors into an EVD allows for sensor
validation in human cerebrospinal fluid under real-world drainage conditions, without causing
significant risk to the patient. For this study, the flow sensor described in Chapter 2 was combined
with a pressure sensors and patency sensor developed by colleagues into a multi-sensor module.
This module was paired with an electronic datalogger designed to record sensor measurements for
up to 14 days. This chapter describes the design of the multi-sensor module and of the first-
generation datalogger electronics. It then describes the results of patient enrollment, and the
disappointing failure modes that were only uncovered after patient enrollment began. The chapter
then describes work towards fixing these failure modes, the most important being the development
of a second-generation datalogger with significantly higher sensitivity. The study is currently
ongoing, and results using the new sensors and electronics are expected soon.
The final chapter describes milestones and future plans for developing a wireless,
chronically implantable multi-sensor module for hydrocephalus shunts. Preliminary work on the
wireless electronics and packaging for implantation and integration with hydrocephalus shunts is
described.

- 5 -

YDROCEPHALUS is an incurable chronic disease which affects over one million people in the
US. Characterized by an accumulation of excess cerebrospinal fluid (CSF) in the brain’s
ventricles, hydrocephalus is most often a congenital condition and will lead to brain damage
and death if not promptly treated. No pharmaceutical treatment exists today to treat hydrocephalus;
surgical treatment involves permanent implantation of a hydrocephalus shunt, which redirects fluid
to the abdomen where it can be reabsorbed. Hydrocephalus shunts have been widely used for the
past 60 years and have saved millions of lives, but shunt failure is extremely common (>40%
failure rate within one year
1
) and difficult to diagnose. Embedding wireless sensors within a shunt
would enable non-invasive shunt failure diagnosis and prediction. A so-called “smart shunt” is
highly desired by both patients and physicians; although some efforts have been made toward the
development of such a shunt in recent years, limitations in sensor technology have prevented any
smart shunt from entering the market.

Chapter 1
HYDROCEPHALUS AND THE NEED FOR
CHRONICALLY IMPLANTABLE SENSORS
H
- 6 -
What is Hydrocephalus?
1.1.1 Pathology of hydrocephalus

Figure 1.1 (Left) A child with hydrocephalus
2
. (Right) Computerized tomography (CT) scan of a
hydrocephalus patient showing cranial tissue compression and ventricular enlargement
3
.
Hydrocephalus, which in Greek means “water on the brain”, refers to an accumulation of
excess CSF in the brain’s ventricles
4,5
. The ventricles are fluid-filled chambers in the center of the
brain which both produce CSF and direct its flow to the subarachnoid space, where it can be
reabsorbed. The ventricles provide stabilization and protect cortical tissue from the effects of
shaking or impacts to the skull, and the flow of CSF through the ventricles flushes metabolic waste
out of the brain. A healthy ventricular system will produce and reabsorb 500 mL of CSF daily.
Even small deviations in the size of the ventricles have been associated with neurodegenerative
diseases such as Alzheimer’s and schizophrenia.
In hydrocephalus, the flow of CSF is disrupted and excess CSF builds up in the ventricles.
This increases intracranial pressure (ICP) and leads to ventricular enlargement, which compresses
cortical tissue. Elevated ICP and compressed cortical tissue cause serious symptoms, which vary
depending on the age of the patient. Neonates and newborns experience characteristic cranial
enlargement, since the cranial plates do not fuse until 4–26 months
6
. Other symptoms include eyes
that constantly gaze downward (a consequence of increased pressure and skull deformation),
excessive irritability, inability to sleep, vomiting, and persistent seizures. In older children with
fused cranial plates, hydrocephalus manifests more insidiously as changes in personality, memory,
or intellectual capability; constant sleepiness; headaches; mood swings; loss of muscular
coordination; blurred vision; and urinary incontinence. If not treated, most hydrocephalus patients
will die, and those who survive will have numerous intellectual, physical, and neurological
disabilities. Despite relative obscurity, hydrocephalus is a common disease: for every 1000
- 7 -
children born, 1 or 2 will acquire congenital hydrocephalus, and over 1 million people in the US
currently live with the condition
4
.

1.1.2 Causes of hydrocephalus
Hydrocephalus can be either congenital or acquired and has a variety of causes, many of
which are not fully understood. Patients with hydrocephalus can be classified into three broad
categories: non-communicating hydrocephalus, communicating hydrocephalus, or normal
pressure hydrocephalus
5,7
.
Non-communicating hydrocephalus is caused by a physical obstruction of the path that
cerebrospinal fluid must take from the choroid plexus where CSF is produced to the subarachnoid
space where it is reabsorbed. This blockage could occur in the foramina of Monro, which connects
the lateral ventricles to the third ventricle; the aqueduct of Sylvius, a narrow channel between the
third and fourth ventricles; the fourth ventricle itself; or the lateral aperture which connects the
fourth ventricle to the subarachnoid space (Fig. 2). Congenital non-communicating hydrocephalus
can result from infection, hemorrhaging, tumors, brain structure deformation, or pressure on a
communicating pathway as a complication of arachnoid cysts, Dandy-Walker syndrome, a neural
tube defect, or a Chiari malformation. Non-communicating hydrocephalus can also be acquired
later in life due to meningitis which leads to swelling and stenosis of a communicating pathway,
blockage by a brain tumor, or traumatic brain injury which causes hemorrhaging and blocks CSF
flow.

Figure 1.2 Diagram showing the ventricles of the brain. After being produced in the choroid plexus, CSF
flows first into the lateral ventricles, then through the foramina of Monro to the third ventricle, then through
- 8 -
the aqueduct of Sylvius to the fourth ventricle. The fourth ventricle contains openings which allow CSF to
drain into the subarachnoid space, where it is reabsorbed. A blockage in any of these passageways, whether
congenital or acquired, can lead to hydrocephalus.
Communicating hydrocephalus does not involve a pathway obstruction, and instead results
from and inability to properly reabsorb CSF. The exact cause of communicating hydrocephalus is
not known for certain, but may involve impairment or destruction of arachnoidal granulations,
which allow CSF to diffuse into the bloodstream. These granulations may be absent congenitally
due to genetic or developmental errors, or they may be destroyed or scarred over due to
intraventricular hemorrhage, meningitis, or other inflammatory events.
Both communicating and non-communicating hydrocephalus are generally seen in
newborns or younger children. Normal pressure hydrocephalus (NPH), on the other hand,
generally affects older patients, and has a peak onset between 60 and 70 years of age. As the name
suggests, normal pressure hydrocephalus is characterized by an increase in ventricular volume
with little or no corresponding increase in ICP. Symptoms of NPH are limited compared to
conventional hydrocephalus and generally manifest as Hakim’s triad of urinary incontinence, gait
disturbance, and dementia. Due to the late age of onset and nonspecificity of symptoms, patients
with NPH are often misdiagnosed with Parkinson’s disease, Alzheimer’s disease, or dementia. The
exact cause of NPH is not known but it is thought to be associated with age-related demyelination
and degeneration of white matter tissue
9,10
.

History of Hydrocephalus Treatment

Figure 1.3 (Left) Diagram of the brain’s ventricles by Leonardo da Vinci, 1510. (Right) Medical pamphlet
describing hydrocephalus, Nuremberg, 1556.
5

- 9 -
Due to the obvious nature of pediatric hydrocephalus symptoms (e.g. cranial enlargement),
hydrocephalus has been recognized numerous times in ancient literature. The earliest existing
description of the disease dates to 466-377 BC by Hippocrates, who coined the word
“hydrocephalus”; he described it as an accumulation of water on the outside of the brain, instead
of in the ventricles
5
. Claudius Galen of Pergamon (130-200 AD) furthered the study of
hydrocephalus by animal dissection and was the first to theorize that the choroid plexus was the
source of CSF production. Abulkassim Al Zahrawi (963-1013 AD) was the first to record an
attempted surgical therapy for hydrocephalus. To treat hydrocephalus, Al Zaharawi says, “we must
open the middle of the skull in three places, make the liquid flow out, then close the wound and
tighten the skull with a bandage.” The success rate of this procedure was not recorded.
Knowledge of the brain’s anatomy advanced through the Renaissance, and in 1555,
Andreas Vesalius identified the ventricles as the actual location of CSF accumulation in
hydrocephalus. Once the ventricles were identified as the location of abnormal CSF accumulation,
puncturing the ventricles to allow excess fluid to drain into the subarachnoid space would
constitute the logical next step in treatment. However, the first documented ventricular puncture
to treat hydrocephalus did not appear until 1744, by Le Cat. Ventricular puncture often had no
long-term benefit, and due to the lack of sterile surgical tools, patients would often die from
infection. However, once sterile equipment and procedures were introduced, ventricular puncture
became a widely used method for acutely draining CSF until the 1970s, when it was supplanted
by the use of external ventricular drains (EVDs)
11
. For treating chronic hydrocephalus, a number
of ad hoc solutions were attempted during the late 19
th
and early 20
th
centuries. In 1893, Mikulicz
implanted a glass wool wick to transport CSF from the ventricles of a 6-month old infant into the
subarachnoid space, successfully alleviating the child’s symptoms for at least two years
5
. The first
ventriculoperitoneal (VP) shunt, which drained fluid from the ventricles to the abdominal cavity,
was attempted in 1905, though the patient did not survive the procedure. VP shunting was only
sporadically attempted through the 1950s, since the procedure had an almost 50% mortality rate.
Other surgical procedures were less dangerous, though not necessarily as effective at treating
hydrocephalus. Destruction of the choroid plexus, which produces CSF, was introduced in the
1930s and was brought down to a 1% mortality rate in the 1960s, thought this procedure is seldom
used now since patients almost always redevelop hydrocephalus and require shunting. Third
ventriculostomy, in which the floor of the third ventricle is pierced to allow CSF to bypass
obstructions, was first attempted in 1923, became a promising therapy with the invention of
microsurgical techniques in the 1970s, and is still used for some hydrocephalus patients today.
Many attempts to construct a shunt from the ventricles directly into the bloodstream were made in
the early 1900s, though none of these were widely adopted due to the risk of blood traveling back
into the ventricles.
The 1950s saw a revolution in hydrocephalus treatment due to the development of silicone
rubber and the invention of implantable one-way valves. The first valves designed for
hydrocephalus shunts were developed by Nulsen and Spitz in 1949
12
, Pudenz and Heyer in 1955
13

(with improvement from Schulte in 1958), and Holter in 1956
5
. Silicone rubber, which was first
- 10 -
used in during World War II as a thermally resistant material in airplane construction, was first
used to make catheters for hydrocephalus shunts in 1956 and was found to be exceptionally
biocompatible and resistant to mechanical stress
14
. These advances lowered mortality rates and
increased effectiveness such that by 1961, VP shunting became the standard of care for
hydrocephalus treatment
15,16
.

Figure 1.4 An early VP shunt made using Silastic. This shunt contains a “flushing chamber” roughly
equivalent to pumping reservoirs in current valves and features a slit valve at the distal end
17
. © 1967 AANS

- 11 -
Current Shunt Technology

Figure 1.5 Modern hydrocephalus shunts consist of a proximal catheter implanted in the ventricle, a valve
which opens in response to a pressure gradient between the brain and the rest of the body, and a distal
catheter which drains fluid into the peritoneal, pleural, or atrial cavities.
Today, VP shunts are used in 98% of pediatric hydrocephalus patients
18
and around 80%
of adult patients
19,20
. Shunts generally consist of a proximal catheter placed in the brain’s
ventricles, a valve placed under the skin on top of the skull, and a distal catheter which routes
through the neck into the peritoneal cavity.
1.3.1 Proximal and Distal Catheters

Figure 1.6 A proximal catheter for hydrocephalus shunts, containing 16 holes of 500 µm diameter.
Proximal catheters are made of silicone rubber and are infused with barium as a contrast agent for x-ray
and CT imaging.
1 cm
- 12 -
Proximal catheters are made of silicone rubber, with rows of small holes (generally 16 total
and 500 µm in diameter) at the ventricular end. Some proximal catheters are impregnated with
various antibiotics to prevent infection
21,22
. Distal catheters are also made of silicone and are
usually simple tubes with a single outlet at the drainage site. Proximal catheters are the most
common source of infection and occlusion, and many researchers have attempted to redesign this
catheter to reduce or eliminate various risk factors. Proximal catheters impregnated with antibiotics
have seen some commercial success. These catheters contain either silver particles or a
combination of clindamycin and rifampin. Shunt infection usually occurs due to contamination
with Staphylococci species present on the patient’s skin, so catheters which can kill these species
on contact were considered a potential solution. However, the efficacy of antibiotic-impregnated
catheters is up for debate, with some studies showing a mild improvement in infection rates
21
and
others showing no difference whatsoever
22
.
Since shunt occlusion occurs most often in the proximal catheter, a few mechanical
changes to the proximal catheter have been proposed to reduce or eliminate blockage. Catheters
generally have 16 holes of 500 µm diameter each, a number and size which was chosen arbitrarily
according to the drill sizes available when these catheters were first introduced
24
. There has been
very little research into how changing hole size and position might affect long-term catheter failure
or occlusion. At least one group has found evidence that reducing the total number of holes may
lead to less occlusion due to more even flow through the remaining holes
25
.

Figure 1.7 Flow through the proximal catheter occurs primarily the top hole or hole pair
26
. © 2016 Springer
Nature
Another mechanical improvement tested on proximal catheters involves adding flanges,
which are silicone rings situated between each hole. Flanged catheters were introduced in 1968
and promised to decrease occlusion by preventing the choroid plexus from entering inlet holes, as
well as by protecting the holes from debris during insertion
27
. However, retrospective studies have
shown that flanged catheters are more likely to become occluded than non-flanged catheters, and
- 13 -
present additional danger to patients during shunt resection surgeries by having additional
attachment points for tissue damage
28
.

Figure 1.8 A flanged proximal catheter, designed by Portnoy in 1971
29
. © 1976 Springer Nature
1.3.2 Valves
Valves are the most complex part of a hydrocephalus shunt and are available in a variety
of different designs. The most common type of valve is a differential pressure valve, which only
opens once a certain pressure differential between the ventricles and the drainage end is reached.
The intent is to prevent both overdrainage and backflow.
There are three main designs for differential pressure valves. Slit valves are the oldest and
simplest valve designs, consisting only of a slit cut into a silicone membrane which requires a
certain pressure to open. Opening pressure is highly dependent on the silicone’s stiffness, so these
valves have a large amount of manufacturing variation and drift substantially as the silicone ages.

Figure 1.9 Diagram of a distal slit valve
30
. Distal slit valves were the first type of valves put in use, but
problems with clogging led to them being quickly phased out.
2 mm
- 14 -
Diaphragm valves operate via a flexible diaphragm that is normally held shut, but which
can be opened once a certain pressure is applied. These valves are known to wear out relatively
quickly due to their reliance on a moving, flexible diaphragm to control the cracking pressure. The
most common type of valve in hydrocephalus shunts is the ball in cone valve, which uses a spring
to hold a ball, usually made of synthetic ruby, tightly against a cone-shaped section of the flow
channel. The spring’s design determines what force is necessary for the ball to move back and
fluid to flow, and there is no possibility of backflow. These valves are less prone to manufacturing
differences and aging than the other types of valves.

Figure 1.10 The operation of a ball and cone valve. Cracking pressure is defined by the spring’s stiffness
30
.
Valves normally come in one of five pressure settings: very low, low, medium, high, and
very high. Despite a long history there are no standards that define what these pressure levels
actually mean, and the definition of “medium” can vary widely between manufacturers. A
relatively recent development in hydrocephalus treatment is the adjustable valve, with an opening
pressure that can be changed after implantation. Valves which could be adjusted via screws or
other percutaneous mechanisms have been available since 1969
31,32
, but in 1984, the first valves
were tested which could be adjusted via an external magnet, and in 1989 these valves, made by
Cordis, were introduced to the U.S. market
33
. Today nearly every shunt manufacturer offers an
adjustable valve
34
. It is worth noting that the magnetic components cause artifacts during magnetic
resonance imaging (MRI) scans, and must be reprogrammed after any exposure to large magnetic
fields
35
. A table of currently used valves is shown below.
Differential pressure valves work well to prevent backflow and overdrainage in many
cases, but intracranial pressure is inherently variable and activities such as coughing or straining
one’s muscles may increase ICP enough to push extra CSF through the shunt, resulting in
overdrainage. Furthermore, when a patient is standing, the gravitational force exerted on the
column of CSF in the distal catheter can lead to an applied pressure on one side of the valve and
can cause excessive CSF to siphon out of the ventricles. To address these problems, several
- 15 -
additional shunt features have been developed. The first anti-siphoning valves used the weight of
the hydrostatic column in the distal catheter to close off the shunt when a patient was standing
36
,
but these devices relied on a flexible membrane near the skin which could be occluded with
pressure and varied with the ambient air pressure
37
. Gravitational anti-siphoning devices improved
upon this concept, and contain one or more metal balls in the flow channel which are pressed down
and block CSF flow when the patient is standing, but fall to the side when the patient is laying
down, allowing normal drainage to occur
38
. Another type of valve which acts to prevent siphoning
is the flow control valve, which increases fluidic resistance based on the flow rate
39
.

- 16 -
Table 1.1 Select valves which are currently available in the US market. Many valves combine a ball-and-
cone differential pressure system with either anti-siphon or gravitational device to prevent overdrainage.
Most valves also contain a pumping reservoir, which can be used for shunt taps or to clear valve occlusion
by manually pumping (though this has been shown to have little value).

Picture Company Name Valve Type Features

Medtronic Strata II
40

Ball-in-cone +
Anti-siphon device
Adjustable,
Pumping reservoir,
Diaphragm-based
overdrainage protection

Aesculap/
Miethke
ProGAV 2.0
41

Ball-in-cone +
gravitational valve
Adjustable,
Gravitational
overdrainage protection

Integra OSV II
42
Diaphragm
Three-stage operation,
Flow control,
Pumping reservoir

Spielberg HPBio Sphera Duo
43
Dual Ball-in-cone
Pumping reservoir,
Dual pressure setting

Codman HAKIM
44
Ball-in-cone
Adjustable,
Pumping reservoir

Sophysa Polaris
45
Ball-in-cone Adjustable

- 17 -
1.3.3 Valve Design Improvements
Since the introduction of hydrocephalus shunts, the valve has been the target of most efforts
at improving hydrocephalus outcomes. The first valves employed to treat hydrocephalus were
known as “distal slit valves”, and simply consisted of a sealed distal catheter with a small slit cut
into its distal end. Due to the cylindrical shape of the catheter and the flexibility of silicone, less
pressure is required for fluid to flow out than in to the shunt. The distal slit valve was therefore the
first differential pressure (DP) valve, which opened or closed in response to the difference in
pressure between the two ends (analogous to a diode with a set threshold voltage). Despite their
simplicity, distal slit valves were occlusion-prone and quickly replaced by separate shunt
components placed between the proximal and distal catheters. The majority of these new valves
(and the majority of valves used today) are DP valves which operate through the movement of
either a flexible diaphragm
40,42
and/or a ball and spring
40,41,43-45
. Additionally, modern valves often
contain a reservoir with a self-healing silicone membrane, which can be used to perform shunt taps
or to sample CSF for infection screening.
DP valves present new and unique problems, such as overdrainage due to siphoning. When
a patient is standing, the weight of the hydrostatic column in the distal catheter can create a
negative pressure on the distal side of the valve, leading to excess fluid draining out of the
ventricles. Two different devices have been developed which attempt to solve this problem. The
first class of devices, known as flow-regulating valves, operate by gradually closing (and
increasing fluidic resistance) based on the difference in pressure between the distal catheter and
atmospheric pressure. Flow regulating valves rely on a flexible diaphragm exposed to atmospheric
pressure through the skin, and are susceptible to additional tissue growth or physical pressure on
the device. However, non-randomized studies seem to suggest that flow-controlled valves are less
prone to mechanical complications than DP valves, at least within the first year
46
.
- 18 -

Figure 1.11 Diagram of a flow-regulating (or anti-siphoning) valve, which uses the difference between
atmospheric pressure and the negative pressure due to flow through the distal catheter to progressively
increase fluidic resistance and close the valve when the flow rate is too high
47
.
The second class of devices to tackle the problem of overdrainage are known as
gravitational or gravimetric valves. These devices contain heavy metal balls that press down and
close the valve when the patient is standing, but allow fluid to flow freely when the patient is lying
down. Gravitational valves have been linked with lower rates of overdrainage
48
, though no
controlled double-blind studies exist. Both flow-regulating and gravimetric devices can be
packaged as part of a DP valve
40,41
or can be purchased as separate components which a surgeon
may attach to the shunt as needed.
- 19 -

Figure 1.12 The ProGAV gravitational valve from Miethke contains both a standard ball-in-cone valve and
a gravitational assist device
41
. When the patient is lying down, the fluid’s flow is not impeded, but when
the patient is standing vertically steel balls fall and block flow through the channel, preventing overdrainage
due to self-siphoning.
Another problem with DP valves is that optimal valve pressure is different for each patient.
DP valves commonly come in five different pressure settings (Very Low, Low, Medium, High,
and Very High), though no standard exists between manufacturers over what these settings should
be. Surgeons must guess what type of valve each patient needs when a shunt is first implanted, a
process which is complicated by the fact that a patient’s needs may change as the patient ages or
disease symptoms are reduced. Adjustable valves have been developed to solve this problem, and
to prevent surgeries to treat under- or overdrainage. The first adjustable valves had a screw which
could be turned to adjust the set pressure along a spectrum (a necessarily invasive procedure, but
much easier on the patient than replacing the valve entirely). Newer valves have been developed
which can be adjusted magnetically through the skin, saving the patient any sort of invasive
procedure. These devices have been a huge benefit in hydrocephalus treatment as they give
physicians a tool to treat some causes of shunt failure in an outpatient setting, instead of in the
operating room. However, the lack of convenient feedback and the difficulty of accessing relevant
physiological parameters, such as ICP or shunt flow rate, means that it takes doctors weeks or
months to know if an adjustment had any effect on the patient’s disease state. Furthermore,
magnetically adjustable valves can be scrambled during MRI exams (or, in at least one case, while
playing with an iPad
49
), meaning that a patient’s valve must be readjusted after any imaging study.
- 20 -
Shunt Failure
Shunting remains the gold standard for treating hydrocephalus and has saved the lives of
countless patients who would otherwise have no treatment option. However, any implant,
especially one designed to stay in the body for years at a time, has the potential for failure, and
hydrocephalus shunts have alarmingly high failure rates. Retrospective studies show that VP
shunts fail at a rate of 40-50% within the first year, and that number rises by 3-5% each year until
it levels off at around 80% after 12 years
1,18,20,28,50-54
. Shunt failure can have a number of causes;
the most common being occlusion or blockage of part of the shunt, infection due to the implant,
and mechanical failure or separation of one of the shunt components.
Shunt occlusion is the most common failure mode, accounting for 30-50% of all shunt
failures
28,50,52
. Studies of shunt failures show that most blockages occur in the proximal catheter
(55%), but blockage can also occur in the distal catheter (25%) or in the valve or other accessory
(6.5%)
52
. Occlusion of the proximal catheter may result from ingrowth of the choroid plexus or
protein buildup and stenosis of the catheter lumen. Occlusion of the distal catheter is most often
caused by kinks or overly tight sutures, which narrow the flow channel and instigate protein
buildup
55
. Distal catheter occlusion can also occur due to infection or abdominal detritus. Valve
occlusion occurs from particulate debris or blood accumulating within the valve mechanism.
Repeated shunt taps, which cause material wear on the valve, may accelerate valve occlusion.
While a systematic study of valve blockage has not been performed, evidence suggests that there
is very little difference in occlusion rates between valve types (except for old-style distal slit
valves, which are not generally used due to their higher rate of occlusion in the distal catheter)
18,53
.
Shunt obstruction can occur any time after insertion, and patients with shunt obstruction will
generally experience a return of hydrocephalus symptoms such as nausea, vomiting, irritability,
and headaches, which at first may be hard to distinguish from other common illnesses.
Infection is the second most common shunt failure mode, occurring in 8% of patients.
Infections usually occur within six months of a shunt insertion surgery, pointing to bacterial
contamination on the shunt during implantation as the most probable cause
56
.
Mechanical failure is the third most common cause of shunt failure
28,52
. Distal catheters
can fracture due to a patient’s growth or movement. This can be accelerated by scar tissue around
the catheter, preventing it from sliding normally. After a fracture, CSF may still drain, but patients
often present with a mild increase in ICP and hematomas at the site of fracture. Distal catheter
fracture is a serious condition, but not an emergency situation, and fractures are often not
discovered until a patient goes for a routine doctor’s visit. Another form of mechanical failure
involves disconnection of either the proximal or distal catheter from the valve, which usually
occurs because of improper suturing during the insertion procedure. This is easily diagnosed as
fluid buildup will be evident around the break. The proximal catheter can also undergo migration,
where the catheter shifts to a position devoid of CSF either due to the patient’s growth or due to
cranial tissue expansion after pressure is reduced.
While not necessarily a failure of the shunt itself, overdrainage is one of the most common
problems facing physicians and hydrocephalus patients
55
. Overdrainage refers to any scenario
- 21 -
where the shunt is removing more fluid than necessary from the patient’s ventricles. Acutely,
severe overdrainage immediately following shunt insertion may cause cranial tissue to collapse
and lead to an accumulation of CSF around the outside of the brain, known as extra-axial fluid
collection. This is most common in older patients who are shunted and occurs in 3-4% of
hydrocephalus patients. Treatment involves draining excess fluid via a subdural catheter as well
as adjusting the valve or inserting a second shunt into the extra-axial space. Chronic overdrainage
can result in slit ventricle syndrome, where the brain’s ventricles shrink down to barely perceptible
slits. Slit ventricle syndrome usually manifests years after the original shunt implantation and can
result in headaches, nausea, and vomiting, although symptoms may be intermittent or may vary
with the patient’s position (standing vs lying down). Estimates of the proportion of shunted patients
with slit ventricle syndrome vary widely, from 1% after 6 years
57
to 10% after 16 years
58
, with one
study
59
finding slit ventricle syndrome in 75 out of 141 patients tested. While serious slit ventricle
syndrome usually requires shunt resection, even mild overdrainage can cause headaches and
dizziness and result in substantial treatment cost and patient inconvenience.

1.4.1 Treating shunt failure
Once a shunt failure has been diagnosed and the cause of the failure identified, the method
of treatment depends on the cause of failure. For shunt obstruction, the only option is a surgical
shunt revision
52
. During a revision, a surgeon has the option of removing only the occluded shunt
component, but due to difficulties in identifying the site of the obstruction and the fact that older
shunt components are more prone to failure, many surgeons replace the entire shunt during a
revision. Depending on the cause of the obstruction, surgeons may need to place the new shunt in
a different location, either intercranially or distally. In rare occasions, an endoscopic third
ventriculostomy (where a hole is made in the base of the third ventricle) is attempted instead of a
full shunt revision
60
.
Mechanical failure of a shunt due to fracture or disconnect also requires shunt revision, but
most revisions for mechanical failure can be done on an elective schedule. Fractures in the distal
catheter are treated by replacing only the distal catheter, though complications may arise if pieces
of catheter migrate throughout the abdominal cavity. Disconnection, on the other hand, can be
treated by simply opening up the skin near the valve and reconnecting suture joints.
Out of all shunt failure modes, shunt infections are the most expensive, dangerous, and
time-consuming to treat
18,61
. Proper treatment of an infection involves removing the infected shunt,
placing an external ventricular drain, and giving the patient intravenous antibiotics. CSF is tested
regularly for the presence of pathogens, and once CSF sterility is achieved, a second surgery is
performed to remove the external drain and implant a new shunt
56
.

1.4.2 Diagnosing shunt failure
For any shunt failure mode, treatment involves surgically replacing or correcting the failed
component. However, identifying which component failed, what the failure mode is, or even
whether shunt failure is in fact the cause of a patient’s symptoms is difficult and often leads to
- 22 -
false positives or negatives. For infants and young children, an increase in intracranial pressure
(ICP) causes cranial enlargement, and the size of the skull serves as an obvious, if approximate,
measure of the progression of the disease state or the effectiveness of treatment. However, for
older patients, early stages of shunt failure trigger nonspecific symptoms such as headaches,
dizziness, irritability, and vomiting, which can be erroneously dismissed as the symptoms of a
common cold or flu. A patient who constantly worries about shunt failure will be more likely to
consider any symptom a sign of a failed shunt, and costly and time-consuming medical tests are
required to implicate or exonerate the shunt.
Once a failure is suspected, confirmation and localization is difficult. The first step in
failure diagnosis is to take a shunt series, which involves stitching together multiple x-rays into a
radiographic image of the full shunt
52
. Shunt series are useful tools which reveal disconnects and
some shunt fractures, and are often the only diagnostic tool needed if the shunt failure is
mechanical. However, a shunt series will not reveal obstruction or infection.
Once the cranial plates fuse (at around 36 months) the condition of a hydrocephalus patient
is much more difficult to measure
62
. Once the brain is fully encased by the skull, ICP cannot be
measured non-invasively like blood pressure, and gaining access to the brain’s ventricles involves
drilling a hole through the skull, destroying a portion of cranial tissue, and bypassing the blood-
brain barrier, all of which can lead to complications. Even imaging the brain is difficult since
ultrasound cannot penetrate the skull; the only non-invasive options for imaging the brain of a
hydrocephalus patient are CT and MRI
63
.

Figure 1.13 A CT image of the brain of a shunted hydrocephalus patient. The ventricles are somewhat
enlarged and the shunt’s proximal catheter can be seen inserted into the left ventricle
64
.
CT or MRI studies are the first diagnostic tool used in hydrocephalus patients to assess
symptoms, with MRI being preferred to limit lifetime radiation exposure
65
. By imaging the brain’s
ventricles and comparing the results to previous imaging studies, doctors can determine if the
ventricles are expanding or, for shunted patients, if the proximal catheter is properly placed.
- 23 -
However, an increase in ICP can sometimes cause symptoms without significant ventricular
expansion
66
, and although the position of the proximal catheter can be determined, imaging cannot
assess shunt blockage, malfunction, or overdrainage
67
. To more directly assess shunt status, studies
are occasionally performed where a radiographic tracer is injected into the brain’s ventricles and
tracked via multiple CT scans to determine if cerebrospinal fluid (CSF) is properly draining to the
abdomen
68
. However, even in a properly operating shunt, flow can be slow or absent depending
on the state of the patient, and injecting a tracer into the brain’s ventricles is an invasive procedure
which carries the similar dangers as direct ICP measurement in addition to the increased risk from
radiation exposure.
The other standard of care diagnostic tool for hydrocephalus patients is a shunt tap
69
. Shunt
taps involve inserting a needle into the shunt’s reservoir and using the height of the resulting fluid
column to determine ICP. The procedure is invasive and can provide a possible vector for
infection, though it is less invasive than directly tapping the ventricles
70
. A shunt tap also requires
a neurologist or neurosurgeon who is specifically trained for the procedure, and only provides a
single point measurement; since ICP varies with time of day, patient condition, and other factors
that are not fully understood, a single ICP measurement may understate the severity of
hydrocephalus in a patient or may indicate that a patient has elevated ICP when their shunt is still
intact
71
.
The only reliable method of diagnosing shunt obstruction is a procedure performed during
resection surgery, where a surgeon separates the suture between the proximal catheter and valve
and attempts to drain fluid directly out of the proximal catheter, then injects a column of saline
directly into the valve and measures the speed at which it drains through the distal catheter into
the abdomen. For obvious reasons, this method is not the first choice for diagnosing shunt failure,
but in some cases, the results may drive changes to which shunt component is replaced or even
elimination of the resection altogether.

- 24 -
Smart Shunts

Figure 1.14 A conceptual diagram of a smart shunt system, which would relay physiological and diagnostic
information wirelessly to a physician. A more advanced smart shunt could also include closed-loop
feedback which adjusted the valve position in response to changes in pressure in the brain.
Since the introduction of the shunt to hydrocephalus treatment, many devices have been
developed which promise to improve outcomes, remove complications, and provide a better
quality of life for hydrocephalus patients. Some of these have been at least partially successful.
However, each patient is different and can have unexpected responses to treatment. On top of this,
the body’s regulation of CSF is still poorly understood and hydrocephalus can have a wide variety
of causes and an almost infinite amount of variations. Shunt failure and complications are still far
too common, and shunted patients must live with the constant worry that their shunt could fail at
any time. Most shunted patients live close to a large medical center and cannot travel for long
periods of time, for fear they will be away from their neurosurgeon when a complication occurs
72
.
Patients also tend to stay with their initial pediatric neurosurgeon their whole life, since no other
doctor is familiar with the details of a particular patient’s disease. And patients who get a cold, flu,
or other common disease which shares symptoms with shunt failure must undergo numerous
imaging studies and invasive procedures before shunt failure is ruled out. On the physician’s side,
costly and time-consuming diagnostic studies are necessary to determine the nature of a patient’s
symptoms, and any change in treatment will take weeks or months to have a noticeable effect.
To address the many issues surrounding current shunt technology, the concept of a “smart
shunt” has been proposed
54
. A smart shunt would include wireless sensors to measure both the
patient’s physiological state and the state of the shunt, eliminating the need for shunt taps or
imaging studies. These sensors could be paired with a mechatronic valve to dynamically change
drainage rate based on the real-time pressure in the brain. Some groups have also proposed a self-
- 25 -
cleaning system which could be automatically actuated in response to clogging or protein buildup.
A successful smart shunt would eliminate overdrainage, bring shunt failure rates down to
acceptable levels (doctors have targeted a reduction to 5% from the current first-year failure rate
of ~40%
51
), and give patients peace of mind. The introduction of real-time wireless shunt
monitoring, even without a closed-loop valve or self-cleaning system, would do wonders for
patients and doctors by allowing quick, non-invasive diagnosis of shunt issues. Wireless sensors
would also lead to a greater understanding of hydrocephalus, inform more comprehensive
treatments.
Unfortunately, despite constant pressure from the medical establishment and a clear market
need, nothing approaching a smart hydrocephalus shunt has been developed. This failure is in part
due to the current state of sensor technology. A sensor imbedded in a hydrocephalus shunt must
survive for as long as the shunt is in use, and shunts can be active in the body for a decade or more.
Traditional MEMS sensors are fabricated by micromachining silicon, which thanks to the
microprocessor industry can be precisely machined with features down to the nanoscale. However,
silicon will slowly dissolve in the warm, saline environment inside the body, and the moving parts
found in many microfabricated sensors will undergo significant material fatigue if operated for
many years
73
. Most commercial MEMS sensors are designed for flow or pressure sensing in air,
and cannot interface with physiological fluid. For example, the most common type of
microfabricated pressure sensor is based on a flexible diaphragm on top of a vacuum cavity
74
.
Failure modes during fluidic use include vapor penetration into the vacuum cavity, biofouling on
the membrane, corrosion of the silicon substrate, and fatigue of the membrane itself, all of which
cause drift and eventual sensor failure (Fig. 1.15). Similarly, a commercial MEMS thermal flow
sensor could be coated with a polymer to allow the sensor to function at all in an aqueous
environment
75
, but encapsulation insulates the heater and temperature sensors from the fluid,
significantly lowering sensitivity and raising the necessary power consumption (Fig. 1.16).
Common MEMS failure modes, such as biofouling, silicon degradation, and insulation failure,
would still be issues.

- 26 -

Figure 1.15 A diaphragm-based silicone pressure sensor can fail if used chronically in vivo from multiple
factors, including corrosion of the silicon substrate, biofouling, fatigue of the flexible sensing diaphragm,
and compromised integrity of the vacuum cavity used as a reference pressure
76
. © 2016 IEEE

Figure 1.16 Similar to the pressure sensor, a traditional MEMS thermal flow sensor can undergo failure
due to corrosion or biofouling. In addition, to operate in a saline environment requires polymer
encapsulation of the heater and temperature sensitive elements to protect from short circuit and corrosion.
This encapsulation increases thermal resistance between the sensor and the fluid under consideration and
can increase thermal leakage through the substrate.

1.5.1 Prospective smart shunt technology
In academic literature, several technologies have been proposed to transduce pressure or
flow in hydrocephalus shunts. These include variations on the silicon diaphragm pressure sensor
77-
81
and silicon-based thermal flow sensor
82
with packaging designed to withstand the rigors of in
vivo operation. One device of note is the OSAKA Telesensor, which transduces pressure using a
flexible bellows topped with a magnet that tunes and de-tunes an implanted coil according to the
bellows height, which changes with intracranial pressure
83
. This sensor was tested in 94 patients
for up to 44 months, and shunt failure detection and failure localization via pulse wave monitoring
was successfully demonstrated. These sensors did experience drift of up to 25 mmHg/year,
requiring recalibration via shunt tap. A capacitive pressure sensor from Telemeasurement GmbH
was also tested in seven hydrocephalus patients for up to 17 months, demonstrating the correlations
- 27 -
of hydrocephalus symptoms with intracranial pressure, which was independent of ventricular size
as measured by CT scan
78
. However, the low resolution of these sensors (±5 mmHg) limited
clinical applicability, and to the best of this author’s knowledge no follow-up studies were ever
performed.
Other potential components of a smart shunt include autoregulating valves, self-cleaning
proximal catheters, and algorithms for closed-loop ICP control. Several autoregulating valves or
micropumps designed specifically for hydrocephalus shunts have been reported in academic
literature
77,79,84-86
, and Christopher Miethke currently holds a patent on a mechatronic valve for
hydrocephalus shunts
87
. The major challenges in designing an electronically actuated valve for
hydrocephalus include passively holding the valve’s current state, so that no power is consumed
during normal operation, and preventing further increases in shunt failure rates due to valve
complexity. Algorithms for closed-loop hydrocephalus care based on an ICP sensor and a
mechatronic adjustable valve, with varying degrees of physician involvement, have also been
widely explored in literature
85,88-93
. Several methods of physically removing proximal catheter
obstruction have been explored in the clinic, including ultrasound-induced cavitation
94,95
, laser
ablation
96
, and electrocautery
97,98
; the tools used for these methods are percutaneous and amenable
to easy insertion into a proximal catheter, and future proximal catheter designs could integrate any
of these modalities as self-cleaning solutions. In addition, the development of magnetically-
actuated MEMS devices for removing tissue occlusion on proximal catheters has recently been
reported
99-101
.
So far, only one device has received FDA approved for noninvasively monitoring
hydrocephalus shunt status. ShuntCheck, made by NeuroDX Development, Inc., consists of a
temperature sensor, a custom tablet computer, and an ice pack
102
. To evaluate shunt flow, a doctor
attaches the temperature sensor to patient’s neck on top of the shunt (Fig. 1.17). An ice pack is
then placed on the shunt upstream of the temperature sensor, and if the temperature decreases,
fluid is assumed to be flowing through the shunt. ShuntCheck is simple and non-invasive, but it
can only give a binary yes/no answer to whether fluid is currently flowing through the shunt, and
even in a working shunt there is not always significant flow
102,103
.
- 28 -

Figure 1.17 The ShuntCheck system involves (A) placing a temperature sensor on the skin above the
shunt’s distal catheter and (B) placing an ice pack on the neck upstream of the sensor. If the sensor registeres
a decrease in temperature, it is assumed that fluid is flowing through the shunt
104
.
Other proposed commercial approaches to non-invasive hydrocephalus monitoring have
focused on ICP, which closely correlates with patient symptoms. Since the eyes are directly
connected to the brain via the optic nerve, methods have been proposed to evaluate ICP by
measuring intraocular pressure (IOP). One device which has been developed to non-invasively
measure IOP is the Triggerfish contact lens-based pressure sensor, from Sensimed, Inc
105
. This
device measures the radius of the eye, which changes in response to changes in IOP, and transmits
the data to an external reader unit placed around the eye. Unfortunately, a Triggerfish lens is only
approved to be used for 24 hours of continuous use and can only measure changes in pressure from
when it is first activated
106
. It also requires bulky external hardware to measure which limits its
use
105
.
Another commercial device under development for wirelessly monitoring ICP is the
Miethke Sensor-Reservoir
107
. This device, which is currently undergoing trials in Europe, consists
of a silicon-based pressure sensor placed into a fluidic reservoir which is attached to a
hydrocephalus shunt between the proximal catheter and the valve. The sensor-reservoir can
measure pressure on demand and communicates wirelessly with a handheld device operated by a
physician, allowing for rapid measurement of ICP. This device, if approved, will be a huge step
forward in treating hydrocephalus patients. However, the fact that it employs a membrane-based
pressure sensor may limit its long-term viability in vivo, and since it will only be used for point
measurements of a single variable (ICP), it cannot give a complete picture of a patient’s disease
state. Early clinical studies of the Sensor-Reservoir for managing hydrocephalus have not found a
statistically significant improvement in care, though this is most likely due to low sample size
108
.
- 29 -

Figure 1.18 (Left) The Triggerfish contact lens device measures changes in IOP via changes in the eye’s
radius and transmits pressure measurements to an external reading coil
109
. (Right) The Miethke Sensor-
Reservoir allows doctors to place a traditional MEMS pressure sensor in line with a hydrocephalus shunt.
Pressure data can be wirelessly transmitted on demand to an external receiver
110
.
The recent commercialization and approval of the CardioMEMS intracardial pressure
sensor is an encouraging sign for groups seeking to develop chronically implantable MEMS
sensors
111
. The CardioMEMS sensor uses two inductively coupled coils separated by a glass
vacuum cavity to passively transduce pulmonary artery pressure. In 2013, this device became the
first MEMS sensor approved by the FDA for chronic implantation, and the company was acquired
by St. Jude Medical (now Abbott) shortly thereafter.

Conclusion
Hydrocephalus has been recognized as a disease since the beginning of recorded history,
but its cause and pathology were not understood until relatively recently. We now know that
hydrocephalus can have many different causes, from occlusion of a fluidic channel in the brain to
overproduction of CSF by the choroid plexus to swelling caused by infection. Current treatment
for hydrocephalus involves implanting a shunt, which drains excess CSF from the brain’s
ventricles into the peritoneal, pleural, or atrial cavities where it can be reabsorbed. Shunts consist
of a proximal catheter implanted in the brain, a valve which controls the rate of fluid flow, and a
distal catheter which travels down the neck and toward the abdomen; despite some innovations in
valve technology, shunts have in general not changed since their introduction in the 1960’s.
Although implanted shunts are the first treatment for hydrocephalus to achieve more than
incidental success, they still have many issues: 40% of shunts fail within their first year of use, and
that number rises to almost 90% within the first decade. Shunts can fail due to obstruction by
protein or tissue, mechanical disconnection, or by draining too much or too little fluid from the
brain. The initial symptoms of shunt failure are dizziness, headaches, irritability, blurred vision,
and other nonspecific symptoms, but if left untreated, a patient with a failed shunt can undergo
brain damage and become comatose. Diagnosing a shunt failure is difficult, and doctors must insert
a needle into the brain to measure ICP or perform multiple MRI or CT scans with radioisotopes to
measure ventricular enlargement or flow through the shunt, neither of which can provide definitive
results. To improve treatment and increase the quality of life for patients a “smart shunt” has been
- 30 -
proposed which can measure the integrity of the shunt and automatically clear any occlusions,
while simultaneously allowing doctors to wirelessly monitor the patient’s condition. However, no
sensors currently exist which can provide long-term measurements of physical parameters
associated with hydrocephalus. Therefore, improving hydrocephalus treatment involves
developing sensor technology which can be used for long periods of time in the human body.

