University of Southern California Dissertations and Theses

Parylene-based biomems sensors for multiple physiological systems

(USC Thesis Other)

Parylene-based biomems sensors for multiple physiological systems

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

PARYLENE-BASED BIOMEMS SENSORS FOR MULTIPLE PHYSIOLOGICAL SYSTEMS

by

Xuechun Wang

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)

December 2022

Copyright 2022 Xuechun Wang
ii

To my parents and grandparents

© 2022

Xuechun Wang

All Rights Reserved
iii

A CK NOW L E DGM E NTS
First, I want to thank my parents for their support and encouragement whenever I
struggle. Thank you for all the trust in allowing me to make hard decisions when I was a little
girl. When I sit down and reflect on how you shaped my life and made me who I am today, I
appreciate many things now that I complained about a lot back when I was a kid.
When I struggled with homework problems during primary school, my mom spent much
time explaining where and how to find the solution and then asking me to figure out the answer
myself. I could not understand why she could not just tell me the answer. Now when I look back,
I appreciate her way of educating me to think independently, be self-motivated about learning
new things and solve problems. My dad always taught me to think positively, even in a bad
situation. I complained that he had a “big heart” and did not care about things. Now I understand
what he taught me: always look at things from both the “good” and “bad” sides. When I fail, I
now know the focus is on positively solving the problem instead of sadly being trapped in the
situation and complaining about everything that possibly caused the problem. There are many
more lessons they taught me about life; I cannot say how much I appreciate them now.
Besides my parents, I want to thank my grandparents, especially my grandmother, who
day cared for me for many years when my parents were busy with work. My grandmother is
energetic and always proud of herself, her family, and anything related to her, including me.
Like my parents and grandparents, the mentors, colleagues, and friends I met at school
and work also helped and supported me along the academic and Ph.D. path. Thank you, Prof.
Hong Wang, for the opportunity of allowing me to work in your lab to get early exposure to
iv

research as an undergraduate. The research experience in your lab grew my interest in working
on interdisciplinary research projects with the applications of knowledge from multiple fields.
Thank you, Prof. Gu Zhen, who served as an academic mentor and encouraged me to apply for
Ph.D. programs directly from undergrad when I knew little about graduate school applications.
Thank you, Russell Thomas, Dr. Bruce Pitner, and Dr. Rose Evans-Storms. They provided me the
opportunity to work in a startup company grown from a university lab and taught me the
importance of market research and the power of translational research, specifically how much
the technology developed in a research lab can solve real-life problems.
Translating directly from the undergraduate program to the Ph.D. program was difficult
for me. I want to thank my advisor, Prof. Ellis Meng, for her patience, support, and guidance
throughout my Ph.D. journey. When I joined the lab as a fresh undergraduate graduate, I knew
nothing about MEMS, but she still offered me the opportunity to join the lab and learn new skills
from zero. I enjoyed the projects I have been working on for the past few years. I appreciate her
introducing me to the MEMS and medical device fields. Her supervising and teaching helped me
grow quickly as an independent researcher. Thank you for always being available whenever I
needed help and giving me a lot of freedom and flexibility to figure out the “puzzles” myself.
Besides my PI, others in Meng’s lab have helped me a lot through the past years. Thank
you, Dr. Ahuva Weltman Hirschberg, for taking care of me during the lab rotation. You are a
good teacher who teaches me many new skills, including soldering, making brain probes…
Thank you, Dr. Kee Scholten, who is always available whenever I need help with research and
writing. When I am stuck on research projects and come to you, you always inspire me with new
directions and help to lead to potential solutions. Thank you for being able to help with many
v

last-minute requests and for all the patience. Thank you, Dr. Alex Baldwin, who is always
available to answer questions and is willing to help with writing and poster preparation for
Grodin’s symposium. Thank you, Dr. Jessica Ortigoza, who put much effort into studying
Parylene material for implantable applications and sharing the knowledge with the lab. Thank
you, Dr. Trevor Hudson, who taught me everything about the hydrocephalus flow sensor projects
and is available to answer all my questions when you are busy preparing your thesis defense.
Thank you, Dr. Eugene Yoon, for helping me quickly translate the new bubble sensor project. You
are a good teacher who is very good at explaining things with many details, and you are always
eager to teach me new stuff, including AutoCAD, Gamry, and many others. I am also very
impressed with your technique of making a good latte with beautiful art. Thank you, Chris Larson,
who always likes to step ahead to handle the hard things related to lab maintenance (ex., building
the lab website and setting up the new server) to ensure the lab can operate smoothly. Thank you,
James Yoo, for organizing all the birthday surprises and many lab events. Thank you, Brianna
Thielen, Ping Hu, and Yingyi Gao, for all the help with the projects. Thanks to all the visiting
scholars and undergraduates working with me on varied projects, including but not limited to
Zachary Smith, Gong Chen, Luciana Custer, Janelle Baker, Neha Yadav, Elliot Myong, Alex
Castaneda, Divya Narayanan, Alexis Foroozan, Grace Yakutis, Pablo Tayun-Mazariegos, Nelly
De Leon, Edgardo Castro Ibarra, and Kwayera