- 31 -
References
1. S. Tuli, J. Drake, J. Lawless, M. Wigg, and M. Lamberti-Pasculli, "Risk factors for repeated
cerebrospinal shunt failures in pediatric patients with hydrocephalus," Journal of
neurosurgery, vol. 92, pp. 31-38, 2000.
2. "Children's doctor improves hydrocephalus treatment in Africa," ed: Boston Children's
Hospital, 18 November 2010.
3. "Case courtesy of Dr Paul Simkin, Radiopaedia.org, rID: 30453," ed.
4. HydrocephalusAssociation, "Hydrocephalus Fact Sheet," in
http://www.hydroassoc.org/docs/FactSheet_Stats.pdf, H. Association, Ed., ed, 2015.
5. A. Aschoff, P. Kremer, B. Hashemi, and S. Kunze, "The scientific history of hydrocephalus
and its treatment," Neurosurgical review, vol. 22, pp. 67-93, 1999.
6. M. R. Aisenson, "Closing of the anterior fontanelle," Pediatrics, vol. 6, pp. 223-226, 1950.
7. K. T. Kahle, A. V. Kulkarni, D. D. Limbrick, and B. C. Warf, "Hydrocephalus in children,"
The Lancet, vol. 387, pp. 788-799, 2016.
8. . Brain Physiology Relevant to Hydrocephalus. Available:
http://www.hydroassoc.org/brain-physiology-relevant-to-hydrocephalus/
9. J. Vanneste, P. Augustijn, C. Dirven, W. Tan, and Z. Goedhart, "Shunting normal ‐
pressure hydrocephalus: Do the benefits outweigh the risks? A multicenter study and
literature review," Neurology, vol. 42, pp. 54-54, 1992.
10. R. Adams, C. Fisher, S. Hakim, R. Ojemann, and W. Sweet, "Symptomatic occult
hydrocephalus with normal cerebrospinal-fluid pressure: a treatable syndrome," New
England Journal of Medicine, vol. 273, pp. 117-126, 1965.
11. J. E. Scarff, "Treatment of obstructive hydrocephalus by puncture of the lamina terminalis
and floor of the third ventricle," Journal of neurosurgery, vol. 8, pp. 204-213, 1951.
12. F. E. NULSEN and E. B. SPITZ, "Treatment of hydrocephalus by direct shunt from
ventricle to jugular vain," in Surgical forum, 1950, pp. 399-403.
13. R. H. Pudenz, F. E. Russell, A. H. Hurd, and C. H. Shelden, "Ventriculo-auriculostomy. A
Technique for Shunting Cerebrospinal Fluid into the Right Auricle* Preliminary Report,"
Journal of neurosurgery, vol. 14, pp. 171-179, 1957.
14. S. Kalousdian, M. S. Karlan, and M. A. Williams, "Silicone elastomer cerebrospinal fluid
shunt systems," Neurosurgery, vol. 42, pp. 887-892, 1998.
15. M. S. Turner, "The treatment of hydrocephalus: A brief guide to shunt selection," Surgical
neurology, vol. 43, pp. 314-323, 1995.
16. J. M. Drake and C. Sainte-Rose, The shunt book: Blackwell Science, 1995.
17. R. H. Ames, "Ventriculo-peritoneal shunts in the management of hydrocephalus," Journal
of neurosurgery, vol. 27, pp. 525-529, 1967.
18. J. M. Drake, J. R. Kestle, R. Milner, G. Cinalli, F. Boop, J. Piatt Jr, et al., "Randomized
trial of cerebrospinal fluid shunt valve design in pediatric hydrocephalus," Neurosurgery,
vol. 43, pp. 294-303, 1998.
19. A. Conen, L. N. Walti, A. Merlo, U. Fluckiger, M. Battegay, and A. Trampuz,
"Characteristics and treatment outcome of cerebrospinal fluid shunt-associated infections
in adults: a retrospective analysis over an 11-year period," Clinical infectious diseases, vol.
47, pp. 73-82, 2008.
20. R. V. Patwardhan and A. Nanda, "Implanted ventricular shunts in the United States: the
billion-dollar-a-year cost of hydrocephalus treatment," Neurosurgery, vol. 56, pp. 139-145,
2005.
- 32 -
21. D. M. Sciubba, R. M. Stuart, M. J. McGirt, G. F. Woodworth, A. Samdani, B. Carson, et
al., "Effect of antibiotic-impregnated shunt catheters in decreasing the incidence of shunt
infection in the treatment of hydrocephalus," Journal of Neurosurgery: Pediatrics, vol.
103, pp. 131-136, 2005.
22. P. Kan and J. Kestle, "Lack of efficacy of antibiotic-impregnated shunt systems in
preventing shunt infections in children," Child's Nervous System, vol. 23, pp. 773-777,
2007.
23. A. Pattavilakom, D. Kotasnas, T. M. Korman, C. Xenos, and A. Danks, "Duration of in
vivo antimicrobial activity of antibiotic-impregnated cerebrospinal fluid catheters,"
Neurosurgery, vol. 58, pp. 930-935, 2006.
24. C. A. Harris and J. P. McAllister, "Does drainage hole size influence adhesion on
ventricular catheters?," Child's Nervous System, vol. 27, pp. 1221-1232, 2011.
25. U. W. Thomale, H. Hosch, A. Koch, M. Schulz, G. Stoltenburg, E.-J. Haberl, et al.,
"Perforation holes in ventricular catheters—is less more?," Child's Nervous System, vol.
26, pp. 781-789, 2010.
26. B. J. Kim, W. Jin, A. Baldwin, L. Yu, E. Christian, M. D. Krieger, et al., "Parylene MEMS
patency sensor for assessment of hydrocephalus shunt obstruction," Biomedical
microdevices, vol. 18, p. 87, 2016.
27. H. D. Portnoy, "New ventricular catheter for hydrocephalic shunts: technical note," Journal
of neurosurgery, vol. 34, pp. 702-703, 1971.
28. C. Sainte-Rose, J. Piatt, D. Renier, A. Pierre-Kahn, J. Hirsch, H. Hoffman, et al.,
"Mechanical complications in shunts," Pediatric neurosurgery, vol. 17, pp. 2-9, 1991.
29. J. Haase and R. Weeth, "Multiflanged ventricular Portnoy catheter for hydrocephalus
shunts," Acta Neurochirurgica, vol. 33, pp. 213-218, 1976.
30. S. Hakim, F. D. Roche, and J. Burton, "A critical analysis of valve shunts used in the
treatment of hydrocephalus," Developmental Medicine & Child Neurology, vol. 15, pp.
230-255, 1973.
31. F. Kuffer and D. Strub, "Ein ligaturfreier Konnektor für Hydrocephalus-
Ventile.(Vorschlag zu einem neuen Ventil)," Z Kinderchir, vol. 9, pp. 293-301, 1971.
32. D. Voth and N. Nakayama, "Ein neues ventilgesteuertes System für die präliminare
Ventrikeldrainage:(Technische Beschreibung und klinische Erfahrungen),"
Neurochirurgia, vol. 19, pp. 196-201, 1976.
33. H. Trost, C. Sprung, W. Lanksch, D. Stolke, and C. Miethke, "Dual-switch valve: clinical
performance of a new hydrocephalus valve," in Intracranial Pressure and
Neuromonitoring in Brain Injury, ed: Springer, 1998, pp. 360-363.
34. A. Aschoff, P. Kremer, C. Benesch, K. Fruh, A. Klank, and S. Kunze, "Overdrainage and
shunt technology," Child's Nervous System, vol. 11, pp. 193-202, 1995.
35. G. Zemack and B. Romner, "Seven years of clinical experience with the programmable
Codman Hakim valve: a retrospective study of 583 patients," Journal of neurosurgery, vol.
92, pp. 941-948, 2000.
36. R. Gruber, P. Jenny, and B. Herzog, "Experiences with the anti-siphon device (ASD) in
shunt therapy of pediatric hydrocephalus," Journal of neurosurgery, vol. 61, pp. 156-162,
1984.
37. P. Kremer, A. Aschoff, and S. t. Kunze, "Risks of using siphon-reducing devices," Child's
Nervous System, vol. 10, pp. 231-235, 1994.
- 33 -
38. C. Miethke and K. Affeld, "A New Valve for the Treatment of Hydrocephalus-Ein neues
Ventil zur Behandlung des Hydrocephalus," Biomedizinische Technik/Biomedical
Engineering, vol. 39, pp. 181-187, 1994.
39. C. Sainte-Rose, M. D. Hooven, and J.-F. Hirsch, "A new approach in the treatment of
hydrocephalus," Journal of neurosurgery, vol. 66, pp. 213-226, 1987.
40. Medtronic. (December 2016). Strata Adjustable Pressure Valve. Available:
http://www.medtronic.com/us-en/healthcare-
professionals/products/neurological/shunts/strata-adjustable-pressure-valve.html
41. A. Inc., "Miethke proGAV 2.0 Brochure," ed, 2015.
42. I. L. Corporation. (2016). Integra Flow Regulating Valves. Available:
http://www.integralife.com/index.aspx?redir=detailproduct&Product=584&ProductName
=Integra%20Flow%20Regulating%20Valves&ProductLineName=Hydrocephalus+Mana
gement&ProductLineID=5&PA=Neurosurgeon
43. . HPBio Sphera Duo Hydrocephalus Shunt System. Available:
http://www.spiegelberg.de/products/documents/sphera_duo_20141215_en.pdf
44. . Codman HAKIM Programmable Valve. Available:
https://www.depuysynthes.com/hcp/codman-neuro/products/qs/CODMAN-HAKIM-
Programmable-Valve
45. . Sophysa Polaris. Available: http://www.sophysa.com/en/page/polaris-r
46. P. Decq, J.-L. Barat, E. Duplessis, C. Leguerinel, P. Gendrault, and Y. Keravel, "Shunt
failure in adult hydrocephalus: flow-controlled shunt versus differential pressure shunts—
a cooperative study in 289 patients," Surgical neurology, vol. 43, pp. 333-339, 1995.
47. H. D. Portnoy, R. R. Schulte, J. L. Fox, P. D. Croissant, and L. Tripp, "Anti-siphon and
reversible occlusion valves for shunting in hydrocephalus and preventing post-shunt
subdural hematomas," Journal of neurosurgery, vol. 38, pp. 729-738, 1973.
48. M. Kiefer, R. Eymann, and U. Meier, "Five years experience with gravitational shunts in
chronic hydrocephalus of adults," Acta neurochirurgica, vol. 144, pp. 755-767, 2002.
49. Y. He, R. K. Murphy, J. L. Roland, and D. D. Limbrick, "Interactions between
programmable shunt valves and the iPad 3 with Smart Cover," Child's Nervous System,
vol. 29, pp. 531-533, 2013.
50. Y. Wu, N. L. Green, M. R. Wrensch, S. Zhao, and N. Gupta, "Ventriculoperitoneal shunt
complications in California: 1990 to 2000," Neurosurgery, vol. 61, pp. 557-563, 2007.
51. J. Drake, J. Kestle, and S. Tuli, "CSF shunts 50 years on–past, present and future," Child's
Nervous System, vol. 16, pp. 800-804, 2000.
52. S. R. Browd, B. T. Ragel, O. N. Gottfried, and J. R. Kestle, "Failure of cerebrospinal fluid
shunts: part I: obstruction and mechanical failure," Pediatric neurology, vol. 34, pp. 83-92,
2006.
53. S. C. Stein and W. Guo, "Have we made progress in preventing shunt failure? A critical
analysis," 2008.
54. B. Lutz, P. Venkataraman, and S. Browd, "New and improved ways to treat hydrocephalus:
Pursuit of a smart shunt," Surgical neurology international, vol. 4, p. 38, 2013.
55. S. R. Browd, O. N. Gottfried, B. T. Ragel, and J. R. Kestle, "Failure of cerebrospinal fluid
shunts: part II: overdrainage, loculation, and abdominal complications," Pediatric
neurology, vol. 34, pp. 171-176, 2006.
- 34 -
56. R. T. Schreffler, A. J. Schreffler, and R. R. Wittler, "Treatment of cerebrospinal fluid shunt
infections: a decision analysis," The Pediatric infectious disease journal, vol. 21, pp. 632-
636, 2002.
57. C. Di Rocco, E. Marchese, and F. Velardi, "A survey of the first complication of newly
implanted CSF shunt devices for the treatment of nontumoral hydrocephalus," Child's
Nervous System, vol. 10, pp. 321-327, 1994.
58. S. Sgouros, C. Malluci, A. R. Walsh, and A. Hockley, "Long-term complications of
hydrocephalus," Pediatric neurosurgery, vol. 23, pp. 127-132, 1995.
59. W. Serlo, A.-L. Saukkonen, E. Heikkinen, and L. Von Wendt, "The incidence and
management of the slit ventricle syndrome," Acta neurochirurgica, vol. 99, pp. 113-116,
1989.
60. R. F. Jones, B. C. Kwok, W. A. Stening, and M. Vonau, "The current status of endoscopic
third ventriculostomy in the management of non-communicating hydrocephalus," min-
Minimally Invasive Neurosurgery, vol. 37, pp. 28-36, 1994.
61. A. L. Albright, I. F. Pollack, and P. D. Adelson, Principles and practice of pediatric
neurosurgery: Thieme New York, 1999.
62. X. Nie, "Cranial base in craniofacial development: developmental features, influence on
facial growth, anomaly, and molecular basis," Acta Odontologica Scandinavica, vol. 63,
pp. 127-135, 2005.
63. D. A. Graeb, W. Robertson, J. Lapointe, R. Nugent, and P. Harrison, "Computed
tomographic diagnosis of intraventricular hemorrhage. Etiology and prognosis,"
Radiology, vol. 143, pp. 91-96, 1982.
64. "Case courtesy of Dr Henry Knipe, Radiopaedia.org, rID: 39615," ed.
65. M. Hashimoto, M. Ishikawa, E. Mori, and N. Kuwana, "Diagnosis of idiopathic normal
pressure hydrocephalus is supported by MRI-based scheme: a prospective cohort study,"
Cerebrospinal fluid research, vol. 7, p. 18, 2010.
66. M. Eisha Christian, "Personal Communication," ed, 2016.
67. B. J. Iskandar, C. McLaughlin, T. B. Mapstone, P. A. Grabb, and W. J. Oakes, "Pitfalls in
the diagnosis of ventricular shunt dysfunction: radiology reports and ventricular size,"
Pediatrics, vol. 101, pp. 1031-1036, 1998.
68. P. Graham, R. Howman-Giles, I. Johnston, and M. Besser, "Evaluation of CSF shunt
patency by means of technetium-99m DTPA," Journal of neurosurgery, vol. 57, pp. 262-
266, 1982.
69. S. Sood, S. Kim, S. Ham, A. Canady, and N. Greninger, "Useful components of the shunt
tap test for evaluation of shunt malfunction," Child's Nervous System, vol. 9, pp. 157-161,
1993.
70. M. J. Noetzel and R. P. Baker, "Shunt fluid examination: risks and benefits in the
evaluation of shunt malfunction and infection," Journal of neurosurgery, vol. 61, pp. 328-
332, 1984.
71. M. Czosnyka and J. D. Pickard, "Monitoring and interpretation of intracranial pressure,"
Journal of Neurology, Neurosurgery & Psychiatry, vol. 75, pp. 813-821, 2004.
72. T. M. Family, "Personal Communication," 2019.
73. L. Bowman and J. D. Meindl, "The Packaging of Implantable Integrated Sensors," IEEE
Transactions on Biomedical Engineering, vol. BME-33, pp. 248-255, 1986.
74. L. Yu, B. J. Kim, and E. Meng, "Chronically implanted pressure sensors: challenges and
state of the field," Sensors, vol. 14, pp. 20620-20644, 2014.
- 35 -
75. J. T. W. Kuo, L. Yu, and E. Meng, "Micromachined Thermal Flow Sensors—A Review,"
Micromachines, vol. 3, pp. 550-573, 2012.
76. L. Yu, C. A. Gutierrez, and E. Meng, "An Electrochemical Microbubble-Based MEMS
Pressure Sensor," Microelectromechanical Systems, Journal of, vol. 25, pp. 144-152, 2016.
77. D. Leung, "Implantable Devices for Hydrocephalus Management," ResearchSpace@
Auckland, 2014.
78. K. E. Richard, F. R. Block, and R. R. Weiser, "First clinical results with a telemetrie shunt-
integrated ICP-sensor," Neurological research, vol. 21, pp. 117-120, 1999.
79. H. J. Yoon, J. M. Jung, J. S. Jeong, and S. S. Yang, "Micro devices for a cerebrospinal fluid
(CSF) shunt system," Sensors and Actuators A: Physical, vol. 110, pp. 68-76, 2004.
80. A. Ginggen, Y. Tardy, R. Crivelli, T. Bork, and P. Renaud, "A telemetric pressure sensor
system for biomedical applications," IEEE Transactions on Biomedical Engineering, vol.
55, pp. 1374-1381, 2008.
81. E. Cosman, N. Zervas, P. Chapman, B. Cosman, and M. Arnold, "A telemetric pressure
sensor for ventricular shunt systems," Surgical Neurology, vol. 11, pp. 287-294, 1979.
82. T. Bork, A. Hogg, M. Lempen, D. Müller, D. Joss, T. Bardyn, et al., "Development and in-
vitro characterization of an implantable flow sensing transducer for hydrocephalus,"
Biomedical microdevices, vol. 12, pp. 607-618, 2010.
83. H. Miyake, T. Ohta, Y. Kajimoto, and M. Matsukawa, "A new ventriculoperitoneal shunt
with a telemetric intracranial pressure sensor: clinical experience in 94 patients with
hydrocephalus," Neurosurgery, vol. 40, pp. 931-935, 1997.
84. E. Morganti and G. Pignatel, "Microfluidics for the Treatment of the Hydrocephalus," in
Proceedings of International Conference on Sensing Technology, Palmerston North, 2005.
85. L. Momani, W. Al-Nuaimy, M. Al-Jumaily, and C. Mallucci, "A mechatronic valve in the
management of hydrocephalus: methods and performance," Medical & biological
engineering & computing, vol. 49, pp. 121-132, 2011.
86. S. Szczesny, S. Jetzki, and S. Leonhardt, "Review of current actuator suitability for use in
medical implants," in Engineering in Medicine and Biology Society, 2006. EMBS'06. 28th
Annual International Conference of the IEEE, 2006, pp. 5956-5959.
87. C. Miethke, "Hydrocephalus valve," ed: Google Patents, 2005.
88. L. Momani, A. R. Alkharabsheh, and W. Al-Nuaimy, "Design of an intelligent and
personalised shunting system for hydrocephalus," in Engineering in Medicine and Biology
Society, 2008. EMBS 2008. 30th Annual International Conference of the IEEE, 2008, pp.
779-782.
89. L. Momani, A. R. Alkharabsheh, N. Al-Zuibi, and W. Al-Nuaimy, "Instantiating a
mechatronic valve schedule for a hydrocephalus shunt," in Engineering in Medicine and
Biology Society, 2009. EMBC 2009. Annual International Conference of the IEEE, 2009,
pp. 749-752.
90. I. M. Elixmann, M. Walter, M. Kiefer, and S. Leonhardt, "Simulation of existing and future
electromechanical shunt valves in combination with a model for brain fluid dynamics," in
Hydrocephalus, ed: Springer, 2012, pp. 77-81.
91. A. R. Alkharabsheh, L. Momani, N. Al-Zu'bi, and W. Al-Nuaimy, "An expert system for
hydrocephalus patient feedback," in Engineering in Medicine and Biology Society
(EMBC), 2010 Annual International Conference of the IEEE, 2010, pp. 1166-1169.
92. S. Jetzki, M. Kiefer, M. Walter, and S. Leonhardt, "Concepts for a mechatronic device to
control intracranial pressure," IFAC Proceedings Volumes, vol. 39, pp. 25-29, 2006.
- 36 -
93. A. R. Alkharabsheh, L. Momani, N. Al-Zubi, and W. Al-Nuaimy, "A Multi-Agent
Approach for Self-Diagnosis of a Hydrocephalus Shunting System," in Developments in
E-systems Engineering (DESE), 2010, 2010, pp. 39-43.
94. H. J. Ginsberg and J. Drake, "Unblocking cerebrospinal fluid shunts using low frequency
ultrasonic cavitation," in Ultrasonics Symposium, 2001 IEEE, 2001, pp. 1381-1384.
95. H. J. Ginsberg, J. M. Drake, T. M. Peterson, and R. S. Cobbold, "Recanalization of
obstructed cerebrospinal fluid ventricular catheters using ultrasonic cavitation," Operative
Neurosurgery, vol. 59, pp. ONS-403-ONS-412, 2006.
96. T. C. Hutchens, D. A. Gonzalez, L. A. Hardy, C. S. McLanahan, and N. M. Fried, "Thulium
fiber laser recanalization of occluded ventricular catheters in an ex vivo tissue model,"
Journal of biomedical optics, vol. 22, p. 048001, 2017.
97. K. K. Gnanalingham, J. Lafuente, S. Brew, D. Thompson, W. Harkness, and R. Hayward,
"Percutaneous coagulation of choroid plexus to unblock the ventricular catheter using the
Seldinger technique: preliminary report," World Neurosurgery, vol. 64, pp. 440-443, 2005.
98. J. V. Pattisapu, E. R. Trumble, K. R. Taylor, P. D. Howard, and T. M. Kovach,
"Percutaneous endoscopic recanalization of the catheter: a new technique of proximal
shunt revision," Neurosurgery, vol. 45, pp. 1361-1367, 1999.
99. S. A. Lee, D. J. Vasquez, M. Bergsneider, and J. W. Judy, "Magnetic microactuators for
MEMS-enabled ventricular catheters for hydrocephalus," in Engineering in Medicine and
Biology Society, 2006. EMBS'06. 28th Annual International Conference of the IEEE, 2006,
pp. 2494-2497.
100. S. A. Lee, H. Lee, J. R. Pinney, E. Khialeeva, M. Bergsneider, and J. W. Judy,
"Development of microfabricated magnetic actuators for removing cellular occlusion,"
Journal of Micromechanics and Microengineering, vol. 21, p. 054006, 2011.
101. S. A. Lee, The design and implementation of magnetic microactuators for MEMS-enabled
ventricular catheters for hydrocephalus: University of California, Los Angeles, 2008.
102. J. R. Madsen, G. S. Abazi, L. Fleming, M. Proctor, R. Grondin, S. Magge, et al.,
"Evaluation of the ShuntCheck noninvasive thermal technique for shunt flow detection in
hydrocephalic patients," Neurosurgery, vol. 68, pp. 198-205, 2011.
103. T. P. Boyle, "Comparison of Shuntcheck to Neuroimaging for Diagnosing Ventricular
Shunt Malfunction in the Emergency Department," in 2015 AAP National Conference and
Exhibition, 2015.
104. . ShuntCheck Overview, © 2017 NeuroDx Development. Available:
http://neurodx.com/shuntcheck/
105. G. Sunaric ‐Megevand, P. Leuenberger, and P. R. Preußner, "Assessment of the
Triggerfish™ contact lens sensor for measurement of intraocular pressure variations," Acta
ophthalmologica, vol. 92, 2014.
106. R. Hubanova, F. Aptel, C. Chiquet, B. Mottet, and J. P. Romanet, "Effect of overnight wear
of the Triggerfish® sensor on corneal thickness measured by Visante® anterior segment
optical coherence tomography," Acta ophthalmologica, vol. 92, pp. e119-e123, 2014.
107. C. Miethke, "Implant for measuring the intracorporeal pressure with telemetric
transmission of the measured value," ed: Google Patents, 2013.
108. S. Antes, A. Stadie, S. Müller, S. Linsler, D. Breuskin, and J. Oertel, "Intracranial Pressure–
Guided Shunt Valve Adjustments with the Miethke Sensor Reservoir," World
neurosurgery, vol. 109, pp. e642-e650, 2018.
109. "SENSIMED Triggerfish, © 2014 - SENSIMED AG."
- 37 -
110. Miethke, "SENSOR RESERVOIR Telemetric Shunt Control - Reading Inner Values.," ed:
Aesculap AG.
111. W. T. Abraham, P. B. Adamson, R. C. Bourge, M. F. Aaron, M. R. Costanzo, L. W.
Stevenson, et al., "Wireless pulmonary artery haemodynamic monitoring in chronic heart
failure: a randomised controlled trial," The Lancet, vol. 377, pp. 658-666, 2011.

- 38 -

LOW measurement is important for numerous industrial and biomedical applications, and
many commercial flow sensors exist which can accurately transduce the flow velocity of
gasses. However, measuring flow rate in liquids presents additional challenges, especially
when a sensor must operate inside the warm, corrosive environment found inside the human body.
A flow sensor which could operate chronically in vivo would be especially beneficial to
hydrocephalus patients, allowing the diagnosis of under- or overdrainage within a shunt and giving
doctors a vital piece of data for identifying progressive shunt failure. Several microfabricated flow
sensors have been reported in literature which boast biocompatible construction, but all of them
use high overheat temperatures which are not compatible with implantable use and none are able
to withstand chronic implantation in the body. Here, a novel impedimetric thermal flow sensor is
presented which attempts to address the requirements necessary for chronic use in hydrocephalus
shunts
1-3
. The flow sensor operates by tracking the time of flight of a heat pulse, using a novel
method for accurately measure temperature via high-frequency electrochemical impedance. After
investigating impedimetric temperature measurement and fully characterizing a time of flight flow
sensor, impedimetric thermal flow sensing is extended to the calorimetric mode, allowing for
highly sensitive measurement of ultra-low flow rates using electrode pairs upstream and
downstream of a central microfabricated heater. Construction using only platinum in a flexible
Parylene C substrate leads to long-term stability in vivo, and the use of the novel impedimetric
sensing mechanism leads to a sensitive, biocompatible sensor for cerebrospinal fluid.

Chapter 2
IMPEDIMETRIC TEMPERATURE AND FLOW
SENSORS FOR PHYSIOLOGICAL FLUIDS
F

- 39 -

Background
2.1.1 Measuring Flow in the Body
Precise and accurate knowledge of flow rate is a ubiquitous concern of both medicine and
biology. Accurate flow measurement is critical in applications such as drug delivery
4-6
, drug
synthesis
7,8
, and diagnostic tests
9,10
. In clinical practice, respiratory flow rate measurement is
widely used to measure lung capacity for the diagnosis of respiratory failure or asthma
11-13
. Most
respiratory flow monitoring is done using commercial gas flow sensors, since chronic operation or
compatibility with physiological fluid is not required. Clinical measurement of physiological fluid
flow is more difficult, but non-invasive measurement of flow inside the body is possible using
Doppler ultrasound or MRI. Ultrasound has been used for decades to non-invasively measure
blood flow in peripheral vessels, and can accurately transduce flow velocity in the laminar,
transitional, and turbulent flow regimes
14,15
. Percutaneous catheter-based ultrasound transducers
can also be used to measure blood flow through the aorta or within the heart
16-18
. However,
ultrasound cannot penetrate through the skull, so ultrasound-based flow sensors cannot be used to
measure cerebrospinal fluid dynamics, and flow rates within hydrocephalus shunts are in general
too low for accurate flow transduction via ultrasound. MRI-based flow transduction techniques
can overcome the problem of ultrasound reflection by the skull, and have been clinically used to
measure cerebrospinal fluid dynamics in hydrocephalus patients
19-22
. By spin-tagging water
molecules and measuring their relaxation signal over the course of several seconds, fluid
movement and thus flow rate can be transduced. Since cerebrospinal fluid pulses with heartbeat
and breathing, MRI imaging can be used to track pulsatile CSF dynamics in the brain and identify
blockages in the ventricles. MRI has also been used to measure blood flow in the heart or outer
vessels
23,24
. However, MRI studies are expensive, time-consuming, and impractical for routine
check-ups.
For in vivo applications, implantable microscale flow sensors have been proposed to
monitor the flow of blood, cerebrospinal fluid, or other bodily fluids, provide tracking of a disease
state, or deliver diagnostic information to clinicians. Unfortunately, few flow sensors possess all
the capabilities necessary for medical applications, which include operational compatibility with
physiological fluid, low detection limit, low power consumption, and biocompatibility for in vivo
implantation. Mechanical MEMS flow sensors often mimic macroscale turbine or cantilever flow
sensors
34,35
. These include microfabricated beams made of piezoresistive
36,37
or fiber-optic
38,39

material that are placed perpendicular to the direction of flow and measure the stress generated
from deflection due to flow. A system of pressure sensors can also be used to sense fluid flow, by
measuring the pressure drop across a channel with a known fluidic resistance
40,41
. However,
mechanical designs possess moving parts at risk of biofouling and failure when chronically used
in physiological fluids. Designs with no moving parts are preferred for physiological
environments. Microfluidic thermal flow sensors have been successfully used for gas and liquid

- 40 -
sensing in several applications and operate by measuring convective heat transfer in a fluid as a
function of flow rate
42
. These sensors are composed of a microfabricated heater and one or more
temperature sensors, and can employ one of several methods to transduce flow rate. The most
widely used transduction methods are time-of-flight flow sensing, in which the heater is pulsed
and the time necessary for the resultant heat pulse to pass by the temperature sensors is used to
transduce flow, and calorimetric sensing, in which the heater is held at a constant overheat
temperature and the difference in temperature upstream and downstream of the heater is used to
transduce flow (Fig. 2.1).

Figure 2.1 Thermal flow sensors operate by measuring the effects of convective heat transfer in a flowing
fluid. Several methods of thermal flow sensing exist; the most popular methods are time-of-flight sensing,
where a heater generates a heat pulse in the fluid and the pulse travels downstream past one or more
temperature sensors, and calorimetric sensing, where the difference in temperature upstream and
downstream of a heater is used to transduce flow.
Although many microfabricated thermal flow sensors have been successfully developed
and commercialized for gas flow, there are additional challenges involved in designing sensors for
in vivo measurement of physiological liquids. Due to the high thermal conductance of water and
physiological fluids, sensors must be smaller and operate at a higher overheat temperature to
minimize heat dissipation losses. Sensors designed for implantation must also minimize heating to
avoid damaging the body, and must be constructed from biocompatible materials that will not
corrode or degrade in vivo. Most microfabricated thermal flow sensors are made from
semiconductor-based materials, typically as silicon-based sensors employing a thin-film
microheater (often relying on fragile silicon nitride membranes for thermal isolation) with joule
heaters and resistance temperature detectors (RTDs) fabricated from polysilicon
28,29,43
,
germanium
44-46
, or metals
47
. These sensors have achieved high accuracies and sensitivities over a
wide range of flow velocities thanks to the high temperature sensitivity of doped polysilicon and
germanium. However, semiconductor sensors are not suitable for chronic implantation due to
corrosion under chronic exposure to physiological fluid
48
. In addition, the high thermal
conductivity of semiconductor materials requires the use of delicate micromachined thermal
isolation structures, to prevent heat transfer from the heater to the temperature sensors through the
substrate. Sensors designed for use in water or physiological fluids have been explored in
literature, with metal heaters and temperature sensors fabricated on polyimide
49,50
, Parylene C
51-

- 41 -
53
, or other biocompatible polymer substrates
54,55
. Unfortunately, metal temperature sensors have
an order of magnitude lower sensitivity to temperature changes than semiconductor-based sensors,
resulting in low sensitivity and requiring high overheat temperatures.
For hydrocephalus patients, monitoring flow rate through an implanted shunt would
provide an important tool for early diagnosis of shunt failure, since flow through the shunt
progressively decreases before an increase in intracranial pressure can be measured. In addition,
constant monitoring of flow through the shunt would allow doctors to identify and treat
overdrainage, which affects a large portion of shunted patients. Flow through hydrocephalus
shunts is pulsatile with heartbeat and breathing
25-29
and can vary widely based on the time of day
and the patient’s orientation (standing, sitting, or laying down). The average flow rate through a
shunt is 20 mL/hr
30,31
, though instantaneous rates of up to several hundred mL/hr have been
measured
32,33
, so any implantable flow sensor must be able to measure small variations over this
range.

Table 2.1 A comparison of several thermal flow sensors, both commercial and reported in
literature
2,3,28,29,49,51,54,56,57
. Work described in this thesis is bolded. (Acronyms: PSG = polysilicon glass, SiN
= silicon nitride)
Table 1 shows select parameters of several commercial and laboratory flow sensors. ∆T is
the overheat temperature used and range is either the total range over which effective flow
measurements can be made or the largest range tested. Resolution for commercial sensors is
usually defined as 3× the standard deviation, but in literature the value usually reported is the
smallest flow rate difference at which a difference in measurements was detected. For this table,
resolution is reported as one standard deviation, to maintain parity with values reported in
academic literature. There are very few commercial thermal flow sensors which are able to sense
liquids traveling at low flow velocities, and of these the Sensirion sensor is the most effective,
reporting a minimum detectible flow velocity of ~14 µm/s with a very wide detectable flow range.
However, all available commercial liquid flow sensors use silicon-based sensing elements and
high overheat temperatures, rendering them incompatible with chronic in vivo use. Many other
thermal flow sensors designed for water or other liquids have been reported in literature, but these

Sensing
Method
Substrate
Material
∆T
(°C)
Resolution
(µm/s)
Range
(mm/s)
Sensirion SLI-0430 — Silicon — 13.8 ±9.2
POSiFA PTFD10 Calorimetric Silicon 65 — 0-1850
Ashauer 1998 Time of Flight Silicon, SiN 14 100 0-140
Wu 2001 Anemometer Silicon, PSG 13 1670 0-66.7
Meng 2008 Time of Flight Parylene C 24 10.4 0-0.6
Ahrens 2009 Anemometer Polyimide 12 2890 0-57.9
Vilares 2010 Calorimetric SU-8 60 66.7 ±41.7
Baldwin 2016 Time of Flight Parylene C 1 43.3 ±0.8
Baldwin 2018 Calorimetric Parylene C 1 9.55 ±0.2

- 42 -
have high overheat temperatures, low sensitivities, or non-biocompatible materials which prevent
their use in vivo. The low sensitivity of current polymer-based thermal flow sensors and the lack
of biocompatibility in semiconductor-based thermal flow sensors has left a need for a flow sensor
which is both highly sensitive and biocompatible in design and material.

2.1.2 Impedimetric Thermal Flow Sensing
In pursuit of a sensor which can accurately measure low flow rates found in hydrocephalus
shunts and which can operate chronically in vivo, a thermal flow sensor has been developed which
not only survives in a physiological environment, but exploits the ionic conductivity of
physiological fluids themselves to enable highly sensitive flow rate measurements. This sensor
uses temperature-mediated changes in the conductivity of physiological solutions, which is
monitored via high-frequency electrochemical impedance, to deduce flow rate, with a
corresponding temperature sensitivity an order of magnitude larger than the temperature
coefficient of resistance (TCR) in metallic RTDs. Since the impedance of the surrounding fluid
itself is being used to transduce temperature changes, there is no need to encapsulate the sensor or
protect it from the physiological environment inside the body. A major issue with state of the art
flow sensors designed for use in liquids is that encapsulation of the electrical components is
necessary to prevent corrosion, and this encapsulation is both prone to failure due to water
penetration and serves to thermally isolate the temperature-sensitive elements from the fluid,
decreasing sensitivity and increasing the overheat temperature necessary for flow transduction. By
directly interfacing with physiological fluid, the impedimetric flow sensor eliminates the most
common failure mode of MEMS flow sensors. The high sensitivity of electrochemical impedance
to temperature changes allows flow transduction using overheat temperatures an order of
magnitude less than in the current state of the art, enabling operation while implanted in the human
body.
The impedimetric thermal flow sensors described in this thesis are constructed from a
platinum resistive heater and platinum sensing electrodes on a thin film Parylene C substrate.
Parylene C’s high thermal resistivity reduces thermal conduction across the substrate and
eliminates the need for trenches or other structures used to isolate the heater in silicon-based
thermal flow sensors. A novel time-of-flight based method for flow transduction was also
developed which measures the rate of heat transfer due to both diffusion and convection, enabling
accurate measurements of low flow rates and bidirectional flow detection using only a single
temperature sensor.
The remainder of this chapter presents the theory and experimental verification behind
temperature measurement using electrochemical impedance, the design and characterization of a
time of flight thermal flow sensor composed only of a resistive heater and a single pair of
impedance electrodes, and the further development of a calorimetric flow sensor through the
addition of a second upstream electrode pair. The next section will discuss the microfabrication
techniques used in this chapter, followed by a theoretical and experimental investigation of
temperature measurement using high-frequency electrochemical impedance. Then, an

- 43 -
impedimetric thermal flow sensor which uses a novel time of flight method will be presented and
fully characterized. The following section extends impedimetric thermal flow sensing to the
calorimetric mode via the addition of a second pair of electrodes upstream of the microfabricated
heater. The final experimental section discusses three experiments, involving electrodeposited
platinum-iridium, human cerebrospinal fluid, and device failure under chronic soaking conditions,
which were performed to further develop flow sensors into practical devices for medical use.
Finally, the sensors’ advantages and limitations compared to the state of the art are discussed.

2.1.3 Previous work: a bubble-based flow sensor
A previous attempt at designing a flow sensor which could withstand the rigors of chronic
implantation without damaging the body used an electrolytically-generated bubble as a time-of-
flight tracer
58
. The bubble flow sensor’s design consisted of three parallel pairs of platinum
electrodes sandwiched between two 10 µm layers of Parylene C (Fig. 2.2).

Figure 2.2 (a) Micrograph of sensor as fabricated and released from wafer and (b) attached to flat flexible
cable for testing via zero insertion force connector. Additional electrodes on die were utilized for other
electrochemical impedance measurements. © 2015 IEEE
Flow transduction was accomplished by applying a constant current of 50-70 µA across
the first pair of electrodes, which led to electrolysis and the generation of a bubble. As the bubble
began to flow downstream with the fluid, it passed two more pairs of electrodes, each of which
was connected to an LCR meter which measured electrochemical impedance at 10 kHz. The
bubble acted as an insulator, causing the electrochemical impedance of the downstream electrode
pairs to rise as it flowed passed them and generating a time of flight signal which could be used to
transduce flow rate (Fig. 2.3).

- 44 -

Figure 2.3 Left: Bubble flow sensor operates by (a) generating a bubble upstream via electrolysis. The
bubble then flows downstream past the second (b) and third (c) electrode pairs, generating an impedance
signal (d) which can be used to measure flow rate. Right: the time of flight (TOF) of an electrolyzed bubble
was found to be inversely proportional to flow rate. © 2015 IEEE
This device initially showed promise, and flow transduction was achieved at flow rates
between 0.83-83 µL/min. However, devices operated inconsistently and in many cases, bubbles
adhered to the hydrophobic Parylene substrate and did not flow downstream. Devices also required
the creation of large bubbles which did not dissolve after flow transduction, potentially posing a
problem for in vivo use. These issues motivated the development of new in vivo flow sensing
techniques.

- 45 -
Sensor Fabrication
All sensors discussed in this chapter were microfabricated out of Parylene C and platinum,
and share fabrication procedures that have been developed in Dr. Ellis Meng’s lab over the last
several years
59-61
. These processes take advantage of techniques and tools which were first used
for silicon integrated circuit fabrication and then adapted to manufacture silicon MEMS devices.
Our lab has worked to adapt these techniques to polymer micromachining, to enable batch
fabrication of microscale polymer sensors which are biocompatible and suitable for chronic
implantation in the human body. Since both Parylene C and platinum are for the most part
chemically inert, many of the techniques used are additive; these include chemical vapor
deposition (CVD) of Parylene C and sputter deposition or electron-beam deposition of platinum
thin films. Dry etching of Parylene C is achieved using high-energy oxygen plasma, which can
slowly remove Parylene both mechanically and chemically. Microscale device patterns are made
possible via photolithography. Various photo-sensitive polymers known as photoresists can be
spin-coated onto a silicon or Parylene C-coated wafer; by selectively exposing parts of the
photoresist to ultraviolet light using a photomask, precise patterns can be produced with a
resolution as good as 5 µm. Patterned photoresist can be used to selectively etch Parylene C or
deposit platinum, and excess photoresist can be washed away using an acetone bath, leaving a
device consisting of just Parylene C and platinum.

2.2.1 Parylene C as a Substrate Material
Sensors designed for chronic in vivo use must be fabricated using materials that do not
harm the body or corrode in saline solutions. Parylene C is a clear, flexible polymer which has
recently become popular for MEMS fabrication, particularly in the biomedical field. Parylene C
consists of a monochlorinated poly(para-xylylene) chain (Fig. 2.4) and possesses a high chemical
inertness, high thermal resistance, high dielectric strength, high moisture barrier properties, and
low Young’s modulus (high flexibility)
62
. Parylene C has been shown to be stable and inert in vivo
and has been designated as a USP Class VI biocompatible material.

Figure 2.4 Parylene C is a monochlorinated poly(para-xylylene) chain which possesses properties such as
flexibility, biocompatibility, inertness in vivo, and micromachinability. These properties have led to its use
in a number of biomedical and MEMS devices.

- 46 -
After techniques for chemical vapor deposition of Parylene C were developed in 1966, the
polymer was used to coat medical devices to protect them from corrosion in vivo. More recently,
surface micromachining techniques have been developed which allow Parylene C to be used as a
structural material for MEMS devices. Though normally inert, Parylene C can be etched using
oxygen plasma, and by using standard lithography techniques, complex shapes can be formed from
a flat Parylene substrate
60
. Three dimensional shapes can also be produced via the use of sacrificial
layers which are conformally coated with Parylene, or by thermoforming, which involves placing
a Parylene device in a mold and vacuum annealing at temperatures above its glass transition point
to impart a shape change. Using these techniques, novel Parylene-based sensors
63-65
, neural
probes
66-68
, and implantable devices
69,70
have been produced. Parylene C was leveraged to develop
flow sensors which must survive for long periods of time in vivo without corrosion or drift.

2.2.2 Platinum as a Sensing Material
Platinum is an excellent material for constructing implantable medical devices, and
possesses unique properties which make it ideal for use in an electrochemical sensor. Platinum is
biologically inert and passive
61
. The unique pseudocapacitive reactions which occur in platinum
also make it an ideal material for electrochemistry. For neurostimulation applications, a large
electric field must be generated by an electrode in order to elicit action potentials in nearby
neurons, the limit of which is governed by the electrode’s double-layer capacitance
71
. If too much
charge is injected into an electrode, faradaic reactions will occur, leading to electrolytic gas
generation, chlorine formation, and electrode corrosion
72
. Platinum, has the unique ability to
reversibly absorb and desorb hydrogen ions onto its surface during charge injection, increasing the
effective double-layer capacitance and allowing larger electric fields to be applied to a platinum
electrode before faradaic reactions occur
73
. Due to this pseudocapacitance, platinum is commonly
used for chronic neurostimulation devices such as pacemakers
74
, spinal cord stimulators
75
,
cochlear implants
76
, and deep brain stimulators
77
.
Platinum is also amenable to polymer microfabrication. Several techniques exist to deposit
nanometer-thick platinum coatings onto wafers or devices, including sputter deposition and
electron-beam deposition. Low temperature deposition is important when manufacturing polymer
devices in order to minimize thermal stress and avoid premature annealing. Despite platinum’s
high melting point, our tests have shown that it is possible to deposit platinum thin films by both
sputtering and electron-beam deposition while keeping the substrate temperature under Parylene
C’s glass transition temperature of ~70°C. In addition, atomistic modeling indicates that platinum
may form covalent bonds when deposited on Parylene C
78
, allowing it to adhere to the substrate
without adhesion promoters. Although many groups have used titanium to promote adhesion
between Parylene C and platinum, we have found that Parylene-platinum adhesion is sufficient for
our devices.

- 47 -
2.2.3 Fabrication Details
The following section describes the fabrication of the impedimetric thermal flow sensor.
First, a 12 µm thick Parylene C layer was deposited on a 4” silicon carrier wafer via CVD, using
a SCS Labcoter 2 machine (Fig. 2.5a). To pattern platinum, a 2 µm thick layer of AZ5214
photoresist was spin-coated onto the Parylene-covered wafer and patterned using UV lithography
and an infrared-switching recipe, which resulted in features with negative sidewalls. Wafers were
then cleaned using oxygen plasma in a reactive ion etcher (RIE), followed by the deposition of
2000 Å platinum either by electron beam deposition or sputter deposition. After deposition, metal
was removed from non-desired areas via liftoff in an acetone bath (Fig. 2.5b). After liftoff, the
wafers were again cleaned with oxygen plasma (a procedure known as “descum”) and a second 12
µm layer of Parylene C was CVD deposited. Electrode surfaces and contact pads were exposed
using a switched chemistry deep reactive ion etching (DRIE) process, with alternating steps of
inductively coupled oxygen plasma for etching and C4F8 for passivation, using an Oxford
PlasmaLab 100 ICP machine (Fig. 2.5c). A 10 µm layer of AZ4620 positive photoresist was used
as an etch mask. A second cutout etch was used to separate devices. Free-film devices were
released from their carrier wafer by gently peeling while immersed in deionized water (Fig. 2.5d),
and any residual photoresist was removed by soaking in room-temperature acetone, isopropyl
alcohol, and deionized water for 10 minutes each.

Figure 2.5 The sensor fabrication process involved (a) deposition of 12 µm Parylene C on a silicon carrier
wafer, (b) electron-beam deposition of 2000 Å platinum and patterning using AZ5214 liftoff resist, (c)
exposure of contact pads and electrodes with a switch-chemistry deep RIE in oxygen plasma, and (d) cutout
etch and removal from the carrier wafer by peeling while immersed in DI water. © 2016 IEEE
Electrical connections to sensors were made using a zero insertion force (ZIF) connector
(12 channel, 0.5 mm pitch; Hirose Electric Co., Simi Valley, CA) soldered onto a flat flexible
cable (FFC; Molex Inc., Lisle, IL)
79
. The contact pads of the released device were first attached to
a 250 µm thick PEEK (poly-ether-ether-ketone) backing using cyanoacrylate glue, to achieve the
thickness necessary for insertion into the ZIF connector (250 µm). Sensor were then inserted into

- 48 -
the ZIF connector, and the ZIF connector was encapsulated with biocompatible EpoTek 353-NDT
epoxy for fluidic testing.

2.2.4 Electron-Beam Deposited Platinum vs. Sputtered Platinum
Initial sensor fabrication used electron-beam deposited platinum from a Temescal machine,
located in the Keck cleanroom at USC. E-beam deposited platinum was easy to liftoff when using
AZ5214 photoresist (5-10 minutes in room temperature acetone for complete liftoff) and exhibited
desirable electrochemical properties, with a resistive region beginning below 10 kHz. However,
after maintenance was done to the Temescal deposition system in 2014, devices with e-beam metal
began cracking and lost electrical connection within traces (Fig. 2.6). The problem appeared to be
due to Parylene’s thermal expansion during the deposition process, which for unknown reasons
increased enough after maintenance to cause large cracks to form. Crack formation due to thermal
cycling during metal deposition has been previously reported in literature
80
.

Figure 2.6 Electron-beam deposited platinum contained cracks which rendered devices useless. This most
likely occurred due to thermal expansion of the Parylene C substrate during platinum deposition.
After attempts to fix the Keck e-beam machine failed, sensor fabrication shifted to the use
of sputter-deposited platinum from LGA Thin Films, a third-party company. The sputtering
process resulted in a much more conformal coat, leading to more difficult liftoff (1-2 hours in 40°C
acetone with scrubbing) which damaged the Parylene and increases the propensity for
delamination. In addition, impedance spectroscopy across a pair of electrodes with sputtered
platinum revealed a much higher impedance across all frequencies than electrodes fabricated from
e-beam platinum, and resistive heaters on sputtered devices possessed around half the DC
resistance as heaters on e-beam devices (600-800 Ω for sputtered heaters vs 1.2-1.5 kΩ for e-beam
devices) for identical device designs and metal thicknesses (2000 Å). It is possible that before the
300 µm

- 49 -
e-beam machine began failing, devices still cracked due to thermal expansion, but these cracks
were microscopic and did not cause full device failure. Rather, they served to increase the trace
resistance for heaters and increased the surface area of electrodes, which led to lower impedances.
It also may be possible that sputtered electrodes were covered with acetone residue, redeposited
photoresist, or platinum oxide as a result of more aggressive DRIE etching used on devices
81
. To
avoid the device damage and annoyance which resulted from liftoff on sputtered devices, we
examined an alternate source of electron-beam deposited metal from UCLA. The UCLA e-beam
machine has a much higher throw distance (distance between the evaporating metal and the
wafers), which resulted in minimal thermal cycling and prevented cracks from forming during
cooling, while also maintaining the directional deposition necessary for quick lift-off. All sensor
fabrication shifted towards using UCLA e-beam metal in 2016. Thus far, devices fabricated using
e-beam deposited platinum at UCLA show no visible cracking, with trace resistances and
electrochemical properties similar to that of sputtered platinum. Devices also appear to have better
adhesion between the platinum and the Parylene substrate, which may be a result of the gentler
liftoff process.

- 50 -
Temperature Measurement using Electrochemical Impedance
Using electrochemical impedance to measure temperature in physiological fluids is
desirable both for developing a thermal flow sensor and for standalone temperature sensing
applications. Here, the theoretical basis for the temperature sensitivity of high-frequency
electrochemical impedance is investigated, and experimental results from a microfabricated
impedimetric temperature sensor are shown. This sensor could accurately detect sub-degree
temperature changes in aqueous solutions with vastly different ionic concentration, with a
sensitivity and accuracy higher that the current state of the art.

2.3.1 Background: Measuring Temperature in Physiological Fluids
Developing a thermal flow sensor for chronic use in physiological fluids requires
developing a novel method to measure fluid temperature in the body. This method must be accurate
enough to quickly detect temperature changes of a tenth of a degree or less, since overheat
temperatures would ideally be less than 2°C for an implantable flow sensor. The method must also
be resistant to corrosion and drift in the body, and must be fully biocompatible. Currently, methods
for measuring fluid temperature include thermocouples and resistant temperature detectors
(RTDs)
82,83
.
Thermocouples utilize the Seebeck effect, in which a temperature-dependent voltage
occurs at the junction of two dissimilar conducting elements
84
. They are generally made using
nickel alloys, though occasionally platinum and rhodium junctions are used for stability, and can
have temperature sensitivities up to 60 µV/°C. However, thermocouples have low accuracy and
precision, and generally cannot distinguish temperature changes less than 2°C. Semiconductor
junctions can also be used as thermocouples, with sensitivities up to 110 mV/°C and resolutions
as good as ±0.8°C
82
. Any thermocouple, however, would require insulation to both protect the
conducting elements from water intrusion and to protect the body from the metals, which are
generally not biocompatible.
RTDs are conductive or semiconductive elements whose resistance changes with
temperature. The most common type of RTD, known as PT100, is made from bulk platinum wire
with a nominal resistance of 100 Ω at 0°C and a temperature coefficient of 0.385%/°C
83
. Platinum
RTDs can also be microfabricated on a thin-film substrate, although the sensitivity of thin-film
platinum is lower than that of the bulk material. Platinum is the most common metal for RTDs due
to its highly linear relationship between resistance and temperature. The best laboratory-grade
platinum RTDs can measure temperatures from -200°C to 1000°C with a precision approaching
±0.001°C, though industrial RTDs have precisions around ±0.03°C and thin-film devices are even
worse. The main drawbacks of platinum RTDs are their low sensitivity, which requires highly
sensitive electronics to measure temperature changes with better than ±0.1°C accuracy, which is
necessary for thermal flow sensing. In addition, RTDs require encapsulation to prevent unwanted
faradaic reactions due to voltage drop between the platinum resistor and the surrounding fluid. The
other major category of RTDs, known as thermistors, consist of a semiconductor element whose

- 51 -
electrical resistance changes with temperature
82,85
. In materials such as silicon, germanium, and
metal oxides, higher temperatures cause more electrons to jump into the conduction band, lowering
electrical resistance. This effect results in temperature coefficients an order of magnitude higher
than that of metal RTDs, with comparable precision. However, semiconductor materials will
corrode and disintegrate under chronic soaking conditions
41,86
.
An alternate method to measuring temperature changes in physiological fluid is to measure
the resistivity of the fluid itself. The resistivity of an aqueous solution is highly temperature
dependent, due to increases in ionic mobility that occur as temperature increases. The temperature
coefficient of resistivity of aqueous solutions can be as high or higher than semiconductor RTDs,
and the temperature coefficient of human cerebrospinal fluid has been measured to be
approximately -1.98%/°C
87
(Table 2). Solution resistivity can be accurately transduced via the
high-frequency impedance between two electrodes directly exposed to solution. Using impedance
spectroscopy to transduce temperature has been successfully demonstrated in lithium ion batteries,
where changes in low-frequency impedance were used to measure temperature-dependent changes
in charge transfer resistance
88,89
. However, until now, no one has demonstrated the use of
impedance measurement to transduce temperature in bulk solution, either as part of a thermal flow
sensor or to measure bulk fluid temperature.
Table 2.2 Temperature coefficient of resistance (TCR) of metals and semiconductors which have been used
in resistive temperature sensors, as well as several ionic solutions including human cerebrospinal
fluid.
33,87,90
© 2016 IEEE

2.3.2 Theory: Temperature Sensitivity of Electrolyte Conductivity
The conductance, and thus the electrochemical impedance, of an aqueous electrolyte
solution is highly dependent on temperature. Electrical conduction through an electrolyte solution
Material TCR (%/ºC)
Platinum 0.392
Copper 0.430
Gold 0.390
Polysilicon -2.5 to 0.1
Germanium -2.0
1.32M (5%) NaOH -2.01
7.9M (30%) NaOH -4.50
2.95M (20%) KCl -1.68
Human Cerebrospinal
Fluid
-1.98

- 52 -
(1)
(2)
(3)
(4)
arises from the movement of charged ions, and the solution’s ionic conductance describes the
movement of these ions in an electric field. This movement is determined by the sum of forces
between ions and the surrounding solution. Bulk conductance is directly proportional to ionic
concentration but a molar conductivity can be defined as 𝛬 =𝐺 /𝑐 , where G is the bulk
conductance, and c is the molarity of ions
91
. The limit of conductivity as concentration goes to
zero is known as the state of ‘infinite dilution’ and is useful to consider since in this state
conductivity depends on interactions between a single ion and the solution. If the mobility of an
ion is defined as the limiting velocity per applied force, the conductivity at infinite dilution is
𝜆 0
=𝑢𝑞𝐹
where u is ionic mobility, q is the charge on the ion, and F is Faraday’s constant
92
. The simplest
model for conductivity considers the ion as a charged sphere surrounded by water as a continuous
dielectric fluid. In this model, the mobility of the ion is determined by viscous forces and is given
by Stoke’s law as
𝑢 −1
=4𝜋𝜂𝑟
where  is water’s viscosity and r is the radius of the ion. According to Stoke’s law, any change
in conductivity with temperature would come from changes in water’s viscosity and not from
changes to the ion itself, giving us the following expression for the temperature coefficient:
1
𝑢 𝜕 𝑢 𝜕𝑇
=
−1
𝜂 𝜕𝜂
𝜕𝑇

While theoretical expressions for water’s viscosity are sparse and inaccurate, precise
measurements have been made over a large range of temperatures and pressures. At atmospheric
pressure (0.1 MPa) and for temperatures between 0-100 C, the following equation fits to empirical
data within 1% accuracy
93
:
𝜂 10
−6
𝑃𝑎 𝑠 =∑𝑎 𝑖 (
𝑇 300
)
𝑏 𝑖 4
𝑖 =1

i a b
1 280.68 -1.9
2 511.45 -7.7
3 61.131 -19.6
4 0.45903 -40.0
where T is temperature in Kelvin,  is in Pa·s, and a and b are dimensionless, experimentally
determined constants. When using this formula, the temperature sensitivity of viscosity, defined

- 53 -
(5)
(6)
(7)
as the percent change at a specific temperature over the value at that temperature, can be shown to
vary between -1.5 and -3 %/°C, with values of -2.42 %/°C at 21 C and -1.95 %/°C at 37 C.
Stoke’s law gives the first approximation for ionic mobility as a product of the drag
between a spherical ion and a continuous solution, but this law can be extended by including
dielectric losses produced by moving charges to give
𝑢 −1
=4𝜋𝜂𝑟 +
3
8
𝑞 2
𝜀 0
−𝜀 ∞
𝜀 0
2
𝑟 3
𝜏 0

where 0 and   are the high frequency and low frequency limits of the solution permittivity and
𝜏 0
is the Debye relaxation time
94
, which is the frequency-dependent relaxation time of the dipole
generated by a water molecule in response to an applied electric field. The permittivity of water,
both under static and high-frequency electric fields, increases with the solution’s temperature. Data
for the static and high frequency permittivity of water below 100 GHz has been modeled for
temperatures between -20 and 60 C with the following equations
95
:
𝜀 0
=77.66−103.3(1−
300
𝑇 )
𝜀 ∞
=0.066𝜀 0

The permittivity shows a temperature sensitivity of -0.449 %/°C at 21 C and -0.434 %/°C
at 37 C, which, while less than that of viscosity, is still significant. The Debye relaxation time is
also sensitive to temperature, and can be modeled for the same temperature and frequency range
using the inverse polynomial function
𝜏 0
−1
=2𝜋 (20.27+146.5𝜃 +314𝜃 2
)
𝜃 =1−
300
𝑇
where 0 is in nanoseconds. The temperature sensitivity of the Debye relaxation time then becomes
-2.66 %/°C at 21 C and -2.05 %/°C at 37 C, which is within a few percent of the temperature
sensitivity for viscosity. This similarity likely arises from the fact that viscosity and relaxation time
are both emergent properties that arise from the same fundamental phenomenon, namely thermal
changes in the strength of hydrogen bonds between water molecules. All temperature-dependent
terms in this expression are parameters of water, and viscosity is the term which has the largest
effect, indicating that the effect of changing temperatures on ionic mobility is largely independent
of the ionic species present and is closely related to changes in water’s viscosity.
Viscosities, permittivities, and Debye relaxation times of water at temperatures from 10-
60C were used to calculate conductivity at infinite dilution of a common ion (Na
+
)
93,95,96
. Figure
2.7 shows the percent change in this conductivity per C, which is equivalent the temperature
sensitivity of the solution and analogous to TCR of a solid conductor.

- 54 -
(8)

Figure 2.7 Temperature sensitivity of conductivity, which is comparable to TCR, of an aqueous solution
of sodium ions at infinite dilution. © 2016 IEEE
It is worth noting that the effective radius of an ion in water does not consist of just the
atomic radii of the ion’s component atoms. Any charged particle in water develops a hydration
shell composed of one or more layers of water molecules bound to the particle’s surface via
electrostatic attraction. For small and medium sized ions in liquid water, the electrostatic force
binding the first layer of water molecules to an ion is huge compared to the thermal energy of the
solution, leading to any strong ion having essentially a permanent shell of attached water
molecules. The radius of this shell is largely insensitive to changes in temperature and should be
used to determine ionic mobility. With smaller ions, a second layer of water molecules can form
on the shell’s surface, and the size of this shell has been shown to increase with at higher
temperatures due to breakdown in the structure of bulk water
97
. However, the change in radius per
degree Celsius has been shown to be on the order of 5x10
-4
Å
98
, which is small enough compared
to any hydrated ion’s radius that we can consider it negligible for small temperature changes.
The temperature coefficient of conductivity at infinite dilution is a good indicator of the
response of an electrolyte solution to changes in temperature, but forces exist between ions that
must be taken into account for real solutions. The two major interionic forces which exist in strong
(fully disassociating) ions at low (<0.1 M) concentrations are the electrophoretic effect, in which
solute ions dragged with a moving ion have either an assisting or an opposing effect on the
movement of surrounding ions, and the relaxation effect, in which the electric field of surrounding
ions provides a force which opposes the movement of any one ion. Debye, Huckel, and Onsager
summarized these forces in the expression
Λ=Λ
0
−
(

𝑧 2
𝑒 𝐹 2
3𝜋𝜂 (
2
𝜀𝑅𝑇 )
3
2
+
𝑞 𝑧 3
𝑒𝐹
24𝜋𝜀𝑅𝑇 √
2
𝜀𝑅𝑇 Λ
0
)

√𝑐
where c is the electrolyte concentration in moles per liter, 
0
is the infinite dilution conductivity, z
is the valency number of each ionic species, R is the gas constant, e is the electron charge, and q

- 55 -
describes the symmetry of an ionic species
99
. Calculating this conductivity for various ionic
species reveals that the temperature sensitivity of solution conductivity decreases as concentration
increases, as seen in Figure 2.8, which was generated via MATLAB.

Figure 2.8 The temperature sensitivity of a Na
+
ion in relation to ionic concentration according to the
Debye-Huckel-Onsager equation, at 25°C, where 0 is infinite dilution. The blue arrow indicates the ionic
concentration of cerebrospinal fluid.
Despite a clear change, the magnitude of the difference in temperature coefficient between
infinite dilution and most physiological fluids is minimal. For example, cerebrospinal fluid and
blood have approximate ionic molarities of 295 mM
100
and 345 mM
101
respectively, and
calculations based on equation 8 reveal a drop in temperature sensitivity of 0.19% for CSF and
0.21% for blood, relative to a solution at infinite dilution.
According to theory, one would expect any solution composed only of strong electrolytes
(or strong acids and bases) to have a temperature coefficient at sufficiently low concentrations of
between 1.5%/°C and 3%/°C depending on the ambient temperature, with slightly lower values as
the concentration increases. Solutions with weak electrolytes are much more complex and more
difficult to categorize, but in general their temperature coefficient should be much higher at normal
concentrations and approach that of strong electrolytes at lower concentrations. This is due to the
temperature dependence of the disassociation constant of a weak electrolyte, leading to large
changes in the concentration of disassociated ions when the temperature is altered. Experimental
data shows that at 25°C, 5% dilutions of NaOH, HCl, and KCl vary by 2.01, 1.58, and 2.01%/°C
respectively with higher concentration solutions of the strong electrolytes 30% HCl and 20% KCl
having slightly reduced variations of 1.52 and 1.68%/°C, while a highly concentrated solution of
30% NaOH is more than twice as sensitive at 4.50%/°C
102
.
Intuitively, one would assume that ultrapure water would be either effectively non-
conductive and thus insensitive to changes in temperature or would have a similar sensitivity as
ions at infinite dilution. However, despite resistivities greater than 18 M ∙cm at room temperature,
pure water does indeed conduct electricity, with this conduction mediated by the small fraction of

- 56 -
water molecules spontaneously disassociating into hydroxide and hydronium ions. Ultrapure water
acts similar to a very weak acid or a very weak base and exhibits a much higher sensitivity to
temperature changes than for ionic solutions, with a temperature coefficient ranging from 7.4%/°C
at 0°C to 2.3%/°C at 100°C
103
. However, the addition of ppb levels of impurities will cause the
conductivity of water to change by a much higher factor than changes in temperature, and impure
water will soon have the temperature sensitivity of the impurity.
Changes in conductivity of an electrolyte solution can be detected by measuring the
electrochemical impedance between two immersed electrodes. Typically, three electrodes are used
for electrochemical measurements, with one electrode providing a stable reference voltage in the
electrolyte. This allows for sensitive measurements at the electrode-electrolyte interface, and has
been used for many amperometric sensors which detect analytes such as glucose. However, to
measure changes in the bulk conductivity of the electrolyte, a high frequency signal is passed
between electrodes, and at the proper frequency, the electrode-electrolyte interface is completely
bypassed, meaning that a reference electrode is no longer necessary. All temperature
measurements described here use a two-electrode system (Fig. 2.9).

Figure 2.9 The electrochemical impedance between two electrodes consists of the solution resistance R S,
as well as a charge transfer resistance R ct and double layer capacitance C dl at each electrode interface
104
.

2.3.3 Experimental Design
Experiments were carried out to demonstrate the use of electrochemical impedance to
transduce temperature changes, both as part of a thermal flow sensor and for wider temperature
measurement applications. For these experiments, a prototype flow sensor design was used
consisting of a pair of platinum electrodes and a platinum resistor on a Parylene C substrate (Fig.
2.10). The electrodes had an exposed area of 150 x 250 µm
2
and were spaced 750 µm apart. A
platinum resistor, located 1 mm away from the electrodes on the same Parylene C substrate, was
used as an RTD for benchmarking. The RTD was composed of a serpentine trace 25 µm wide and
had a nominal resistance of ~450 Ω.

- 57 -

Figure 2.10 Impedimetric temperature sensing was tested using a pair of platinum electrodes on a Parylene
C substrate. A microfabricated platinum RTD, fabricated on the same substrate, was used for
benchmarking. © 2016 IEEE
Sensors were primarily tested in phosphate-buffered saline (1× PBS), a common
physiological fluid analog. 1× PBS is isotonic with cerebrospinal fluid and has a resistivity of 50
Ω∙cm. Temperature measurement was also evaluated in concentrated PBS (10× PBS, resistivity 5
Ω∙cm) and deionized (DI) water (resistivity 18 MΩ∙cm). For electrochemical impedance
spectroscopy (EIS) testing, sensors were sealed in glass vials filled with 1× PBS and placed in a
water bath, which was heated to temperatures between 30°C and 50°C. A Gamry Reference 600
potentiostat was used for EIS measurement. For single-frequency impedance tests, solution was
heated from room temperature (~20°C) to 50°C using a hot plate or cooled to 15°C with ice packs,
while a thermocouple/multimeter (Newport TrueRMS Supermeter, 2°C accuracy) was used to
monitor fluid temperature. Data was simultaneously collected from impedance electrodes using an
Agilent E4980A precision LCR meter at 0.1 VPP and from the RTD using a Keithley SourceMeter
with 10 µA bias current. All data was collected and analyzed using a custom LabVIEW program.

2.3.4 Results
EIS between the two impedance electrodes was measured from 1 Hz to 1 MHz at 5°C
temperature intervals (Fig. 2.11). The results showed a clear temperature dependency of
impedance at frequencies between 1-100 kHz, which corresponds to the frequencies at which the
double layer capacitance of the electrodes is bypassed. The temperature coefficient is
approximately -2%/°C, which matches reported temperature coefficient measurements in
cerebrospinal fluid
87
. Discontinuities in the EIS graphs are due to inadequate shielding by the
chicken-wire Faraday cage, and although 10 kHz appears to be the optimal measurement frequency
due to its position in the center of the resistive range, single-frequency benchtop testing showed
that measuring impedance at 100 kHz improved consistency and yielded better results.

- 58 -

Figure 2.11 Electrochemical impedance spectroscopy of a pair of platinum electrodes at temperatures
between 30°C and 50°C. Within the resistive range, as the temperature increases the impedance magnitude
decreases.
The platinum RTD was calibrated by comparing its resistance to thermocouple
measurements in 1× PBS, revealing a temperature coefficient of resistance of 1.21 Ω/°C.
Temperature was then cycled between 15 and 40°C while simultaneously measuring impedance at
100 kHz and temperature using the RTD. The real part of impedance correlated extremely well
with temperature over this range, with a temperature coefficient of -58.3 Ω/°C. Minimal hysteresis
was observed over multiple cycles. Based on noise, the resolution of impedance-based temperature
sensing was measured to be 0.02°C, compared with 0.08°C for the platinum RTD (Fig. 2.12).

Figure 2.12 Left: The real part of impedance at 100 kHz was roughly linear with temperature, with a
temperature coefficient of -58.29 Ω/°C and resolution of 0.02°C. This compared favorably to the RTD,
which showed a temperature coefficient of 1.21 Ω/°C and a resolution of 0.08°C. Right: Minimal hysteresis
was observed after multiple cycles between 15°C and 40°C.
Temperature sensing was subsequently tested in DI water and 10× PBS. EIS measurements
were taken in DI water and 10× PBS at room temperature (Fig. 2.13), and impedance

- 59 -
measurements were taken in the resistive range (the frequency range within which phase is
minimized). For DI water, this meant measuring impedance at 1 kHz, and for 10× PBS this meant
measuring impedance at 1 MHz. The baseline real impedances were 1.77 MΩ at 1 kHz for DI
water and 803 Ω at 1 MHz for 10× PBS, compared to 5.55 kΩ at 100 kHz for 1× PBS, but
temperature coefficients were found to be remarkably constant, even over a >10
3
change in
solution resistivity (Fig. 2.14).

Figure 2.13 EIS showing the magnitude and phase of temperature sensing electrodes in 1× PBS, 10× PBS,
and DI water. Higher solution resistivity resulted in a shift of the resistive range towards lower frequencies.

Figure 2.14 Temperature coefficients of real impedance in the resistive range stayed relatively constant
between 1× PBS, 10× PBS, and DI water.
Finally, a sensor was placed in 1× PBS at room temperature for 14 hours to evaluate drift.
Temperature was measured using the calibrated RTD and compared to the impedance between

- 60 -
sensing electrodes. The results show sensor drift of around 0.25%/hr compared to the RTD, and
demonstrated that impedance electrodes can detect sub-degree fluctuations in room temperature
(Fig. 2.15). The electrodes’ improvement in resolution over the platinum RTD is also apparent.

Figure 2.15 Electrode impedance was measured in 1× PBS for 14 hours and compared to the temperature
as measured by the RTD.
These experimental results confirmed the theoretical predictions that electrochemical
impedance is highly sensitive to changes in temperature, and that this effect can be used to measure
sub-degree temperature fluctuations at a much higher resolution than platinum RTDs, which are
considered the gold standard for temperature transduction. The next step was to apply this
technology to the development of thermal flow sensors, which detect the time of flight of a heat
pulse using a pair of electrodes.

- 61 -
(9)
A Time-of-Flight Flow Sensor Using Electrochemical Impedance
The high sensitivity of electrochemical impedance to changes in temperature was used to
develop a thermal flow sensor for physiological fluids. This flow sensor specifically targeted
chronic in vivo measurement of cerebrospinal fluid flow in hydrocephalus shunts; due to the low
flow rate expected, a novel hybrid time-of-flight method was developed for measuring flow
velocity. Here, the theory behind flow transduction, sensor design and testing methods, and sensor
characterization are shown. The sensor successfully transduced flow rates appropriate for
hydrocephalus treatment.

2.4.1 Theory: Flow Transduction via Rate of Change of Impedance
To enable bidirectional flow measurement at low flow rates, a time of flight method was
invented which only required a heater and a single pair of impedance electrodes. The rate at which
the heated solution reaches the electrodes is used to transduce flow velocity. Figure 2.16 shows
the basic design of the sensor.

Figure 2.16 The sensor transduces flow through the transfer of heat from a resistive heater to flowing fluid
and from the fluid to a pair of electrodes. The electrodes sense changes in temperature via changes in
electrochemical impedance (Z sense). Orange represents the temperature distribution being distorted by flow,
blue represents the flowing liquid and light blue represents the polymer substrate. © 2016 IEEE
The transfer of heat away from a heater can be described using the convection-diffusion
equation:
𝜕𝑇
𝜕𝑡
=∇∗(α∇T)−∇∗(VT)+𝐻 (𝑥 ,𝑡 )
where T is temperature,  is the fluid’s thermal diffusivity, V is the fluid’s velocity vector at any
point, and H(x, t) is any change in temperature forced upon the system
105
. The impulse response
to the convection-diffusion equation is a constantly widening and shifting Gaussian curve, and
when convection is the dominant method of heat transfer, the peak of this Gaussian can be tracked
to derive the flow velocity. However, for low flow rates and over short distances, diffusion is the
dominant method of heat transfer, and the Gaussian will tend to dissipate before the peak moves a

- 62 -
(10)
significant distance. The relative dominance of convection over diffusion is described by the Peclet
number
𝑃𝑒 =
𝐿𝑣
𝛼
where L is length and v is the scalar flow velocity
106
. When the Peclet number is greater than 1,
the system is dominated by convection. The highest flow velocity we tested was around 800 m/s,
which with a temperature sensor spaced 1 mm away from the heater, and using the thermal
diffusivity of pure water (  = 143x10
-9
m
2
/s) gives a Peclet number of 5.59. This indicates that the
system is slightly convection-dominated but that reducing flow velocity would lead to a diffusion-
dominated state. Therefore, a method of flow measurement is needed which operates in both
diffusion-dominated and convection-dominated states.
The maximum rate of change of temperature at electrodes can be used to transduce flow
velocity in both convection and diffusion dominated states, since the heating profile around a
heater is affected by both convection and diffusion. Measuring the maximum rate of change allows
flow transduction down to arbitrarily low flow velocities and for both positive and negative flow
directions. To validate this approach, a 1-dimensional finite difference simulation was used.
Assuming perfect insulation and an infinitely long fluid cylinder, the explicit finite difference
equation for heat transfer via the convection-diffusion equation in one dimension is
𝑇 𝑗 𝑛 −𝑇 𝑗 𝑛 −1
∆𝑡 =𝛼 𝑇 𝑗 +1
𝑛 −1
−2𝑇 𝑗 𝑛 −1
+𝑇 𝑗 −1
𝑛 −1
(∆𝑥 )
2
−𝑉 𝑇 𝑗 +1
𝑛 −1
−𝑇 𝑗 −1
𝑛 −1
2∆𝑥 +𝐻
where 𝑇 𝑗 𝑛 is the temperature at time nt and position j x
107
. The stability criteria for this method
is ∆𝑥 <
2𝛼 𝑉 and ∆𝑡 <
∆𝑥 2
2𝛼 . Choosing 50 m for x and 5 ms for t satisfies these criteria. The initial
temperature was set to 25 C, and after 1 second, the heat at x = 0 increased by 2 C to simulate
constant current delivered to the heater. Heat conduction through the polymer substrate was
considered to be negligible, since the polymer used (Parylene C) has a thermal conductivity an
order of magnitude lower than the surrounding fluid
31
(0.084 W/m·K, versus 0.596 W/m·K for
water).
Figure 2.17 shows the simulation results, which confirm that the heating profile 1 mm away
from the heater is altered by changes in the flow velocity and that the maximum rate of change of
temperature can be used to transduce flow. This method of flow transduction overcomes some
limitations of traditional time of flight measurements and allows the temperature measurement
electrodes to be placed close to the heater, allowing for smaller devices with lower overheat
temperatures and enabling the accurate measurement of low flow velocities.
(11)

- 63 -

Figure 2.17 (A) Simulations show that the temperature profile 1 mm away from the heater is dependent
on flow and (B) that the maximum rate of change of temperature is related to flow velocity. © 2016 IEEE

2.4.2 Sensor Design
Sensors which used the above principles to transduce flow velocity were designed and
fabricated, each sensor consisting of a resistive heater and a pair of exposed platinum electrodes
deposited on a thin-film Parylene C substrate, as show in Figure 2.18. The heater is 1 mm long and
250 µm wide and consists of a serpentine platinum trace with a trace width and spacing of 25 µm.
Each heater has a DC resistance of 750-800 Ω, depending on process variations. The measurement
electrodes are spaced 750 µm apart perpendicular to the direction of flow, with each electrode
having a geometric surface area of 20,000 µm
2
. Two pairs of electrodes were fabricated on each
sensor to enable accurate measurement over a large range of flow rates. Two different sensor
variations were designed, one with electrodes spaced 0.5 and 2 mm away from the center of the
heater and one with electrodes spaced at 1 and 3 mm. Each die also contained a pressure sensor
63

and patency sensor
64
previously developed in our lab. Both the pressure and patency sensors are
designed for use in cerebrospinal fluid, and the fabricated device thus constituted a multi-sensor

- 64 -
system specifically designed to be used in hydrocephalus shunts.

Figure 2.18 Fabricated sensor die showing the resistive heater and two sets of impedance electrodes. The
die also contains microfabricated pressure and patency sensors. © 2016 IEEE
For flow testing, the sensors were packaged into a luer lock compatible device, shown in
Figure 2.19. Using a luer lock connector enables devices to easily attach to shunts and catheters
currently used in hospitals, including external ventricular drains for cerebrospinal fluid, which is
our primary target application. A slot was milled into the top of a luer lock connector with an inner
diameter of 3.25 mm and the sensor was inserted and sealed with epoxy such that one end of the
sensor is fixed to the connector and one is free-standing. Care was taken to place the sensor films
as close to the center of the luer lock connector as possible so that the resistive heater and electrodes
are in the region of maximum flow velocity. A custom acrylic jig was used to hold the sensors in
place while the epoxy dried.

Figure 2.19 A sensor die (A) just after being released from its silicon carrier wafer and (B) packaged in a
luer lock connector for fluidic testing. © 2016 IEEE

- 65 -

2.4.3 Testing Methods
The heater resistance and electrode impedance were characterized in phosphate-buffered
saline (PBS), an isotonic solution commonly used to mimic biomedical solutions (pH 7.4, ionic
molarity 280 mM), at temperatures between 25 and 50 °C. Tests were also performed in artificial
cerebrospinal fluid (aCSF), which has a slightly higher ionic molarity and includes calcium and
magnesium to better mimic the ionic environment in vivo. aCSF was prepared by combining 8.66
g NaCl, 0.224 g KCl, 0.206 g CaCl2·2H2O, 0.163 g MgCl2·6H2O, 0.214 g Na2HPO4·7H2O, and
0.027 g NaH2PO4·H2O in 1 L of double-distilled water. The heater’s DC resistance was measured
using a Keithley 2400 SourceMeter and electrochemical impedance spectra was acquired between
1 Hz and 1 MHz using a Gamry R600 potentiostat. Fluidic testing was performed by flowing PBS
through packaged sensors using a Watson Marlow 120U peristaltic pump with a flow rate range
of 0 to 400 µL/min. Constant current pulses were delivered to the heater using a Keithley 2400
SourceMeter and electrochemical impedance was measured at 100 kHz and 1 VPP using an Agilent
E4980A LCR meter. Sensor performance was first characterized under static conditions using
heater currents between 0.5 and 6 mA, corresponding to overheat temperatures between 0 and
9 °C, and was then characterized at flow rates between -400 and 400 µL/min using a 2 mA heater
current. Sensors were also characterized in 0.25, 0.5, 1, and 2× dilutions of PBS, which have ionic
concentrations of 0.07, 0.14, 0.28, and 0.56 mol/L respectively, as well as tap water. Additional
characterization was performed in PBS at temperatures between 19 and 30 °C, with the exact
temperature at the heater confirmed through measurement of the DC resistance of the heater.
To transduce flow velocity, the electrochemical impedance across a pair of sensing
electrodes was recorded by a LabVIEW program at a rate of 5 Hz. The baseline impedance without
heating was measured, and any deviations from this baseline due to heating were normalized as a
percent change. The instantaneous rate of change was then calculated while the heater was active,
and the minimum rate of change, corresponding to the fastest increase in temperature at the
electrodes, was taken as a measurement of the flow velocity (Fig. 2.20).

- 66 -

Figure 2.20 To transduce flow rate, the impedance during heating was normalized to a baseline value and
the minimum rate of change in impedance, which occurred 1-2 s after heater activation, was recorded. ©
2016 IEEE

2.4.4 Heater Characterization
Characterization of sputtered electrodes at different temperatures showed that the largest
temperature-based changes in electrochemical impedance occurred at frequencies greater than ~60
kHz, corresponding to the frequency range at which phase approaches zero. To ensure that all
measurements were in this range, all impedance measurements were performed at 100 kHz. Heater
characterization showed that the TCR of the resistive heater was approximately 0.16 %/°C, which
is within the expected range for thin-film platinum
108
. From the TCR, the overheat temperature
was calculated during flow testing by recording the heater’s resistance during current injections.
Figure 2.21 shows the heater overheat temperature and the impedance of electrodes 1 mm away
normalized to the pre-heating value. A heater current of 2 mA provided an adequate impedance
response with an overheat temperature of only 1.04 °C. To minimize power consumption and
reduce the risk of damaging tissue or other structures due to excessive heating, all subsequent tests
used a heater current of 2 mA.

- 67 -

Figure 2.21 (A) The overheat temperatures at the heater and (B) the response of an electrode pair 1 mm
away for various current levels at no flow. © 2016 IEEE

2.4.5 Flow Transduction
A volumetric flow rate of 1 µL/min will result in a flow velocity of 2 µm/s when sent
through a tube with a diameter of 3.25 mm. Figure 2.22A shows the impedance response of
electrodes 1 mm away from the heater operating at an overheat temperature of 1°C under different
flow velocities. The response profile and the minimum rate of change of impedance was found to
vary with flow velocity and to be consistent across multiple trials (Fig. 2.22B).

Figure 2.23 shows the minimum rate of change of impedance at flow velocities between -
800 and 800 µm/s. The rate of change versus flow velocity takes the form of a steadily increasing
Figure 2.22 (A) The percent change in impedance at electrodes 1 mm away from heater during constant
1°C heating and (B) the instantaneous rate of change. The minimum (peak) of the rate of change is used to
transduce flow velocity. © 2016 IEEE

- 68 -
curve at small negative and positive flow velocities, but begins to level off at large negative flow
velocities. The maximum standard deviation between -200 and 200 µm/s is 43.3 µm/s.

Figure 2.24 shows the response of sensors with electrodes spaced 0.5, 1, and 2 mm away
from the heater. Changing the heater-electrode spacing changes the dynamic range of the sensor,
and electrodes spaced 0.5 mm away from the heater showed a higher dynamic range than the
electrodes at 1 mm. However, measurements from electrodes at a 0.5 mm spacing were observed
to be noisier than measurements at 1 mm. Sensors with a 4 mm electrode-heater spacing showed
no response. Based on the previously-measured temperature coefficients of the platinum
electrode’s impedance, the difference in temperature at the electrodes are all much less than 1°C
during sensor operation, and for electrodes at 2 mm the change in temperature during heater
activation at 0 flow is approximately 0.02°C, which is the limit of detection measured during
temperature sensing experiments. Any change in temperature on electrodes 4 mm away from the
heater must be obscured by noise.
Figure 2.23 The minimum rate of change of impedance 1 mm away at flow velocities from -800 to 800
µm/s. © 2016 IEEE

- 69 -

Figure 2.24 The sensor response of electrodes spaced 0.5 mm, 1 mm, and 2 mm away from the heater.
The response of electrodes at 4 mm was negligible. © 2016 IEEE

2.4.6 Fluid Temperature and Ionic Concentration Experiments
To ensure that results using PBS can be generalized to physiological fluids, a sensor was
tested in both PBS and artificial cerebrospinal fluid (aCSF), which contains 150 mM Na
+
, 3.0 mM
K
+
, 1.4 mM Ca
2+
, 0.8 mM Mg
2+
, 1.0 mM PO4
3-
, and 155 mM Cl
-
. Figure 2.25 shows the results,
which reveal no significant differences in the response.

- 70 -

Figure 2.25 A comparison of sensor response in 1x PBS and aCSF, which shows no significant differences.
© 2016 IEEE
The baseline impedance of electrodes in different dilutions of PBS ranged from 3.67 ±0.04
kΩ in 2× PBS to 20.7 ±0.18 kΩ in 0.25× PBS, and electrodes in tap water showed a baseline
impedance of 82.5 ±8.9 kΩ (n = 15, mean ±SD). Figure 2.26 shows the sensor’s response in these
solutions. As concentration decreased, the magnitude of the sensor’s response increased slightly,
though this change is only greater than standard deviation between tap water and the most
concentrated PBS solution.

Figure 2.26 (A) The minimum rate of change versus flow velocity at various dilutions of PBS. 0.25, 0.5,
1, and 2× PBS have ionic concentrations of 70, 140, 280, and 560 mM respectively. (B) The sensor response
versus the baseline impedance of the temperature sensing electrodes. © 2016 IEEE

- 71 -
Figure 2.27 shows the sensor’s response in 1× PBS at ambient fluid temperatures between
19 and 30°C. The baseline rate of change of the sensor decreases slightly over this range.

Figure 2.27 The sensor response using 1°C overheat temperature at ambient temperatures between 19°C
and 30°C. The minimum rate of change becomes slightly less sensitive at higher temperatures. © 2016
IEEE

2.4.7 Increasing Overheat Temperature
Using a 1°C overheat temperature is desirable to reduce power consumption and avoid
tissue damage in in vivo applications, but higher overheat temperatures would provide greater
dynamic range and potentially increase sensor resolution, which would be useful for non-
implanted applications. A sensor was characterized using a 1, 2, and 10 °C overheat temperatures,
which correspond to 2, 2.9, and 6.2 mA currents delivered to the resistive heater. Figure 2.28 shows
the sensor response results at these different operating regimes. When the overheat temperature is
increased from 1 to 2 °C, the resolution improves significantly, improving from 28.6 to 17.3 µm/s
between -200 and 200 µm/s. Increasing the overheat temperature to 10 °C further decreases the
resolution to 13.1 µm/s, though this improvement is not proportional to the change in temperature
due to higher measurement variations seen during higher temperature operation.

- 72 -

Figure 2.28 Comparison of sensor response between (A) 1°C and 2°C and (B) 1°C and 10°C. Higher
overheat temperatures are somewhat noisier but possess a significantly higher dynamic range, resulting
in a net improvement in measurement resolution. © 2016 IEEE

- 73 -
A Calorimetric Flow Sensor Using Electrochemical Impedance
Calorimetric thermal flow sensors operate by measuring the difference in temperature
upstream and downstream of a central heater. An increase in flow rate causes preferential heat
transfer downstream of the heater, increasing this temperature difference. After demonstrating
flow transduction in aqueous ionic solutions using a heater and a single temperature-sensitive
electrode pair, a calorimetric flow sensor which used electrochemical impedance was designed
and tested
3
. This calorimetric/impedimetric flow sensor achieved a 4× increase in sensitivity at
ultra-low flow rates compared to sensors which used time of flight transduction.

2.5.1 Theory: Flow Transduction via Difference in Impedance Dip
Calorimetric flow sensors consist of a central heater with one or more pairs of temperature-
sensitive elements upstream and downstream. To implement calorimetric flow sensing using
electrochemical impedance measurements, a pair of electrodes was placed perpendicular to flow
both upstream and downstream of a resistive heater (Fig. 2.29). Traditional calorimetric flow
sensors utilize semiconductor or metal RTDs to measure temperature, and hold the central heater
at a constant overheat temperature while periodically measuring temperature at each sensing
element. However, in a two-electrode electrochemical system absolute impedance will drift
unpredictably with time, so a method of normalizing electrode responses is required.

Figure 2.29 The calorimetric/impedimetric thermal flow sensor consisted of a central heater with pairs of
impedance electrodes upstream and downstream. The uneven diffusion of heat due to flow resulted in
higher temperatures, and larger impedance dips, downstream of the heater. © 2018 IEEE
To provide consistent results from multiple electrode pairs, power was delivered to the central
heater in square pulses, with 30 seconds of cooldown time after each pulse. Impedance magnitude
was measured before and during pulsing, and impedance values were normalized as a percent of
the baseline impedance just before each pulse. The impedance dip, defined as the percent
difference in impedance between baseline and the impedance at the end of the heat pulse, was used
to transduce flow rate (Figure 2.30).

- 74 -

Figure 2.30 To measure flow rate, the impedance across an electrode pair is measured in response to an
applied heat pulse. The impedance dip, defined as the percent change in impedance at the end of heater
activation, is used to transduce flow velocity. © 2018 IEEE

2.5.2 Sensor Design
The calorimetric flow sensor consisted of a central resistive heater and two pairs of exposed
impedance-measurement electrodes, constructed out of thin-film platinum on a Parylene C
substrate (Fig. 2.31). The heater shared the design of the heater used in the time-of-flight flow
sensor from the previous section (25 µm wide traces, 2000 Å thick, 1 mm long), and had a nominal
resistance of 600 Ω. Two electrode pairs were placed perpendicular to the direction of flow
upstream and downstream of the heater, with a separation distance of 500 µm and a heater-
electrode separation of 1 mm. Each electrode had an exposed area of 150 x 200 µm
2
.

- 75 -

Figure 2.31 Image of microfabricated flow sensor, which consists of a platinum resistive heater and two
pairs of impedance-sensing electrodes on a Parylene C substrate. © 2018 IEEE

2.5.3 Experimental Methods
Calorimetric flow sensors were packaged in luer lock spacers (ID 3.25 mm) for testing (Fig.
2.32). Sensors were centered inside the luer lock spacers and sealed in with EpoTek 353NDT
biocompatible epoxy. To enable calorimetric flow sensing, the impedance across upstream and
downstream electrodes had to be measured simultaneously, so a custom multiplexer board was
designed which allowed simultaneous measurement of impedance magnitude and phase from
multiple electrode pairs at a rate >5 Hz each. The multiplexer board consisted of four ADG1206
multiplexer ICs on a custom PCB and was controlled using LabVIEW. Impedance was measured
at 100 kHz and 0.1 VPP using an Agilent 4980A precision LCR meter, the resistive heater was
activated using 3.3 V square pulses from a Keithley SourceMeter, and flow was delivered via
syringe pump (Fig. 2.33). Phosphate-buffered saline (1× PBS) was again used for benchtop testing.

Figure 2.32 Calorimetric flow sensor packaged in a luer lock spacer and sealed with EpoTek 353NDT
biocompatible epoxy.

- 76 -

Figure 2.33 A custom multiplexer (MUX) setup was designed to simultaneously measure impedance from
two electrode pairs with a sampling rate >5 Hz each. Phosphate-buffered saline (PBS) was flowed using a
syringe pump, electrode impedance at 100 kHz was measured with an LCR meter, and a Keithley
SourceMeter delivered 3.3 V to the heater. © 2018 IEEE
2.5.4 Calorimetric Results
To determine the ideal heat pulse width for flow transduction, the impedance dip of a single
downstream electrode pair was measured at flow rates between 0 and 200 µm/s during 1, 5, 10,
and 20 second heat pulses (Fig. 2.34). The sensitivity of impedance dip to flow velocity increased
with pulse width up to 10 seconds, but no significant difference in sensitivity was observed
between 10 and 20 second pulses, so 10 second pulses were used for all subsequent tests.

Figure 2.34 Data from a single pair of electrodes downstream of the heater revealed that sensitivity
increased with heat pulse length up to 10 s, but there was no significant improvement between 10 and 20 s.

- 77 -
© 2018 IEEE

Tests using the custom multiplexer showed that accurate impedance measurements could be
achieved from upstream and downstream electrodes simultaneously at a rate >5 Hz for each
electrode pair; differences in heat transfer due to flow were clearly visible as differences in the
impedance dip between the upstream and downstream electrodes (Fig. 2.35). Sensors were tested
under flow velocities between ±400 µm/s, revealing that as flow velocity was increased, the
magnitude of the impedance dip increased for downstream electrodes and decreased for upstream
electrodes, as would be expected.

Figure 2.35 Left: Using a custom multiplexer setup, both upstream and downstream electrode response to
heat was simultaneously measured. Heat was asymmetrically distributed upstream and downstream of the
heater as a function of flow velocity. Right: Both upstream and downstream electrodes can be used to
simultaneously measure flow rate. Using multiple, independent flow measurements increased accuracy and
added redundancy. These are important for chronic in vivo applications in which electrodes may experience
biofouling or degradation. © 2018 IEEE
The difference in impedance dip between upstream and downstream electrodes was used to
transduce flow velocity. Figure 2.36 shows the relationship between the impedance dip difference
(upstream dip minus downstream dip) and the flow velocity. The profile is symmetric around 0
µm/s flow velocity and has a sigmoidal shape. Between ±200 µm/s, the response is linear with a
sensitivity of 0.035%/µm/s and a 2σ resolution of 19.1 µm/s.

- 78 -

Figure 2.36 The difference between upstream and downstream impedance dip response allows highly
sensitive flow measurement, with a 2σ resolution of 19.1 µm/s for ultra-low flow velocities (-200 to 200
µm/s). © 2018 IEEE

One of the major advantages of this calorimetric method of flow transaction is that it can
be used alongside the time of flight method, allowing sensors to be constructed which are
extremely sensitive at the low end of their measurement range but which can still transduce flow
rates over the full range expected for hydrocephalus shunts. For a sensor with electrodes both
upstream and downstream of the heater, time of flight and calorimetric signals can be recorded
from both electrode pairs simultaneously, allowing truly bidirectional flow measurement by
multiple transduction modes. Only one heater-electrode spacing was tested for calorimetric flow
sensing, so future test devices will include electrodes closer and farther from the heater, and multi-
modal flow transduction will be optimized for different flow velocity ranges. In addition to time
of flight and calorimetric flow transduction, a third method of measuring flow rate is also possible
using the second-generation flow sensor design. Thermal anemometers sense gas flow by
delivering a constant amount of power to a heater and measuring the heater’s temperature, since
higher flow rates will increase self-cooling. Use of the microfabricated platinum heater as a
thermal anemometer was tested and found to not be viable for the flow rate range expected within
hydrocephalus shunts. However, by measuring the impedance across the heater, a much more
sensitive measure of heater self-cooling could be possible. Future work will test this method of
flow transduction and evaluate it for sensitivity over a wide flow velocity range; impedance-based
anemometry could be useful for high flow velocities, where calorimetric and time of flight flow
transduction are not viable. The next generation flow sensor design could consist of a central heater
with multiple electrode pairs both upstream and downstream, which is able to simultaneously
measure flow rate using calorimetric, time of flight, and anemometric flow transduction.

- 79 -
Additional Experiments

During sensor development and testing, several experiments were carried out to evaluate
the practical use of impedimetric flow sensors as chronically implantable devices. The first
experiment involved coating the flow sensors electrodes with electrodeposited platinum-iridium
(Pt-Ir), in the hopes that the fractal nature of the coating would lead to increased electrode surface
area and improve sensor sensitivity. Delamination between Parylene layers during chronic soaking
was also investigated, since this is the major failure mode expected during chronic in vivo use, and
a circuit model was developed and evaluated to explain the effects of delamination on
electrochemical impedance measurements. Finally, sensors were tested for 80 days in real human
CSF collected from patients with external ventricular drains at Los Angeles County Hospital.
2.6.1 Platinum-Iridium Coatings on Flow Sensors
Platinum-iridium coatings have been widely used in retinal and neural electrodes to lower
impedance and increase the charge storage capacity
109
. The measurement electrodes of flow
sensors were coated to determine if Pt-Ir coating would lower impedance and improve sensitivity.
Devices were coated by collaborators in Dr. James Weiland’s lab in partnership with Platinum
Group Coatings, LLC; Figure 2.37 shows microscope images of an electrode before and after
coating. Electrochemical impedance spectroscopy was performed before and after coating to gauge
the effectiveness of the coating in increasing surface area and reducing electrochemical impedance.
Impedance magnitude was seen to drop over all frequencies, and the beginning of the flat resistive
range was reduced from around 50 kHz to less than 100 Hz (Fig. 2.38).

Figure 2.37 Microscope image of a flow sensor electrode (A) before the coating process and (B) after being
coated with Pt-Ir. The rough, fractal nature of electroplated Pt-Ir causes it to appear dark under the
microscope.

- 80 -

Figure 2.38 Impedance magnitude and phase of a flow electrode before and after coating with platinum-
iridium. The lowest frequency available for sensor transduction dropped from 100 kHz to less than 100 Hz,
and impedance magnitude decreased by 81× at low frequencies.
Sensors with both flow electrodes coated with platinum-iridium were tested at flow
velocities between -400 and 800 µm/s. Due to the lower impedance across the electrodes, flow
measurements showed a higher sensitivity at the same frequency compared to sensors with
platinum electrodes (Fig. 2.39). Additionally, flow transduction was achieved at much lower
frequencies (1 kHz with Pt-Ir compared to 100 kHz with Pt), which could potentially reduce
measurement noise and simplify electronics used to record measurements. Drift during 12 hours
of consecutive measurements also decreased significantly, from 2.08%/hr before coating to
0.16%/hr after coating.

Figure 2.39 Left: Flow measurements taken with a sensor coated in platinum-iridium at 100 kHz, 10 kHz,
and 1 kHz, compared to measurements taken with a bare platinum sensor at 100 kHz. Right: Drift in flow
sensor measurements over a 12-hour period decreased from 2.1%/hr before Pt-Ir coating to negligible
(0.16%/hr) after Pt-Ir coating.

- 81 -
2.6.2 Delamination
One advantage in using Parylene C for chronically implantable devices is that it is passive
and does not naturally bond to things, which results in reduced biofouling. Unfortunately, it’s
reluctance to bond extends to other layers of Parylene C. Device constructed out of platinum and
Parylene layers can delaminate and fail if exposed to mechanical stress or chronic soaking (Fig.
2.40). In dry devices, delamination was often seen around the point where the potting epoxy met
the Parylene die. Delamination may preferentially occur at this location due to motion-induced
stress at the Parylene-epoxy interface or due to stress induced by epoxy shrinkage during curing.
In extreme cases, or in devices which have intrinsic stress due to thermal annealing, delamination
can propagate and cause complete device destruction. Delamination between platinum and
Parylene is often seen when devices are soaked in saline (Fig. 2.41). Water naturally penetrates
through Parylene and can pool at defect points between layers, weakening the Van der Waals
forces that hold adjacent layers together. Extreme delamination can be seen under the microscope
as wrinkling platinum on exposed electrode pads, shifting and disconnected platinum traces, or
Newton rings between Parylene layers.

Figure 2.40 Delamination occurs when adhesion is lost between adjacent Parylene C layers. (A) This can
occur due to movement causing stress at a mechanical interface, such as around the point that a Parylene
device is in contact with epoxy. (B) This delamination can propagate throughout the device and cause
complete separation of layers.
(A)
(B)

- 82 -

Figure 2.41 Delamination also occurs during soaking due to water penetration disrupting the Van der Waals
forces keeping the Parylene and platinum layers together. Here you can see platinum traces becoming
detached and moving around between Parylene layers in a severely delaminated device.
Even before delamination can be detected visually, electrical changes can be observed
using EIS. For a platinum electrode insulated between Parylene layers, early delamination is often
detectable via the presence of a second time constant in the EIS spectrum, appearing as a phase
peak at low frequencies (Fig. 2.42). This second time constant is due to an alternating conduction
path which develops through the Parylene insulation, and the peak’s position will shift with the
degree of delamination.

Figure 2.42 EIS across flow sensor electrodes with early signs of delamination, a low-frequency peak in
the phase, which results from signal transmission through liquid between traces.
To model electrode delamination during chronic soaking conditions, a circuit model was
developed which could be fit to EIS data (Fig. 2.43). An electrode under normal conditions was
modeled as a Randles circuit
104
, with a solution resistance RS, charge transfer resistance Rct and a
constant phase element Ydl to represent the double layer capacitance. A constant phase element is
1 mm

- 83 -
a circuit element which has impedance 𝑍 =
1
𝑌 (𝑗𝜔 )
𝑎 ⁄
where a is between 0 and 1. Constant phase
elements have been shown to more accurately model the double-layer effects of the electrode-
electrolyte interface than a capacitor, although modeling as a capacitor is sufficient for many
applications. Delamination is represented as an alternate conduction path from the electrode to the
solution, with Rdelam representing resistance through the insulation and constant phase element
Ydelam representing the capacitance through the insulation between electrode trace and electrolyte.
A capacitance Cwire was also included to model parasitic capacitances within connecting cables,
which causes a phase rolloff at high frequencies. This model was constructed in Gamry Echem
Analyst and fit to EIS data from eight platinum electrodes soaked in 37°C saline for 14 days. All
electrodes measured 300 x 1500 µm
2
and were constructed from 2000Å thick platinum insulated
between 10 µm thick Parylene layers on the same die. EIS measurements were recorded between
each electrode and a large platinum counter electrode using a Gamry Reference 600 potentiostat,
with an Ag/AgCl electrode used as a reference.

Figure 2.43 Circuit model of the electrical response of a delaminated electrode. The solution and electrode-
electrolyte interface are modeled as a Randles circuit, with a constant phase element (Y dl & a dl) used for the
double layer capacitance. Delamination is modeled as an alternate conducting path to the solution
resistance, with a delamination resistance (R delam) modeling the resistive path and another constant phase
element (Y cross & a cross) modeling the capacitance across the insulation. A wire capacitance (C wire) is also
included to model parasitic capacitance within connection cables, which is responsible for the phase rolloff
observed at high frequencies.
Modeling results show that both solution resistance and delamination resistance drop
dramatically after only one day of soaking, while the interface capacitances gradually increase in
magnitude over the testing period (Fig. 2.44). The dramatic, 2 order of magnitude drop in
delamination resistance may be due to either water penetration through the Parylene or the early
stages of delamination, and the small drop in solution resistance may be due to delamination
around the exposed electrode, which would increase the electrode’s effective surface area. More
studies are needed to isolate the electrochemical effects of delamination on thin-film devices.

- 84 -

Figure 2.44 The results of modeling EIS spectra between a thin-film platinum electrode insulated between
10 µm Parylene C layers and a large platinum counter electrode during a 14-day soak in PBS at 37°C. Both
R S and R delam drop after the first day, while the magnitudes of both the double layer capacitance and the
cross-insulation capacitance steadily increase over the course of the test.
The effects of delamination are readily apparent when operating impedimetric thermal flow
sensors. The earliest stages of delamination cause small spikes in impedance to appear when the
heater is turned on or off, which can obscure sensor measurement or cause drift. Later stages of
delamination result in a large impedance increase and complete loss of useable signal. The reason
for these impedance spikes is unclear, and could be to thermal expansion of delaminated Parylene,
(A) (B)
(C) (D)

- 85 -
electrostatic forces between parts of the sensor, or crosstalk between delaminated heater traces and
electrodes.

Figure 2.45 Flow electrodes responding to heater activation on (left) a partially delaminated device and
(right) a completely delaminated device.
2.6.3 Human CSF Tests
To ensure that benchtop tests were representative of in vivo conditions, and to ensure that
sensors would operate correctly in real physiological fluid, sensors were tested in human CSF
collected from patients with external ventricular drains at LA County Hospital. Human CSF was
obtained using USC Institutional Review Board (IRB) approved protocols and all experiments
were conducted per USC Institutional Biosafety Committee (IBC) approved procedures (IRB#
HS-14-00,608, USC IBC# BUA-15-00,026). Two flow sensors with electrodes spaced 0.5 mm
away from the heater were tested in both 1x PBS and human CSF at room temperature using flow
velocities between 0 and 200 µm/s (Fig. 2.46). There was no significant difference in either the
slope response (used for time-of-flight flow sensing) or the impedance dip response (used for
calorimetric flow sensing).

- 86 -

Figure 2.46 Flow response curve of an R-type time-of-flight flow sensor (electrodes spaced 0.5 mm away
from the heater) in both human CSF and 1x PBS. There was no significant difference in sensor response.
To test chronic use of flow sensors in physiological fluid, a luer lock spacer containing a flow
sensor was filled with human CSF and sealed closed. This allowed a flow sensor to be evaluated
for drift in a physiological fluid, as well as the difference between drift of the baseline electrode
impedance and drift of the normalized sensor response. Due to concerns that impedance
spectroscopy or cyclic voltammetry might accelerate device failure, the electrodes on the flow
sensor were not electrochemically evaluated beyond recording their baseline impedance and phase
at 100 kHz. The response of electrode pairs 0.5 mm and 2 mm away from the heater to heat pulses
was measured periodically at room temperature for 80 days. When testing the electrode pair 0.5
mm away from the heater (R2-1), the slope response showed little to no drift over the testing
period, while the impedance dip response increased for the first week before leveling off. The
response from the electrode pair 2 mm away from the heater (R2-2) showed much more variability
but little to no net drift (Fig. 2.47).

- 87 -

Figure 2.47 Response from a flow sensor tested in human CSF at room temperature for 80 days. (A) the
slope response of electrodes 0.5 mm away from the heater showed no significant drift, while (B) the
impedance dip response only showed drift during the first week of testing. (C, D) Electrodes 2 mm away
from the heater showed much higher relative variability but no net drift.
The baseline impedances of the flow electrodes were also measured during chronic soaking
in human CSF. Impedance magnitude steadily decreased throughout the experiment, while the
phase decreased over the first 14 days and then leveled off (Fig. 2.48). The change in baseline
impedance did not affect flow measurements, since the algorithm for transducing flow involves
normalizing all data to the baseline impedance before calculating slope or impedance dip. The
reason behind the steady drop in impedance is not clear. It is possible that changes in baseline
impedance result from water and ion penetration into the Parylene, but other groups have shown
that this happens over a much shorter time scale that what was observed here. It is therefore likely
that the observed baseline drift is due to progressive delamination between the exposed platinum
electrode and the Parylene substrate, which in its early stages would increase the effective
electrode surface area and result in a downward drift in impedance magnitude.
(A)
(B)
(C)
(D)

- 88 -

Figure 2.48 The impedance magnitude and phase at 100 kHz of flow sensor electrodes in human CSF
during an 80 day soak test at room temperature.

Discussion and Comparison to State of the Art
High-frequency electrochemical impedance in a two-electrode system was shown to be
highly sensitive to temperature, with a temperature coefficient an order of magnitude higher than
metal RTDs. However, the measured temperature coefficient of our electrodes (-1.2 %/°C) was
lower than the theoretical temperature coefficient of physiological fluids (-1.9%/°C), which may
be due to the influence of cable or trace impedances, which increase at higher frequencies. High-
frequency impedance tracked temperature over a physiologically relevant range (15 to 40°C) with
minimal hysteresis over five cycles, and the noise level for impedance measurements was
significantly lower that the noise for a platinum RTD fabricated on the same substrate, leading to
a 4× improvement in measurement resolution. A 14-hour test did show a small but linear drift in
baseline impedance which was unconnected to changes in ambient temperature, so further work is
necessary to characterize the use of impedance for chronic temperature sensing. Measuring
impedance in “pulses”, where the impedance measurement signal is only applied for a few seconds
at a time and the electrode-electrolyte interface is allowed to relax between measurements, may be
sufficient to prevent drift in the system. Alternatively, a reference electrode could be used to
stabilize the signal. Even without additional signal stabilization, the high sensitivity and low noise
seen when measuring impedance allows detection of sub-degree temperature fluctuations, and
could be used to implement the impedimetric flow sensor.
Results from time-of-flight flow sensors tracked closely with simulation results (Fig. 2.49).
However, experimental results showed a significantly lower dynamic range compared to the
simulation, most likely due to heat loss through the walls of the luer lock connector, which is
assumed to be zero in the simulation. At negative and low positive velocities, the minimum rate
of change was related to flow velocity as predicted by the simulation, but at high positive velocities

- 89 -
the rate of change began to deviate slightly from what the simulation predicted. This may have
been due to the use of constant current for heater activation, which would result in a small decrease
in heater temperature as the flow velocity increased. The dynamic range at higher flow velocities
could therefore be improved by using a constant temperature at the resistive heater, though this
would require more complex feedback circuitry.

Figure 2.49 A comparison between simulated results and experimental results, both with approx. 2°C
heating. The results show a similar pattern over the majority of the tested flow rate range, though due to
heat loss through the walls of the flow channel the experimental results have a lower dynamic range. ©
2016 IEEE
The decrease in sensor response with both increasing ambient temperature and increasing
ionic concentration also followed theoretical predictions, although the changes in sensitivity were
found to be insignificant within an order of magnitude difference in ionic concentration. However,
when used in an environment where large changes in ionic concentration are expected, the baseline
impedance of the sensing electrodes without heater activation could be used to calibrate sensor
response, though this would require a priori calibration over the expected ionic concentration
range. Changes in sensitivity with ambient temperature were small but significant compared to the
sensor’s dynamic range, but temperature compensation is easily accomplished by using the
baseline impedance or the heater resistance to measure ambient temperature.
The dynamic range of the time-of-flight sensor was relatively small when using a 1 C
overheat temperature, but was improved during operation with 2 and 10 °C overheat temperature.
Slightly increasing the overheat temperature results in noticeable improvements in sensor
resolution, and for ex vivo applications, overheat temperatures even higher than 10 °C may be
useful, as long as those temperatures do not affect the surrounding environment. Association for
the Advancement of Medical Instrumentation standards limit temperature increases in implants to

- 90 -
between 1 and 2 C
110
, so any implantable device will need to stay within this range, but even with
a minimal overheat temperature, the sensor can transduce flow velocity ranges relevant to
cerebrospinal fluid flow through hydrocephalus shunts
57
or external ventricular drains
111
as well
as capillary blood flow
112
. Arterial blood flow has a much higher average velocity, but even
without increasing overheat temperature, the sensor can be tailored to measure higher flows by
increasing the distance between electrode and heater.
When operated in the calorimetric mode, flow sensors were able to transduce ultra-low
flow rates with high precision. While the addition of an upstream electrode pair increases device
size and complexity, simultaneous impedance measurement from two electrode pairs significantly
improves measurement resolution. Using the custom multiplexer, a 2σ measurement resolution of
19.1 µm/s was achieved, which is significantly higher than any other reported impedance-based
flow sensor and a 4× improvement over sensors operated in time-of-flight mode.
Table 2.3 The calorimetric flow sensor compared with other EI flow sensors reported in literature
2,58,113,114
.
The sensor exhibits a 4× higher resolution at ultra-low flow velocities when operating in the calorimetric
mode. Resolution has been reported as twice the standard deviation (2σ) for all sensors. (ToF = Time of
Flight) © 2018 IEEE
Paper
Sensing
Method
Resolution
(µm/s)
Range
(µm/s)
Wu
2001
ToF, dissolved
O 2 tracer
- 16-250
Ayliffe
2003
Ionic pathway
changes
2x10
4
0-4.6x10
5

Yu 2015
ToF,
microbubble
tracer
896 0-2780
Baldwin
2016
ToF, thermal
tracer
86.6 ±800
Baldwin
2018
Calorimetric 19.1 ±200

Simultaneous measurement from multiple electrode pairs has the potential to improve sensor
robustness against biofouling, drift, or degradation, which are concerns for chronic in vivo
implementation. The range available for calorimetric sensing is limited for this design to ±200
µm/s, but both the measurement range and sensor resolution have potential for further
improvement via the addition of more electrode pairs at varying distances upstream and
downstream from the heater. No simulations were performed to model calorimetric sensing in this
flow regime, and the causes behind differences in flow response between maximum rate of change
and impedance dip have not been explored, but future work should include a robust investigation
of these effects and lead to an optimized flow sensor which uses multiple measurement modalities
to achieve highly accurate bidirectional flow measurement over a wide range of flow rates.

- 91 -
The small size of our sensor and its construction on a flexible substrate allow it to be easily
integrated into shunts, catheters, and other equipment, and this design could be further
miniaturized to access additional applications. Variation between devices was of course present,
which may have been a result of small deviations in the placement of the Parylene sensor die from
the center of the flow channel. Non-centered placement of the sensor would result in a lower flow
velocity at the sensor die, which may have caused a baseline shift. Before packaging an assumption
was made that, due to water’s high thermal conductivity, heat would spread to the region of highest
flow velocity quickly enough that small deviations from center would not have a noticeable effect
on sensor performance. Future work will test this assumption and evaluate whether an alternative
packaging scheme could decrease inter-device variability.
To be practical for chronic in vivo use, a sensor must have minimal drift over its projected
lifetime. Although these flow sensors were designed to avoid many of the causes of drift which
occur in silicon sensors, drift may still occur due to delamination of the Parylene C substrate,
cracking or failure of metal traces, or occlusion of the electrode sites themselves. Preliminary long-
term tests in room-temperature human CSF showed minimal drift over 80 days, but elevated
temperatures have been shown to accelerate delamination and the robustness of sensors to
vibration or high flow rates have not been evaluated. Future tests will involve placing sensors in a
mock hydrocephalus system and performing long-term tests using physiologically relevant
pressures, flow rates, and temperatures. Accelerated lifetime testing will also be performed to
evaluate material integrity for chronic applications, and adhesion promoters such as AdPro Plus or
AdPro Poly will be evaluated for improving device lifetime and eliminating long-term drift.
Biofouling is also a problem for devices in chronic contact with physiological fluid, so cyclic
voltammetry will be evaluated for potential in vivo cleaning.

- 92 -
Conclusion
Chronic in vivo flow sensing is an important unsolved problem with applications well
beyond smart hydrocephalus shunts. Current microfabricated flow sensors consist of
semiconductor junctions on silicon, which cannot last for long periods of time in the body, and use
high overheat temperatures which may damage tissue. I have attempted to solve this problem by
developing flow sensors consisting of platinum on a Parylene C substrate, which has a proven
track record of biocompatibility and long-term viability in vivo. By using the electrochemical
impedance across an electrode pair we were able to accurately measure fluid temperature via
changes in ionic mobility, with high sensitivities and resolutions that matched or even surpassed
that of semiconductor junctions without requiring any sort of encapsulation. We have also
developed a modified time-of-flight measurement method which uses the rate of change of
temperature at a point downstream of the heater to transduce flow velocity, and allows for
bidirectional flow measurement with only one temperature sensitive element. The addition of a
second upstream electrode pair allows for calorimetric sensing and further increases precision at
low flow rates. The high temperature sensitivity and novel method of transducing flow, combined
with the high thermal conductivity of physiological fluids, allowed us to measure flow with only
a 1°C overheat temperature. Flow sensors was tested under a wide range of ambient conditions
and are stable in fluids with many orders of magnitude difference in ionic concentration. Due to
high sensitivity, low temperature and low power operation, and fully biocompatible construction,
impedimetric flow sensors are ideal for chronic integration into hydrocephalus shunts.

- 93 -
References
1. A. Baldwin, L. Yu, and E. Meng, "An electrochemical-based thermal flow sensor," in
Micro Electro Mechanical Systems (MEMS), 2016 IEEE 29th International Conference
on, 2016, pp. 341-344.
2. A. Baldwin, L. Yu, and E. Meng, "An Electrochemical Impedance-Based Thermal Flow
Sensor for Physiological Fluids," Journal of Microelectromechanical Systems, vol. 25,
pp. 1015-1024, 2016.
3. A. Baldwin, T. Hudson, and E. Meng, "A calorimetric flow sensor for ultra-low flow
applications using electrochemical impedance," presented at the 31st IEEE Conference
on Micro Electro Mechanical Systems (MEMS), Belfast, North Ireland, 2018.
4. C. Eskey, N. Wolmark, C. McDowell, M. Domach, and R. Jain, "Residence time
distributions of various tracers in tumors: implications for drug delivery and blood flow
measurement," JNCI: Journal of the National Cancer Institute, vol. 86, pp. 293-299,
1994.
5. M. Weiss, O. Bänziger, T. Neff, and S. Fanconi, "Influence of infusion line compliance
on drug delivery rate during acute line loop formation," Intensive care medicine, vol. 26,
pp. 776-779, 2000.
6. F. Theeuwes and S. Yum, "Principles of the design and operation of generic osmotic
pumps for the delivery of semisolid or liquid drug formulations," Annals of biomedical
engineering, vol. 4, pp. 343-353, 1976.
7. H. Eerikäinen, W. Watanabe, E. I. Kauppinen, and P. P. Ahonen, "Aerosol flow reactor
method for synthesis of drug nanoparticles," European Journal of Pharmaceutics and
Biopharmaceutics, vol. 55, pp. 357-360, 2003.
8. Q. Xu, M. Hashimoto, T. T. Dang, T. Hoare, D. S. Kohane, G. M. Whitesides, et al.,
"Preparation of monodisperse biodegradable polymer microparticles using a microfluidic
flow ‐focusing device for controlled drug delivery," Small, vol. 5, pp. 1575-1581, 2009.
9. Z. Darzynkiewicz, S. Bruno, G. Del Bino, W. Gorczyca, M. Hotz, P. Lassota, et al.,
"Features of apoptotic cells measured by flow cytometry," Cytometry Part A, vol. 13, pp.
795-808, 1992.
10. S. Shoji and M. Esashi, "Microflow devices and systems," Journal of Micromechanics
and Microengineering, vol. 4, p. 157, 1994.
11. M. R. Miller, J. Hankinson, V. Brusasco, F. Burgos, R. Casaburi, A. Coates, et al.,
"Standardisation of spirometry," European respiratory journal, vol. 26, pp. 319-338,
2005.
12. F. W. Dekker, A. C. Schrier, P. J. Sterk, and J. H. Dijkman, "Validity of peak expiratory
flow measurement in assessing reversibility of airflow obstruction," Thorax, vol. 47, pp.
162-166, 1992.
13. M. Hetzel and T. Clark, "Comparison of normal and asthmatic circadian rhythms in peak
expiratory flow rate," Thorax, vol. 35, pp. 732-738, 1980.
14. K. K. Shung, Diagnostic ultrasound: Imaging and blood flow measurements: CRC press,
2015.
15. R. W. Gill, "Measurement of blood flow by ultrasound: accuracy and sources of error,"
Ultrasound in Medicine and Biology, vol. 11, pp. 625-641, 1985.
16. S. Nishino, N. Watanabe, T. Nakama, T. Takei, K. Ashikaga, N. Kuriyama, et al., "Direct
measurement of coronary flow during a vasospastic angina attack by transthoracic
Doppler echocardiography," Journal of echocardiography, vol. 15, pp. 88-90, 2017.

- 94 -
17. H. Kaneda, T. Irino, T. Minami, and M. Taneda, "Diagnostic reliability of the
percutaneous ultrasonic Doppler technique for vertebral arterial occlusive diseases,"
Stroke, vol. 8, pp. 571-579, 1977.
18. R. W. Martin and D. Watkins, "An ultrasonic catheter for intravascular measurement of
blood flow: Technical details," IEEE Transactions on Sonics and Ultrasonics, vol. 27,
pp. 277-286, 1980.
19. A. A. Linninger, M. Xenos, D. C. Zhu, M. R. Somayaji, S. Kondapalli, and R. D. Penn,
"Cerebrospinal fluid flow in the normal and hydrocephalic human brain," IEEE
Transactions on Biomedical Engineering, vol. 54, pp. 291-302, 2007.
20. P. H. Luetmer, J. Huston, J. A. Friedman, G. R. Dixon, R. C. Petersen, C. R. Jack, et al.,
"Measurement of cerebrospinal fluid flow at the cerebral aqueduct by use of phase-
contrast magnetic resonance imaging: technique validation and utility in diagnosing
idiopathic normal pressure hydrocephalus," Neurosurgery, vol. 50, pp. 534-543, 2002.
21. P. Gideon, F. Ståhlberg, C. Thomsen, F. Gjerris, P. S. Sørensen, and O. Henriksen,
"Cerebrospinal fluid flow and production in patients with normal pressure hydrocephalus
studied by MRI," Neuroradiology, vol. 36, pp. 210-215, 1994.
22. W. G. Bradley Jr, D. Scalzo, J. Queralt, W. N. Nitz, D. J. Atkinson, and P. Wong,
"Normal-pressure hydrocephalus: evaluation with cerebrospinal fluid flow measurements
at MR imaging," Radiology, vol. 198, pp. 523-529, 1996.
23. H. L. Liu, Y. Pu, Y. Liu, L. Nickerson, T. Andrews, P. T. Fox, et al., "Cerebral blood
flow measurement by dynamic contrast MRI using singular value decomposition with an
adaptive threshold," Magnetic resonance in medicine, vol. 42, pp. 167-172, 1999.
24. J. Lotz, C. Meier, A. Leppert, and M. Galanski, "Cardiovascular flow measurement with
phase-contrast MR imaging: basic facts and implementation," Radiographics, vol. 22, pp.
651-671, 2002.
25. N. Nguyen, "Micromachined flow sensors—a review," Flow measurement and
Instrumentation, vol. 8, pp. 7-16, 1997.
26. N. Svedin, E. Stemme, and G. Stemme, "A static turbine flow meter with a
micromachined silicon torque sensor," Journal of microelectromechanical systems, vol.
12, pp. 937-946, 2003.
27. Y.-H. Wang, C.-Y. Lee, and C.-M. Chiang, "A MEMS-based air flow sensor with a free-
standing micro-cantilever structure," Sensors, vol. 7, pp. 2389-2401, 2007.
28. S. Zhang, L. Lou, and C. Lee, "Piezoresistive silicon nanowire based
nanoelectromechanical system cantilever air flow sensor," Applied Physics Letters, vol.
100, p. 023111, 2012.
29. L. Yuan, J. Yang, and Z. Liu, "A compact fiber-optic flow velocity sensor based on a
twin-core fiber Michelson interferometer," IEEE sensors journal, vol. 8, pp. 1114-1117,
2008.
30. Y. Zhao, K. Chen, and J. Yang, "Novel target type flowmeter based on a differential fiber
Bragg grating sensor," Measurement, vol. 38, pp. 230-235, 2005.
31. K. Takahata, A. DeHennis, K. D. Wise, and Y. B. Gianchandani, "A wireless
microsensor for monitoring flow and pressure in a blood vessel utilizing a dual-inductor
antenna stent and two pressure sensors," in Micro Electro Mechanical Systems, 2004.
17th IEEE International Conference on.(MEMS), 2004, pp. 216-219.

- 95 -
32. R. Oosterbroek, T. S. Lammerink, J. W. Berenschot, G. J. Krijnen, M. C. Elwenspoek,
and A. van den Berg, "A micromachined pressure/flow-sensor," Sensors and Actuators
A: Physical, vol. 77, pp. 167-177, 1999.
33. J. T. W. Kuo, L. Yu, and E. Meng, "Micromachined Thermal Flow Sensors—A Review,"
Micromachines, vol. 3, pp. 550-573, 2012.
34. S. Wu, Q. Lin, Y. Yuen, and Y.-C. Tai, "MEMS flow sensors for nano-fluidic
applications," Sensors and Actuators A: Physical, vol. 89, pp. 152-158, 2001.
35. M. Ashauer, H. Glosch, F. Hedrich, N. Hey, H. Sandmaier, and W. Lang, "Thermal flow
sensor for liquids and gases," in Micro Electro Mechanical Systems, 1998. MEMS 98.
Proceedings., The Eleventh Annual International Workshop on, 1998, pp. 351-355.
36. T. K. Hsiai, S. K. Cho, P. K. Wong, M. H. Ing, A. Salazar, S. Hama, et al., "Micro
sensors: linking real-time oscillatory shear stress with vascular inflammatory responses,"
Annals of biomedical engineering, vol. 32, pp. 189-201, 2004.
37. F. Kohl, A. Jachimowicz, J. Steurer, R. Glatz, J. Kuttner, D. Biacovský, et al., "A
micromachined flow sensor for liquid and gaseous fluids," Sensors and Actuators A:
Physical, vol. 41, pp. 293-299, 1994.
38. A. Glaninger, A. Jachimowicz, F. Kohl, R. Chabicovsky, and G. Urban, "Wide range
semiconductor flow sensors," Sensors and Actuators A: Physical, vol. 85, pp. 139-146,
2000.
39. H. Ernst, A. Jachimowicz, and G. A. Urban, "High resolution flow characterization in
Bio-MEMS," Sensors and Actuators A: Physical, vol. 100, pp. 54-62, 2002.
40. T. S. Lammerink, N. R. Tas, M. Elwenspoek, and J. H. Fluitman, "Micro-liquid flow
sensor," Sensors and Actuators A: Physical, vol. 37, pp. 45-50, 1993.
41. A. Vanhoestenberghe and N. Donaldson, "Corrosion of silicon integrated circuits and
lifetime predictions in implantable electronic devices," Journal of Neural Engineering,
vol. 10, p. 031002, 2013.
42. R. Ahrens and K. Schlote-Holubek, "A micro flow sensor from a polymer for gases and
liquids," Journal of Micromechanics and Microengineering, vol. 19, p. 074006, 2009.
43. S. Shibata, Y. Niimi, and M. Shikida, "Flexible thermal MEMS flow sensor based on Cu
on polyimide substrate," in IEEE SENSORS 2014 Proceedings, 2014, pp. 424-427.
44. E. Meng, P.-Y. Li, and Y.-C. Tai, "A biocompatible Parylene thermal flow sensing
array," Sensors and Actuators A: Physical, vol. 144, pp. 18-28, 5/28/ 2008.
45. C. Li, P.-M. Wu, J. A. Hartings, Z. Wu, C. H. Ahn, and R. K. Narayan, "Cerebral blood
flow sensor with in situ temperature and thermal conductivity compensation," in Micro
Electro Mechanical Systems (MEMS), 2012 IEEE 25th International Conference on,
2012, pp. 1021-1024.
46. H. Yu, L. Ai, M. Rouhanizadeh, D. Patel, E. S. Kim, and T. K. Hsiai, "Flexible polymer
sensors for in vivo intravascular shear stress analysis," Microelectromechanical Systems,
Journal of, vol. 17, pp. 1178-1186, 2008.
47. R. Vilares, C. Hunter, I. Ugarte, I. Aranburu, J. Berganzo, J. Elizalde, et al., "Fabrication
and testing of a SU-8 thermal flow sensor," Sensors and Actuators B: Chemical, vol. 147,
pp. 411-417, 2010.
48. C. Li, P.-M. Wu, J. Han, and C. H. Ahn, "A flexible polymer tube lab-chip integrated
with microsensors for smart microcatheter," Biomedical microdevices, vol. 10, pp. 671-
679, 2008.
49. Sensirion, "SLI Liquid Flow Meter Series," ed: Sensirion, 2015.

- 96 -
50. I. POSiFA Microsystems, "PTFD10 Thermal Flow Sensor Die," ed, 2016.
51. H. Stephensen, N. Andersson, A. Eklund, J. Malm, M. Tisell, and C. Wikkelsö,
"Objective B wave analysis in 55 patients with non-communicating and communicating
hydrocephalus," Journal of Neurology, Neurosurgery & Psychiatry, vol. 76, pp. 965-970,
2005.
52. D. C. Zhu, M. Xenos, A. A. Linninger, and R. D. Penn, "Dynamics of lateral ventricle
and cerebrospinal fluid in normal and hydrocephalic brains," Journal of Magnetic
Resonance Imaging, vol. 24, pp. 756-770, 2006.
53. T. Matsumoto, H. Nagai, Y. Kasuga, and K. Kamiya, "Changes in intracranial pressure
(ICP) pulse wave following hydrocephalus," Acta neurochirurgica, vol. 82, pp. 50-56,
1986.
54. H. Kimelberg, "Water homeostasis in the brain: basic concepts," Neuroscience, vol. 129,
pp. 851-860, 2004.
55. M. Edsbagge, M. Tisell, L. Jacobsson, and C. Wikkelso, "Spinal CSF absorption in
healthy individuals," American Journal of Physiology-Regulatory, Integrative and
Comparative Physiology, vol. 287, pp. R1450-R1455, 2004.
56. M. Matsumae, O. Sato, K. Itoh, T. Fukuda, and Y. Suzuki, "Quantification of
cerebrospinal fluid shunt flow rates," Child's Nervous System, vol. 5, pp. 356-360, 1989.
57. M. Hara, C. Kadowaki, Y. Konishi, M. Ogashiwa, M. Numoto, and K. Takeuchi, "A new
method for measuring cerebrospinal fluid flow in shunts," Journal of neurosurgery, vol.
58, pp. 557-561, 1983.
58. L. Yu, B. J. Kim, and E. Meng, "An implantable time of flight flow sensor," in 2015 28th
IEEE International Conference on Micro Electro Mechanical Systems (MEMS), 2015,
pp. 620-623.
59. E. Meng, X. Zhang, and W. Benard, "Additive processes for polymeric materials," in
MEMS Materials and Processes Handbook, ed: Springer, 2011, pp. 193-271.
60. B. J. Kim and E. Meng, "Micromachining of Parylene C for bioMEMS," Polymers for
Advanced Technologies, 2015.
61. E. Meng, Biomedical microsystems: CRC Press, 2011.
62. "SCS Parylene Properties," ed: Specialty Coating Systems.
63. L. Yu, C. A. Gutierrez, and E. Meng, "An Electrochemical Microbubble-Based MEMS
Pressure Sensor," Microelectromechanical Systems, Journal of, vol. 25, pp. 144-152,
2016.
64. B. J. Kim, W. Jin, A. Baldwin, L. Yu, E. Christian, M. D. Krieger, et al., "Parylene
MEMS patency sensor for assessment of hydrocephalus shunt obstruction," Biomedical
microdevices, vol. 18, p. 87, 2016.
65. C. A. Gutierrez and E. Meng, "Impedance-based force transduction within fluid-filled
parylene microstructures," Journal of Microelectromechanical Systems, vol. 20, pp.
1098-1108, 2011.
66. J. T. Kuo, B. J. Kim, S. A. Hara, C. D. Lee, C. A. Gutierrez, T. Q. Hoang, et al., "Novel
flexible Parylene neural probe with 3D sheath structure for enhancing tissue integration,"
Lab on a Chip, vol. 13, pp. 554-561, 2013.
67. B. Kim, J. Kuo, S. Hara, C. Lee, L. Yu, C. Gutierrez, et al., "3D Parylene sheath neural
probe for chronic recordings," Journal of neural engineering, vol. 10, p. 045002, 2013.

- 97 -
68. S. Takeuchi, D. Ziegler, Y. Yoshida, K. Mabuchi, and T. Suzuki, "Parylene flexible
neural probes integrated with microfluidic channels," Lab on a Chip, vol. 5, pp. 519-523,
2005.
69. P.-J. Chen, D. C. Rodger, S. Saati, M. S. Humayun, and Y.-C. Tai, "Microfabricated
implantable parylene-based wireless passive intraocular pressure sensors," Journal of
Microelectromechanical Systems, vol. 17, pp. 1342-1351, 2008.
70. P.-Y. Li, J. Shih, R. Lo, S. Saati, R. Agrawal, M. S. Humayun, et al., "An
electrochemical intraocular drug delivery device," Sensors and Actuators A: Physical,
vol. 143, pp. 41-48, 2008.
71. Y. Nam, "Material considerations for in vitro neural interface technology," MRS bulletin,
vol. 37, pp. 566-572, 2012.
72. L. S. Robblee and T. Rose, "The electrochemistry of electrical stimulation," in
Engineering in Medicine and Biology Society, 1990., Proceedings of the Twelfth Annual
International Conference of the IEEE, 1990, pp. 1479-1480.
73. S. F. Cogan, "Neural stimulation and recording electrodes," Annu. Rev. Biomed. Eng.,
vol. 10, pp. 275-309, 2008.
74. C. WALTON, S. GERGELY, and A. P. ECONOMIDES, "Platinum Pacemaker
Electrodes: Origins and Effects of the Electrode ‐Tissue Interface Impedance," Pacing
and Clinical Electrophysiology, vol. 10, pp. 87-99, 1987.
75. S. Tesfaye, J. Watt, S. J. Benbow, K. A. Pang, J. Miles, and I. A. MacFarlane, "Electrical
spinal-cord stimulation for painful diabetic peripheral neuropathy," The Lancet, vol. 348,
pp. 1698-1701, 1996.
76. T. Stöver and T. Lenarz, "Biomaterials in cochlear implants," GMS current topics in
otorhinolaryngology, head and neck surgery, vol. 8, pp. Doc10-Doc10, 2009.
77. W. M. Grill, "Safety considerations for deep brain stimulation: review and analysis,"
Expert review of medical devices, vol. 2, pp. 409-420, 2005.
78. A. V. Vasenkov, "Atomistic modeling of parylene-metal interactions for surface micro-
structuring," Journal of molecular modeling, vol. 17, pp. 3219-3228, 2011.
79. C. A. Gutierrez, C. Lee, B. Kim, and E. Meng, "Epoxy-less packaging methods for
electrical contact to parylene-based flat flexible cables," in 2011 16th International Solid-
State Sensors, Actuators and Microsystems Conference, 2011, pp. 2299-2302.
80. A. Volynskii, S. Bazhenov, O. Lebedeva, A. Ozerin, and N. Bakeev, "Multiple cracking
of rigid platinum film covering polymer substrate," Journal of applied polymer science,
vol. 72, pp. 1267-1275, 1999.
81. X. C. Ong, M. Forssell, and G. K. Fedder, "Processing of platinum electrodes for
parylene-C based neural probes," in Micro Electro Mechanical Systems (MEMS), 2016
IEEE 29th International Conference on, 2016, pp. 509-512.
82. P. Childs, J. Greenwood, and C. Long, "Review of temperature measurement," Review of
scientific instruments, vol. 71, pp. 2959-2978, 2000.
83. A. Tong, "Improving the accuracy of temperature measurements," Sensor Review, vol.
21, pp. 193-198, 2001.
84. S. Kasap, "Thermoelectric effects in metals: thermocouples," Canada: Department of
Electrical Engineering University of Saskatchewan, 2001.
85. H.-M. Chuang, K.-B. Thei, S.-F. Tsai, and W.-C. Liu, "Temperature-dependent
characteristics of polysilicon and diffused resistors," Electron Devices, IEEE
Transactions on, vol. 50, pp. 1413-1415, 2003.

- 98 -
86. Y. Kanda, R. Aoshima, and A. Takada, "Blood compatibility of components and
materials in silicon integrated circuits," Electronics Letters, vol. 17, pp. 558-559, 1981.
87. S. B. Baumann, D. R. Wozny, S. K. Kelly, and F. M. Meno, "The Electrical Conductivity
of Human Cerebrospinal Fluid at Body Temperature," IEEE Transactions on Biomedical
Engineering, vol. 44, 1997.
88. L. Raijmakers, D. Danilov, J. Van Lammeren, M. Lammers, and P. Notten, "Sensorless
battery temperature measurements based on electrochemical impedance spectroscopy,"
Journal of Power Sources, vol. 247, pp. 539-544, 2014.
89. R. R. Richardson, P. T. Ireland, and D. A. Howey, "Battery internal temperature
estimation by combined impedance and surface temperature measurement," Journal of
Power Sources, vol. 265, pp. 254-261, 2014.
90. J. J. Barron and C. Ashton, "The effect of temperature on conductivity measurement,"
TSP, vol. 7, 2005.
91. R. A. Robinson and R. H. Stokes, Electrolyte solutions: Courier Corporation, 2002.
92. M. Berkowitz and W. Wan, "The limiting ionic conductivity of Na+ and Cl− ions in
aqueous solutions: Molecular dynamics simulation," The Journal of chemical physics,
vol. 86, pp. 376-382, 1987.
93. M. Huber, R. Perkins, A. Laesecke, D. Friend, J. Sengers, M. Assael, et al., "New
international formulation for the viscosity of H2O," Journal of Physical and Chemical
Reference Data, vol. 38, pp. 101-125, 2009.
94. R. Zwanzig, "Dielectric friction on a moving ion. II. Revised theory," The Journal of
Chemical Physics, vol. 52, pp. 3625-3628, 1970.
95. H. J. Liebe, G. A. Hufford, and T. Manabe, "A model for the complex permittivity of
water at frequencies below 1 THz," International Journal of Infrared and Millimeter
Waves, vol. 12, pp. 659-675, 1991.
96. W. Ellison, "Permittivity of pure water, at standard atmospheric pressure, over the
frequency range 0–25 THz and the temperature range 0–100 C," J. Phys. Chem. Ref.
Data, vol. 36, pp. 1-18, 2007.
97. E. Nightingale Jr, "Phenomenological theory of ion solvation. Effective radii of hydrated
ions," The Journal of Physical Chemistry, vol. 63, pp. 1381-1387, 1959.
98. R. Schmid, A. M. Miah, and V. N. Sapunov, "A new table of the thermodynamic
quantities of ionic hydration: values and some applications (enthalpy–entropy
compensation and Born radii)," Physical Chemistry Chemical Physics, vol. 2, pp. 97-102,
2000.
99. L. Onsager, "Theories of Concentrated Electrolytes," Chemical Reviews, vol. 13, pp. 73-
89, 1933.
100. E. R. Kandel, J. H. Schwartz, T. M. Jessell, S. A. Siegelbaum, and A. Hudspeth,
Principles of neural science vol. 4: McGraw-hill New York, 2000.
101. C. Waymouth, "Osmolality of mammalian blood and of media for culture of mammalian
cells," In Vitro, vol. 6, pp. 109-127, 1970.
102. J. J. Barron and C. Ashton, "The Effect of Temperature on Conductivity Measurement,"
Reagecon, vol. TSP-07.
103. T. S. Light, S. Licht, A. C. Bevilacqua, and K. R. Morash, "The Fundamental
Conductivity and Resistivity of Water," Electrochemical and Solid-State Letters, vol. 8,
p. E16, 2005.

- 99 -
104. J. E. B. Randles, "Kinetics of rapid electrode reactions," Discussions of the faraday
society, vol. 1, pp. 11-19, 1947.
105. T. Stocker, Introduction to climate modelling: Springer Science & Business Media, 2011.
106. R. Goldstein, Fluid mechanics measurements: CRC Press, 1996.
107. E. Majchrzak, "The Finite Difference Method for transient convection-diffusion
problems," Scientific Research of the Institute of Mathematics and Computer Science,
vol. 11, pp. 63-72, 2012.
108. F. Mailly, A. Giani, R. Bonnot, P. Temple-Boyer, F. Pascal-Delannoy, A. Foucaran, et
al., "Anemometer with hot platinum thin film," Sensors and Actuators A: Physical, vol.
94, pp. 32-38, 10/31/ 2001.
109. A. Petrossians, J. J. Whalen, J. D. Weiland, and F. Mansfeld, "Electrodeposition and
characterization of thin-film platinum-iridium alloys for biological interfaces," Journal of
The Electrochemical Society, vol. 158, pp. D269-D276, 2011.
110. P. D. Wolf, "Thermal Considerations for the Design of an Implanted Cortical Brain-
Machine Interface (BMI)," in Indwelling Neural Implants: Strategies for Contending with
the In Vivo Environment, W. M. Reichert, Ed., ed Boca Raton (FL): CRC Press/Taylor &
Francis, 2008.
111. J. M. Drake, C. Sainte-Rose, M. DaSilva, and J.-F. Hirsch, "Cerebrospinal fluid flow
dynamics in children with external ventricular drains," Neurosurgery, vol. 28, pp. 242-
250, 1991.
112. E. N. Marieb and K. Hoehn, The Cardiovascular System: Blood Vessels, 9th Edition ed.:
Pearson Education, Inc., 2013.
113. J. Wu and W. Sansen, "Electrochemical time of flight flow sensor," Sensors and
Actuators A, pp. 68-74, 18 October 2001 2001.
114. H. E. Ayliffe and R. Rabbitt, "An electric impedance based microelectromechanical
system flow sensor for ionic solutions," Meas Sci Technol, vol. 14, pp. 1321-1327, Aug
2003.

- 100 -

HE flow sensor presented in the previous chapter is ideal for integration into hydrocephalus
shunts, but any implantable sensor requires additional electronics for wireless power and data
transfer. Typical electronics consist of silicon integrated circuits or discrete electronic components
that are susceptible to corrosion and failure in the human body. In this chapter, a new method of
wirelessly transducing electrochemical impedance, without integrated circuits or discrete electrical
components, is described and characterized
1
. The resonant frequency and impedance magnitude at
resonance of a planar inductive coil is affected by the load on a coupled secondary coil. If the
secondary coil’s load is composed of the electrochemical impedance between a pair of sensor
electrodes, wireless transduction of electrochemical impedance is possible. This reflected
impedance method could allow passive, wireless measurement from impedimetric sensors while
maintaining a flexible, biocompatible form factor. To test reflected impedance, flexible coils
terminating in exposed electrodes were fabricated from gold and Parylene C. Characterization
involved measuring the reflected impedance of these coils in phosphate-buffered saline solutions
of varying concentration. Both the resonant frequency and impedance at resonance were highly
sensitive to changes in solution conductivity, and additional experiments explored vertical
separation, lateral misalignment, and temperature changes. To apply reflected impedance to
hydrocephalus treatment, proximal catheter patency sensing using reflected impedance was
investigated, and prototype reflected impedance patency sensors are described and their
performance and limitations discussed. Applications of reflected impedance for biomarker
detection are also explored, and preliminary data on using thin-film reflected impedance coils to
detect glucose is included.

Chapter 3
PASSIVE, WIRELESS IMPEDIMETRIC
SENSORS USING REFLECTED IMPEDANCE
T

- 101 -
Background
3.1.1 Impedance measurement in the body
Numerous biomedical applications require electrochemical impedance measurement. In
vivo impedance measurement has been explored to predict heart failure
2
and monitor the
effectiveness of deep brain stimulation
3
, and impedance-based biosensors exist that monitor
biomarkers
4,5
or physical parameters such as the flow rate
6,7
and pressure
8,9
of cerebrospinal fluid
or the force applied to an intracortical probe
10,11
. However, standard in vivo impedance
measurement requires percutaneous leads for connection to external impedance measurement
electronics. A number of integrated circuits have been reported which measure impedance and
wirelessly transmit data out of the body
12-14
, but integrated circuit technology is fundamentally
incompatible with the aqueous ionic environment found in vivo
15
and requires bulky hermetic
packaging. Encapsulation by a biocompatible polymer such as Parylene C
16,17
could protect against
integrated circuit corrosion, but even the best polymer coatings experience water penetration and
failure under chronic in vivo conditions. In any case, discrete and integrated electronics add cost
and complexity to an implantable device. A passive method of measuring electrochemical
impedance without integrated circuits or discrete electronic components would allow for a wide
range of simple, long-lasting implantable devices.
3.1.2 Techniques for wireless signal transmission
The most popular methods for powering and communicating with implanted medical
devices include radiofrequency identification (RFID)-based techniques and inductive coupling.
RFID communication uses a custom integrated circuit to harvest energy from a radio signal for
device power
18-25
. Digital telemetry is achieved by modulating antenna admittance, creating a
backscatter signal which can be detected through careful signal processing. RFID devices can
transmit data over long distances in air (up to 10 meters
26
), but RF absorption by tissue (~0.4 W/kg)
causes significant signal attenuation
27,28
; power available to sensors is on the order of single
microwatts
29
, and devices require integrated circuits for power rectification and digital antenna
modulation
30,31
. Despite these limitations, RFID-based sensors have been widely explored for skin-
based
32-34
and implantable
20,35,36
applications.
Inductive coupling is another common method for powering and communicating with
implantable medical devices
14,22,37-42
. Inductive power and data transfer involves modulating the
magnetic field between two coils, allowing significant energy transfer from one coil to the other.
Similar to RFID, admittance modulation of the secondary coil allows digital signal transmission
back to the primary coil through the inductive link. Significant power transfer is possible for
inductively coupled systems, but the technique requires close proximity (less than the radius of the
transmitting coil
43,44
) and complex power electronics must still be implanted inside the body. Still,
inductive coupling has found widespread commercial use in cochlear
45,46
and retinal
40,47

stimulators.

- 102 -
A third method for wireless power and data transmission to an implanted sensor involves
the use of ultrasound
48-53
. Piezoelectric energy harvesters can utilize an externally applied
ultrasound signal to power implantable sensors, with a second piezoelectric actuator used to
transmit data. Ultrasound shows promise for in vivo use due to the low attenuation of ultrasound
in the body
54
, but mechanical stress and failure of piezoelectric transducers, the low amounts of
total power available for ultrasound devices, miniaturization, and issues protecting piezoelectric
components from corrosion present limitations that have thus far prevented commercial use.
RFID, inductive coupling, and ultrasound all transmit digital signals. Digital signal
transmission requires implanted integrated circuits to digitize the analog sensor signals and
modulate the transmission antenna. However, a passive wireless sensing technique exists which
directly transduces analog signals from an implantable sensor. By placing an inductive coil in
series with a capacitive sensor (such as a diaphragm-based pressure sensor), resonance frequency
shifts with sensor output
55-58
. If the coil’s quality factor is high enough, a phase dip representing
the device’s resonance can be detected externally via radio frequency or inductive coupling. This
principle has been used commercially to develop a passive implantable sensor for detecting atrial
pressure
59
, and has been widely explored in literature to measure intraocular pressure
60-62
, among
other applications. However, this method cannot be used with impedimetric sensors, since high-
frequency impedance is primarily resistive, instead of capacitive. This high resistive component
of impedance will reduce the quality factor of the phase dip by several orders of magnitude,
rendering it undetectable.
Table 3.1. A comparison of several methods of data transmission within the body. Work described in this
thesis is bolded.
Method Signal
Max Transmission
Distance (mm)
Frequency
Range
Avg. Antenna
Diameter (mm) Ref.
Inductive
Coupling
Digital 100
125 kHz –
25 MHz
15 14, 22, 32-37
RFID Digital 10,000
800 MHz –
5 GHz
71 18-25
Ultrasound Digital 100
720 kHz –
6 MHz
11 43-48
RF Coupling Analog 30
100 MHz –
1 GHz
3 50-53
Reflected
Impedance
Analog 6 5.4 MHz 30 1

3.1.3 Reflected Impedance
In this chapter, a novel method for passive wireless transduction of high-frequency
electrochemical impedance is presented. The resonance of an external primary coil is affected
through inductive coupling with a secondary coil in series with an impedimetric sensor. By placing
the sensor’s electrodes on the terminals of the secondary coil, multiple parameters of the primary

- 103 -
coil’s self-resonance peak can be used to transduce impedance. Since the primary coil is outside
the body, the primary resonance peak can be directly measured, enabling highly sensitive
transduction and avoiding many issues inherent in the device which require measuring the
secondary coil’s resonance. Sensor transduction involves measuring the height of the primary
coil’s self-resonance peak. A second, parallel method of reflected impedance transduction is
possibly via measuring frequency shifts in the primary resonance frequency, though this requires
the primary and secondary coil’s self-resonant frequencies to be close but not identical. A high
quality factor is not required for accurate and sensitive electrochemical impedance transduction
using reflected impedance, and this work achieves impedance transduction using secondary coils
with a quality factor <0.1. Limitations of reflected impedance include the device’s relatively large
size compared to the possible signal transmission distance (Fig. 3.1), and shifts in received signal
which occur due to vertical separation or lateral offset. However, this method remains the only
reported technique for passive wireless transduction of signals from resistive sensors. The
following chapter presents the theory behind reflected impedance and the construction and
characterization of passive, biocompatible, thin-film devices which can transduce electrochemical
impedance over a wide range of solution resistances. Reflected impedance is then applied to
patency sensing in hydrocephalus shunts and wireless glucose sensing.

Figure 3.1 Transmission distance versus antenna diameter for the reflected impedance test coils, compared
to passive, wireless sensors reported in literature that employ inductive coupling, RFID, ultrasound, or RF
coupling
1,14,22-25,40-42,51-54,57,60,62
. There is a roughly linear correlation between size and transmission distance
which is maintained across techniques. Reflected impedance devices perform worse than average, but
further developments in miniaturization and high-frequency operation promise future improvements.

- 104 -
Reflected Impedance Theory
3.2.1 Circuit models

Figure 3.2 The circuit model for the wireless sensing system: a primary coil is magnetically coupled to a
secondary coil which terminates in a pair of sensor electrodes. On the primary side (left), L P is the primary
coil inductance, C P is the parallel capacitance between windings, and R P is primary coil winding resistance.
On the secondary side (right), L S is the secondary coil inductance, R S is the trace resistance of the secondary
coil windings, C dl is the double layer capacitance of each electrode, and R Sol is the solution resistance
between electrodes. M represents the mutual inductance between primary and secondary coils. © 2017
Springer Nature
When a system of conductive coils is stimulated with a time-varying electric field, a
coupled magnetic field is induced which can be used for energy transfer. Inductive energy transfer
is commonly used in transformers, which transfer energy with minimal attenuation through a
common ferrite core
63
. More recently, energy transfer between planar coils separated by an air gap
has become common for charging smartphones and other consumer electronics, though significant
attenuation can occur and coils must be parallel and separated by a distance less than their radius
64
.
The magnitude of inductive energy transfer between planar coils depends on the geometry of the
coils, their orientation relative to each other, the frequency at which energy is transferred, and the
load on the secondary side. For most applications, the resonant frequencies of the primary and
secondary planar coils are matched and the two coils are positioned parallel and concentrically.
A system of two coupled coils can be modeled as a pair of RLC circuits, with mutual
induction between the two coil inductances. For an ideal case with no trace resistance on the coil,
the only loss in the system is through the load on the secondary coil. If the coil terminates in a pair
of electrodes exposed to solution, then at high frequencies, the load will consist of the solution
resistance between the electrodes. Therefore, the amount of energy absorbed by the secondary coil
will vary with changes in solution resistance. Since the primary and secondary coils are inductively
coupled, changes in the energy absorbed by the secondary coil can be transduced by measuring
the impedance reflected on to the primary coil.

- 105 -

Figure 3.3 The linearized circuit model of Figure 3.2, showing the leakage inductances (L P-M and L S-M).
Z represents the complex impedance of the circuit as seen from the primary coil terminals. © 2017 Springer
Nature
We used reflected impedance to wirelessly transduce the solution resistance between a pair
of electrodes in series with the secondary coil (Fig. 3.2). The circuit model for this system consists
of a primary side parallel RLC circuit with a resistance in series with the inductance to model the
winding resistance of the coil. The inductance is mutually coupled with a secondary circuit
consisting of a secondary coil inductance and winding resistance and a simplified Randles circuit
65

for the electrodes and solution; the Randles circuit includes the double layer capacitance of each
electrode and the solution resistance between them. This circuit can be linearized as shown in
Figure 3.3, where the primary and secondary coils are coupled by a mutual inductor of value 𝑀 =
𝑘 √ 𝐿 𝑃 𝐿 𝑆 , where k is a coupling coefficient which varies from 0 to 1 based on coil geometry and
orientation, and the leakage inductance due to incomplete coupling is represented as inductors of
value LP-M and LS-M. The impedance of this circuit as seen from the primary coil can be written
as
𝑍 =
(
𝐿 𝑃 𝐶 𝑃 ⁄ +
𝑅 𝑃 𝑗𝜔 𝐶 𝑃 ⁄ ) 𝑍 𝑆 − 𝑗 𝜔 𝑀 2
𝐶 𝑃 (
1
𝑗𝜔 𝐶 𝑃 ⁄ + 𝑗𝜔 𝐿 𝑃 + 𝑅 𝑃 ) 𝑍 𝑆 + 𝜔 2
𝑀 2

𝑍 𝑆 = 𝑗𝜔 𝐿 𝑠 + 𝑅 𝑆 +
2
𝑗𝜔 𝐶 𝑑𝑙
⁄ + 𝑅 𝑠 𝑜𝑙

To explore the effects of the secondary coil’s reflected impedance on the primary-side
impedance, the primary winding resistance can be assumed to be negligible (RP = 0) and the
impedance Z can be evaluated at the primary coil’s natural frequency 𝜔 =
1
√ 𝐿 𝑃 𝐶 𝑃 ⁄
. Under these
conditions the impedance is
𝑍 =
𝐿 𝑃 2
𝑍 𝑆 − 𝑗 𝑀 2
𝐿 𝑃 √ 𝐿 𝑃 𝐶 𝑃 𝑀 2

Thus, the magnitude of impedance at the primary resonant frequency is related to the
secondary coil’s impedance (ZS). As the coupling between the coils goes to zero (M = 0) the
impedance would become infinite. It is also clear that with the introduction of a secondary coil,
(1)
(2)

- 106 -
the resonant frequency of the system is no longer equal to the resonant frequency of the primary
coil alone, since there is an imaginary component inherent to the secondary impedance.
Equation (2) can be expanded under the condition of non-zero RP, and under the assumption
that the natural frequencies of the primary and secondary coils are equal, i.e. 𝜔 =
1
√ 𝐿 𝑃 𝐶 𝑃 ⁄
=
√
2
𝐿 𝑆 𝐶 𝑑𝑙
⁄ .

𝑍 =
𝐿 𝑃 2
( 𝑅 𝑆 + 𝑅 𝑆 𝑜𝑙 ) − 𝑗 (
𝐿 𝑃 𝑀 2
√ 𝐿 𝑃 𝐶 𝑃 + ( 𝑅 𝑆 + 𝑅 𝑆 𝑜𝑙 ) 𝑅 𝑃 𝐿 𝑃 √ 𝐿 𝑃 𝐶 𝑃 )
𝑀 2
+ 𝑅 𝑃 𝐿 𝑃 𝐶 𝑃 ( 𝑅 𝑆 + 𝑅 𝑆 𝑜𝑙 )

This expression reveals that both the real and imaginary components of the system’s
impedance are dependent on the load of the secondary coil, which for our system consists of the
secondary coil resistance (RS) and the solution resistance (Rsol). This suggests that the resonant
frequency, where the imaginary component of impedance is zero, can be used to transduce the
secondary load. By measuring impedance over a range of frequencies near the expected resonant
frequency of the primary coil, the resonant frequency of the system and the magnitude of the
impedance at resonance can both be identified, and either of these values can be used to transduce
solution resistance.
3.2.2 MATLAB simulations
The effect of changing the solution resistance on both the resonant frequency and the
impedance of the system at resonance was evaluated using MATLAB. A Simulink model of the
equivalent circuit in Figure 3.3 was developed, and the resonant frequency and impedance
magnitude at resonance were evaluated over a range of solution resistances, shown in Figure 3.4.
The relationship between solution resistance and both impedance magnitude and resonant
frequency over a range of coupling coefficients, secondary inductances, and secondary
capacitances were also evaluated. These simulations elucidated a direct relationship between both
the resonant frequency and impedance at resonance of the system and the solution resistance across
the electrodes. The sensitivity of this relationship increases with the coupling between the two
coils. The relationship between impedance at resonance and solution resistance is most sensitive
when the resonant frequencies (with load) of the primary and secondary coils are equal. More
interestingly, the relationship between resonant frequency and solution resistance displays an
inflection point where the primary and secondary resonance frequencies match. When the
secondary resonance is lower than the primary, a lower solution resistance will cause the system’s
resonant frequency to drop, and when the secondary resonance is higher, the full system’s resonant
frequency will rise. The secondary coil can be imagined as “pulling” the system’s resonant
frequency toward itself with a strength inversely proportional to the secondary load.
(3)

- 107 -

Figure 3.4 Results of simulating the effects of changing solution resistance on the reflected impedance as
seen from the primary coil. The (i) impedance at resonance and (ii) resonant frequency were both found to
change with solution resistance under order-of-magnitude estimates of circuit element values (R P = 1Ω, L P
= 30 µH, C P = 2 nF, L S = 20 µH, R S = 100Ω, C dl = 5 nF, k = 0.4). In another simulation, increasing the
coupling coefficient k was found to increase the sensitivity of both (iii) impedance magnitude and (iv)
resonant frequency to changes in solution resistance (R P = 1Ω, L P = 24 µH, C P = 2 nF, L S = 4 µH, R S = 1Ω,
C dl = 5 nF), and by varying the secondary inductance L S it was found that (v) the sensitivity of impedance
magnitude is maximized and (vi) the sensitivity of resonant frequency has an inflection point where the
primary and secondary resonant frequencies are the same. © 2017 Springer Nature

Test Coils
3.3.1 Design
To explore the use of reflected impedance with thin-film impedimetric sensors for wireless
in vivo sensing, test coils were designed and fabricated (Fig. 3.5 & 3.6). Coils were fabricated from
gold film on a thin substrate of Parylene C. Each coil had an outer diameter of 30 mm and consisted
of 100 µm wide and 200 nm thick gold traces which terminated in square electrodes with an
exposed area of 300 x 300 µm
2
. Three different types of coils were fabricated, with 1, 5, or 16
turns each.

- 108 -

Figure 3.5 (i) Thin-film gold coils were fabricated on a flexible Parylene C substrate. (ii) Each coil
terminated in two exposed electrodes of area 300 x 300 µm
2
. © 2017 Springer Nature

Figure 3.6 Three variations of the secondary coil, with 1, 5, or 16 turns each, were fabricated and tested.
All coils had an outer radius of 30 mm. © 2017 Springer Nature

3.3.2 Fabrication
Fabrication followed standard processes for Parylene-based MEMS devices
66
, which are
for the most part described in the preceding chapter. First, 10 µm of Parylene C was chemical
vapor deposited onto a silicon carrier wafer. 200 nm Au, with a 20 nm Ti adhesion layer, was
electron-beam deposited onto the Parylene and patterned via liftoff using a 2 µm layer of AZ 5214
image reversal photoresist (EMD Performance Materials). Gold was chosen instead of platinum
for the test coils due to its lower resistivity, which increases coil’s quality factor, and because
electron-beam gold deposition could be performed in-house. A second 10 µm layer of Parylene C
was deposited on top of the gold layer, and electrode pads were exposed using reactive ion etching
(RIE) in oxygen plasma, with 15 µm of AZ 4620 photoresist (EMD Performance Materials) used
as an etch mask. Finally, devices were cut out of the wafer by hand and released by gently peeling
while the wafer was immersed in deionized water.

- 109 -
3.3.3 Testing and characterization
The electrical properties of the thin-film test coils were first characterized, and then their
effectiveness for transducing solution resistance via the reflected impedance method was tested.
First, 1, 5, and 16 turn coils were directly characterized by measuring DC resistance and the
inductance and quality factor at the expected measurement frequency of 5 MHz. Next, the effects
of changing solution resistance on the coil’s reflected impedance was characterized using
phosphate-buffered saline (PBS) solutions of concentration 0.5×, 1×, 5×, and 10×, with >18
MΩ•cm deionized (DI) water used as a control. Further characterization was performed on a 16-
turn coil using 1× PBS to evaluate the effects of vertical separation, lateral offset, and changes in
solution temperature on reflected impedance. Vertical separation was achieved by stacking 1 mm
thick glass slides between the primary and secondary coils.
DC resistance was measured using a digital multimeter and inductance and quality factor
were measured using an HP 4285A precision LCR meter (75 kHz–30 MHz, 1 VPP). For reflected
impedance measurement, the precision LCR meter was attached to a ferrite-cored planar coil from
Wurth Electronics (24 µH, Q = 180 @125 kHz, 50 mm diameter, 22 turns), which served as a
primary coil. A custom LabVIEW program was used to sweep the measurement frequency while
measuring complex impedance. The microfabricated secondary coils were held in a petri dish with
a 1 mm thick base which was positioned directly on top of the primary coil (Fig. 3.7). For each
measurement, 100 µL of PBS or DI water was placed on the electrodes using a calibrated syringe
and impedance was measured using the LCR meter. The coil was triple-rinsed with DI water and
re-centered between each individual measurement. PBS concentrations of 0.5×, 1×, 5×, and 10×
correspond to solution conductivities of 8, 16, 80, and 160 mS/cm respectively, and 1× PBS is a
common analog for physiological fluids and is isotonic with cerebrospinal fluid.

Figure 3.7 (i) Microfabricated coils were placed in a petri dish with varying concentrations of PBS and
centered over a commercial primary coil connected to an LCR meter (ii) A 22-turn primary coil with a rated
resonant frequency of 5 MHz (Wurth Electronics, 24 µH, Q = 180@125kHz, 50mm diameter). © 2017
Springer Nature
For each measurement, the complex impedance was measured over a range of frequencies
around the expected primary resonant frequency, with a resolution of 1 kHz. Both the resonant
frequency and the magnitude of impedance at resonance were evaluated as possible impedance
transduction methods. The resonant frequency was defined as the frequency at which the system’s

- 110 -
response is fully real, i.e. the phase is zero, and the impedance at resonance is the magnitude of
impedance at the resonant frequency.

Test Coil Results
3.4.1 Basic coil properties
The DC resistance, quality factor and inductance values for the 1, 5 and 16 turn secondary
coils are presented in Table 2. At high frequencies, the skin effect can cause a significant increase
in the resistance of a wire or trace
67
, but as the thickness (200 nm) of the traces used here were
much thinner than the skin depth in gold at the MHz frequencies of interest (~10s of µm), the skin
effect can be considered negligible. Coils were tested flat, and the inductances and DC resistances
of curved coils have not yet been established.

Table 3.2 The DC resistance, and the inductance and quality factor at 5 MHz, of 1, 5, and 16 turn
microfabricated thin-film coils. © 2017 Springer Nature
Turns
DC
Resistance Inductance Q
1 136.4 Ω 0.121 µH 0.05
5 605.9 Ω 1.625 µH 0.08
16 3890 Ω 35.9 µH 0.33

To identify the resonant frequency of the coupled system, the primary coil’s impedance
was measured from 3 to 10 MHz, with DI water used to complete the circuit on the secondary coil.
A peak in system impedance and an inflection point in phase was seen around 5.4 MHz, which
corresponds to the expected resonant frequency of the primary coil (Figure 3.8).

- 111 -

Figure 3.8 (i) The impedance of the primary coil – secondary coil system between 3 and 10 MHz shows a
clear resonance peak close to 5.4 MHz. (ii) Both the resonant frequency and the impedance magnitude at
resonance were used to transduce solution resistance across the measurement electrodes on the secondary
coil. © 2017 Springer Nature

3.4.2 Conductivity measurement
A significant drop in the system impedance at resonance with increasing conductivity was
observed for all three coils; the 16 turn secondary coil showed the largest sensitivity. Despite
rinsing and re-centering coils by hand between measurements, measurement error was minimal;
the average standard deviation between measurements was 13.0 Ω.

Figure 3.9 The system impedance magnitude at resonance is related to the conductivity of solution across
the electrodes, which is inversely proportional to measured impedance at the sensing electrodes. Error bars
are present but not visible in most cases; the average standard deviation is 13.0 Ω. © 2017 Springer Nature

- 112 -
Resonant frequency was also sensitive to solution resistance changes, though for the 1-turn
coil, the effect was not significant and for the 5-turn coil the effect was very small. When testing
the 16-turn coil, however, the resonant frequency significantly increased with an increase in
conductivity. The average standard deviation between measurements was 659 Hz.

Figure 3.10 The resonant frequency is also related to conductivity across the electrodes of the 16-turn and
5-turn devices, but the effect is only substantial for the 16-turn device. Error bars are present; the average
standard deviation is 659 Hz. © 2017 Springer Nature

3.4.3 Coil offset and misalignment
Due to the high sensitivity of the 16-turn coil, this coil was used to assess the effects of
primary to secondary vertical separation and misalignment. When 1× PBS was applied to the
secondary coil, the system impedance at resonance increased in a somewhat linear manner as the
coils were separated, while the resonant frequency was found to drop between 1 mm and 2 mm of
separation but did not significantly change for greater distances (Fig. 3.11-i). However, when the
impedance at resonance and resonant frequency were tested across the full range of PBS
concentrations, the sensitivity of resonant frequency to secondary electrode impedance was found
to decrease with increasing separation between the primary and secondary coils (Fig. 3.11-ii, iii).

- 113 -

Figure 3.11 Reflected impedance of a 16-turn secondary coil was measured as the secondary coil was
separated vertically from the primary coil using 1 mm thick glass slides. (i) When 1× PBS was applied to
the secondary coil, the impedance at resonance increased linearly as the coils were separated, while the
resonant frequency saw a slight decrease and then relative stability. However, the sensitivity of both (ii) the
impedance at resonance and (iii) the resonant frequency to changes in solution conductivity decreased when
separated from the primary coil. © 2017 Springer Nature
Misalignment tests also showed a change in response for both the impedance at resonance
and the resonant frequency (Fig. 3.12). Both the resonant frequency and the impedance at
resonance increased during misalignment, leveling off at around 15 mm displacement, at the radius
of the secondary coil.

- 114 -

Figure 3.12 When the coil was offset from center, both the (i) impedance magnitude and the (ii) resonant
frequency increased. This increase began to level off once the secondary coil was offset 15 mm, which is
equal to its radius. © 2017 Springer Nature
3.4.4 Temperature measurement
Additional tests were performed on a 16-turn test coil to evaluate the response of reflected
impedance to changes in solution temperature. The solution resistance of an aqueous ionic solution
is proportional to temperature, and changes in temperature can be measured via the
electrochemical impedance between two electrodes
6
. When measuring reflected impedance, the
impedance magnitude at resonance decreased and the resonant frequency increased with increasing
temperature, which is consistent with the expected decrease in solution resistance (Fig. 3.13).

Figure 3.13 The (i) impedance magnitude at resonance decreased and the (ii) resonant frequency increased
when the solution temperature increased, which is consistent with a decrease in solution resistance due to
changes in ionic mobility when an ionic solution is heated. © 2017 Springer Nature

- 115 -
Reflected Impedance Patency Sensor
Protein and debris buildup in implanted catheters can cause progressive occlusion and
failure. Progressive occlusion due to buildup in the proximal catheter is the most common cause
of failure for hydrocephalus shunts
68-70
, but catheter occlusion is also a significant failure mode in
central venous catheters
71,72
, hemodialysis catheters
73
, and insulin pumps
74
. A patency sensor
which directly transduces partial catheter occlusion using the electrochemical impedance through
a ventricular catheter was previously developed, allowing doctors to monitor progressive blockage
and predict hydrocephalus shunt failure
75
. Here, a wireless patency sensor using reflected
impedance is presented. The reflected impedance patency sensor can transduce progressive
blockage without the use of integrated circuits or discrete electronic components.

3.5.1 Design and testing of first-generation sensor
Instead of thin-film microfabricated coils, the reflected impedance patency sensor used a
3D printed module designed to fit in line with currently-used hydrocephalus shunts (Fig. 3.14).
Modules were designed in SolidWorks and printed out of VeroClear material by Modern Tech,
Inc. 34 AWG copper wire was wrapped around a slit in its center, creating a 12-turn secondary
coil (L = 4.31 µH @ 1 MHz, Q = 23.4, RDC = 1.2 Ω). To simulate proximal catheters with varying
degrees of occlusion, silicone tubes 1 mm in diameter were sealed at one end and bored with holes
500 µm in diameter, matching commercial proximal catheters. These tubes were placed in a beaker
containing 100 mL of 1x PBS and filled. Each end of the coil was connected to a platinum
electrode, with one electrode inserted into a silicone catheter above the holes and one electrode
placed in the beaker of PBS. The 3D printed module was centered above a commercial primary
coil and reflected impedance was measured non-consecutively through catheters with 12, 8, 4, 2,
and 0 open holes.

Figure 3.14 (i) A secondary coil for measuring catheter patency in hydrocephalus shunts, made by
wrapping 34 AWG copper wire around a 3D-printed module compatible with current shunt technology. L

- 116 -
= 4.31µH, Q = 23.4@1MHz, R DC = 1.2Ω. (ii) A mock proximal catheter, consisting of a silicone tube 1 mm
wide with 12 open holes of 500 µm diameter each. © 2017 Springer Nature

3.5.2 Experimental results
The impedance through these catheters ranged from 26.5 kΩ for a catheter with 12 open
holes to 31.5 kΩ for a catheter with 2 open holes, which according to simulations is outside the
region of maximum sensitivity for reflected impedance. Despite this, progressive blockage could
be clearly transduced by measuring the impedance magnitude at resonance (Fig. 3.15). There was
no significant change in the resonant frequency between catheters with different numbers of holes.

Figure 3.15 The impedance at resonance increased with decreasing number of holes, indicating that
reflected impedance could be used to predict proximal catheter occlusion

3.5.3 Second-generation patency sensor design
A second-generation reflected impedance patency sensor was designed and
tested as an attempt to increase sensitivity to progressive blockage and
improve compatibility with hydrocephalus shunts. The second-generation
sensor was designed as a drop-in replacement for existing CSF reservoirs,
which sit in series with hydrocephalus shunts at the point where the
proximal catheter exits the skull. The module consisted of a wide, flat
reservoir meant to sit above the skull with two connection points, one
pointed downwards through the skull toward the brain’s ventricles and one
pointed outwards toward the shunt valve and distal catheter. The module contained a groove
Figure 3.16 Diagram
of a patency sensing
reservoir design.

- 117 -
around the outside for the secondary coil, a pass-through hole so that an electrode could be placed
inside the reservoir, and an indentation on the bottom to place the second electrode as close to the
ventricles as possible. A silicone cap was also designed, to allow CSF sampling.
After designing the module in SolidWorks, a prototype for initial testing was 3D printed
out of ABS plastic in the USC undergrad fab lab (Figs. 3.16, 3.17). 30 loops of 34 AWG copper
wire were used to create the secondary coil. Each end of the coil was attached to a platinum wire
electrode, which could be placed in a beaker of PBS or inserted into a mock catheter.

Figure 3.17 SolidWorks model (left) and 3D-printed prototype (right) of a second generation reflected
impedance patency sensor. The second-generation prototype consisted of a 30-loop coil of 34 AWG copper
wire, which terminated in platinum wire electrodes (not shown).

3.5.4 Preliminary second-generation results
For the reflected impedance method to work, the primary and secondary coils must have
the same or very close resonant frequencies. The primary coil’s resonance frequency is set by the
coil inductance and the parasitic capacitances between traces, but for the thin-film test coils the
double layer capacitance across the electrodes determined the resonance frequency. Unfortunately,
platinum wire electrodes have too large a double-layer capacitance and prevents resonance near
the 5.4 MHz primary coil resonance frequency. Therefore, the use of the parasitic capacitance
within the secondary coil to tune the resonant frequency was explored. Coils with varying numbers
of loops were tested on the 3D printed second-generation module, revealing that 30 loops resulted
in a resonant frequency around 5.4 MHz. However, patency transduction tests showed no
significant difference in signal when the electrodes were placed in mock proximal catheters with
varying numbers of holes, although a large difference in the reflected impedance signal was
observed between the bare primary coil, the dry secondary coil, and the secondary coil with
electrodes in PBS (Fig. 3.18).
5 mm

- 118 -

Figure 3.18 (A) The second-generation sensor showed a large shift in the reflected impedance signal when
electrodes were placed in saline. (B) However, when electrodes were placed in the mock proximal catheters
there was no significant difference in signal between catheters with different numbers of open holes.

The low response of the patency sensor tuned using parasitic capacitance is most likely due
to signals preferentially traveling between coil traces and bypassing the electrodes entirely. To
avoid this problem and to force signals to travel through both the coil and the electrodes, resonance
frequency tuning using a discrete tuning capacitor in series with the coil and electrodes was tested
(A)
(B)

- 119 -
(Fig. 3.19). The secondary coil was re-wound with 10 turns so that parasitic capacitance would be
negligible between 1-10 MHz, and a discrete capacitor was chosen to give a resonant frequency of
5.4 MHz. Tests with this system again showed a significant difference in reflected impedance
signal between dry catheters and catheters filled with PBS, but there was no significant difference
between catheters with different numbers of holes (Fig. 3.20).

Figure 3.19 To avoid the resonance produced by the parasitic inter-trace capacitance within the coil,
secondary coil tuning was attempted using a discrete tuning capacitor in series with the coil and electrodes.

Figure 3.20 Tests with the second-generation prototype + tuning capacitor showed that the reflected
impedance signal was highly sensitive to the presence or absence of fluid in the mock catheters, but there
was still no significant difference between catheters with different numbers of holes.
As a final test using a discrete tuning capacitor, an idealized system was constructed and
tested using two store-bought coils. A Wurth ferrite-cored coil with a 25 MHz self-resonance was
used as a secondary coil and was tuned down 5.4 MHz using a series capacitor (Fig. 3.21). Instead
of electrodes, resistors were placed in series to provide a more controlled load for characterization.

- 120 -
Testing with resistive loads showed that the impedance at resonance was highly sensitive to load
resistance up to around 6 kΩ, after which the signal saturated (Fig. 3.22). Strangely, the resonant
frequency did not correlate with load resistance at all. This suggests that changes in resonant
frequency observed on the thin-film test coils may have been due to changes in the double-layer
capacitance, rather than solution resistance changes.

Figure 3.21 To test the performance limit of using a tuning capacitor, the 5.4 MHz commercial primary
coil was paired with a 25 MHz secondary coil, which was then tuned down to 5.4 MHz using a discrete
capacitor in series. Load resistors of various values were used to simulate solution resistances.

- 121 -

Figure 3.22 Testing the “ideal” commercial coil system with resistive loads showed that the impedance at
resonance was highly sensitive to load resistance until around 6 kΩ, at which point the signal saturated.
Strangely, the resonant frequency did not correlate with load resistance at all, suggesting that the shifts in
resonant frequency seen with the thin-film test coils may have been due to variations in the double-layer
capacitance instead of in solution resistance.
It may be necessary to re-examine the previously-discussed MATLAB model, and further
tests are necessary to determine the exact nature of the reflected impedance effect seen using the
thin-film coils. Work on the wireless patency sensor is now focused on using a coil with no tuning
capacitor and decreasing the impedance across the patency electrodes such that progressive
catheter blockage changes impedance within the sensitive range for reflected impedance

- 122 -
measurements. Implementation of this method may involve using a proximal catheter with built-
in electrodes that is directly attached to the reservoir and wireless coil.

Reflected Impedance Glucose Sensor
Along with its potential to detect catheter blockage, the reflected impedance method also
holds significant promise for biomarker detection. Passive, wireless detection of biomarkers could
lead to chronic implant for disease monitoring and treatment or to flexible, low-cost temporary
biomarker sensors that are skin-based or percutaneous. Alongside catheter patency sensing,
glucose sensing using reflected impedance was investigated, and preliminary experiments were
performed towards designing functionalized reflected impedance glucose sensors.

3.6.1 Diabetes
Diabetes mellitus is a disease characterized by the inability of body tissue to absorb and use
glucose
76,77
. Glucose is a sugar used as the primary energy source for muscle, adipose tissue, and
liver tissue; insulin is the primary hormone regulating uptake of glucose into these tissues. The
destruction of insulin-producing cells or an acquired insensitivity to insulin will prevent tissues
from absorbing glucose, leading to glucose accumulating in the bloodstream. Tissue which cannot
absorb glucose will have trouble synthesizing proteins, leading to persistent ulcers and a delayed
wound healing response. In addition, accumulation of glucose in the blood osmotically forces
water out of cells, leading to both persistent dehydration and increased urination. Diabetes patients
often swing between hyperglycemic (excess glucose in blood) and hypoglycemic (too little glucose
in blood) states during their treatment; hypoglycemia can cause numerous symptoms such as
fatigue, confusion, memory loss, seizures, or coma. The long-term effects of diabetes are mainly
consequences of gradual damage to blood vessels which occurs throughout the body. Blood vessel
degradation in the eye (diabetic retinopathy) will lead to blindness
78
; damage of the coronary
arteries leads to cardiovascular disease
79
; damage of small vessels in the kidneys (diabetic
nephropathy) leads to kidney failure and may require dialysis or transplant
80
. In addition to blood
vessel failure, nerve damage (diabetic neuropathy) can also occur in long-term diabetes
81
; the
symptoms include itching and tingling for damaged nerves in the body’s extremities, loss of pain
sensation which can lead to ulceration or injuries, muscle weakness due to decreased muscle
enervation, and cognitive decline due to nerve damage in the central nervous system.
Diabetes is a common disease, affecting 8.5% of adults worldwide, and its prevalence is
increasing. In 2016 there were 1.6 million deaths from diabetes, and another 2.2 million could be
traced to diabetes complications
77
. There are two main types of diabetes, along with a handful of
one-off pathologies. Type I diabetes, also known as juvenile diabetes, is an autoimmune disease
where T-cells attack and destroy insulin-producing β-cells in the pancreas’ Islets of Langerhans
82
.
This type of diabetes can develop suddenly (over the course of weeks or a few months) and often
emerges in children, though it can develop in adults as well. Although the causes of Type I diabetes
are not fully understood, the disease is believed to have both genetic and environmental factors.

- 123 -
On the genetic side, there are over 50 genes which have been shown to be associated with Type I
diabetes; however, concordance of type I diabetes among identical twins is only 20-40%
83,84
,
suggesting that environmental factors play a large role. Several environmental factors are being
investigated as triggers for Type I, including viral infections
85
, gut bacteria
86
, gliadin (a protein in
gluten)
87
, and exposure to environmental toxins
88
. There is no correlation between Type I diabetes
and being obese or overweight, and Type I patients tend to be at or below average body mass
index. Type I is the less common type of diabetes; about 10% of diabetes patients are Type I.
Type II diabetes is more common, comprising 90% of diabetes patients worldwide. Type II is
characterized by slow-onset insulin resistance in tissue, causing a decrease in insulin production
by β-cells. Type II most often occurs in adults, and is primarily triggered by obesity, lack of
physical activity, overconsumption of simple sugars, and long-term stress. However, there is a
strong genetic component of Type II, as evidenced by a 58% concordance between identical
twins
89
. Lack of sleep
90
, changes in gut bacteria
91
, and smoking
92
have also been linked to
increased risk of Type II diabetes.

3.6.2 Diabetes diagnosis and treatment
Diabetes diagnosis has traditionally involved measuring blood glucose levels
76,93,94
. This can
be done either using a fasting glucose test or a glucose tolerance test. In a glucose tolerance test,
the patient drinks 75 g of glucose and their blood glucose levels over the next two hours are
compared to a pre-test baseline. Normal glucose levels less than 7.8 mmol/L after 2 hours are
considered normal; levels between 7.8 and 11.1 mmol/L are considered “impaired”, and levels
greater than 11.1 mmol/L are considered indicative of diabetes. A fasting glucose test simply
measures blood glucose and HbA1C (glycated hemoglobin) levels after several hours of fasting;
fasting blood glucose levels less than 6.1 mmol/L and HbA1C levels less than 42 mmol/mol are
considered normal, and blood glucose levels greater than 7.0 mmol/L or HbA1C levels greater than
48 mmol/mol are indicative of diabetes. Two positive tests conducted several days apart are
generally necessary for a diabetes diagnosis.
Once a patient has been diagnosed with diabetes, treatment generally consists of constant
monitoring of blood glucose levels and self-administration of insulin if hyperglycemia is
detected
93
. The most common types of glucose monitors require patients to draw blood, usually
through a finger prick, and place the sample on a disposable test strip which is then placed into an
electronic reader. Strips consist of multiple electrodes coated in glucose oxidase, which breaks
down glucose into gluconolactone and hydrogen peroxide
95
. Commercial glucose sensors are
required to have accuracies ±15 mg/dL (±1 mmol/L), which is sufficient for diagnosing
hyperglycemia but can be problematic when measuring hypoglycemia
96
. The finger-prick method
of blood collection can be painful, reducing patient compliance and treatment efficacy, and
scarring can form due to repeated measurements. Several companies, such as Medtronic
97
,
Dexcom
98
, and Senseonics
99
, now market continuous blood glucose monitors. Most of these
monitors use subcutaneous needles with attached electronics to monitor glucose levels for up to
14 days, though Senseonics utilizes a partially implantable system which is rated for up to 90 days

- 124 -
of continuous monitoring. However, continuous glucose monitors must be frequently calibrated
using traditional finger-prick tests, as there is not a perfect correlation between blood glucose and
glucose levels in interstitial fluid
100
. Strategies to develop truly non-invasive glucose sensors by
using transdermal iontophoresis
101
or sonophoresis
102
to draw interstitial fluid to the surface of the
skin have also been investigated.
If hyperglycemia is suspected, a patient can administer insulin through hypodermic needle
injection or subcutaneous insulin infusion pumps. Inhalable insulin (MannKind Corporation
103
,
Novo Nordisk
104
) or oral insulin (Oramed
105,106
, BioLingus) are also available but have not been
widely adopted, in part due to concerns over increased risk of lung cancer
107
. Because administered
insulin is not regulated by the body’s natural feedback mechanisms, severe hypoglycemia can
occur as a side effect of treatment. Hypoglycemia can also be detected using at-home blood glucose
monitors, though with limited accuracy
108
, and is usually treated by eating a snack high in sugar
or carbohydrates. However, since the symptoms of hypoglycemia can include cognitive changes,
seizures, and coma, self-testing is not always possible. Despite the risks and difficulties, studies
such as the Diabetes Control and Complications Trial in 1993
109
and the United Kingdom
Prospective Diabetes Study in 1999
110
have shown that close monitoring of glucose and self-
administration of insulin multiple times a day significantly reduces symptoms and improves long-
term outcomes of both Type I and Type II patients.
Companies and research groups are attempting to develop a closed-loop diabetes treatment
system which combine a continuous glucose monitor and automatic insulin pump to mimic the
actions of the pancreas; both individuals and research groups have reportedly developed such a
closed-loop system, though none has reached the market yet. Closed-loop diabetes therapy is
complicated by the fact that blood glucose measurement is not always sufficient to determine the
appropriate insulin response. Glucose homeostasis is mainly controlled by four hormones: insulin,
amylin, glucagon, and incretin
111
. Self-administration of insulin does not mimic the natural
response of the pancreas to high glucose levels, and this mismatch can lead to weight gain and
dangerous hypoglycemia. In addition to these four hormones, substances such as GIP, GLP-1,
epinephrine, cortisol, and growth hormones also affect blood glucose levels and can be affected
by both the disease and its treatment. In particular, chronic increases in cortisol have been linked
to the increase in cardiovascular disease that serves as diabetes’ most common comorbidity
112
.

3.6.3 State of the art in glucose sensing

Figure 3.23 Glucose oxidase catalyzes the reaction of D-glucose with oxygen to produce D-gluconolactone
and hydrogen peroxide. Amperometric glucose sensors use an electrode held at a 0.7V bias to break down

- 125 -
hydrogen peroxide into water and oxygen, a process which results in a measurable current through the bias
electrode. Even without a voltage bias, however, increasing the concentration of hydrogen peroxide in
solution lowers solution resistance due to increasing the local ionic concentration.

First generation glucose sensors operate via amperometric detection of hydrogen peroxide after
a sample is exposed to the glucose oxidase (GOx) enzyme
95
. An amperometric system consists of
a working, counter, and reference electrodes, with the working electrode having GOx coating.
Hydrogen peroxide is produced via the GOx-mediated conversion of glucose to gluconolactone
(Fig. 3.23), and by holding the working electrode at a constant positive voltage (0.7 V) the
hydrogen peroxide is converted to oxygen, a reaction which liberates electrons and causes a
measurable current between the working and counter electrodes
113
. Compounds normally found in
blood, such as uric acid, ascorbic acid, and acetaminophen, can also generate free electrons in the
presence of a DC voltage, so a fourth working electrode which is uncoated with GOx is often used
as a reference
95
. Further improvements on GOx-based glucose biosensors can be made by adding
a semipermeable membrane, such as Nafion, to electrode surfaces to prevent interference agents
from coming in contact with electrodes
114
, and various strategies have been investigated to increase
the oxygen available at the electrodes, since the amount of dissolved oxygen is the limiting reagent
for most physiological fluids. Second-generation glucose sensors forego the high electrode
potentials required for hydrogen peroxide breakdown and instead use mediators which oxidize at
lower potentials, improving sensor resilience against interference agents
95,115
. These second-
generation sensors currently see wide use in external finger-prick glucose sensors, but concerns
about mediator leaching and toxicity have prevented widespread in vivo use. Third-generation
GOx sensors are currently under development which wire GOx directly to an electrode surface,
completely eliminating the need for oxidation reactions and DC electrode potentials
95,116
. In
addition to sensors using GOx, numerous other methods have been investigated for glucose
measurement
117
, including infrared spectroscopy
118-120
, fluorescence-based sensors
99,121,122
, and
high-frequency electrochemical impedance measurement
123,124
. In particular, high-frequency
electrochemical impedance showed great promise as a truly non-invasive, label-free diagnostic
tool for glucose measurement, and a sensor utilizing this principle was introduced to the European
market in 2003 by Pendragon Medical
125
. However, the device was subsequently pulled from the
market due to low accuracy and the manufacturer filed for bankruptcy in 2004
126
.
Despite their widespread use, transcutaneous glucose sensors share many problems with
finger-prick sensors, the largest one being persistent scarring and loss of sensitivity at the
measurement site. Senseonics attempts to avoid this problem by using an implantable
subcutaneous sensor with a skin-mounted reader that measures the fluorescent signal through the
skin, but the implant must be replaced by a doctor every 90 days, causing scarring and requiring
frequent doctors’ visits. Another issue which limits transcutaneous and subcutaneous glucose
measurement, especially for the development of closed-loop diabetes therapies, is that there is not
a 1-1 correlation between blood and interstitial glucose and a time delay for glucose level changes
to be detected in interstitial fluid. Implantable wireless measurement of blood glucose directly

- 126 -
would remove these obstacles and could allow accurate, long-term blood glucose monitoring, but
current technology would require the implantation of integrated circuits or electronic components
to transmit the glucose signal to the surface, which would add bulk and increase failure modes in
vivo
15
. Several methods have been proposed for passive microfabricated sensors, such as RF
coupling to a coil with a capacitive pressure sensor
60,62,127
or interrogation of a piezoelectric sensor
via ultrasound
48,102
, but no method for passive, wireless measurement of blood glucose or other
blood biomarkers has yet been proposed. We hypothesized that the buildup of hydrogen peroxide
due to the glucose-GOx reaction could be measured using high-frequency impedance, and we
tested wireless transduction of glucose levels using reflected impedance, without additional
electronics.

Figure 3.24 MATLAB simulations show that both the secondary (double-layer) capacitance and the
solution resistance are related to the impedance at resonance and the resonant frequency. Therefore, an
electrode coating which changes either the double-layer capacitance or the local solution resistance should
allow biomarker detection via reflected impedance.

3.6.4 Test coil functionalization experiments
In a sensor which uses GOx, electrodes are typically as large and as close together as
possible in order to minimize baseline impedance and maximize the area available for the glucose
reaction. The 16-turn thin-film coils characterized previously have electrodes spaced far apart and
relatively small electrode areas, limiting their potential sensitivity. However, a preliminary attempt
was made to functionalize these coils. The electrode coating process followed previously reported
methods for adhering GOx to electrodes with a polymer substrate
128,129
. 100 µL of a solution of 10
mg glucose oxidase type VII per mL of 0.1× PBS was placed on the 16-turn coil’s electrodes, and
the coil was suspended in a sealed container above titanium isopropoxide at 30°C for 6 hours to
form a sol-gel. The device was gently rinsed with DI water after coating, and 5% Nafion was added
to the surface to increase sensitivity
129
. The device was stored in 1× PBS at 2°C.

- 127 -
During testing, the resonant frequency of the primary coil alone was found and subsequent
measurements were taken at this frequency. 100 µL of either 100 mM D-glucose dissolved in 1×
PBS or 1× PBS without glucose were alternately placed on the electrodes, with the device rinsed
with DI water between measurements. Impedance magnitude and phase was measured after 30
seconds of settling.

3.6.5 Test coil results
The device was tested for four days after electrode coating. There was a large difference in
the baseline impedance between the four days, but a change in the reflected impedance signal was
observed on each day. Figure 3.25 shows the difference in impedance and phase when the
electrodes were exposed to PBS containing glucose versus plain PBS on the fourth day after
coating. The impedance increased with the addition of glucose, and a small drop in phase was
observed. These results indicate that reflected impedance may be useful for wirelessly measuring
the presence of biomarkers, and provide motivation for future sensor development.

Figure 3.25 A 16-turn coil coated with a GOx/titanium isopropoxide sol-gel was alternatingly exposed to
PBS with and without 100 mM d-glucose. When impedance was measured at the primary coil’s resonant
frequency, (i) the impedance magnitude increased and (ii) the phase slightly decreased on exposure to the
glucose-containing solution. This effect was still present up to 4 days after coating (data shown). © 2017
Springer Nature

3.6.6 Second-generation glucose sensor
Reflected impedance sensors coated with glucose oxidase were shown to be sensitive to
the presence or absence of glucose in solution, but the sensitivity was several orders of magnitude
lower than what is necessary for diagnostic purposes. Glucose levels in blood range from 2 to 40
mM for diabetes patients, and glucose sensors must have a limit of detection of 1 mM or less to be
practical in the clinic. The low sensitivity of the test coils coated with GOx was in part due to the
low surface area of the electrodes, which limits the area available for the glucose-GOx reaction,
and the relatively large distance between electrodes, which adds resistance and decreases

- 128 -
sensitivity. To overcome these issues, the next iteration of the reflected impedance glucose sensor
was designed with large interdigitated electrodes similar to those used in commercial
amperometric glucose sensors. The use of interdigitated electrodes requires multiple metal layers
and at least one via between adjacent Parylene layers, since a trace will have to cross over the coil
to minimize electrode spacing. A fabrication process involving multiple metal layers was tested
where the first metal layer was deposited and patterned via liftoff, a thin Parylene insulation layer
was deposited on top of the metal, and then a second metal layer was deposited and patterned on
top of the first (Fig. 3.26). Electrodes and contact pads could be exposed by reactive ion etching
in an Oxford Plasmalab 100 ICP; as long as the RF voltage was below 70 V, neither gold nor
platinum should sputter during over-etching, allowing both metal layers to be exposed in a single
etch step. Instead of using a via to connect metal layers, a bridge capacitor was used. A capacitor
can transmit high-frequency signals with little to no loss; by creating a capacitor of large enough
area using metal on either side of the Parylene insulation layer, signal transmission is achieved
without the fabrication complexity introduced by a directly connected via.

Figure 3.26 Fabrication of coils with two metal layers involves (a) deposition of metal on a Parylene C
substrate and patterning via liftoff, (b) deposition of a second, thin (1-5 µm) Parylene layer, (c) patterning
and liftoff of the second metal layer, deposition of a final Parylene insulation layer, and etching in oxygen
plasma, and (d) release from the silicon carrier wafer by peeling while immersed in DI water.
Another limitation of the test coils is the lack of optimization of the secondary inductance
and capacitance. According to the MATLAB simulations, sensitivity is optimized when the
resonant frequencies of the primary and secondary coils are matched and the secondary coil
inductance is maximized. Secondary inductance can be increased by increasing the number of
turns on the secondary coil, but the resonance frequency of the first-generation test coils was set
by the double-layer capacitance of the electrodes, which is hard to predict or measure. For the
interdigitated electrodes, large electrode surface areas lead to large double-layer capacitances,
requiring a very low secondary inductance to match resonant frequency with the primary coil. To
achieve both a high secondary inductance and a matched resonant frequency, second-generation

- 129 -
glucose sensors used the bridge capacitor to tune the secondary resonance. The size of the bridge
capacitor can be tightly controlled and will not change during electrode coating, enabling more
consistent device fabrication.
Prototype second generation glucose sensors were fabricated out of Parylene C and gold.
Devices consisted of 3000Å thick gold traces with a 200Å thick Ti adhesion layer. The top and
bottom insulation layers were 10 µm thick Parylene C, and a 4 µm thick Parylene layer separated
the first and second metal layers. Devices contained round interdigitated electrodes composed of
25 µm wide traces with 25 µm spacing. Several different designs were fabricated and are currently
being tested (Fig. 3.27).

Figure 3.27 Prototype second generation reflected impedance designs, constructed of two layers of gold
within three Parylene C layers.

3.6.7 Future biomarker sensing work
Future work will focus on optimizing the GOx coating process for our sensors and on
further improving the sensitivity of thin-film reflected impedance coils to ionic concentration
changes. Coil miniaturization will also be explored, since smaller coils allow for less obtrusive
medical devices, and electrode size and design will be considered for improving sensitivity.
Assuming the sensitivity of reflected impedance coils can approach a clinically relevant range, the
next steps will be to design a system for practical use as a continuous glucose monitor. One
potential concept would be to place the electrodes on a tab which could be inserted through the
skin into interstitial fluid, with the coil adhering to the skin’s surface. An external reader could be
passed over the coil any time a glucose measurement was required. Developing this system would
require designing an electrode tab which could be reversibly adhered to a sharp insertion tool
(possibly using a thermoformed Parylene cone on the reverse side as the electrodes) and testing
insertion in a skin phantom.

3 mm 10 mm
10 mm

- 130 -

Figure 3.28 Proposed continuous glucose monitor using reflected impedance. Placing the electrodes on a
Parylene tab allows a three-dimensional device to be constructed out of a two-dimensional substrate.

Figure 3.29 The reflected impedance glucose monitor would consist of an electrode tab, which would sit
in the interstitial fluid, and a coil adhered to the surface of the skin. (A) The electrode tab would be attached
to an insertion tool for placement in the skin. (B) The insertion tool would be used to punch the electrodes
through the skin into the interstitial fluid and then (C) removed, leaving a highly flexible, low-profile
implant. (D) An external reader could be used to measure glucose levels on demand.
In addition to measuring glucose, the reflected impedance method could also be used for
other analytes in blood or interstitial fluid. In particular, insulin and cortisol are important markers
of diabetes progression and state, and continuous insulin or cortisol monitoring could fill the gaps
which exist in diabetes treatment via glucose monitoring and allow closed-loop diabetes control.
Normal levels of insulin in blood range from 50-80 pM
130
, and cortisol levels range from 100-500
nM
131
, so sensors will have to detect analytes with extremely high sensitivity. Antibody-based
coatings for insulin and cortisol detection will be tested on wired electrodes and then on reflected
impedance devices towards developing percutaneous sensors for these analytes, and reflected
impedance devices with multiple tuning capacitors and electrode pairs could be constructed to
explore multi-modal analyte detection using a single reflected impedance device.

- 131 -
The Reflected Impedance Method: Discussion and Comparison to
State of the Art
Despite their low quality factor, 1, 5, and 16 turn thin-film coils were able to accurately
transduce a wide range of solution resistances using the reflected impedance method. For all coils,
the system impedance magnitude at the primary resonant frequency varied with solution resistance,
and for the 16-turn coil an additional transduction method using the resonant frequency value was
developed. The variance of impedance magnitude at resonance clearly matches up with theoretical
and simulation results, and the large variance in primary resonant frequency seen when testing the
16-turn coil suggests that the resonant frequency of this coil, which is determined by the coil’s
inductance and the double layer capacitance of the two electrodes, is slightly higher than the
primary coil’s resonant frequency.
Further testing using the 16-turn coil showed that accurate transduction using this reflected
impedance method requires close vertical proximity and precise lateral alignment. As the
secondary coil is separated vertically from the primary coil, the sensitivity of the system to changes
in solution resistance decreases linearly, following the decrease in the coupling constant. Lateral
misalignment also decreases coupling, and once the secondary coil’s center is one secondary radius
(15 mm) away from the primary coil’s center, the signal stabilizes, indicating that by this point the
two coils are uncoupled. This shift in signal with misalignment means that close proximity and
confidence in primary coil alignment are required for any practical application of this technique,
which for an implantable sensor means that the secondary coil must be placed just under the skin.
The shift in resonant frequency with vertical misalignment presents a major disadvantage over
passive capacitive sensors, which measure the resonance of a high-Q secondary coil directly and
only experience a decrease in the sensor’s signal to noise ratio when their primary coil is vertically
separated from their sensor.
Despite its limitation, this method presents a number of advantages over currently-used
implantable sensors. No other sensing method has been developed which can wirelessly transduce
a resistive load without the use of discrete electrical components or integrated circuits. Our method
does not require a high secondary quality factor and in fact showed remarkable accuracy and
sensitivity using coils with Q<0.5, which allows for thin-film, fully flexible sensors and removes
the need for the complex fabrication steps and thick, electroplated metal layers used in previously
reported thin-film coils. The two parallel methods of transduction available from a single coil allow
for redundancy and the detection of certain failure modes.
The optimal applications for reflected impedance, therefore, are implantable sensors which
sit near the surface of the skin and transduce loads based on high-frequency impedance or
resistance. Two applications were explored which met these requirements; catheter patency
sensing in hydrocephalus shunts and glucose sensing. The valves of hydrocephalus shunts typically
sit just under the skin above the skull, and a method has been previously demonstrated to transduce
catheter patency using high-frequency electrochemical impedance, making this application ideal
for reflected impedance. A module designed to sit in-line with existing shunts was developed and

- 132 -
progressive detection of catheter blockage using reflected impedance was demonstrated, though
the sensitivity was lower than expected due to the high impedance magnitude through the catheter.
Future improvements could include increasing inductance by increasing turns or adding a ferrite
core, and further work is necessary to determine the best way to tune the secondary patency coil
to the optimal frequency for reflected impedance transduction. Glucose measurement is a daily
routine for millions of diabetes patients worldwide and require measuring the low-frequency
impedance across coated electrodes just under the skin’s surface. A 16-turn coil coated in glucose
oxidase showed a small but significant change in reflected impedance when exposed to glucose-
containing saline, but work needs to be done to increase sensitivity. Toward this end, devices have
been designed and fabricated with large, interdigitated electrodes, and more effective coating
methods will be explored to enable transduction of physiologically-relevant glucose levels. Further
applications for reflected impedance sensors could include passive dose sensing in implanted
micropumps. Dose sensing based on high-frequency electrochemical impedance has already been
demonstrated in an electrolysis-based implantable micropump
132,133
, and applications such as
passive drug delivery to ocular tissue
134
could also benefit from a flexible, passive, thin-film dose
sensor.

Conclusion
A new method for passive, wireless sensing of high-frequency electrochemical impedance
based on the reflected impedance across inductively coupled coils was developed and
characterized. Tests performed using microfabricated thin-film coils showed that this method can
transduce a wide range of impedance values by monitoring either the primary resonant frequency
or the impedance at resonance, with extremely high sensitivities achieved for both methods despite
very low quality factors on the secondary coils. Several disadvantages do exist with reflected
impedance, namely that vertical or lateral misalignment will cause progressive coil decoupling and
will interfere with signal transduction, but to the best of our knowledge this method is the only
reported method of passively measuring resistive loads without integrated circuits or discrete
electronic components. This method lends itself well to the development of wireless impedance-
based sensors for implantation in the human body; two potential applications in hydrocephalus
treatment and glucose sensing were explored, and future work will continue to apply and optimize
reflected impedance for the development of biomedical sensors.

- 133 -
References
1. A. Baldwin, L. Yu, M. Pratt, K. Scholten, and E. Meng, "Passive, wireless transduction
of electrochemical impedance across thin-film microfabricated coils using reflected
impedance," Biomedical Microdevices, vol. 19, p. 87, 2017.
2. C.-M. Yu, L. Wang, E. Chau, R. H.-W. Chan, S.-L. Kong, M.-O. Tang, et al.,
"Intrathoracic Impedance Monitoring in Patients With Heart Failure," Correlation With
Fluid Status and Feasibility of Early Warning Preceding Hospitalization, vol. 112, pp.
841-848, 2005.
3. F. L. Scott, M. Svjetlana, D. J. Matthew, L. V. Jerrold, and C. M. Cameron, "In vivo
impedance spectroscopy of deep brain stimulation electrodes," Journal of Neural
Engineering, vol. 6, p. 046001, 2009.
4. R. Elshafey, A. C. Tavares, M. Siaj, and M. Zourob, "Electrochemical impedance
immunosensor based on gold nanoparticles–protein G for the detection of cancer marker
epidermal growth factor receptor in human plasma and brain tissue," Biosensors and
Bioelectronics, vol. 50, pp. 143-149, 2013/12/15/ 2013.
5. L. Yang, Y. Li, and G. F. Erf, "Interdigitated Array Microelectrode-Based
Electrochemical Impedance Immunosensor for Detection of Escherichia coli O157:H7,"
Analytical Chemistry, vol. 76, pp. 1107-1113, 2004/02/01 2004.
6. A. Baldwin, L. Yu, and E. Meng, "An Electrochemical Impedance-Based Thermal Flow
Sensor for Physiological Fluids," Journal of Microelectromechanical Systems, vol. PP,
pp. 1-10, 2016.
7. A. Baldwin, L. Yu, and E. Meng, "An electrochemical-based thermal flow sensor," in
Micro Electro Mechanical Systems (MEMS), 2016 IEEE 29th International Conference
on, 2016, pp. 341-344.
8. L. Yu, C. A. Gutierrez, and E. Meng, "An Electrochemical Microbubble-Based MEMS
Pressure Sensor," Microelectromechanical Systems, Journal of, vol. 25, pp. 144-152,
2016.
9. L. Yu and E. Meng, "A microbubble pressure transducer with bubble nucleation core," in
Micro Electro Mechanical Systems (MEMS), 2014 IEEE 27th International Conference
on, 2014, pp. 104-107.
10. B. J. Kim, C. A. Gutierrez, and E. Meng, "Parylene-based electrochemical-MEMS force
sensor for studies of intracortical probe insertion mechanics," Journal of
Microelectromechanical Systems, vol. 24, pp. 1534-1544, 2015.
11. B. Kim, C. Gutierrez, and E. Meng, "IN VITRO CHARACTERIZATION OF A PROBE-
MOUNTED PARYLENE-BASED PRESSURE SENSOR ARRAY FOR
INTRACORTICAL APPLICATIONS," Sensors, vol. 2, p. 7.
12. M. Min, T. Parve, V. Kukk, and A. Kuhlberg, "An implantable analyzer of bio-
impedance dynamics: mixed signal approach," in IMTC 2001. Proceedings of the 18th
IEEE Instrumentation and Measurement Technology Conference. Rediscovering
Measurement in the Age of Informatics (Cat. No.01CH 37188), 2001, pp. 38-43 vol.1.
13. Y.-T. Li, C.-H. Chang, J.-J. J. Chen, 陳家進, C.-C. Wang, and C.-K. Liang,
"Development of implantable wireless biomicrosystem for measuring electrode-tissue
impedance," 2005.

- 134 -
14. M. M. Ahmadi and G. A. Jullien, "A wireless-implantable microsystem for continuous
blood glucose monitoring," IEEE Transactions on Biomedical Circuits and Systems, vol.
3, pp. 169-180, 2009.
15. A. Vanhoestenberghe and N. Donaldson, "Corrosion of silicon integrated circuits and
lifetime predictions in implantable electronic devices," Journal of Neural Engineering,
vol. 10, p. 031002, 2013.
16. T. Stieglitz, S. Kammer, K. Koch, S. Wien, and A. Robitzki, "Encapsulation of flexible
biomedical microimplants with parylene C," IFESS 2002, vol. 5, 2002.
17. C. Hassler, R. P. von Metzen, P. Ruther, and T. Stieglitz, "Characterization of parylene C
as an encapsulation material for implanted neural prostheses," Journal of Biomedical
Materials Research Part B: Applied Biomaterials, vol. 93, pp. 266-274, 2010.
18. W. Yao, C.-H. Chu, and Z. Li, "The adoption and implementation of RFID technologies
in healthcare: a literature review," Journal of medical systems, vol. 36, pp. 3507-3525,
2012.
19. I. Jones, L. Ricciardi, L. Hall, H. Hansen, V. Varadan, C. Bertram, et al., "Wireless RF
communication in biomedical applications," Smart Materials and Structures, vol. 17, p.
015050, 2008.
20. P. Burke and C. Rutherglen, "Towards a single-chip, implantable RFID system: is a
single-cell radio possible?," Biomedical microdevices, vol. 12, pp. 589-596, 2010.
21. M. Ghovanloo and S. Atluri, "An integrated full-wave CMOS rectifier with built-in back
telemetry for RFID and implantable biomedical applications," IEEE Transactions on
Circuits and Systems I: Regular Papers, vol. 55, pp. 3328-3334, 2008.
22. A. J. Faul, "Next generation RFID telemetry design for biomedical implants," 2014.
23. M. D. Weiss, J. L. Smith, and J. Bach, "RF coupling in a 433-MHz biotelemetry system
for an artificial hip," IEEE Antennas and Wireless Propagation Letters, vol. 8, pp. 916-
919, 2009.
24. S. Manzari, S. Pettinari, and G. Marrocco, "Miniaturized and tunable wearable RFID tag
for body-centric applications," in RFID-Technologies and Applications (RFID-TA), 2012
IEEE International Conference on, 2012, pp. 239-243.
25. E. Moradi, K. Koski, L. Ukkonen, Y. Rahmat-Samii, T. Björninen, and L. Sydänheimo,
"Embroidered RFID tags in body-centric communication," in Antenna Technology
(iWAT), 2013 International Workshop on, 2013, pp. 367-370.
26. A. Sani, M. Rajab, R. Foster, and Y. Hao, "Antennas and propagation of implanted
RFIDs for pervasive healthcare applications," Proceedings of the IEEE, vol. 98, pp.
1648-1655, 2010.
27. P. Röschmann, "Radiofrequency penetration and absorption in the human body:
limitations to high ‐field whole ‐body nuclear magnetic resonance imaging," Medical
physics, vol. 14, pp. 922-931, 1987.
28. P. Bernardi, M. Cavagnaro, S. Pisa, and E. Piuzzi, "Specific absorption rate and
temperature elevation in a subject exposed in the far-field of radio-frequency sources
operating in the 10-900-MHz range," IEEE Transactions on biomedical engineering, vol.
50, pp. 295-304, 2003.
29. H. J. Visser and R. J. Vullers, "RF energy harvesting and transport for wireless sensor
network applications: Principles and requirements," Proceedings of the IEEE, vol. 101,
pp. 1410-1423, 2013.

- 135 -
30. S. Mandal and R. Sarpeshkar, "Low-power CMOS rectifier design for RFID
applications," IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 54, pp.
1177-1188, 2007.
31. R. Want, "Enabling ubiquitous sensing with RFID," Computer, vol. 37, pp. 84-86, 2004.
32. X. Yi, T. Wu, Y. Wang, R. T. Leon, M. M. Tentzeris, and G. Lantz, "Passive wireless
smart-skin sensor using RFID-based folded patch antennas," International Journal of
Smart and Nano Materials, vol. 2, pp. 22-38, 2011.
33. D. P. Rose, M. E. Ratterman, D. K. Griffin, L. Hou, N. Kelley-Loughnane, R. R. Naik, et
al., "Adhesive RFID sensor patch for monitoring of sweat electrolytes," IEEE
Transactions on Biomedical Engineering, vol. 62, pp. 1457-1465, 2015.
34. S. Milici, S. Amendola, A. Bianco, and G. Marrocco, "Epidermal RFID passive sensor
for body temperature measurements," in RFID Technology and Applications Conference
(RFID-TA), 2014 IEEE, 2014, pp. 140-144.
35. C. Occhiuzzi, G. Contri, and G. Marrocco, "Design of implanted RFID tags for passive
sensing of human body: The STENTag," IEEE Transactions on Antennas and
Propagation, vol. 60, pp. 3146-3154, 2012.
36. Z. Xiao, X. Tan, X. Chen, S. Chen, Z. Zhang, H. Zhang, et al., "An implantable RFID
sensor tag toward continuous glucose monitoring," IEEE journal of biomedical and
health informatics, vol. 19, pp. 910-919, 2015.
37. X. Wei, Z. Wang, and H. Dai, "A critical review of wireless power transfer via strongly
coupled magnetic resonances," Energies, vol. 7, pp. 4316-4341, 2014.
38. T. Bork, A. Hogg, M. Lempen, D. Müller, D. Joss, T. Bardyn, et al., "Development and
in-vitro characterization of an implantable flow sensing transducer for hydrocephalus,"
Biomedical microdevices, vol. 12, pp. 607-618, 2010.
39. M. Ghovanloo and S. Atluri, "A wide-band power-efficient inductive wireless link for
implantable microelectronic devices using multiple carriers," IEEE Transactions on
Circuits and Systems I: Regular Papers, vol. 54, pp. 2211-2221, 2007.
40. G. Wang, W. Liu, M. Sivaprakasam, M. Zhou, J. D. Weiland, and M. S. Humayun, "A
dual band wireless power and data telemetry for retinal prosthesis," in Engineering in
Medicine and Biology Society, 2006. EMBS'06. 28th Annual International Conference of
the IEEE, 2006, pp. 4392-4395.
41. M. Catrysse, B. Hermans, and R. Puers, "An inductive power system with integrated bi-
directional data-transmission," Sensors and Actuators A: Physical, vol. 115, pp. 221-229,
2004.
42. S. Lee, J. Yoo, H. Kim, and H.-J. Yoo, "A dynamic real-time capacitor compensated
inductive coupling transceiver for wearable body sensor network," in VLSI Circuits, 2009
Symposium on, 2009, pp. 42-43.
43. C. M. Zierhofer and E. S. Hochmair, "Geometric approach for coupling enhancement of
magnetically coupled coils," IEEE transactions on Biomedical Engineering, vol. 43, pp.
708-714, 1996.
44. K. Fotopoulou and B. W. Flynn, "Wireless power transfer in loosely coupled links: Coil
misalignment model," IEEE Transactions on Magnetics, vol. 47, pp. 416-430, 2011.
45. H. McDermott, "An advanced multiple channel cochlear implant," IEEE Transactions on
Biomedical Engineering, vol. 36, pp. 789-797, 1989.

- 136 -
46. C. M. Zierhofer, I. J. Hochmair-Desoyer, and E. Hochmair, "Electronic design of a
cochlear implant for multichannel high-rate pulsatile stimulation strategies," IEEE
Transactions on Rehabilitation Engineering, vol. 3, pp. 112-116, 1995.
47. L. S. Theogarajan, "A low-power fully implantable 15-channel retinal stimulator chip,"
IEEE Journal of Solid-State Circuits, vol. 43, pp. 2322-2337, 2008.
48. D. Seo, J. M. Carmena, J. M. Rabaey, E. Alon, and M. M. Maharbiz, "Neural dust: An
ultrasonic, low power solution for chronic brain-machine interfaces," arXiv preprint
arXiv:1307.2196, 2013.
49. S. Arra, J. Leskinen, J. Heikkila, and J. Vanhala, "Ultrasonic power and data link for
wireless implantable applications," in Wireless Pervasive Computing, 2007. ISWPC'07.
2nd International Symposium on, 2007.
50. S.-n. Suzuki, S. Kimura, T. Katane, H. Saotome, O. Saito, and K. Kobayashi, "Power and
interactive information transmission to implanted medical device using ultrasonic,"
Japanese journal of applied physics, vol. 41, p. 3600, 2002.
51. S. Ozeri and D. Shmilovitz, "Simultaneous backward data transmission and power
harvesting in an ultrasonic transcutaneous energy transfer link employing acoustically
dependent electric impedance modulation," Ultrasonics, vol. 54, pp. 1929-1937, 2014.
52. D. Seo, J. M. Carmena, J. M. Rabaey, M. M. Maharbiz, and E. Alon, "Model validation
of untethered, ultrasonic neural dust motes for cortical recording," Journal of
neuroscience methods, vol. 244, pp. 114-122, 2015.
53. D. Shmilovitz, S. Ozeri, C.-C. Wang, and B. Spivak, "Noninvasive control of the power
transferred to an implanted device by an ultrasonic transcutaneous energy transfer link,"
IEEE Transactions on Biomedical Engineering, vol. 61, pp. 995-1004, 2014.
54. A. Denisov and E. Yeatman, "Ultrasonic vs. inductive power delivery for miniature
biomedical implants," in Body Sensor Networks (BSN), 2010 International Conference
on, 2010, pp. 84-89.
55. M. A. Fonseca, M. G. Allen, J. Kroh, and J. White, "Flexible wireless passive pressure
sensors for biomedical applications," in Tech. Dig. Solid-State Sensor, Actuator, and
Microsystems Workshop (Hilton Head 2006), 2006, pp. 37-42.
56. Y. Bäcklund, L. Rosengren, B. Hök, and B. Svedbergh, "Passive silicon transensor
intended for biomedical, remote pressure monitoring," Sensors and Actuators A:
Physical, vol. 21, pp. 58-61, 1990/02/01/ 1990.
57. L. Y. Chen, B. C.-K. Tee, A. L. Chortos, G. Schwartz, V. Tse, D. J. Lipomi, et al.,
"Continuous wireless pressure monitoring and mapping with ultra-small passive sensors
for health monitoring and critical care," Nature communications, vol. 5, p. 5028, 2014.
58. O. Akar, T. Akin, and K. Najafi, "A wireless batch sealed absolute capacitive pressure
sensor," Sensors and Actuators A: Physical, vol. 95, pp. 29-38, 2001.
59. W. T. Abraham, P. B. Adamson, R. C. Bourge, M. F. Aaron, M. R. Costanzo, L. W.
Stevenson, et al., "Wireless pulmonary artery haemodynamic monitoring in chronic heart
failure: a randomised controlled trial," The Lancet, vol. 377, pp. 658-666, 2011.
60. P.-J. Chen, S. Saati, R. Varma, M. S. Humayun, and Y.-C. Tai, "Wireless intraocular
pressure sensing using microfabricated minimally invasive flexible-coiled LC sensor
implant," Journal of Microelectromechanical Systems, vol. 19, pp. 721-734, 2010.
61. K. C. Katuri, S. Asrani, and M. K. Ramasubramanian, "Intraocular Pressure Monitoring
Sensors," IEEE Sensors Journal, vol. 8, pp. 12-19, 2008.

- 137 -
62. N. Xue, S.-P. Chang, and J.-B. Lee, "A SU-8-based microfabricated implantable
inductively coupled passive RF wireless intraocular pressure sensor," Journal of
Microelectromechanical Systems, vol. 21, pp. 1338-1346, 2012.
63. P. Horowitz, W. Hill, and I. Robinson, The art of electronics vol. 2: Cambridge
university press Cambridge, 1980.
64. S. Hui, "Planar wireless charging technology for portable electronic products and Qi,"
Proceedings of the IEEE, vol. 101, pp. 1290-1301, 2013.
65. J. E. B. Randles, "Kinetics of rapid electrode reactions," Discussions of the faraday
society, vol. 1, pp. 11-19, 1947.
66. B. J. Kim and E. Meng, "Micromachining of Parylene C for bioMEMS," Polymers for
Advanced Technologies, 2015.
67. H. A. Wheeler, "Formulas for the skin effect," Proceedings of the IRE, vol. 30, pp. 412-
424, 1942.
68. A. Aschoff, P. Kremer, B. Hashemi, and S. Kunze, "The scientific history of
hydrocephalus and its treatment," Neurosurgical review, vol. 22, pp. 67-93, 1999.
69. J. Drake, J. Kestle, and S. Tuli, "CSF shunts 50 years on–past, present and future,"
Child's Nervous System, vol. 16, pp. 800-804, 2000.
70. S. Tuli, J. Drake, J. Lawless, M. Wigg, and M. Lamberti-Pasculli, "Risk factors for
repeated cerebrospinal shunt failures in pediatric patients with hydrocephalus," Journal of
neurosurgery, vol. 92, pp. 31-38, 2000.
71. P. S. Shah and V. S. Shah, "Continuous heparin infusion to prevent thrombosis and
catheter occlusion in neonates with peripherally placed percutaneous central venous
catheters," The Cochrane Library, 2008.
72. B. Mirza, V. W. Vanek, and D. T. Kupensky, "Pinch-off syndrome: case report and
collective review of the literature," The American surgeon, vol. 70, pp. 635-644, 2004.
73. S. B. Palder, R. L. Kirkman, A. D. Whittemore, R. M. Hakim, J. M. Lazarus, and N. L.
Tilney, "Vascular access for hemodialysis. Patency rates and results of revision," Annals
of surgery, vol. 202, p. 235, 1985.
74. A. C. van Bon, D. Dragt, and J. H. DeVries, "Significant time until catheter occlusion
alerts in currently marketed insulin pumps at two basal rates," Diabetes technology &
therapeutics, vol. 14, pp. 447-448, 2012.
75. B. J. Kim, W. Jin, A. Baldwin, L. Yu, E. Christian, M. D. Krieger, et al., "Parylene
MEMS patency sensor for assessment of hydrocephalus shunt obstruction," Biomedical
microdevices, vol. 18, p. 87, 2016.
76. A. D. Association, "Diagnosis and classification of diabetes mellitus," Diabetes care, vol.
37, pp. S81-S90, 2014.
77. W. H. Organization and W. H. Organization, "Diabetes fact sheet," Geneva, Switzerland
URL: http://www.who.int/mediacentre/factsheets/fs312/en/, 2016.
78. C. Wilkinson, F. L. Ferris, R. E. Klein, P. P. Lee, C. D. Agardh, M. Davis, et al.,
"Proposed international clinical diabetic retinopathy and diabetic macular edema disease
severity scales," Ophthalmology, vol. 110, pp. 1677-1682, 2003.
79. H. Ginsberg, J. Plutzky, and B. E. Sobel, "A review of metabolic and cardiovascular
effects of oral antidiabetic agents: beyond glucose-level lowering," Journal of
cardiovascular risk, vol. 6, pp. 337-346, 1999.

- 138 -
80. J. L. Gross, M. J. De Azevedo, S. P. Silveiro, L. H. Canani, M. L. Caramori, and T.
Zelmanovitz, "Diabetic nephropathy: diagnosis, prevention, and treatment," Diabetes
care, vol. 28, pp. 164-176, 2005.
81. P. J. Dyck, P. Thomas, A. Asbury, A. Winegrad, and D. Porte, "Diabetic neuropathy,"
Diabetic neuropathy, pp. 255-295, 1999.
82. J. A. Bluestone, K. Herold, and G. Eisenbarth, "Genetics, pathogenesis and clinical
interventions in type 1 diabetes," Nature, vol. 464, p. 1293, 2010.
83. J. Kaprio, J. Tuomilehto, M. Koskenvuo, K. Romanov, A. Reunanen, J. Eriksson, et al.,
"Concordance for type 1 (insulin-dependent) and type 2 (non-insulin-dependent) diabetes
mellitus in a population-based cohort of twins in Finland," Diabetologia, vol. 35, pp.
1060-1067, 1992.
84. V. Hyttinen, J. Kaprio, L. Kinnunen, M. Koskenvuo, and J. Tuomilehto, "Genetic
liability of type 1 diabetes and the onset age among 22,650 young Finnish twin pairs,"
Diabetes, vol. 52, pp. 1052-1055, 2003.
85. H.-S. Jun and J.-W. Yoon, "A new look at viruses in type 1 diabetes," ILAR journal, vol.
45, pp. 349-374, 2004.
86. L. Wen, R. E. Ley, P. V. Volchkov, P. B. Stranges, L. Avanesyan, A. C. Stonebraker, et
al., "Innate immunity and intestinal microbiota in the development of Type 1 diabetes,"
Nature, vol. 455, p. 1109, 2008.
87. D. J. Smyth, V. Plagnol, N. M. Walker, J. D. Cooper, K. Downes, J. H. Yang, et al.,
"Shared and distinct genetic variants in type 1 diabetes and celiac disease," New England
Journal of Medicine, vol. 359, pp. 2767-2777, 2008.
88. H. K. Åkerblom, O. Vaarala, H. Hyöty, J. Ilonen, and M. Knip, "Environmental factors in
the etiology of type 1 diabetes," American Journal of Medical Genetics Part A, vol. 115,
pp. 18-29, 2002.
89. B. Newman, J. Selby, M.-C. King, C. Slemenda, R. Fabsitz, and G. Friedman,
"Concordance for type 2 (non-insulin-dependent) diabetes mellitus in male twins,"
Diabetologia, vol. 30, pp. 763-768, 1987.
90. F. P. Cappuccio, L. D'elia, P. Strazzullo, and M. A. Miller, "Quantity and quality of sleep
and incidence of type 2 diabetes," Diabetes care, vol. 33, pp. 414-420, 2010.
91. N. Larsen, F. K. Vogensen, F. W. van den Berg, D. S. Nielsen, A. S. Andreasen, B. K.
Pedersen, et al., "Gut microbiota in human adults with type 2 diabetes differs from non-
diabetic adults," PloS one, vol. 5, p. e9085, 2010.
92. C. Willi, P. Bodenmann, W. A. Ghali, P. D. Faris, and J. Cornuz, "Active smoking and
the risk of type 2 diabetes: a systematic review and meta-analysis," Jama, vol. 298, pp.
2654-2664, 2007.
93. A. D. Association, "Standards of medical care in diabetes—2015 abridged for primary
care providers," Clinical diabetes: a publication of the American Diabetes Association,
vol. 33, p. 97, 2015.
94. N. D. D. Group, "Classification and diagnosis of diabetes mellitus and other categories of
glucose intolerance," Diabetes, vol. 28, pp. 1039-1057, 1979.
95. J. Wang, "Glucose biosensors: 40 years of advances and challenges," Electroanalysis,
vol. 13, p. 983, 2001.
96. G. Freckmann, A. Baumstark, N. Jendrike, E. Zschornack, S. Kocher, J. Tshiananga, et
al., "System accuracy evaluation of 27 blood glucose monitoring systems according to
DIN EN ISO 15197," Diabetes technology & therapeutics, vol. 12, pp. 221-231, 2010.

- 139 -
97. N. Sachedina and J. Pickup, "Performance assessment of the Medtronic ‐MiniMed
continuous glucose monitoring system and its use for measurement of glycaemic control
in type 1 diabetic subjects," Diabetic medicine, vol. 20, pp. 1012-1015, 2003.
98. A. Garcia, A. L. Rack-Gomer, N. C. Bhavaraju, H. Hampapuram, A. Kamath, T. Peyser,
et al., "Dexcom G4AP: an advanced continuous glucose monitor for the artificial
pancreas," ed: SAGE Publications Sage CA: Los Angeles, CA, 2013.
99. M. Mortellaro and A. DeHennis, "Performance characterization of an abiotic and
fluorescent-based continuous glucose monitoring system in patients with type 1
diabetes," Biosensors and Bioelectronics, vol. 61, pp. 227-231, 2014.
100. V. Lodwig and L. Heinemann, "Continuous glucose monitoring with glucose sensors:
calibration and assessment criteria," Diabetes technology & therapeutics, vol. 5, pp. 572-
586, 2003.
101. A. Sieg, R. H. Guy, and M. B. Delgado-Charro, "Electroosmosis in transdermal
iontophoresis: implications for noninvasive and calibration-free glucose monitoring,"
Biophysical journal, vol. 87, pp. 3344-3350, 2004.
102. J. Kost, S. Mitragotri, R. A. Gabbay, M. Pishko, and R. Langer, "Transdermal monitoring
of glucose and other analytes using ultrasound," Nature medicine, vol. 6, pp. 347-350,
2000.
103. L. W. Fleming, J. W. Fleming, and C. S. Davis, "Afrezza: An inhaled approach to insulin
delivery," Journal of the American Association of Nurse Practitioners, vol. 27, pp. 597-
601, 2015.
104. D. Barlocco, "Insulin detemir. Novo Nordisk," Current opinion in investigational drugs
(London, England: 2000), vol. 4, pp. 449-454, 2003.
105. M. Kidron, I. Raz, M. Wolfensberger, H. Schwob, and C. Schruefer, "Pharmacokinetics
(PK) and pharmacodynamics (PD) of oral insulin in healthy subjects," in American
Diabetes Association 68th Annual Meeting, San Francisco, CA, 2008.
106. R. Eldor, E. Arbit, A. Corcos, and M. Kidron, "Glucose-reducing effect of the ORMD-
0801 oral insulin preparation in patients with uncontrolled type 1 diabetes: a pilot study,"
PLoS One, vol. 8, p. e59524, 2013.
107. J. Kling, "Inhaled insulin's last gasp?," ed: Nature Publishing Group, 2008.
108. Z. Trajanoski, G. Brunner, R. Gfrerer, P. Wach, and T. Pieber, "Accuracy of home blood
glucose meters during hypoglycemia," Diabetes Care, vol. 19, pp. 1412-1415, 1996.
109. T. D. Control and C. T. R. Group, "The Effect of Intensive Treatment of Diabetes on the
Development and Progression of Long-Term Complications in Insulin-Dependent
Diabetes Mellitus," New England Journal of Medicine, vol. 329, pp. 977-986, 1993.
110. U. P. D. S. Group, "Tight blood pressure control and risk of macrovascular and
microvascular complications in type 2 diabetes: UKPDS 38," BMJ: British Medical
Journal, vol. 317, p. 703, 1998.
111. S. L. Aronoff, K. Berkowitz, B. Shreiner, and L. Want, "Glucose metabolism and
regulation: beyond insulin and glucagon," Diabetes Spectrum, vol. 17, pp. 183-190, 2004.
112. I. Chiodini, G. Adda, A. Scillitani, F. Coletti, V. Morelli, S. Di Lembo, et al., "Cortisol
secretion in patients with type 2 diabetes," Diabetes Care, vol. 30, pp. 83-88, 2007.
113. N. C. Foulds and C. R. Lowe, "Enzyme entrapment in electrically conducting polymers.
Immobilisation of glucose oxidase in polypyrrole and its application in amperometric
glucose sensors," Journal of the Chemical Society, Faraday Transactions 1: Physical
Chemistry in Condensed Phases, vol. 82, pp. 1259-1264, 1986.

- 140 -
114. Y. Zhang, Y. Hu, G. S. Wilson, D. Moatti-Sirat, V. Poitout, and G. Reach, "Elimination
of the acetaminophen interference in an implantable glucose sensor," Analytical
Chemistry, vol. 66, pp. 1183-1188, 1994.
115. A. E. Cass, G. Davis, G. D. Francis, H. A. O. Hill, W. J. Aston, I. J. Higgins, et al.,
"Ferrocene-mediated enzyme electrode for amperometric determination of glucose,"
Analytical chemistry, vol. 56, pp. 667-671, 1984.
116. T. J. Ohara, R. Rajagopalan, and A. Heller, "" Wired" enzyme electrodes for
amperometric determination of glucose or lactate in the presence of interfering
substances," Analytical Chemistry, vol. 66, pp. 2451-2457, 1994.
117. N. Oliver, C. Toumazou, A. Cass, and D. Johnston, "Glucose sensors: a review of current
and emerging technology," Diabetic Medicine, vol. 26, pp. 197-210, 2009.
118. H. Heise, R. Marbach, T. Koschinsky, and F. Gries, "Noninvasive blood glucose sensors
based on near ‐infrared spectroscopy," Artificial Organs, vol. 18, pp. 439-447, 1994.
119. S. F. Malin, T. L. Ruchti, T. B. Blank, S. N. Thennadil, and S. L. Monfre, "Noninvasive
prediction of glucose by near-infrared diffuse reflectance spectroscopy," Clinical
chemistry, vol. 45, pp. 1651-1658, 1999.
120. H. Zeller, P. Novak, and R. Landgraf, "Blood glucose measurement by infrared
spectroscopy," The International journal of artificial organs, vol. 12, pp. 129-135, 1989.
121. J. C. Pickup, F. Hussain, N. D. Evans, O. J. Rolinski, and D. J. Birch, "Fluorescence-
based glucose sensors," Biosensors and Bioelectronics, vol. 20, pp. 2555-2565, 2005.
122. H. Fang, G. Kaur, and B. Wang, "Progress in boronic acid-based fluorescent glucose
sensors," Journal of fluorescence, vol. 14, pp. 481-489, 2004.
123. R. K. Shervedani, A. H. Mehrjardi, and N. Zamiri, "A novel method for glucose
determination based on electrochemical impedance spectroscopy using glucose oxidase
self-assembled biosensor," Bioelectrochemistry, vol. 69, pp. 201-208, 2006.
124. A. Caduff, F. Dewarrat, M. Talary, G. Stalder, L. Heinemann, and Y. Feldman, "Non-
invasive glucose monitoring in patients with diabetes: a novel system based on
impedance spectroscopy," Biosensors and Bioelectronics, vol. 22, pp. 598-604, 2006.
125. S. A. Weinzimer, "Analysis: PENDRA: The Once and Future Noninvasive Continuous
Glucose Monitoring Device?," Diabetes technology & therapeutics, vol. 6, pp. 442-444,
2004.
126. I. Wentholt, J. Hoekstra, A. Zwart, and J. DeVries, "Pendra goes Dutch: lessons for the
CE mark in Europe," Diabetologia, vol. 48, pp. 1055-1058, 2005.
127. P.-J. Chen, D. C. Rodger, S. Saati, M. S. Humayun, and Y.-C. Tai, "Microfabricated
implantable parylene-based wireless passive intraocular pressure sensors," Journal of
Microelectromechanical Systems, vol. 17, pp. 1342-1351, 2008.
128. J. Yu, S. Liu, and H. Ju, "Glucose sensor for flow injection analysis of serum glucose
based on immobilization of glucose oxidase in titania sol–gel membrane," Biosensors
and Bioelectronics, vol. 19, pp. 401-409, 2003.
129. H. Yao, A. J. Shum, M. Cowan, I. Lähdesmäki, and B. A. Parviz, "A contact lens with
embedded sensor for monitoring tear glucose level," Biosensors and Bioelectronics, vol.
26, pp. 3290-3296, 2011.
130. G. Bisker, N. M. Iverson, J. Ahn, and M. S. Strano, "A pharmacokinetic model of a tissue
implantable insulin sensor," Advanced healthcare materials, vol. 4, pp. 87-97, 2015.

- 141 -
131. S. K. Arya, G. Chornokur, M. Venugopal, and S. Bhansali, "Antibody functionalized
interdigitated μ-electrode (IDμE) based impedimetric cortisol biosensor," Analyst, vol.
135, pp. 1941-1946, 2010.
132. C. A. Gutierrez, R. Sheybani, and E. Meng, "Electrochemically-based dose measurement
for closed-loop drug delivery applications," in Solid-State Sensors, Actuators and
Microsystems Conference (TRANSDUCERS), 2011 16th International, 2011, pp. 2839-
2842.
133. R. Sheybani, A. Cobo, and E. Meng, "Wireless programmable electrochemical drug
delivery micropump with fully integrated electrochemical dosing sensors," Biomedical
microdevices, vol. 17, p. 74, 2015.
134. R. Lo, P.-Y. Li, S. Saati, R. N. Agrawal, M. S. Humayun, and E. Meng, "A passive
MEMS drug delivery pump for treatment of ocular diseases," Biomedical microdevices,
vol. 11, p. 959, 2009.

- 142 -

ensors discussed in previous chapters were all fabricated on a Parylene C substrate, mostly
due to the material’s high thermal resistivity, high biocompatibility, and amenability to
microfabrication. However, Parylene C only served as a static substrate for metal traces,
and its unique properties as a highly flexible polymer were not put to use. Micromachined sensors
which take advantage of Parylene C’s flexibility could prove beneficial for monitoring a number
of diseases, including hydrocephalus. To that end a kirigami-based strategy was investigated for
strain transduction, using serpentine gold traces embedded in a thin-film Parylene C substrate.
Multiple slit designs were evaluated, and devices with more tightly packed slits stretched up to
17.5 mm (350% strain) before mechanical failure occured. Multiple modes of strain transduction
were evaluated, including DC resistance changes due to stress at trace inflection points, high-
frequency changes in self-impedance, and inter-trace capacitance. DC resistance was found to
increase linearly with stretch distance with sensitivities between 0.04 and 0.16 Ω/mm. Capacitance
was also found to increase linearly with stretch distance after 2 mm, and high-frequency impedance
increased nonlinearly, with different trace routing designs resulting in different patterns of
impedance change with strain. The biocompatible construction and extremely low profiles (20 µm
thick) are attractive for minimally invasive in vivo strain sensing applications. For hydrocephalus
treatment, stretch sensors could be used to directly monitor the size of ventricles, and combining
a thin-film reflected impedance coil with a kirigami sensor could lead to a passive, wireless,
implantable stretch sensor. Outside of sensing, Parylene C kirigami devices could be used to enable
flexible electronics or to provide mechanical filtering and stability for neural probes or other
implantable devices.

Chapter 4
USING KIRIGAMI PRINCIPLES TO DEVELOP
STRETCHABLE PARYLENE C DEVICES
S

- 143 -
Kirigami
Kirigami (a Japanese term translating to ‘cut’ and ‘paper’) is a paper cutting technique
often used to complement origami and enable the creation of three-dimensional shapes from a two-
dimensional sheet of paper (Fig. 4.1). By strategically placing cuts in a planar paper substrate and
then subjecting the substrate to tension or compression, complex three-dimensional patterns can
be produced
1-3
.

Figure 4.1 Kirigami techniques use slits or cuts in a two-dimensional substrate (such as paper) to generate
complex three-dimensional shapes
4
. Credit: McEwen Lab, Nature, DOI: 10.1038/nature14588

Research groups in several different fields have implemented kirigami techniques for
manufacturing, packaging, actuation, or sensing
5-8
. When applied to thin film devices such as
microelectromechanical systems (MEMS), kirigami enables complex three-dimensional structures
using only two-dimensional planar microfabrication techniques
9
. Most MEMS fabrication
techniques involve deposition onto or etching on planar silicon or polymer substrates. This limits
device aspect ratios to the thickness of the substrate, which is tens to hundreds of microns thick.
By combining kirigami patterns and microfabrication, complex three-dimensional designs that
possess microscale or nanoscale features can be produced. Several groups have reported lithium-
ion batteries or supercapacitors fabricated using kirigami patterns
8,10,11
. These devices were first
fabricated on a planar substrate, then released and folded into their final, packaged shape. Kirigami
devices can also generate out-of-plane movement from linear displacement or applied force;
applications include sun-tracking solar cells
12
, dynamic mechanical filters with tunable
properties
13
, and tunable diffraction gratings
14
(Fig 4.2). Combined with piezoelectric or
pneumatic actuation, kirigami can be used to produce complex motion from simple, one-
dimensional forces. Mechanical hands
15
, biomimetic robots
16,17
, artificial muscles
18
, and energy
harvesters
19
that combine kirigami techniques with actuators have all been reported.

- 144 -

Figure 4.2 Left: Solar cells printed on polyimide with kirigami slits can change their angle with linear
motion. Credit: A. Lamoureux, et.al., 2015
12
. Right: Kirigami slits in graphene allow the substrate to stretch
by 267%, compared to 4% for the substrate without slits. Credit: T.C. Shyu, et.al., 2015
20
.
A common kirigami technique is the use of offset rows of slits to enable a material to
stretch far beyond what its normal tensile properties would allow. By converting tensile stress to
torsion, a kirigami slit pattern allows linear movement over a high dynamic range without
compromising the structural integrity of the substrate
14,21
. Kirigami slits have been utilized in
graphene to achieve up to 267% strain before mechanical failure
4,20,22
, while on Parylene C
substrates stretch distances more than 1100% have been reported
23
. Slotted kirigami devices
exhibit minimal stress between slits, suggesting electronic stability in conduction paths. By
introducing non-uniform conduction paths and selectively increasing trace resistance in regions
which tend toward either compressive or tensile strain, a trace’s DC electrical resistance can
correlate with strain
24
. In literature, flexible stretch sensors typically exhibit relatively strain
operation
24,25
. The potential combination of flexible, thin-film Parylene C with kirigami slit
patterns could enable strain sensors with much higher dynamic ranges.

- 145 -

Figure 4.3 A kirigami slit pattern etched into a 10 µm thick Parylene C substrate. The devices has not yet
been released from its silicon carrier wafer.

Parylene C Kirigami Test Devices
4.2.1 Devices design
To test kirigami patterns for stretchable Parylene C electronics, devices with serpentine traces
wound among an etched kirigami slit array were designed and fabricated (Figure 4.4). Each film
contained eight gold traces sandwiched between 10 µm thick Parylene C layers, with each trace
forming a single device. Parylene C is highly flexible but has a low elongation at break (between
20% and 25% unannealed
26,27
), which limits its elasticity. In addition, Parylene C’s compatibility
with oxygen plasma etching allows fabrication of precisely aligned, high aspect ratio kirigami
slits
28
. Each device consisted of a kirigami slit array 5 mm long and 4 mm wide with contact pads
at either end.

Figure 4.4 Parylene strain sensors consist of an actuation section 5 mm long and 4 mm wide containing
gold traces wound through an etched kirigami slit array, with two contact pad regions at either end.

- 146 -
Slit dimensions were held constant at 600 µm long and 20 µm wide. Slit ends were capped
with 50 µm diameter circles to avoid stress concentration
23
. However, lateral etching of Parylene
during deep reactive ion etching caused feature wash-out and resulted in rounded slits 600 µm long
and 50 µm wide. Two variations of the kirigami slit array were designed. A-type devices contained
slits spaced 400 µm apart with inter-row spacing of 200 µm, while slits in B-type devices were
spaced 150 µm apart in offset rows every 100 µm (Figure 4.5). Serpentine gold traces between 5
and 20 µm wide with nominal DC resistance values of 50 and 150 Ω were wound throughout the
slit array. Traces either wound to one side of slits consistently, to enhance proximity between
traces and increase inter-trace capacitance, or alternated sides to isolate self-impedance.

Figure 4.5 (Top) Type A devices contained slits spaced 400 µm apart in offset rows, with 200 µm
separation between rows. (Bottom) Type B devices contained slits spaced 150 µm apart in offset rows with
100 µm separation. Traces either wind around slits consistently (right two traces) or alternated sides (left
trace).

4.2.2 COMSOL simulations
Both A- and B-type kirigami devices were simulated in COMSOL, and the maximum
internal stress during stretching was calculated. Material properties of Parylene C were based on
literature from Specialty Coating Systems, Inc., which manufactures Parylene dimer and
deposition systems
30
. Both A-type and B-type FEM models took their dimensions from the
fabrication masks for the kirigami devices, without considering overetching or the multi-layer
nature of planar microfabricated Parylene devices (Fig. 4.6).
A
B

- 147 -

Figure 4.6 Top view of the A-type and B-type models used for COMSOL simulations
Stress inside A-type devices was simulated between 0 and 3 mm displacement, and
concentrated in four lateral bands between offset rows of slits (Figure 4.7). Maximum stress
quickly exceeded the yield strength of Parylene C (3200 MPa
29
).

Figure 4.7 The von Mises stress within the A-type kirigami device when stretched (i) by 1 mm and (ii) by
3 mm. Stress was concentrated in lateral bands between the rows of offset slits. Strain is shown at a 1:1
scale, with an outline showing the geometry of the unstretched device.
Stress inside B-type devices remained much lower for the same amount of strain (Figure
4.8). Stress was uniformly distributed throughout the device, and B-type devices could stretch

- 148 -
more than 10 mm before stress exceeded the yield strength of Parylene C. Stress in B-type devices
concentrated around the ends of slits (Figure 4.9).

Figure 4.8 The von Mises stress inside B-type devices when stretched by (i) 2 mm and (ii) 14 mm. Stress
levels are much lower in B-type devices than in A-type devices for the same stretch distance. Strain is
shown at a 1:1 scale, with an outline showing the geometry of the unstretched device.

Figure 4.9 A close-up of the stress inside of a B-type device stretched by 6 mm. Stress in these devices is
much more uniform than in A-type devices, and is concentrated around the ends of the slits.
Plotting the maximum von Mises stress acquired from simulation versus strain revealed
the advantage of the B-type kirigami slit design (Figure 4.10). The stress in simulated A-type
devices exceeded the yield strength of Parylene C after only 1.7 mm displacement, while the stress

- 149 -
in simulated B-type devices did not exceed Parylene C’s yield strength until 11.8 mm
displacement.

Figure 4.10 The maximum stress in A and B type kirigami devices versus strain, taken from COMSOL
simulations. Stress exceeded the yield strength of Parylene C after 1.696 mm for A-type devices and after
11.84 mm for B-type devices.

4.2.3 Fabrication
Both A- and B-type kirigami devices consisted of Parylene C layers sandwiching gold
traces, and were fabricated using previously reported micromachining techniques
29,31
(Fig. 4.11).
First, 10 µm of Parylene C was deposited onto a 4” silicon carrier wafer. A 200 Å titanium adhesion
layer and 2000 Å of gold were then deposited using electron-beam metal deposition and patterned
via liftoff using AZ5214 image reversal photoresist. A second 10 µm layer of Parylene was
deposited as insulation, and the slit array was etched with a switched-chemistry deep reactive ion
etch in oxygen plasma
32
using an Oxford PlasmaLab 100 ICP etcher. This etching process can
generate high aspect ratios, with sidewall angles exceeding 70°. Contact pads and device cutouts
were exposed with a second, lower power deep reactive ion etch. To prevent gold sputtering, the
RF power was lowered such that voltages did not exceed 70V. Devices were released from the
silicon carrier wafer by soaking in deionized water. Overetching of slits and the device cutout was
necessary, since peeling devices off the carrier wafer could result in the destruction of the delicate
kirigami slit array.

- 150 -

Figure 4.11 Sensor fabrication involved (a) deposition of Parylene C on a silicon carrier wafer, (b) electron
beam deposition and liftoff patterning of 2000 Å Au with 200 Å Ti adhesion layer, (c) etching of slits via
a switched-chemistry deep reactive ion etch in oxygen plasma and (d) release from the carrier wafer by
peeling while immersed in DI water. © 2017 IEEE

Electrical connection was achieved by attaching a PEEK (polyether ether ketone) backing to
contact pads using cyanoacrylate glue and inserting both ends in zero insertion force (ZIF)
connectors (12 channel, 0.5 mm pitch; Hirose Electric Co., Simi Valley, CA) connected to flat
flexible cables (FFC; Molex Inc., Lisle, IL)
31
(Fig. 4.12).

Figure 4.12 Devices were attached to PEEK backing and inserted into ZIF connectors during electrical
testing. A stretched device is shown between the ZIF connectors. © 2017 IEEE

- 151 -
4.2.4 Testing methods
Stretch testing of devices was accomplished by immobilizing one end and displacing the
other end at 0.2 mm/s using a ThorLabs Z812 motorized stage. For mechanical testing, force was
measured using a calibrated force gauge. For electrical testing, DC resistance was measured using
a Keithley 2400 SourceMeter. All data was collected with a custom LabVIEW program. After
mechanically and electrically testing devices to failure, three methods of strain transduction were
evaluated. The DC resistance of traces was tracked over multiple stretch cycles to monitor stress-
mediated resistance changes along the conduction path. The high frequency impedance of single
traces was measured to evaluate changes in self-impedance. Finally, inter-trace capacitance was
tracked as a measure of net changes in the distance between traces as well as changes in dielectric
properties due to stress within the Parylene substrate. High-frequency impedance and capacitance
were measured using an Agilent E4980A precision LCR meter.

Characterization
4.3.1 Mechanical characterization
Devices with A-type and B-type slit arrays were stretched until mechanical failure while
measuring the applied force. Identical devices without any etched kirigami slits were tested as a
control. Displacement was defined as the linear distance from the unstretched sensor along the axis
between contact pad regions, and strain was defined relative to the length of the slit array (5 mm
displacement = 100% strain). Both A- and B-type kirigami slits drastically increased the distance
that devices could be stretched before breaking. However, A-type devices appeared to plastically
deform after only 6 mm displacement (120% strain), while B-type devices appeared to maintain
elastic deformation to between 10 mm and 12 mm displacement (200% to 240% strain) (Figure
4.13). On average, A-type devices failed at 101% strain, corresponding to 0.7 N force, while B-
type devices failed at 350% strain and 0.3 N of force (Figure 4.14). Parylene devices without slits
broke almost immediately at approximately 2 N of force.

- 152 -

Figure 4.13 Representative force vs displacement curves for device with A-type and B-type kirigami slit
patterns, compared with an identical Parylene device without kirigami slits. A-type devices begin plastically
deforming at around 6 mm displacement, while B-type devices stay in the elastic deformation range until
around 10-12 mm displacement, though this transition is not well defined. Parylene devices without slits
deform and break almost immediately.

Figure 4.14 Break force and break distance for devices with A-type kirigami slits, B-type kirigami slits,
and no slits (bare Parylene). A-type devices break at around 101% strain, while B-type devices break at
around 350% strain. Devices without slits break almost immediately (14% strain). Break force was
inversely proportional to break distance.

- 153 -

4.3.2 DC electrical characterization
Electrical integrity of traces during stretching was evaluated by measuring DC resistance while
stretching to failure. DC resistance on A-type device traces was stable through 3.2 mm
displacement (64% strain), then increased exponentially until open circuit failure occurred at 3.8
mm displacement (78% strain). B-type trace resistance stayed stable through 6 mm displacement
(120% strain) and failed at 9 mm displacement (180% strain) (Figure 4.16). Microcracks in the
Parylene C were observed at inflection points near the ends of slits (Figure 4.15).

Figure 4.15 Microscope image of a B-type device stretched to 9 mm (180% strain). Stress is concentrated
at two inflection points near the end of the slit. © 2017 IEEE

- 154 -

Figure 4.16 The DC resistance of a trace on A- and B-type devices stretched at 0.2 mm/s until failure.
Tighter slit spacing increased the stretch distance to failure from 78% to 180% strain and showed that wider
spacing between slits did not improve the stability of embedded traces. A-type traces have a lower DC
resistance than B-type traces due to shorter path length; the baseline resistance of A-type traces is
approximately 50 Ω versus 108 Ω for B-type traces.

Strain Transduction
4.4.1 DC strain sensitivity
Kirigami devices were evaluated for strain sensitivity, with the goal of developing a thin-film
Parylene C strain sensor. Because of their greater mechanical and and electrical stability, only B-
type devices were evaluated for strain transduction. DC trace resistance was measured during
cycling from 0-4 mm displacement, revealing a positive linear correlation between resistance and
strain (sensitivity 0.10-0.16 Ω/mm) (Figure 4.17). However, measurement noise was significant
compared to the sensor’s sensitivity, resulting in 3σ resolutions between 0.2 mm and 1.6 mm (4%
to 32% strain). A small amount of hysteresis was observed over multiple stretch cycles (Figure
4.18).

- 155 -

Figure 4.17 The DC resistance of two different traces on a B-type device over three displacement cycles.
The top trace shows a strain sensitivity of 0.131 Ω/mm with 3σ resolution of 1.07 mm (21.4% strain), while
the bottom traces has a strain sensitivity 0.100 Ω/mm with 3σ resolution of 0.81 mm (16.2% strain). Both
traces are of the same design; differences in baseline resistance are likely due to process variations.

Figure 4.18 The average DC resistance change versus strain for a trace over three displacement cycles
between 0 and 4 mm. This trace showed a sensitivity of 0.156 Ω/mm with a resolution of 0.21 mm. A small
amount of hysteresis is observed, most likely due to stress relaxation when stretched. The baseline
resistance for this trace was 118.7 Ω.

- 156 -

4.4.2 High-frequency trace impedance
Complex impedance at 2 MHz was recorded for a single trace during multiple displacement
cycles from 0-5 mm. The complex impedance of a trace includes the trace resistance R, the self-
inductance L, and the self-capacitance CS, and has the relationship
𝑍 =
𝑅 1 + 𝑗𝜔 𝑅 𝐶 𝑠 + 𝑗𝜔 𝐿
where ω is the measurement frequency and j is √ − 1. The impedance magnitude showed a
nonlinear dependence on strain, which was different depending on whether the trace wound to one
side of slits consistently or alternated winding side. Consistent traces showed a sigmoidal increase
in impedance with an inflection point at approximately 2.5 mm displacement and a total impedance
increase of 1%, while alternating traces showed a simpler curved trend and a total impedance
increase of 0.4% (Fig. 4.19).

Figure 4.19 The high-frequency impedance magnitude shows a non-linear relationship with displacement,
with (left) a curved relationship for alternating traces and (right) a sigmoidal relationship for consistent
traces.

4.4.3 Inter-trace capacitance
The capacitance between adjacent consistently-wound traces (CP) was evaluated at 100 kHz.
Capacitance was relatively stable up to 2 mm displacement and to increase linearly at a rate of 4.7
fF/mm between 2 and 5 mm displacement (Fig. 4.20).

- 157 -

Figure 4.20 The capacitance between two adjacent traces measured at 100 kHz. At around 2 mm
displacement, the capacitance begins to increase linearly with strain.

Discussion
Experimental results showed that kirigami devices fabricated out of Parylene C and gold
stretched to 4.5x their original size before mechanical failure occurred. However, the electrical
integrity of traces failed much earlier, at a point which appeared to coincide with the transition
between elastic and plastic deformation as seen in force/displacement curves. Microcracks in
Parylene are visible near the ends of each slit after stretching, and are the likely cause of electrical
failure. Electrical and mechanical tests confirm that increasing slit density enables high strain
operation, but increasing inter-trace width does not improve the electrical stability of traces.
The DC resistance of traces was observed to correlate linearly with strain. This strain
sensitivity likely results from stress within the thin-film gold traces, which concentrates at
inflection points near the end of slits. Even without optimization, the sensitivity of resistance-based
strain transduction is competitive with previously reported polyimide devices. Parylene devices
possessed a higher dynamic range despite a conservative testing protocol (0.16 Ω/mm over 80%
strain vs -0.18 Ω/mm over 20% strain for polyimide sensors
25
).

- 158 -

Table 4.1. This work compared to other proposed strain sensors. CNT = carbon nanotubes, EGaIn = eutectic
gallium-indium (liquid metal).

Max Strain Gauge Factor Thickness Active Material
Park 2012 250% 3.6 3500 µm EGaIn
Lee 2013 25% 2-22 1000 µm CNT
Firouzeh 2015 20% 0.004 60 µm Constantin
This work

350% 0.004 20 µm Gold

Measurement noise was higher than desirable for precise strain measurements. High noise
levels may have been due to thermal noise in the conductive trace or environmental noise in the
testing setup. A small amount of hysteresis was observed, possibly due to polymer reorganization
during device displacement. There was also a large difference in sensitivity between traces, even
within the same device; the trace’s position on the device did not appears to be the cause of
sensitivity differences. Further work is needed to determine the causes of sensitivity differences
and to optimize trace designs for DC strain sensing.
Both self-impedance and inter-trace capacitance increased with strain. Increases in trace
impedance were likely a result of variations in self-inductance within a trace. Changes in inter-
trace capacitance may have resulted from varying either the separation between traces or the
dielectric properties of Parylene C under stress. Construction out of Parylene C and gold renders
sensors not only resistant to corrosion but biocompatible and suitable for use in the human body.
This contrasts with other reported strain sensors that utilize semiconductor-based piezoresistive
materials or carbon nanotubes, neither of which is compatible with chronic implantation in the
human body.

Potential Applications
There are several potential applications for stretchable Parylene C electronics enabled by
kirigami designs. Improvements in strain transduction would enable extremely low profile (~20
µm) stretch sensors which are compatible with chronic implantation in the human body. These
sensors could be used to measure bladder filling for incontinence patients
32-35
or to measure the
size of ventricles in hydrocephalus patients
36,37
. Outside the body, a kirigami strain sensor could
be used to monitor muscle or joint position, and a thin-film kirigami strain sensor adhered to the
skin using biocompatible adhesive would be barely noticeable by a user. Outside of strain sensing,
kirigami techniques could be used to increase the compliance of neural probes and increase the
longevity of neural interfaces. For penetrating neural probes, the difference in compliance between
the probe material and the brain causes strain and inflammation due to the brain’s micromotion,
leading to glial encapsulation and scarring and reducing the fidelity of neural signals. Constructing
probes out of soft materials such as PDMS or Parylene C can improve this problem
38
, but these

- 159 -
materials still have Young’s moduli several orders of magnitude higher than brain tissue, and
neural probes which are fixed to the skull will still experience high shear stresses due to relative
motion between the skull and brain. Kirigami patterns in flexible Parylene C neural probes could
be used to mechanically isolate neural probes from the skull and could provide a mechanical filter
for micromotion, reducing inflammation and scarring (Fig. 4.21).

Figure 4.21 Kirigami slits could be used to mechanically isolate neural probes from relative motion
between the skull and brain, increasing effective compliance and leading to less inflammation and glial
encapsulation.

Conclusions and Future Work
Using kirigami slit designs, devices were fabricated out of Parylene C and gold which could
stretch to multiple times their length while maintaining electrical integrity of traces. The devices
were found to have multiple electrical parameters which vary predictably with strain. DC
resistance’s sensitivity to strain was found to be of comparable sensitivity to previously reported
polymer sensors but with a higher dynamic range and simpler trace geometry. Inter-trace
capacitance and high-frequency inductance also varied with strain, which could lead to additional
sensing modalities based off of frequency transduction as opposed to resistive transduction.
Construction out of only Parylene and gold renders these sensors not only resistant to corrosion
but fully biocompatible and suitable for chronic implantation in the human body. This contrasts
with other common strain sensors which utilize semiconductor-based piezoresistive materials or
carbon nanotubes for strain sensing, neither of which is amenable to chronic implantation. A high
dynamic range, low stress strain sensor could potentially be used in numerous biomedical
applications.

- 160 -
Future work will involve optimizing slit size and orientation in order to maximize stretch
distance and minimize internal stress, as well as exploring different trace geometries which may
improve the sensitivity of strain transduction. Capacitive and impedance-based strain sensing will
continue to be explored, and kirigami strain sensors paired with thin-film inductive coils
34
will be
tested to develop passive, wireless strain sensors.

References
1. T. Castle, Y. Cho, X. Gong, E. Jung, D. M. Sussman, S. Yang, et al., "Making the cut:
Lattice kirigami rules," Physical review letters, vol. 113, p. 245502, 2014.
2. S. J. Callens and A. A. Zadpoor, "From flat sheets to curved geometries: Origami and
kirigami approaches," Materials Today, 2017.
3. R. M. Neville, F. Scarpa, and A. Pirrera, "Shape morphing Kirigami mechanical
metamaterials," Scientific reports, vol. 6, p. 31067, 2016.
4. M. K. Blees, A. W. Barnard, P. A. Rose, S. P. Roberts, K. L. McGill, P. Y. Huang, et al.,
"Graphene kirigami," Nature, 2015.
5. Y. Liu, K. He, G. Chen, W. R. Leow, and X. Chen, "Nature-Inspired Structural Materials
for Flexible Electronic Devices," Chemical reviews, vol. 117, pp. 12893-12941, 2017.
6. M. Weston, "Anisotropic Operations," International Journal of Architectural Computing,
vol. 10, pp. 105-120, 2012.
7. Y. Zhang, F. Zhang, Z. Yan, Q. Ma, X. Li, Y. Huang, et al., "Printing, folding and
assembly methods for forming 3D mesostructures in advanced materials," Nature
Reviews Materials, vol. 2, p. 17019, 2017.
8. Z. Song, Studies of Origami and Kirigami and Their Applications: Arizona State
University, 2016.
9. Y. Zhang, Z. Yan, K. Nan, D. Xiao, Y. Liu, H. Luan, et al., "A mechanically driven form
of Kirigami as a route to 3D mesostructures in micro/nanomembranes," Proceedings of
the National Academy of Sciences, vol. 112, pp. 11757-11764, 2015.
10. Z. Song, X. Wang, C. Lv, Y. An, M. Liang, T. Ma, et al., "Kirigami-based stretchable
lithium-ion batteries," Scientific reports, vol. 5, 2015.
11. K. K. Fu, J. Cheng, T. Li, and L. Hu, "Flexible batteries: From mechanics to devices,"
ACS Energy Letters, vol. 1, pp. 1065-1079, 2016.
12. A. Lamoureux, K. Lee, M. Shlian, S. R. Forrest, and M. Shtein, "Dynamic kirigami
structures for integrated solar tracking," Nature communications, vol. 6, p. 8092, 2015.
13. M. Ouisse, M. Collet, and F. Scarpa, "A piezo-shunted kirigami auxetic lattice for
adaptive elastic wave filtering," Smart Materials and Structures, vol. 25, p. 115016,
2016.
14. N. A. Kotov, L. Xu, X. Wang, Y. Kim, T. C. Shyu, and J. Lyu, "Kirigami
Nanocomposites as Wide-Angle Diffraction Gratings," ACS nano, 2016.
15. Y. H. Chan, Z. Tse, and H. Ren, "Design evolution and pilot study for a kirigami-inspired
flexible and soft anthropomorphic robotic hand," in Advanced Robotics (ICAR), 2017
18th International Conference on, 2017, pp. 432-437.
16. J. Rossiter and S. Sareh, "Kirigami design and fabrication for biomimetic robotics," in
Bioinspiration, Biomimetics, and Bioreplication 2014, 2014, p. 90550G.

- 161 -
17. K. Zhang, C. Qiu, and J. S. Dai, "Helical Kirigami-Inspired Centimeter-Scale Worm
Robot With Shape-Memory-Alloy Actuators," in ASME 2014 International Design
Engineering Technical Conferences and Computers and Information in Engineering
Conference, 2014, pp. V05BT08A039-V05BT08A039.
18. S. Sareh and J. Rossiter, "Kirigami artificial muscles with complex biologically inspired
morphologies," Smart Materials and Structures, vol. 22, p. 014004, 2012.
19. C. Wu, X. Wang, L. Lin, H. Guo, and Z. L. Wang, "Based Triboelectric Nanogenerators
Made of Stretchable Interlocking Kirigami Patterns," ACS nano, vol. 10, pp. 4652-4659,
2016.
20. T. C. Shyu, P. F. Damasceno, P. M. Dodd, A. Lamoureux, L. Xu, M. Shlian, et al., "A
kirigami approach to engineering elasticity in nanocomposites through patterned defects,"
Nature materials, 2015.
21. M. Isobe and K. Okumura, "Initial rigid response and softening transition of highly
stretchable kirigami sheet materials," Scientific Reports, vol. 6, p. 24758, 2016.
22. Z. Qi, D. K. Campbell, and H. S. Park, "Atomistic simulations of tension-induced large
deformation and stretchability in graphene kirigami," Physical Review B, vol. 90, p.
245437, 2014.
23. Y. Morikawa, S. Yamagiwa, H. Sawahata, M. Ishida, and T. Kawano, "An origami-
inspired ultrastretchable bioprobe film device," in 2016 IEEE 29th International
Conference on Micro Electro Mechanical Systems (MEMS), 2016, pp. 149-152.
24. A. Firouzeh and J. Paik, "The design and modeling of a novel resistive stretch sensor
with tunable sensitivity," IEEE Sensors Journal, vol. 15, pp. 6390-6398, 2015.
25. C. Lee, L. Jug, and E. Meng, "High strain biocompatible polydimethylsiloxane-based
conductive graphene and multiwalled carbon nanotube nanocomposite strain sensors,"
Applied Physics Letters, vol. 102, p. 183511, 2013.
26. R. P. Von Metzen and T. Stieglitz, "The effects of annealing on mechanical, chemical,
and physical properties and structural stability of Parylene C," Biomedical microdevices,
vol. 15, pp. 727-735, 2013.
27. C. Hassler, R. P. von Metzen, P. Ruther, and T. Stieglitz, "Characterization of parylene C
as an encapsulation material for implanted neural prostheses," Journal of Biomedical
Materials Research Part B: Applied Biomaterials, vol. 93, pp. 266-274, 2010.
28. B. J. Kim and E. Meng, "Micromachining of Parylene C for bioMEMS," Polymers for
Advanced Technologies, 2015.
29. "SCS Parylene Properties," ed: Specialty Coating Systems.
30. E. Meng, P.-Y. Li, and Y.-C. Tai, "Plasma removal of Parylene C," Journal of
Micromechanics and Microengineering, vol. 18, p. 045004, 2008.
31. C. A. Gutierrez, C. Lee, B. Kim, and E. Meng, "Epoxy-less packaging methods for
electrical contact to parylene-based flat flexible cables," in 2011 16th International Solid-
State Sensors, Actuators and Microsystems Conference, 2011, pp. 2299-2302.
32. E. L. Koldewijn, P. E. Van Kerrebroeck, E. Schaafsma, H. Wijkstra, F. M. Debruyne, and
G. S. Brindley, "Bladder pressure sensors in an animal model," The Journal of urology,
vol. 151, pp. 1379-1384, 1994.
33. S. R. Lacey, L. A. Carris, A. J. Beyer, and R. G. Azizkhan, "Bladder pressure monitoring
significantly enhances care of infants with abdominal wall defects: a prospective clinical
study," Journal of pediatric surgery, vol. 28, pp. 1370-1375, 1993.

- 162 -
34. W. Sansen, R. Vereecken, R. Puers, G. Folens, and T. Van Nuland, "A closed loop
system to control the bladder function," in Proceedings of the Ninth Annual Conference
of the IEEE Engineering in Medicine and Biology Society, Boston, November, 1987, pp.
13-16.
35. R. Styles, D. Neal, C. Griffiths, and P. Ramsden, "Long-term monitoring of bladder
pressure in chronic retention of urine: the relationship between detrusor activity and
upper tract dilatation," The Journal of urology, vol. 140, pp. 330-334, 1988.
36. A. Linninger, S. Basati, R. Dawe, and R. Penn, "An impedance sensor to monitor and
control cerebral ventricular volume," Medical Engineering and Physics, vol. 31, pp. 838-
845, 2009.
37. J. Jovicich, S. Czanner, X. Han, D. Salat, A. van der Kouwe, B. Quinn, et al., "MRI-
derived measurements of human subcortical, ventricular and intracranial brain volumes:
reliability effects of scan sessions, acquisition sequences, data analyses, scanner upgrade,
scanner vendors and field strengths," Neuroimage, vol. 46, pp. 177-192, 2009.
38. G. Lind, C. E. Linsmeier, and J. Schouenborg, "The density difference between tissue
and neural probes is a key factor for glial scarring," Scientific reports, vol. 3, p. 2942,
2013.

- 163 -

ESPITE progress in medical electronics and sensors, no implantable sensor for hydrocephalus
shunts has been commercialized. One difficulty in bringing hydrocephalus sensors to market
involves the inherent risks and regulatory hurdles surrounding clinical studies of an implantable
device. In Chapter 2, a flow sensor designed for chronic implantable use in cerebrospinal fluid
(CSF) was presented, and similar sensors for intracranial pressure and catheter patency have also
been developed at USC. To investigate sensor operation under real-world conditions, a multi-
sensor module comprised of these three sensors was tested in pediatric patients with external
ventricular drains (EVDs). EVDs are used for acute hydrocephalus or brain trauma to quickly drain
excess CSF to an external collection bag. EVDs provide an excellent clinical analog for
hydrocephalus shunts; by attaching sensors to EVDs, sensor performance can be evaluated in
human CSF with realistic drainage dynamics with minimal patient risk or regulatory oversight.
Multi-sensor modules were packaged in luer lock spacers and connected to a custom electronic
datalogger. Ten sensor modules were tested in the clinic between 2015 and 2016. Unfortunately,
several issues with device integrity, datalogger accuracy, and doctor-engineer miscommunication
resulted in suboptimal data. Patient enrollment was paused while these problems were identified
and fixed. To improve device integrity, the metal deposition and thermal annealing processes were
changed, and a new electronic datalogger with much higher precision was developed. Tests with
new sensors and dataloggers showed accurate data collection during two weeks of constant
operation in human CSF on the benchtop. Data collection in patients using the second-generation
devices is currently underway.

Chapter 5
CLINICAL TESTS OF A MULTI-SENSOR
MODULE IN PEDIATRIC EVD PATIENTS
D
- 164 -
A Multi-Sensor Module for Hydrocephalus Shunts
Novel sensor technology is necessary to chronically monitor the fluidic state inside
hydrocephalus shunts. Several implantable MEMS devices constructed only from thin-film
Parylene C and biocompatible metals have been developed at USC
1-12
. Recently, a new class of
impedimetric sensors that interact directly with the in vivo fluidic environment have been
developed
13-19
. Three impedimetric sensors in particular may be useful chronic measurement of
hydrocephalus shunt status.

5.1.1 Impedimetric sensors for hydrocephalus shunts
Previous work has led to the development of a flow sensor
13-15
, pressure sensor
18,19
, and
patency sensor
16,17
that are relevant to hydrocephalus treatment. Chapter 2 describes the
development of the flow sensor, which consists of a platinum resistive heater and one or more
pairs of impedance measurement electrodes. The flow rate affects heat transfer between the heater
and electrodes, which can be measured as a change in high-frequency electrochemical impedance.
The pressure sensor consists of platinum electrodes placed inside a microfabricated channel
with fluidic ports at either end
18
. When the microchannel fills with fluid, the resistance through
the channel can be measured using high-frequency impedance between the electrodes at each end.
To transduce pressure, a microbubble is created inside the microchannel via electrolysis (Fig. 5.1).
The bubble remains confined inside the microchannel, and the size of this bubble correlates with
impedance. Several methods have been developed to transduce pressure using this microbubble.
For a stable bubble, rapid changes in pressure such as those produced by breathing or a heartbeat
cause the bubble to compress and expand, which is reflected in the impedance magnitude. Over
longer timescales, both the initial size and dissolution rate of the bubble can be used to transduce
absolute pressure. Additionally, by orienting the fluidic channel parallel to the direction of flow a
pressure difference develops between the two fluidic ports, allowing flow rate measurement
19
. The
first-generation pressure sensor used a single pair of electrodes for both bubble generation and
impedance measurement, but to increase the consistency of electrolysis the pressure sensor used
in the multi-sensor module includes a nucleation core with separate bubble generation electrodes
(Fig. 5.2).
- 165 -

Figure 5.1 The pressure sensor operates by creating a confined microbubble. The microbubble is contained
in a Parylene C fluidic chamber, and its size is measured using electrochemical impedance. The initial size
and dissolution rate of the microbubble can be used to transduce absolute pressure, and high-frequency
pressure changes can be transduced by measuring the compression and expansion of a stable bubble.

Figure 5.2 The pressure sensor consists of two pairs of platinum electrodes placed inside a microfabricated
fluidic chamber (filled with red photoresist in this image), with open ports at each end. Electrolysis produces
a bubble in the nucleation core, which then fills the bubble chamber, where its size is measured by electrodes
at either end.
The patency sensor directly measures the occlusion of a hydrocephalus shunts’s proximal
catheter
16
. The sensor consists of an electrode inside the catheter and an electrode placed outside
the catheter but still exposed to CSF (Fig. 5.3). By measuring the high-frequency electrochemical
impedance between these two electrodes, occlusion of the proximal catheter can be directly
transduced.
- 166 -

Figure 5.3 The patency sensor operates by sensing the electrochemical impedance between an electrode
inside and an electrode outside of the proximal catheter. Any occlusion of the catheter end will cause a
corresponding increase in impedance.

5.1.2 A multi-sensor module
The flow, pressure, and patency sensors were combined on the same substrate to create the
multi-sensor module. Two variations of the multi-sensor module were fabricated (Fig. 5.4). Z type
devices contain two pairs of flow sensing electrodes placed 1 mm and 3 mm away from the center
of the heater, and a pressure sensor oriented parallel to the direction of flow. R type devices contain
flow electrodes 0.5 mm and 2 mm away from the heater, and a pressure sensor oriented
perpendicular to the direction of flow.

- 167 -

Figure 5.4 The full sensor system combines pressure, flow, and patency sensors on a single die. Two sensor
designs were fabricated; Z type devices contain a flow sensor with electrodes spaced 1 mm and 3 mm
downstream of the heater and a pressure sensor oriented parallel to the direction of flow, and R type devices
contain flow sensor electrodes spaced 0.5 mm and 2 mm downstream of the heater and a pressure sensor
normal to the direction of flow.
Fabrication of the multi-sensor system closely followed the fabrication procedure for flow
sensors, which was discussed in Chapter 2
20
(Fig. 5.5). First, 10 µm of Parylene C was deposited
onto a silicon carrier wafer using a SCS LabCoter 2 CVD deposition system. A mask for metal
deposition was defined using a 2 µm layer of AZ5214 negative sidewall resist, and 2000 Å of
platinum was sputter deposited by LGA Thin Films, Inc. Metal features were then defined using
liftoff in hot (~60°C) acetone for 1-4 hours with gentle scrubbing. A second 10 µm layer of
Parylene C was deposited, and contact pads were etched out using a switch-chemistry deep reactive
ion etch with oxygen plasma. All etching steps relied on a 12 µm thick etch mask of AZ4620
photoresist and was performed in an Oxford Plasmalab 100 ICP Etcher. To create the fluidic
chambers for the pressure sensor, a 12 µm thick AZ4620 sacrificial layer was patterned, and a 4
µm layer of Parylene C was deposited on top. Fluidic ports and contact pads were then etched, and
a third etch defined device outlines. Devices were released from the silicon carrier wafer by gently
peeling while submerged in deionized water. Sacrificial photoresist in the fluidic chambers was
dissolved by soaking in warm (~40°C) acetone for 12-18 hours. To improve adhesion between
Parylene C layers, devices were annealed post-release in a vacuum oven at 200°C for 48 hours
21
.
- 168 -

Figure 5.5 Fabrication of our sensors involves (a) CVD deposition of 10 µm Parylene C on a silicon carrier
wafer, (b) electron-beam depositing platinum and patterning via liftoff, (c) depositing a second 10 µm
Parylene C layer and exposing electrodes via DRIE, (d) defining the fluidic chamber using sacrificial
photoresist and a 5 µm Parylene C layer, (e) DRIE etching of fluidic ports, electrodes, and contact pads,
and (f) release from the carrier wafer, followed by dissolving the sacrificial photoresist in an acetone bath.

Designing a Sensor System for External Ventricular Drains
5.2.1 Background on EVDs
External ventricular drains (EVDs) are used to acutely relieve swelling or excess fluid
buildup within the brain
22
. EVDs contain a proximal catheter identical to those used in
hydrocephalus shunts which is inserted into the brain’s ventricles. This proximal catheter is
connected to a drainage setup which allows doctors to manually set drainage pressure, sample
CSF, and measure ICP (Fig. 5.6). Excess CSF drains into a removable bag. To set drainage
pressure, a physician raises the drainage column a certain height above the patient’s head, and the
difference in pressure between the ventricles and the raised column controls the flow rate.
Similarly, to measure ICP a doctor raises the column until fluid ceases to flow
23
. Some EVDs have
additional electronic pressure sensors attached to allow continuous ICP monitoring
24
. Under
normal clinical procedures a patient’s ICP and the total volume of fluid collected is measured every
hour. EVDs are often used in cases of acute hydrocephalus caused by brain trauma or infection,
and can be used in chronic hydrocephalus to quickly decrease ICP before a permanent shunt is
implanted
25
.
- 169 -

Figure 5.6 The Becker External Drainage and Monitoring System from Medtronic, which allows pressure
measurement, draining pressure setting, and CSF sampling
EVDs have been used in several studies to investigate CSF production and drainage in
hydrocephalus patients
23,26-28
, and ICP measurement using an EVD is a common clinical
procedure
24
. Therefore, EVDs offer an ideal testing platform for sensors designed for implantation
in hydrocephalus shunts. By attaching a device to an EVD, CSF dynamics can be monitored
without implanting sensors or electronics in the human body. Sensors in an EVD would be exposed
to similar pressures and flow rates as those seen in hydrocephalus shunts
28
, so sensors could be
tested under real-world physiological conditions. In addition, the presence of blood and protein in
CSF
29
would allow biofouling and material robustness to be evaluated. Patients do not typically
have EVDs inserted for more than 14 days
30
, so such a study would not allow for long-term testing
of drift or biofouling, but the study would allow the investigation of many failure modes which
could occur in an implanted sensor system. Differences between sensor implementation in EVDs
versus in hydrocephalus shunts include lower temperature at the sensors, since fluid will be
exposed to room temperature after exiting the skull, and differences in sensor orientation during
operation, since sensors are not mechanically connected to the patient’s skull. In addition, CSF
inside an EVD tends to have higher levels of blood and protein than CSF inside a hydrocephalus
shunt
31,32
. EVDs have been previously used to evaluate prospective hydrocephalus shunts sensors,
including a catheter-based pressure sensor
33
and a commercial ultrasonic flow sensor
34
.
- 170 -

5.2.2 Integrating the multi-sensor module into EVDs
In 2015, IRB approval was granted to test a Parylene C multi-sensor module in patients
with EVDs at Children’s Hospital Los Angeles (CHLA). The proposal involved attaching sensors
to the EVDs such that the system’s normal operation is not affected, and no part of the device
comes in contact with the patient. The sensors were to be attached in-line with the EVD’s sample
port, which is at head height and thus experiences the same fluidic pressure as in the brain. Sensors
would be attached to a custom electronics/datalogger module to power the sensors and
continuously collect data for the duration of the study (Fig. 5.7).

Figure 5.7 The multi-sensor module was integrated into hospital EVDs to test sensor performance under
real-world, clinical conditions. Sensors were packaged into luer lock modules which can interface directly
with EVD hardware. These sensor modules were attached to a custom datalogger box which measures flow,
pressure, and patency at set intervals and saves data to a microSD card for later analysis.

Sensors were packaged into plastic luer lock spacers to enable integration with EVDs. The
luer lock (or luer taper) attachment system is a twist-on, leakless attachment standard which is
widely used in syringes, catheters, and hospital equipment, including EVDs
35
. Male-to-female luer
lock spacers 15 mm long with inner diameters of 3.25 mm (80,379, QOSINA, Edgewood, NY)
were purchased and modified by adding a slit approximately 8 mm long by 1 mm wide using a
drill press and end mill. Electrical connection to the sensors was achieved by attaching a 250 µm
thick polyether ether ketone (PEEK) backing to the sensor module’s contact pads using
cyanoacrylate glue, followed by insertion into a zero insertion force (ZIF) connector (12 channel,
0.5 mm pitch; Hirose Electric Co., Simi Valley, CA) attached to a flat flexible cable (FFC; Molex
Inc., Lisle, IL)
36
. The connected sensor was inserted into the slit in the luer lock spacer, and all
electronics and openings were sealed using EpoTek 353-NDT biocompatible epoxy (Epoxy
Technology, Inc., Billerica, MA).

- 171 -
5.2.3 Sterilization
Any clinical device used with CSF must be sterilized before use. There are several
sterilization methods available for surgical devices: autoclaving, which exposes devices to steam
at temperatures up to 150°C and pressures up to 3.5 atm; ethylene oxide sterilization, which uses
vaporized ethylene oxide to kill any pathogens and deproteinate biofouling
37
; hydrogen peroxide
plasma sterilization, which uses radiofrequency energy to create free radicals from gas-phase
hydrogen peroxide, inactivating microorganisms
38
; and gamma sterilization, which uses gamma
rays to kill bacteria. Parylene C’s glass transition point is between 60 and 90 °C
7
, which rules out
autoclaving as a viable sterilization option. To ensure that devices were not damaged during
sterilization, devices were sterilized at CHLA using a Sterrad 100NX hydrogen peroxide plasma
sterilizer (Advanced Sterilization Products, Irvine, CA) with a maximum temperature of 55°C
39
.
The EpoTek 353-NDT glue used to protect electrical connections changed color after
sterilization, turning from a semitransparent brown to an opaque white (Fig. 5.8). No other part of
the device showed discoloration, and no leakage or flaking was observed after sterilization. To
assess the effects of sterilization on sensors, cyclic voltammetry and electrochemical impedance
spectroscopy was performed on patency sensor electrodes before and after sterilization (Fig. 5.9).
No difference in electrochemical properties were observed. Additional patency transduction tests
were performed before and after sterilization. Patency was measured through silicone test catheters
with varying number of holes to simulate progressive blockage. Catheter tips were placed in a
beaker of phosphate buffered saline (PBS) to simulate the brain’s ventricles, and a platinum
counter electrode with surface area 1 cm
2
was placed in the beaker as an external electrode. No
significant differences in sensor response after sterilization were observed (Fig. 5.10).

Figure 5.8 H 2O 2 plasma sterilization caused discoloration on the surface of the EpoTek epoxy used to
protect electrical connections, but did not affect the integrity of the device.

- 172 -

Figure 5.9 The electrochemical properties of an electrode were tested via (a) electrochemical impedance
spectroscopy and (b) cyclic voltammetry before and after sterilization. No significant differences were
found. © 2016 Springer Nature

Figure 5.10 Patency was transduced before and after sterilization, revealing no significant differences in
sensor performance. © 2016 Springer Nature

5.2.4 Full system packaging
The custom datalogger was packaged in a white plastic case with dimensions 9.5 cm x 7.7
cm x 2.2 cm (1593BB, Hammond Mfg. Ltd., Guelph, Ontario, Canada) for protection in the clinic.
An on-off switch and a power indicator LED were placed on the front (Fig. 5.11). The box was
designed so that users simply had to turn it on, ensure that the LED was lit, and record the time of
activation. The box was connected to the sensors via a custom flexible cable, and the connection
between the cable and the FFC was achieved with a Hirose connector (LX40-12P, Hirose Electric
Co. Ltd., Osaka, Japan), which is directional, robust to stress, and compatible with H2O2 plasma
sterilization, though not compatible with MRI imaging.
- 173 -

Figure 5.11 For hospital use the datalogger electronics were packaged in a white case with a power-
indicating LED and an on-off switch. A custom flexible cable connected the electronics to the luer-lock
packaged sensor module.

Clinical Study Round 1
5.3.1 First-generation electronics design
All of the impedimetric sensors in the multi-sensor module require external electronics to
measure electrochemical impedance, and the pressure and flow sensor require power for the
microfabricated heater and nucleation electrodes. For benchtop characterization, a Keithley 2400
SourceMeter and an Agilent E4980A LCR meter are used to control sensors. Both are operated
using a custom LabView program, which records impedance magnitude and phase information as
well as currents and voltages produced by the Keithley at a rate of five times per second. This
setup allows precise data recording for long periods of time and direct control of sensors, but due
to the size and cost of the equipment involved is not practical for use in a clinical environment
(despite our efforts). Clinical testing also requires operation by physicians or nurses, who do not
have the time to learn how to operate a complex laboratory measurement system. For clinical
testing, a small, unobtrusive, and easy to use electronic datalogger was required (Figs. 5.12, 5.13).
The core of the datalogger electronics was a Teensy 3.1 microcontroller board (PJRC.com,
LLC, Sherwood, OR). This board consisted of an MK20DX256 microcontroller, power
electronics, and a smaller MINI54 microcontroller pre-loaded with an Arduino bootloader. The
Teensy 3.1 contains 33 digital I/O connections and 20 analog inputs, all of which connect to a 16-
bit analog-digital converter (ADC), and can be run at clock speeds up to 96 MHz. To record data,
the microcontroller was connected to an SD card shield (WIZ820io, PJRC.com, LLC, Sherwood,
OR) via the SPI bus, allowing data to be written to a microSD card (2 GB, Kingston Technology,
Inc.). Unfortunately, data could not be written at the target operating frequency of five
- 174 -
measurements per second, so measurements had to be stored in the microcontroller’s flash memory
and uploaded in bulk to the SD card, requiring some averaging and signal smoothing to be
performed in real-time by the microcontroller.

Figure 5.12 The datalogger electronics consisted of a Teensy 3.1 microcontroller board which was
integrated with an SD card shield. For impedance measurement, a 10 kHz wave was sent through voltage
and current limiting components and into a voltage divider, with the output going directly to a 16-bit ADC.
Electrolysis was accomplished by using an LM334 current source, and flow sensing involved directly
stimulating the heater with a 3.3 V square wave.

Figure 5.13 The microcontroller and other electronic components were integrated into a custom PCB
- 175 -
To measure electrochemical impedance, a digital output line on the microcontroller was
set to output a 50% duty cycle square wave at 10 kHz. This signal, which oscillates between 0 and
+3.3 V, was sent through a current limiting resistor and then voltage-locked via a diode with a
forward voltage drop of 0.7 V, ensuring that the signal did not exceed platinum’s water window
(~1.2 V). The signal then traveled through a voltage divider consisting of the electrodes under
consideration and a resistor equal to the expected impedance magnitude of each electrode pair.
The voltage at the center of the divider was recorded by the 16-bit ADC. With a maximum wave
voltage of 0.7 V this allowed measurement with 13900 bit resolution, with each bit corresponding
to approximately 50 µV.
Unfortunately, the ADC in the ARM microcontroller could not acquire measurements
quickly enough to measure the resulting waveform, which for a 10 kHz signal requires a sampling
rate above its Nyquist frequency of 20 kHz. Instead, the board was programmed to record 1000
randomly sampled points at a sampling rate of approximately 1200 Hz, and then calculate the
average and standard deviation of these values. The average should correspond to the DC offset of
the signal (about 0.35 V) and the standard deviation should be the RMS value of the signal. Using
the RMS value and the known voltage divider resistance, the impedance of the electrode pair could
be determined. Calibrated resistor tests showed that this method is accurate over a wide range of
resistances, with the highest sensitivity occurring when the impedance is equal to the voltage
divider resistor (Fig. 5.14). Unfortunately, phase cannot be measured without additional
electronics or a faster sampling rate.

Figure 5.14 Using calibrated resistors, a good correlation was found between resistance and ADC input.
To activate the heater, a constant 3.3 V was delivered via a digital output line. With a
nominal resistance of around 750 Ω, the heater would generate 14 mW of power and approximately
4°C overheat temperature. For bubble generation, a pulse-wave modulated signal was delivered to
a LM334 precision current source (Texas Instruments, Dallas, TX), which could be set using a
discrete resistor to deliver currents as low as 1 µA. All electronics were powered by an 850 mAh
lithium ion battery (PRT-13854, Sparkfun Electronics, Boulder, CO). Battery life of the full system
- 176 -
was tested by activating the heater and current source and measuring impedance from four pairs
of electrodes (pressure, two flow, and patency) every five minutes, with the processor going into
sleep mode between measurements. The system operated correctly for over 14 days.

5.3.2 Benchtop testing – sensors and first-generation electronics
Each sensor was tested independently with the datalogger electronics prior to use in the
clinic. Patency sensor testing was performed using silicone tubes which were sealed at one end
and contained varying numbers of holes 500 µm in diameter. Fluid was drawn through the catheter
and into the sensor module using a syringe (Fig. 5.15). The impedance was measured by the
electronics board between one of the patency electrodes and an external platinum counter electrode
with approximate surface area of 1 cm
2
, which was placed in the fluid reservoir. A good correlation
was observed between the board reading and the number of open holes in the catheter (Fig. 5.16).

Figure 5.15 Patency transduction was tested by drawing PBS through silicone catheters with varying
numbers of open holes and into the luer-lock module. Two methods of patency transduction were tested:
measuring between the two patency electrodes on the Parylene die, or measuring between one of the
electrodes and an external platinum counter electrode (not shown) placed in the beaker of PBS.

- 177 -

Figure 5.16 When using an external catheter, a clear relationship was observed between the number of
open holes and the impedance measured by the datalogger.
Unfortunately, to reduce patient risk it is not possible to use an external electrode in direct
contact with the patient’s CSF, and options for skin electrodes were also vetoed by clinicians due
to patient contact concerns. An attempt was made to transduce catheter patency using the
impedance between the two patency electrodes on the sensor module. Since impedance is
proportional to the total volume of electrolyte surrounding the electrodes, it was theorized that
reducing the number of open catheter holes might be detectable. Tests using the benchtop LCR
meter seemed to confirm this transduction mode, though the sensitivity was much lower and the
trend was in the opposite of the expected direction (Fig. 5.17). Unfortunately, the accuracy and
noise in the datalogger electronics was much worse than in the LCR meter, and patency could not
be accurately transduced using this method with the datalogger electronics. The electrodes for the
patency sensor were therefore used as test electrodes, to measure the integrity of the sensor die and
the presence or absence of biofouling, bubbles, or other debris.

n=3
Mean±SD
- 178 -

Figure 5.17 (Left) Measurements with an LCR meter revealed that the number of open holes is directly
proportional to the impedance between the two patency electrodes on the sensor die. (Right) When using
the datalogger electronics to measure impedance, accuracy decreases and noise increases, and no significant
relationship between the impedance between patency electrodes and the number of open holes was found.

The flow sensor was tested with the datalogger electronics, and both the minimum rate of
change and impedance dip methods (see Chapter 2) were evaluated (Fig. 5.18). An R-type sensor
with electrodes placed 0.5 mm away from the heater was tested at different flow rates using both
the LCR meter and the datalogger electronics, with impedance measured at 10 kHz and a 3.3 V
square wave being used to activate the heater. Both the rate of change and the difference in
impedance followed expected trends relative to flow rate, though the slope was noisier than has
been previously measured using constant current injections. Unfortunately, there was significant
noise in the datalogger electronics compared to the LCR meter, and slope could not accurately be
related to impedance using the datalogger (Fig. 5.19). Impedance dip showed better sensitivity
both when the LCR meter and the datalogger box were used for measurement, but noise was still
significant in the datalogger. Therefore, the datalogger was programmed to take five flow
measurements at 30 second intervals, which were averaged to form each data point.
- 179 -

Figure 5.18 Raw impedance data from the datalogger electronics (blue) and a 10-point moving average
(black). Voltage was delivered to the heater between 10 and 20 seconds. Two methods for transducing flow
were tested with the electronics; the first, described in chapter 2, involves measuring the slope of impedance
as the heater is activated, while the second involves finding the difference between the heated average and
the baseline average.

Figure 5.19 (Left) Measurements using the LCR meter revealed a direct relationship between flow rate and
the rate of change of impedance, but this was not reflected in measurements by the datalogger electronics.
(Right) A relationship between flow rate and the impedance difference before and after heater activation
was seen when both the LCR meter and the datalogger box were used for measurement.

For previous benchtop studies of the pressure sensor, a pre-wetting step was performed to
ensure that the fluidic chamber was fully filled. This involved dropping isopropyl alcohol (IPA)
onto the fluidic ports, which filled the microchannel and temporarily made the Parylene C
hydrophilic. After filling with IPA, the sensor was immersed in PBS, where after 5-10 minutes,
the IPA diffused out and the fluidic chamber filled with PBS. However, sensors cannot be pre-
filled with IPA or any other fluid in the clinic due to sterilization requirements. To test the potential
- 180 -
time delay when used in vivo, a sensor in a luer lock module was submerged in PBS and attached
to the datalogger box, which injected current and measured impedance every thirty minutes.
Impedance between the measurement electrodes showed that the main fluidic channel filled after
13 hours and 49 minutes (Fig. 5.20). However, even after the channel was filled no bubble
nucleation events were observed. After 24 hours of soaking, electrochemical impedance
spectroscopy was performed on the nucleation electrodes without removing the sensor from
solution, revealing that although the main chamber was filled the nucleation core was still dry.

Figure 5.20 (Left) When immersed in PBS, the fluidic chamber of the pressure sensor filled between 13:05
and 13:49 after immersion. However, even after the sensor was filled, no bubble nucleation events were
observed. (Right) After 24 hours of soaking, EIS was performed on the nucleation core electrodes. The EIS
results were purely capacitive, indicating that fluid had not filled the nucleation core.

5.3.3 Clinical issues
Transitioning from the benchtop to the clinic presented unexpected challenges. Several
components of the system which appeared to be robust during benchtop testing broke or would not
operate reliably in the clinic. The first obstacle discovered was that the connection between the
datalogger electronics box and the luer lock-packaged sensors was not adequate for clinical use.
Originally, the sensor’s electrical connections terminated with the contact pads on the FFC, which
was then directly inserted into a ZIF connector attached to the cable coming from the datalogger.
This was chosen so that all parts of the luer lock module would be MRI compatible, and so the
patient could be separated from the electronics box quickly if an MRI was necessary. However,
this connection was weak and prone to slipping, and the ZIF-FFC connection was directional and
confusing to connect, resulting in several devices being connected backwards. Incorrect
connection resulted in loss of data from several devices used in patients. To create a more robust
connection, the ZIF-FFC connection was replaced with a Hirose connector (LX40-12P, Hirose
Electric Co, Ltd., Osaka, Japan), which is both robust to jerking or slipping and unidirectional.
The downside to the Hirose connectors is that they have not been tested for MRI compatibility, so
patients involved with the study cannot undergo MRI imaging while the sensors are attached.
- 181 -
Another issue involved confusion between R and Z type devices. The different orientation
of electrodes between these two devices makes two different connection schemes necessary; the
preliminary solution for benchtop testing was to make different datalogger cables for each device
design. For clinical testing, tried several different methods to differentiate R and Z type cables
were attempted. These included labeling and color-coding both devices and cables, but our efforts
were largely unsuccessful and some data was lost because of mismatched devices. Eventually,
physicians were only supplied with one design variant at a time, to avoid mix-up. There were also
issues in record-keeping and organization at CHLA due to personnel changes. During one of these
transitions, the storage container for the devices was lost for several months. Luckily, all devices
were eventually recovered, and now multiple people on the clinical side of the project are aware
of where devices are stored. There was also an unexpected issue where the datalogger box was not
turned on at the same time the sensors were attached to the EVD, but was instead activated hours
later. To save battery, the datalogger does not record the time during storage or before activation,
so the difference in time between EVD insertion and box activation prevented correlation between
the measurements using the multi-sensor module and the standard of care measurements taken by
the hospital. Now, the exact time the electronics box is activated is recorded on the back of the box
itself.
On the manufacturing side, there were issues with releasing the fluidic channels on the
pressure sensors. Normally, channels can be released by soaking in room-temperature acetone, but
after switching to a DRIE process for etching release became more difficult, taking days to perform
even an incomplete release. Furthermore, attempts at anneal devices with the photoresist still in
the channel to prevent channel collapse resulted in hardened PR, such that even several days of
soaking in hot acetone would not open the channels. Lowering the DRIE power and released
devices before annealing allowed channel release after about 24 hours of acetone soaking, though
efforts are still underway to improve microchannel release.

5.3.4 Clinical results
Ten total patients were enrolled at CHLA between November 2015 and November 2016.
Data was only recovered from eight patients due to misaligned connectors. Median device
operation time was 17.32 ± 37 hours (median ± SD). Due to problems with filling the fluidic
channels on the pressure sensors, only the flow sensors and impedance measurement between the
test electrodes were active on these devices. Since using an external counter electrode for patency
measurement was not possible due to the non-invasive nature of the study, the datalogger was
programmed to instead record the average impedance over 30 seconds between the two patency
electrodes. By using the impedance between the former patency electrodes as a “test impedance”,
sensor integrity, biofouling, or the presence of bubbles passing through the EVD could be
recorded.
Test electrode measurements (Figs. 5.21 – 5.23) showed that sensors filled with CSF upon
activation and electrode impedance stayed stable for around 12 hours of use. In one patient, there
was an unexplained region of almost zero impedance, which may have corresponded to a short
- 182 -
due to water on a connector (Fig. 5.22). In another patient, there was a region in which the
impedance rose by approximately 200 Ω for 20 hours before returning to the previous baseline
(Fig. 5.23). This may have occurred due to a bubble or debris partially occluding the electrodes.
In three patients, after 12 hours the impedance began to rise quickly, which most likely indicates
failure due to delamination. In one case, however, the electrodes saw very little drift even over 120
hours of use. After use in the patient devices were photographed, which appeared to show
significant delamination on all electrode surfaces (Fig. 5.24).

Figure 5.21 The impedance between the two test electrodes in four patients. For three of these patients,
impedance began to dramatically rise between 10-12 hours after device activation.

Figure 5.22 The impedance data from patient 5 showed an electrical short circuit during ~7 hours of testing.
This may have been caused by water or CSF shorting a connector.
- 183 -

Figure 5.23 Patient 3 had a device active for the longest time (123 hours). Test electrode impedance was
stable for the first 120 hours except for a region of higher impedance between 25 and 40 hours. This could
have been caused by a bubble partially occluding the sensor or a piece of tissue sticking to the surface.

Figure 5.24 Photograph of electrodes on devices recovered from (A) patient 5 and (B) patient 6, showing
severe delamination of the exposed platinum from the Parylene C substrate.

Data from flow sensors also indicated rapid delamination. Data from the first set of flow
electrodes in Patient 5 is shown in Figure 5.25. The average impedance dip and initial slope across
five trials, each spaced 30 seconds apart, was calculated from the raw recorded data using a custom
Python program. After the first flow measurement, rate of change dropped to zero, where it stayed
for approximately the first 8 hours of in vivo use. After this, minimum rate of change began to drop
again, and the variation between measurements increased. For all of these tests the change in
impedance when the heater was activated was positive, indicating that heater activation increased
impedance instead of causing the expected decrease in impedance.
Partial Occlusion
- 184 -

Figure 5.25 (Left) The minimum rate of change of impedance versus time, from the first pair of flow
electrodes in Patient 5. Minimum rate of change stays at zero until 8 hours after device activation, at which
point it begins to trend downward, indicating progressive delamination. (Right) The difference between
heated and non-heated impedance of the first pair of flow electrodes is positive, indicating that impedance
rises when the heater is activated. This suggests that some delamination has occurred before the first flow
measurement, though it becomes much worse after the 12 hour mark.
Looking at raw impedance data from a single flow measurement revealed that impedance
does rise across the board during heater activation, and large spikes in impedance were seen both
at the start and at the end of heating (Fig. 5.26). This signal is characteristic of delaminated
electrodes (see chapter 2), and was most likely caused by current leakage between the heater and
electrodes through the delaminated Parylene. This signal could not be used to transduce flow rate.

Figure 5.26 The impedance data from a single heater activation / flow rate measurement attempt reveals
that impedance jumps sharply upon heater activation, with large spikes when the heater is turned on or off.
This could indicate capacitive coupling through the Parylene due to delamination at the heater.

- 185 -

Figure 5.27 SEM images of a flow electrode from a device which has not been tested on the benchtop or
in vivo. The platinum surface is smooth and free from debris.

Figure 5.28 The surface of a flow sensor electrode from patient 9, active for 12 hours in vivo. Delamination
and cracking of the platinum surface is clearly visible, as well as contamination by what could be salt debris
or bacteria. Higher-magnification images reveal spindle-like structures on the platinum surface, which
could indicate bacterial or fungal growth adhering to the electrode surface.

Figure 5.29 Images of another flow electrode from patient 9 reveal omnipresent delamination and minor
cracking, as well as scum covering a large portion of the electrode surface.
- 186 -
To further investigate device failure after in vivo use, the sensor from patient 9 was
removed from its luer lock module and imaged using a scanning electron microscope (SEM). The
device had been active in the patient for 11 hours and 59 minutes. A newly fabricated device which
had not been used either for benchtop testing or in the clinic was also imaged to provide a control
(Fig. 5.27). Images of electrodes on patient 9’s device revealed that cracking and delamination of
the platinum from the Parylene substrate was present on all electrodes. Possible biofouling was
observed covering electrodes. On one flow electrode, large round protrusions which could
represent salt crystals or bacterial colonies were present (Fig. 5.28). In addition, spindle-like
structures were visible under high magnification which could indicate fungal or bacterial growth
covering the electrode surface. On another electrode, a biofilm was visible which partially covered
the electrode surface (Fig. 5.29).

5.3.5 Conclusion to clinical study round 1
After overcoming many procedural issues involved in operating a clinical study, several
multi-sensor modules were deployed in patients. After recovering and analyzing data from these
sensors, it became clear that devices failed rapidly due to delamination between the platinum
electrodes and the Parylene substrate. At this point the study was paused until device integrity in
the clinic could be guaranteed.

Clinical Study Round 2
5.4.1 Sensor fabrication issues
Delamination between electrode surfaces and the underlying Parylene substrate was a
major failure mode during the first round of the clinical study. It was suspected that delamination
was caused, or at least accelerated, by the rough metal deposition and liftoff process. The first-
generation devices used platinum which was sputter-deposited by LGA Thin Films, Inc. (see
Chapter 2). The liftoff process for this metal required soaking in hot acetone for several hours and
manually scrubbing the wafer with cleanroom swabs, a process which caused visible scratching of
the Parylene surface and which most likely contributed to the rapid failure of our devices. After
switching to platinum deposited at UCLA using a low-temperature electron-beam deposition
process, liftoff became easier, taking only 10-20 minutes of soaking in room-temperature acetone
with little to no scrubbing. After annealing, these sensors appeared to last longer than two weeks
in body-temperature PBS without delaminating.
Another possible accelerant of delamination in first-generation clinical devices was the
small DC bias that was put on electrodes during impedance measurement. This voltage was small
(~0.35 V) but could have resulted in electrostatic forces between the Parylene and platinum or in
faradaic reactions that might weaken adhesion. The elimination of DC bias on sensors provided
yet another motivation for redesigning the datalogger electronics.

- 187 -
5.4.2 Second-generation electronics design
A new electronics platform was designed and built to increase resolution and decrease
noise in impedance measurements. This new datalogger utilized an AD5933 network analyzer IC
(Analog Devices). This IC can record the magnitude and phase of impedance at frequencies
between 100 Hz and 100 kHz and output the result digitally through an I
2
C serial bus. The AD5933
boasts a 12-bit impedance resolution and is rated to measure impedance magnitudes between 1 kΩ
and 10 MΩ. It contains an internal clock and amplifier which can be set using an external resistor.
The AD5933 has two ports to interface with the device under testing, an output port (Vout)
which outputs a sinusoidal voltage signal between 0.1V and 2V and an input port (Vin) connected
to internal ADC and discrete Fourier transform modules. For implementation in the clinical
datalogger, a series capacitor, voltage divider, and buffer amplifier were added to create a virtual
ground at 1.65 V. To interface with multiple impedimetric sensors, a digitally-controlled 2:8
analog multiplexer IC (CD4052B, Texas Instruments) was used. This IC was controlled by digital
outputs from the microcontroller and could connect pairs of electrodes on the device to the network
analyzer’s input and output ports. In front of the Vin port on the network analyzer another op amp
was placed, both for signal amplification and to buffer the signal going into the device. Both the
input and output amps were co-located on the same IC (AD8608, Analog Devices). To read data
from the network analyzer and save it to an SD card, the Teensy 3.1 microcontroller board (PRJC
Inc.) was again used. Programming and software validation of the second-generation
microcontroller board was performed by other members of the Biomedical Microsystems Lab
(Trevor Hudson).

Figure 5.30 Block diagram of the second-generation impedance measurement system, showing the
AD5933 network analyzer IC connected to the Teensy 3.1 microcontroller via an I
2
C port and to electrodes
on the impedimetric sensors through a multiplexer and an analog front-end.
The second-generation electronics were housed on a 3-layer PCB designed to fit in the
same plastic case used for the first-generation electronics. PCBs were designed in Eagle Pro 6 and
manufactured by Gold Phoenix PCB, Inc. To account for the higher power consumption of the
network analyzer IC, a larger 2000mAh lithium ion battery was used (PRT-13855, Sparkfun
Electronics). The battery was fed through a voltage regulator integrated onto the Teensy dev board
- 188 -
to give a stable 3.3V output, which was used to power all ICs on the board. To enable flow sensor
operation, a digital 3.3V output was tied directly to the flow sensor’s heater. Using a voltage-
controlled heating instead of current-controlled eliminated some of the effects of heater self-
cooling under high flow rates and simplified the electronic design. With a nominal heater
resistance of 600 Ω, a 3.3V pulse to the heater was expected to generate around 4°C of overheat
temperature. Due to failures of our pressure sensors to fill under clinical conditions a current
regulator for bubble electrolysis was not included on the board, though a spot for it was designed
in. However, recent tests suggest that voltage-controlled bubble electrolysis is more stable anyway,
so future versions may include a digitally-controlled voltage regulator tied to the pressure sensor’s
nucleation core electrodes.

Figure 5.31 Picture of the second-generation datalogger board, with the network analyzer and multiplexer
chips highlighted. The second-generation electronics were layed out on a 3-layer PCB, which was designed
to fit into the same white plastic package used for the first-generation electronics.
For the first benchtop tests of sensors using the second-generation electronics, power was
delivered to the electronics by directly connecting the Teensy board to wall power via a USB
power converter (microUSB phone charger, Belkin International, Inc.). Using a wall plug for
power allowed testing of sensor drift and long-term operation without worrying about battery life.
However, benchtop tests revealed that switching noise from wall-mounted power converters added
a significant amount of noise to impedance measurements, making flow sensing impossible and
decreasing the resolution of all other impedance measurements (Fig. 5.32). Operating the board
using the lithium-ion battery solved this problem and reduced noise to a level which allowed for
highly accurate flow and impedance measurements.

- 189 -

Figure 5.32 Difference in baseline noise when measuring the impedance of a resistor using (top) wall
power, delivered via a store-bought USB power converter connected directly to the Teensy board, and
(bottom) the lithium-ion battery to power the electronics board. Using wall power adds a significant amount
of noise to measurements, most likely due to switching noise within the power converter. Figure credit:
Trevor Hudson, Biomedical Microsystems Lab

5.4.3 Benchtop testing
A 14-day benchtop test using human CSF was performed to validate the second-generation
electronics performance and to verify that the issues present for the first clinical study had been
solved. Sensors fabricated using electron-beam deposited platinum from UCLA were vacuum
annealed at 200°C for 48 hours, and packaged in luer lock spacers using EpoTek 353NDT
biocompatible epoxy. Three devices connected to datalogger boxes were placed in series, and a
Watson-Marlow 120U peristaltic pump was used to flow human CSF at 14 µL/min, which
approximates the flow rates expected in clinical EVDs. A fourth device was tested in PBS and was
placed parallel on the same pump. Measurements were recorded once per hour. Each measurement
consisted of flow sensor activation using a 10s, 3.3V pulse to the heater while measuring 50 kHz
impedance across electrodes 1 mm and 3 mm away, single-frequency (50 kHz) measurement
across the pressure sensor electrodes to determine channel filling, and electrochemical impedance
spectroscopy between 100 Hz and 100 kHz across the test electrodes. All tests occurred at room
temperature, and datalogger boxes were powered using wall power. After 12 days of measurement
data was read off of the µSD cards.
- 190 -
Data from the control device in PBS revealed that minimal drift occurred after 12 days of
measurement (Fig. 5.33). In particular, the EIS across the test electrodes did not show the
formation of the additional phase peak which usually indicates the beginning of delamination. EIS
across test electrodes in human CSF also showed minimal drift over the tested time period (Fig.
5.34). SEM images did not reveal the wrinkling and cracking seen in the previous clinically-used
sensors, though some biofouling was observed (Fig. 5.35).

Figure 5.33 EIS between 100 Hz and 100 kHz across test electrodes on a device tested in PBS for 12 days.
Despite some drift in impedance at lower frequencies, there is no significant change in magnitude or phase
of impedance at the target measurement frequency (50 kHz).

Figure 5.34 EIS across test electrodes tested in human CSF also revealed little to no impedance drift in the
frequencies of interest.
n=5, mean±SD
PBS
n=5, mean±SD
PBS
n=5, mean±SD
Human CSF
n=5, mean±SD
Human CSF
- 191 -

Figure 5.35 Flow electrode from device 2 under SEM. The electrode, which was tested in human CSF for
14 days, shows no wrinkling or delamination, but cells can be observed adhered to the electrode surface.
The baseline impedance of flow sensing electrodes also stayed remarkably consistent over
the course of the study (Fig. 5.36). Of the devices tested in human CSF, two showed baseline drift
over the first two days and then stable impedance magnitudes between 2 and 14 days, while one
device showed no significant drift over the course of the study. Unfortunately, this study was
conducted using wall-powered electronics, so impedance measurements were too noisy for proper
flow sensor operation.

Figure 5.36 The baseline impedance magnitude at 50 kHz of flow sensing electrodes on three devices tested
in human CSF for >12 days. Two devices showed some baseline drift within the first two days, but all three
devices showed stable baseline impedances between 2 and 14 days of testing.
n=5, mean±SD
Human CSF
- 192 -
Results from the flow sensor tested in PBS also revealed one of the strengths of the new
electronics system. Due to worries that activating the flow sensor’s heater when there is no liquid
on the channel, the board was programmed to evaluate the baseline impedance of the flow sensors
before heater activation and abort measurement if the impedance was too high. While testing the
PBS device, three bubbles entered the channel and became stuck on the sensor. Each time, the
program correctly identified the bubbles and did not deliver voltage to the heater (Fig. 5.37).

Figure 5.37 During benchtop testing of a sensor/board in PBS, three bubbles entered the device and adhered
to the Parylene sensor. This is visible as a jump in impedance magnitude and a drop in the phase on the
flow sensor electrodes. Each time a bubble was detected, the electronics correctly identified the bubble and
did not activate the flow sensor’s heater.
Unfortunately, measurement across the pressure sensor electrodes showed that the pressure
channel did not fill in PBS or human CSF, even after 14 days of soaking. To ensure that our
electronics could properly measure channel filling, a box was connected to a device which was
placed in a PBS-filled acrylic jig and pressurized to 400 mmHg. The impedance across the
measurement electrodes was measured by the box while time-stamped microscope images were
acquired every 30 seconds. A decrease in impedance magnitude was clearly seen when the jig was
filled with PBS, and a second drop in impedance was observed when the air pocket in the pressure
sensor “burst” and a fluidic connection was made between the measurement electrodes through
the channel (Fig. 5.38). Complete channel filling occurred after approximately 1 hour of
pressurization. Since pressurization at 400 mmHg is impossible in a clinical setting, hand
pressurization using a syringe filled with PBS was tested. A new device was placed in the jig and
pressure was delivered by hand-squeezing a syringe while monitoring air pocket size with the
microscope. The air pocket noticeably shrank due to the pressure delivered by the syringe, but did
not dissolve in a reasonable time, and pressurization could not be maintained for more than 5
minutes due to hand fatigue.

n=5, mean±SD
Human CSF
- 193 -

Figure 5.38 The impedance across the pressure measurement electrodes was monitored by a second-
generation electronics box (wall power) while the sensor was pressurized to 400 mmHg in PBS. (A) When
the jig was first filled with PBS, a sharp drop in impedance resulted, due to PBS contacting the electrodes
on each end of the fluidic port. (B) After almost an hour of pressurization the air pocket membrane inside
the pressure channel “burst”, creating a fluidic pathway between the electrodes and causing a second drop
in impedance magnitude.
Further tests of the second-generation electronics under battery power confirmed that flow
rate could be accurately measured at body temperature (37°C) and at the flow rates expected for
external ventricular drains (Fig. 5.39).

Figure 5.39 Several flow measurements at 37°C recorded by the second-generation electronics board on
battery power, with 30 seconds of cool-down time between each measurement. Without the switching noise
from the wall power converter, the electronics are able to measure the heating signal with high precision
and to accurately transduce flow using either the initial slope or the impedance dip method. Figure credit:
Trevor Hudson, Biomedical Microsystems Lab

- 194 -
5.4.4 Preliminary clinical results
The first device with the new electronics and datalogger was delivered to Children’s
Hospital Los Angeles on February 6, 2018. A patient was enrolled in the study, and the datalogger
box collected data for 88 hours before being deactivated. Data was collected once per hour; each
dataset included the magnitude and phase of impedance between the patency electrodes from 1kHz
to 100kHz, the magnitude of impedance between the flow electrodes at 50kHz during four heat
pulses, and the impedance magnitude at 50kHz between the pressure sensor’s measurement
electrodes. The impedance spectroscopy between the two patency electrodes indicates that fluid
was present between the electrodes upon activation, though there was a decrease in impedance
over the first six hours of use (Fig. 5.40). After stabilizing at 6 hours, the magnitude and phase of
impedance stayed constant for the next 40 hours, indicating that no delamination or changes in
electrochemical properties occurred during this time period (Fig. 5.41). Between 40 and 41 hours
of use, there was a drastic increase in impedance across the frequency spectrum and a decrease in
phase towards -90°, indicating open circuit failure (Fig. 5.42). After recovering the device from
the hospital, it was observed that connection between the Hirose connector and the FFC was
damaged; this damage most likely caused the observed open circuit at 41 hours post-activation.
Future studies will reinforce the affected area with epoxy to prevent this failure mode.

Figure 5.40 The magnitude (left) and phase (right) between the two patency electrodes during the first six
hours of ex vivo use. The impedance magnitude decreased during this time period, stabilizing after 6 hours.

- 195 -

Figure 5.41 The impedance magnitude (left) and phase (right) between the two patency electrodes during
72 hours of recording. Data is shown for every 12 hours. Impedance magnitude and phase was stable for
the first 40 hours.

Figure 5.42 The impedance between the patency electrodes went from stable to open circuit between hours
40 and 41. This was most likely caused by damage to the Hirose-FFC connection.

The impedance between the flow sensor electrodes did not indicate open circuit failure
until 89 hours post-activation. Before this time the impedance exhibited relative stability, with a
drift rate around -0.1%/hr (Fig. 5.43). Flow sensor measurements were acquired with high
sensitivity using both the impedance dip and initial slope methods; for both of these methods, no
drift was observed until approximately 72 hours of use, at which time the signal went to zero (Fig.
5.44). Since delamination is a progressive process, this loss of signal is probably not due to
delamination. It may be due to damage at the heater’s electrical connections.
- 196 -

Figure 5.43 The baseline impedance of the flow sensor electrodes was relatively stable for the first 88 hours
of use. Within this time period, the impedance magnitude drifted by an average of -0.1%/hr.

Figure 5.44 Both the initial slope and the impedance dip signals were stable for the first 72 hours, then
jumped to zero.

Due to ongoing development of the microbubble pressure sensor, pressure measurements
were not recorded for this patient. However, the impedance magnitude between the sensor’s
measurement electrodes was monitored to evaluate microchannel filling. Impedance data
indicates that the channel filled after 5 hours of soaking (Fig. 5.45).
- 197 -

Figure 5.45 The impedance magnitude across the pressure sensor’s measurement electrodes. This data
indicates that the sensor’s microchannel filled with fluid after 5 hours of use.

Discussion and Future Work
External ventricular drains provide an excellent platform for evaluating fluidic sensors for
hydrocephalus shunts. EVDs use hardware similar to that of implanted shunts, drain real human
CSF from patients, and experience similar pressures, flow rates, and vasomotoric waves as those
found in shunts. Furthermore, integrating sensors with EVDs does not require invasive surgery or
patient contact, and studies using EVDs can be performed with minimal risk to the patients.
A multi-sensor module composed of three fluidic sensors on a Parylene C substrate was
tested in pediatric patients with EVDs at Children’s Hospital Los Angeles. The sensors were
packaged into luer lock modules, which integrate in-line with the EVD drainage channel. Using
sensors in the clinic required the development of a portable datalogger system, which could power
the sensors and record data for the duration of patient treatment. Such a system was developed,
and was shown on benchtop tests to record sensor data for up to 14 days using an integrated
lithium-ion battery. The sensors and portable datalogger was tested in ten patients; due to
procedural problems, data was only collected from eight. Unfortunately, patient data revealed that
the datalogger’s impedance measurement electronics were not sensitive enough to acquire clinical
flow sensor data. Even worse, sensors underwent catastrophic failure after only 12 hours of clinical
use due to delamination between the platinum electrodes and the Parylene C substrate. The study
was paused after ten patients to address these problems.
Several changes to both the sensors and the datalogger electronics were implemented
during the one year gap in patient enrollment. To improve sensor integrity, electron-beam metal
deposition was utilized to improve fabrication and increase platinum-Parylene C adhesion. In
addition, the datalogger electronics were completely redesigned to improve impedance
- 198 -
measurement resolution as well as allow self-diagnosis and compensation for sensor drift. The
second-generation electronics were able to record flow data with high sensitivity and to measure
impedance over a wide frequency range, though battery life was limited to 4-5 days. Benchtop
testing using the second-generation electronics and sensors with electron-beam deposition showed
two weeks of continuous data collection in human CSF, and no sensor failure or drift was observed.
Patient enrollment was re-opened in 2018, and data was collected from one EVD patient
over a period of 88 hours. Despite some issues with mechanical failure of a connector, impedance
data over this period shows no drift either in impedance spectroscopy data or in flow sensor data.
Bulk fluid drainage data was collected from this patient, and the data is being analyzed to validate
flow measurements. Pressure was not measured for this patient, but future devices will include the
microbubble pressure sensor and will evaluate measurements against a commercial pressure sensor
integrated with the EVD. Patient enrollment is ongoing; future clinical enrollment and data
analysis will be performed by Trevor Hudson, who handled the first new patient’s enrollment and
data collection. The goal of this study is to enroll 10-20 pediatric patients at CHLA, with an option
to enroll additional adult patients at LA County Hospital, and quantify the correlation between
sensor measurements and standard of care fluid collection data. Validation of these sensors under
real CSF drainage conditions will inform implantable electronics development and provide a huge
step forward towards deploying sensors in implanted hydrocephalus shunts.
- 199 -
References
1. A. Baldwin, L. Yu, M. Pratt, K. Scholten, and E. Meng, "Passive, wireless transduction of
electrochemical impedance across thin-film microfabricated coils using reflected
impedance," Biomedical Microdevices, vol. 19, p. 87, 2017.
2. C. A. Gutierrez and E. Meng, "Impedance-based force transduction within fluid-filled
parylene microstructures," Journal of Microelectromechanical Systems, vol. 20, pp. 1098-
1108, 2011.
3. B. Kim, C. Gutierrez, and E. Meng, "IN VITRO CHARACTERIZATION OF A PROBE-
MOUNTED PARYLENE-BASED PRESSURE SENSOR ARRAY FOR
INTRACORTICAL APPLICATIONS," Sensors, vol. 2, p. 7.
4. B. Kim, J. Kuo, S. Hara, C. Lee, L. Yu, C. Gutierrez, et al., "3D Parylene sheath neural
probe for chronic recordings," Journal of neural engineering, vol. 10, p. 045002, 2013.
5. B. J. Kim, C. A. Gutierrez, and E. Meng, "Parylene-based electrochemical-MEMS force
sensor for studies of intracortical probe insertion mechanics," Journal of
Microelectromechanical Systems, vol. 24, pp. 1534-1544, 2015.
6. B. J. Kim and E. Meng, "Micromachining of Parylene C for bioMEMS," Polymers for
Advanced Technologies, 2015.
7. B. J. Kim and E. Meng, "Review of polymer MEMS micromachining," Journal of
Micromechanics and Microengineering, vol. 26, p. 013001, 2015.
8. P.-Y. Li, T. K. Givrad, D. P. Holschneider, J.-M. I. Maarek, and E. Meng, "A parylene
MEMS electrothermal valve," Journal of Microelectromechanical Systems, vol. 18, pp.
1184-1197, 2009.
9. W. Li, D. C. Rodger, E. Meng, J. D. Weiland, M. S. Humayun, and Y.-C. Tai, "Flexible
parylene packaged intraocular coil for retinal prostheses," in Microtechnologies in
Medicine and Biology, 2006 International Conference on, 2006, pp. 105-108.
10. E. Meng, P.-Y. Li, and Y.-C. Tai, "A biocompatible Parylene thermal flow sensing array,"
Sensors and Actuators A: Physical, vol. 144, pp. 18-28, 5/28/ 2008.
11. E. Meng, P.-Y. Li, and Y.-C. Tai, "Plasma removal of Parylene C," Journal of
Micromechanics and Microengineering, vol. 18, p. 045004, 2008.
12. E. Meng and Y.-C. Tai, "A Parylene MEMS Flow Sensing Array," presented at the
Transducers, 2003.
13. A. Baldwin, T. Hudson, and E. Meng, "A calorimetric flow sensor for ultra-low flow
applications using electrochemical impedance," presented at the 31st IEEE Conference on
Micro Electro Mechanical Systems (MEMS), Belfast, North Ireland, 2018.
14. A. Baldwin, L. Yu, and E. Meng, "An electrochemical-based thermal flow sensor," in
Micro Electro Mechanical Systems (MEMS), 2016 IEEE 29th International Conference
on, 2016, pp. 341-344.
15. A. Baldwin, L. Yu, and E. Meng, "An Electrochemical Impedance-Based Thermal Flow
Sensor for Physiological Fluids," Journal of Microelectromechanical Systems, vol. 25, pp.
1015-1024, 2016.
16. B. J. Kim, W. Jin, A. Baldwin, L. Yu, E. Christian, M. D. Krieger, et al., "Parylene MEMS
patency sensor for assessment of hydrocephalus shunt obstruction," Biomedical
microdevices, vol. 18, p. 87, 2016.
17. B. J. Kim, W. Jin, L. Yu, and E. Meng, "MEMS electrochemical patency sensor for
detection of hydrocephalus shunt obstruction," in Micro Electro Mechanical Systems
(MEMS), 2015 28th IEEE International Conference on, 2015, pp. 662-665.
- 200 -
18. L. Yu, C. A. Gutierrez, and E. Meng, "An Electrochemical Microbubble-Based MEMS
Pressure Sensor," Microelectromechanical Systems, Journal of, vol. 25, pp. 144-152, 2016.
19. L. Yu and E. Meng, "A dual mode microbubble pressure and flow sensor," in Micro Electro
Mechanical Systems (MEMS), 2016 IEEE 29th International Conference on, 2016, pp.
325-328.
20. E. Meng, X. Zhang, and W. Benard, "Additive processes for polymeric materials," in
MEMS Materials and Processes Handbook, ed: Springer, 2011, pp. 193-271.
21. R. P. Von Metzen and T. Stieglitz, "The effects of annealing on mechanical, chemical, and
physical properties and structural stability of Parylene C," Biomedical microdevices, vol.
15, pp. 727-735, 2013.
22. N. K. Z. Roitberg, MS Alp, T. Hersonskey, FT Charbel, JI Ausman, B, "Bedside external
ventricular drain placement for the treatment of acute hydrocephalus," British journal of
neurosurgery, vol. 15, pp. 324-327, 2001.
23. G. Gambardella, C. Zaccone, E. Cardia, and F. Tomasello, "Intracranial pressure
monitoring in children: comparison of external ventricular device with the fiberoptic
system," Child's Nervous System, vol. 9, pp. 470-473, 1993.
24. S. Taylor, L. Bayless, M. W. Klein, and J. M. Janzen, "Placement of external ventricular
drains and intracranial pressure monitors by neurointensivists," Neurocritical care, vol. 10,
p. 241, 2009.
25. J. M. Mathew, V. Rajshekhar, and M. J. Chandy, "Shunt surgery in poor grade patients
with tuberculous meningitis and hydrocephalus: effects of response to external ventricular
drainage and other variables on long term outcome," Journal of Neurology, Neurosurgery
& Psychiatry, vol. 65, pp. 115-118, 1998.
26. J. M. Drake, C. Sainte-Rose, M. DaSilva, and J.-F. Hirsch, "Cerebrospinal fluid flow
dynamics in children with external ventricular drains," Neurosurgery, vol. 28, pp. 242-250,
1991.
27. T. Yasuda, T. Tomita, D. G. McLone, and M. Donovan, "Measurement of cerebrospinal
fluid output through external ventricular drainage in one hundred infants and children:
correlation with cerebrospinal fluid production," Pediatric neurosurgery, vol. 36, pp. 22-
28, 2002.
28. P. Raboel, J. Bartek, M. Andresen, B. Bellander, and B. Romner, "Intracranial pressure
monitoring: invasive versus non-invasive methods—a review," Critical care research and
practice, vol. 2012, 2012.
29. S. Rammos, J. Klopfenstein, L. Augsburger, H. Wang, A. Wagenbach, J. Poston, et al.,
"Conversion of external ventricular drains to ventriculoperitoneal shunts after aneurysmal
subarachnoid hemorrhage: effects of site and protein/red blood cell counts on shunt
infection and malfunction," Journal of neurosurgery, vol. 109, pp. 1001-1004, 2008.
30. M. Eisha Christian, "Personal Communication," ed, 2016.
31. J. B. Sommer, C. Gaul, J. Heckmann, B. Neundörfer, and F. J. Erbguth, "Does lumbar
cerebrospinal fluid reflect ventricular cerebrospinal fluid? A prospective study in patients
with external ventricular drainage," European neurology, vol. 47, pp. 224-232, 2002.
32. H. Brydon, R. Hayward, W. Harkness, and R. Bayston, "Does the cerebrospinal fluid
protein concentration increase the risk of shunt complications?," British journal of
neurosurgery, vol. 10, pp. 267-274, 1996.
- 201 -
33. J. Piek and P. Raes, "Pressure-controlled drainage of cerebrospinal fluid: clinical
experience with a new type of ventricular catheter (Ventcontrol MTC) and an integrated
Piezo-resistive sensor at its tip," Neurosurgery, vol. 38, pp. 216-218, 1996.
34. T. Pennell, J. L. Yi, B. A. Kaufman, and S. Krishnamurthy, "Noninvasive measurement of
cerebrospinal fluid flow using an ultrasonic transit time flow sensor: a preliminary study,"
Journal of Neurosurgery: Pediatrics, vol. 17, pp. 270-277, 2016.
35. C. Luer-Lock, "Error-avoidance recommendations for tubing misconnections when using
Luer-tip connectors: a statement by the USP Safe Medication Use Expert Committee,"
2008.
36. C. A. Gutierrez, C. Lee, B. Kim, and E. Meng, "Epoxy-less packaging methods for
electrical contact to parylene-based flat flexible cables," in 2011 16th International Solid-
State Sensors, Actuators and Microsystems Conference, 2011, pp. 2299-2302.
37. G. C. Mendes, T. R. Brandao, and C. L. Silva, "Ethylene oxide sterilization of medical
devices: a review," American journal of infection control, vol. 35, pp. 574-581, 2007.
38. M. N. Bathina, S. Mickelsen, C. Brooks, J. Jaramillo, T. Hepton, and F. M. Kusumoto,
"Safety and efficacy of hydrogen peroxide plasma sterilization for repeated use of
electrophysiology catheters," Journal of the American College of Cardiology, vol. 32, pp.
1384-1388, 1998.
39. P. M. Schneider, "New technologies and trends in sterilization and disinfection," American
journal of infection control, vol. 41, pp. S81-S86, 2013.

- 202 -

HE development of polymer MEMS sensors which can be chronically implanted in the
human body has the potential to significantly improve the treatment of chronic diseases. For
hydrocephalus patients, the ability to record shunt status and remotely access historical
information on a shunt’s hydrodynamics will allow timely, non-invasive diagnosis of shunt failure.
This avoids the expense and patient suffering inherent in the multiple CT scans and shunt taps
currently used to diagnose shunt failure. Collecting data from multiple sensors will also allow
doctors to locate where a shunt has failed and treat the patient accordingly. Furthermore,
integrating multiple sensor modalities onto the same sensor module gives devices a degree of
redundancy, and allows both advanced sensor calibration and continuous shunt monitoring even if
one sensor fails. Future work will implement smart algorithms which automatically predict shunt
failure weeks or months in advance of patient symptoms and alert both patients and doctors.
Development of these algorithms is only possible due to the high time resolution of the sensors
proposed in this thesis, and the collection of multiple physiological parameters, such as flow,
pressure, catheter patency, and fluid temperature.

Figure 6.1 Integrating multiple sensing modalities into a hydrocephalus shunt would enable the detection
and location of multiple shunt failure modes. For example, a patency sensor would be most useful in
identifying proximal catheter occlusion, a pressure sensor could accurately identify valve or distal catheter
occlusion, and a flow sensor would be useful for correcting for overdrainage.
C ONC LUS IO N
T
- 203 -

In chapter 2, a flow sensor was introduced which combined low-temperature operation and
high sensitivity at low flow rates into a flexible, biocompatible package. The sensor operated by
measuring heat transfer from a microfabricated resistive heater to a pair of impedance-
measurement electrodes. Due to the high sensitivity of electrochemical impedance to temperature
changes, this sensor could measure flow-mediated heat transfer using only a 1°C overheat
temperature. The sensor was tested in both physiological saline and human cerebrospinal fluid,
revealing highly sensitive, robust, and chronic operation under conditions found in hydrocephalus
shunts. In addition to flow measurement, this sensor was capable of highly sensitive measurement
of fluid temperature using both the resistive heater (which can pull double duty as a resistant
temperature detector) and the high-frequency impedance between the measurement electrodes.
This flow/temperature sensor can be microfabricated out of platinum and Parylene C, which
renders it flexible, biocompatible, and corrosion-resistant. Work on Parylene C micromachining
also led to the development of micro-patterned strain sensors, discussed in Chapter 4, and flexible
thin-film inductive coils, discussed in Chapter 3.
To integrate sensors into hydrocephalus shunts, wireless electronics are necessary for
power and data transfer. These electronics must record high-frequency electrochemical impedance
from multiple electrode pairs while simultaneously powering active components on the sensors,
and transmit the data out of the body to an external recording unit. Chapter 5 discussed the
development of wired, battery powered electronics which are capable of power and data collection
from a multi-sensor module composed of a flow, pressure, and patency sensor. In the future, the
basic architecture of this unit can be paired with wireless power electronics to develop an
implantable module. These implantable electronics would receive power from an external coil over
an inductive link, and digital data transmission would be accomplished using antenna admittance
modulation. An early prototype of a wireless electronics board has been fabricated, and wireless
power transfer has been confirmed. Future work will develop the external electronics for on-
demand powering and digital data reception, as well as a secure database for data storage which
can be accessed remotely by a physician.

Figure 6.2 A block diagram of a proposed implantable electronics system, which could record data from
multiple sensors and output it digitally via antenna admittance modulation.
- 204 -

Figure 6.3 A proposed power and data telemetry system for the implantable electronics. Power would be
delivered via an inductive coupled coil, and back-telemetry would be achieved by selectively de-tuning the
power coil via a transistor, producing a phase-shift keyed signal in the coupled impedance response.

Figure 6.4 A prototype of the implantable electronics was implemented using a PCB connected to a
secondary coil composed of 34 gauge copper wire wrapped around a 3D printed mock implantable module.
Wireless power delivery via inductive coupling was tested and achieved.
While the sensors discussed in this chapter are fully biocompatible and interact directly
with physiological fluid, implanted electronics would require hermetic packaging to withstand the
corrosive environment in vivo. Chronically implanted electronics for pacemakers and deep brain
stimulators use titanium packages to ensure hermeticity; the hydrocephalus module could use this
approach or pursue a hermetic polymer package. Packaging prototypes can be fabricated using 3D
printing, to allow rapid prototyping. Several prototype designs have already been fabricated and
3D printed for consideration.

Figure 6.5 SolidWorks model of a candidate package for implantation in line with hydrocephalus shunts.
The sensors discussed in Chapter 2, as well as other Parylene C sensors developed in the
Biomedical Microsystems Lab of USC, require implanted digital electronics for power and data
digitization. In Chapter 3, a new paradigm for the development of wireless implantable sensors
was proposed. By directly connecting a thin-film inductive coil to a pair of exposed electrodes, the
high-frequency impedance between the electrodes could be wirelessly transduced. This method,
- 205 -

which was referred to as reflected impedance, was used to sense parameters such as the
conductivity of a solution or the solution’s temperature with high sensitivity. Preliminary work
toward a catheter patency sensor and a blood glucose sensor using reflected impedance was also
presented. The use of reflected impedance allows wireless transduction from implantable sensors
without implanted integrated circuits or discrete electronic components. With future
improvements, this could enable simple, low-cost, low-profile sensors which can operate in the
body indefinitely.
There is still significant work to be done before chronically implantable polymer MEMS
sensors are practical for clinical use. In particular, delamination between successive polymer and
metal layers must be prevented before devices can last for years in the body without drift or failure.
Additional methods must also be developed to clean electrode surfaces, since protein or debris in
the body could obscure electrodes and cause sensor drift. However, the sensors reported in this
thesis show a huge amount of promise for future development since they already avoid the failure
modes which have prevented the chronic implantation of silicon-based sensors. It is this author’s
hope that polymer MEMS sensors will soon help monitor, diagnose, and treat chronic diseases and
alleviate patient suffering.

- 206 -

Appendix A: Fabrication flow and photoresist recipes

Multi-Layered Parylene C Device Process Flow
1. Deposit 10µm of Parylene C onto a 4” silicon carrier wafer
a. Place prime 4” wafer in an oven at 120°C overnight for dehydration
b. Deposit 10-12 µm of Parylene C onto wafer using SCS LabCoter 2 Parylene deposition
system
2. E-beam metal deposition
a. Spin and pattern 2µm layer of AZ5214 image reversal photoresist (see page 217 for
detailed photoresist recipes)
b. O 2 descum metal using reactive ion etcher (RIE): 100 W, 100 mTorr, 60 seconds
c. Place wafers in Temescal e-beam metal deposition system
d. Run e-beam deposition according to Temescal SOP
i. Platinum: deposit 500 Å at a time, with 15 minutes of cooldown between
depositions
ii. Gold: first deposit 200 Å titanium as an adhesion layer, then gold with up to 1000
Å at a time
3. Metal liftoff
a. Soak wafers in acetone until photoresist dissolves and metal lifts off
i. It may be necessary to heat acetone to ~40°C. Be careful of bubbles! If the
acetone boils, bubbles will get between the Parylene and the silicon wafer and
your devices will be ruined
ii. It may also be necessary to “scrub” wafers by gently wiping with a cleanroom
swab. Try to avoid this unless absolutely necessary
iii. Spraying the wafer with acetone from the spray bottle can help remove stubborn
pieces of metal, but watch out because this can damage gold features
b. After metal has lifted off, transfer wafer to a clean acetone bath for 10 minutes, then IPA
bath for 10-15 minutes, then DI water bath for 10 minutes
4. 10µm Parylene C Insulation deposition
a. Descum wafers again (O 2 plasma, 100 W, 100 mTorr, 60 seconds)
b. Deposit 10-12 µm of Parylene C onto wafer using SCS LabCoter 2 Parylene deposition
system
5. Electrode exposure etch
a. Spin and pattern 15µm AZ4620 photoresist
b. Deep Reactive Ion Etch (DRIE) until electrodes are exposed
i. O 2 for etching and C 4F 8 for passivation
ii. Oxford PlasmaLab ICP machine, “Meng” recipe
iii. Etch for 25 loops then rotate wafer ¼ turn
iv. Check under the scope every 25 loops to determine when etch is complete
c. Strip photoresist mask by soaking in acetone, IPA, and DI water for 10 minutes each
6. For 3-layer devices with microchannels:
a. Pattern sacrificial photoresist (AZ4620, however high it needs to be)
b. Descum wafer, then deposit 4µm of Parylene C
c. Spin another layer of AZ4620 and pattern to define etch mask for fluidic ports
- 207 -

d. Use RIE etch in O 2 plasma (300 W, 300 mTorr) to expose fluidic ports
e. Strip photoresist before next mask
7. Cutout etch
a. Spin and pattern 15µm AZ4620 photoresist
b. DRIE etch until device outlines are almost completely etched to the level of the silicon
8. Clean wafers by soaking in acetone, IPA, and DI water for 10 minutes each

AZ5214 Image Reversal Photoresist - 2µm on Parylene C
1. O 2 descum 4” Si wafer coated in 10 µm Parylene (100 W @ 100 mTorr for 60s)
2. Spin AZ5214
a. 15s @ 500 rpm spin-on, 45s @ 1800 rpm (Recipe B) - approximately 2 µm thick
b. At least 2 droppers of resist, maybe 3
3. Softbake: 90°C on hotplate for 70s
4. First exposure (through mask): 25 mW/cm
2
* 1.5s = 37.5 mJ/cm
2

5. Image Reversal bake
a. Use hotplate with smallest gradient over surface
b. Hotplate @ 116°C as confirmed by IR gun for 60s
6. Wait 3 minutes for wafer to return to room temperature
7. Second exposure (flood): 25 mW/cm
2
* 40 s = 1000 mJ/cm
2

8. Develop: 1:4 AZ340:H20 (150 mL total) for 23s with agitation
9. DI water bath with rinsing
10. Blow dry and inspect
This recipe works well for large (>30 µm) structures
If small features fail to appear it is a sign that the IR bake is too hot or too long
If dense features lift off, or if unexposed regions appear white/damaged after development, IR bake may
be too cold
The IR bake is very temperature sensitive, changes in temperature of 2°C can change results drastically

AZ4620 Photoresist - 15µm on Parylene C
1. Spin AZ4620
a. 5s @ 500rpm, 45s @ 1200rpm
b. At least three full droppers of resist
2. Softbake: 5min @ 90°C
3. Hydration: let wafer sit in air for 30 minutes
4. Exposure with mask: 550mJ, hard contact
5. Development: AZ340 1:4 dilution with DI water, 90s while agitation
6. Hard Bake:
a. 5 minutes on 90°C hotplate
b. 15 minutes in vacuum oven at 90°C
- 208 -

Appendix B: Device Release and Packaging

Device release
After fabrication, devices will still be connected to their silicon carrier wafer. For most devices, the final
cutout etch will be incomplete, to prevent against devices flying off the wafer during processing. The
following describe the steps to release a device and prepare it for packaging.
1. Using a razor blade, carefully trace the outline of the device to ensure that the device is fully
separated from its neighbors. It is best to do this under a microscope, preferably the Vision scope.
Recommendation: using the corner of one of the square “cartridge-style” razor blades instead of
the triangular “scalpel-style” blades, as the tips of the scalpels have an annoying tendency to
break off.
2. Place the wafer on a flat surface beneath the microscope and cover the device with deionized (DI)
water.
3. Using tweezers and under a microscope, gently peel the device off the wafer. It is usually best to
peel the device a tiny bit at one corner (whichever corner is furthest from critical features), wait
for the DI water to soak between the device and the silicon wafer, and then gently pull the device
off.
4. Clean devices by soaking in acetone for 10 minutes, isopropyl alcohol (IPA) for 10 minutes, and
then DI water for 10 minutes.
5. Store devices in a petri dish with a Chemwipe lining before packaging.

Attaching PEEK to devices
A polyether-ether-ketone (PEEK) backing must be added to device contact pads so they are thick enough
to fit into one of our ZIF connectors. There are two possible methods for attaching PEEK to devices:
Method 1 – PEEK film and cyanoacrylate
1. Obtain a 250 µm thick PEEK film, preferably with one smooth side and one rough side.
2. Under a microscope, place devices face down on a chemwipe surface (Make sure it is face down!
Superglue on contact pads is not fun).
3. Apply a small amount of cyanoacrylate (superglue) to the edge of a contact pad-sized section of
the PEEK film. Apply superglue to the ROUGH side if possible.
4. QUICKLY dab the superglue on an unused part of the chemwipe, then QUICKLY place onto the
backside of the device’s contact pads (make sure not to cover any active parts of the device with
PEEK – the purpose is only to increase the thickness of the contact pad area so that it will fit in a
ZIF connector).
5. Press down on the peek with a finger for three seconds, then flip the device/PEEK sheet over and
wait ~1 minute for glue to dry.
a. While the glue is drying, look at the contact pads under the scope to make sure there is no
superglue on contact pads. It is super easy to apply too much superglue to the PEEK and
have it leak around the device and prevent electrical contact. If there is superglue on
- 209 -

pads, tt is possible to use a q-tip with acetone to remove it, but this is easier said than
done.
6. After the glue is dry, use scissors or a razor blade (I prefer the razor blade, but others might prefer
the tiny packaging scissors) to cut the device out. Cut with care: it is easy to slip and ruin a
device, but there is also very little margin for cutting if the device is to fit in a ZIF connector.
Method 2 – PEEK tape (recommended)
1. Obtain or create 250 µm thick PEEK tape. It may be necessary to combine 100 µm and 150 µm
thick PEEK tape layers create a 250 µm thick film.
2. Place device face down under the scope
3. Cut a small portion of PEEK tape off of the roll, peel back the tape’s backing, and stick the
adhesive portion of the tape onto the back sides of the device’s contact pads.
4. Press down hard for a second or two. Use a fingernail to “roll out” any potential air pockets
between the tape and the device.
5. Turn over and cut out device using the same procedure as in Method 1

Soldering ZIF connectors to FFCs
Electrical connection to devices is achieved using zero insertion force (ZIF) connectors, which are
soldered to flat flexible cables (FFCs). Soldering a ZIF to an FFC is tricky, as this is one of the smallest
soldering jobs most people would be expected to do by hand. Here are some instructions and tips on ZIF-
FFC soldering:
1. It is best to solder ZIFs to FFCs under the Vision scope. The binocular vision will help with
alignment
2. Tape a large glass slide down for soldering. Glass is a good insulator, and using it as a base for
soldering helps things heat up quicker.
3. Set your soldering iron to 620°F (or at least higher than 500°F)
4. Stick the ZIF connector to a piece of tape, where the electrical contacts are sticking out of one
side of the tape. Use that tape to help align the ZIF contacts with the FFC contacts and to make
sure it stays in place during soldering (Fig. 1)
- 210 -

Figure 1. A ZIF connector lined up with the contact pads of an FFC
5. Make sure that the pins on the ZIF connector are actually touching the FFC pads, and not just
sticking in the air. Try moving the ZIF a little further down on the FFC to get better alignment.
6. Cover pins and contact pads with flux (Fig. 2). Adding flux is vital for soldering small
components.

Figure 2. A ZIF-FFC before soldering, covered with Flux.
7. Take the soldering iron and gently touch the tip to the point where the ZIF pin and the FFC
contact touch. The goal is to heat both the ZIF and the FFC pins simultaneously. Don’t hold the
soldering iron there more than 1-2 seconds, as the ZIF’s plastic will eventually start to melt.
8. Gently touch the tip of the solder wire to the bottom of the FFC’s contact pad, below the point
that the soldering iron is touching. If done properly, a small amount of solder will melt and flow
across both the FFC contact and the ZIF pin, electrically connecting the two (Fig. 3).
- 211 -

Figure 3. Good solder joints between a ZIF connector’s pins and FFC contact pads.
9. Once there is a metallic connection between the FFC contact pads and ZIF pins, quickly remove
the soldering iron.
10. Experienced solderers will be able to solder each pin individually, but during practice it is likely
that solder will get everywhere and there will be lots of shorts. That’s ok; excess solder can be
removed using a solder wick, though the process can be difficult:
a. Step 1: apply more flux!
b. Step 2: place the solder wick on top of the areas with excess solder
c. Step 3: place the soldering iron on top of the solder wick, above the excess solder. Don’t
try to poke through the solder wick!
d. Step 4: wait for the solder wick to heat up, the solder below it to melt, and the melted
solder to soak into the wick. This requires patience, as the process can take up to 30
seconds
e. Step 5: Once the solder is done wicking, quickly remove both the solder wick and the
soldering iron AT THE SAME TIME. If the soldering iron is removed first, the solder
wick will cool down and the wick will be soldered to the ZIF connector!
f. Step 6: re-evaluate and touch up. It may be neccesary to re-solder one or more pins, or to
wick more solder away

Insert device into ZIF connector
After soldering a ZIF to an FFC, a device can be placed into the ZIF:
1. Test the ZIF-FFC connection to make sure there are no shorts or open circuits. Use a multimeter
to confirm electrical connection. Make sure to check that each contact is “shorted” (properly
connected) AND that adjacent contacts are open circuit.
2. Open the latch on the ZIF connector
3. Insert the contact pad region of the device (which already has a PEEK backing on them) into the
ZIF connector. Make sure the contacts are facing DOWN! This means that the PEEK side is
towards the latch of the ZIF connector, and the electrical contacts are on the side away from the
latch of the ZIF connector.
4. Close the latch
- 212 -

5. Look under a microscope and try to determine if the contact pads are lined up with the ZIF pins.
It may be possible to spot incorrect alignment looking from the top of the ZIF connector, and if
pads are obviously misaligned then it is easy to remove and re-insert.
6. If everything looks good visually, use the LCR meter to check if the electrodes are properly
connected:
a. Place one drop of saline onto the electrodes of the device. Make sure that no liquid gets
on the ZIF connector! (Liquid on the ZIF will short all pins and give a “false positive”
connection; if this happens, clean the ZIF with DI water and wait for everything to dry
before trying again the next day).
b. Connect electrodes to an LCR meter (Agilent E4980A or equivalent). This may require
inserting the other end of the FFC into another ZIF connector attached to a breakout
board.
c. Check that the impedance on the LCR meter is close to the correct impedance for the
electrodes under consideration. Mostly pay attention to the phase: if the phase is near -
90° then it is probably open circuit, if the phase is near 0 or positive then it is probably
short circuit.
d. For devices that are not electrodes (heater, resistors, etc.) use a multimeter to check the
DC resistance.
7. Once you are confident that your device is making good electrical connections to the ZIF-FFC,
you can proceed to packaging.

Packaging device in luer lock spacers
Most tests described in this thesis used devices packaged in luer lock spacers. Luer lock spacers are small
plastic modules which can be clipped on to many pieces of lab equipment or clinical catheters (including
EVDs). Packaging a device in a luer lock spacer also protects it somewhat from physical damage during
storage. Below are the steps necessary for luer lock packaging:
1. Obtain a plastic luer lock spacer
2. Using an endmill, mill a small slot longways into the side of the luer lock spacer. The slot should
be as thin as possible while still fitting the device.
a. It may be tempted to use a Dremel tool to create this slit. I tried this too, and it ended up
melting part of the plastic. Attempt at your own risk.
3. Place the device into the slit on the luer lock spacer. You will need to find a way to hold the luer
lock spacer and device/FFC in place for a long period of time. I used a custom acrylic jig, which
might still be somewhere in the Meng lab depending on how long it’s been since I wrote this.
a. Make sure the device is oriented the proper direction before applying epoxy.
4. Cover the slit and electrical connections with epoxy:
a. EpoTEK 353-NDT: Mix the main part of the epoxy and the curing agent in a 10:1 ratio
by weight. Mix well, until the mixture is uniform. Using a toothpick, paint epoxy onto the
luer lock module until the slit is sealed shut and the ZIF/FFC contact pads are completely
covered in epoxy. To cure, place in an oven at 50°C for 6 hours, then air dry for 18 more
hours before use.
b. Marine epoxy: Mix the two parts of the marine epoxy in a 1:1 ratio until the mixture is
uniform. Cover the luer lock slit, the ZIF connector, and the FFC contact pads
completely. Air dry for 24-48 hours before use.
- 213 -

Appendix C: Fabrication Masks

Reflected Impedance Test Coils:

- 214 -

- 215 -

Double Metal Layer Reflected Impedance Devices:

- 216 -

- 217 -

- 218 -

Kirigami Strain Sensors and Experimental Devices:

- 219 -

- 220 -

- 221 -

- 222 -

- 223 -

Flow and Temperature Sensors:

- 224 -

- 225 -

- 226 -

- 227 -

Abstract (if available)

Abstract The ability to record physiological signals from inside the body without invasive surgery or dangerous testing was a revolutionary development in medicine that has accelerated our understanding of the human body and the ability to diagnose, manage, and cure diseases. However, many physiological signals cannot be accurately measured using non-invasive imaging techniques. These signals include hormones and biomarkers in blood, the neural signals which make up cognition, reason, and perception, and cerebrospinal fluid dynamics. Even for diseases which can be non-invasively monitored, several barriers exist that limit the usefulness of medical imaging. Imaging studies can only be performed during doctor’s visits, which generally occur once every six months or less. This limits a doctor’s ability to understand and manage chronic conditions. Patients can experience very different physiological responses in a doctor’s office compared to normal activity, a phenomenon colloquially known as “white coat syndrome”. Measurements made in a doctor’s office also preclude disease monitoring during exercise or strenuous activities. To improve medical care and advance our understanding of chronic diseases, techniques are needed to remotely monitor patients outside of the clinic, to acquire data on a disease state during periods of physical activity, to take measurements at time scales of minutes or hours instead of months, and to record biosignals that are not currently detectable using non-invasive medical imaging. This would give us unprecedented access to a patient’s disease state, enabling the detection or prediction of adverse events before the patient has any symptoms and leading to rapid, targeted treatment. It would also advance our scientific understanding of chronic diseases, allow high-precision trialing of therapies, and may lead to cures for diseases which are now considered life-long conditions. Continuous glucose monitoring in diabetes patients acts as early validation of continuous biosignal monitoring outside the clinic

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Strategies for improving mechanical and biochemical interfaces between medical implants and tissue

PDF

Liquid flow rate sensing via solution resistance for in vivo applications

PDF

Magnetic induction-based wireless body area network and its application toward human motion tracking

PDF

A series of longitudinal analyses of patient reported outcomes to further the understanding of care-management of comorbid diabetes and depression in a safety-net healthcare system

Asset Metadata

Creator Baldwin, Alexander Barnes (author)

Core Title Thin-film impedimetric sensors for chronic in vivo use: design and application to hydrocephalus treatment

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 08/05/2018

Defense Date 04/17/2018

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag bioMEMS,coil,Diabetes,electrochemical,flow sensor,glucose sensor,Gold,hydrocephalus,impedance,inductive,kirigami,MEMS,microelectromechanical systems,microfabrication,OAI-PMH Harvest,Parylene C,patency sensor,platinum,platinum-iridium,polymers,pressure sensor,reflected impedance,RF induction,RFID,sensors,thin-film sensor,wireless

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Meng, Ellis (committee chair), Gupta, Malancha (committee member), Shung, Kirk (committee member)

Creator Email abbaldwi@usc.edu,axbaldwin@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-50915

Unique identifier UC11668785

Identifier etd-BaldwinAle-6634.pdf (filename),usctheses-c89-50915 (legacy record id)

Legacy Identifier etd-BaldwinAle-6634.pdf

Dmrecord 50915

Document Type Dissertation

Format application/pdf (imt)

Rights Baldwin, Alexander Barnes

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA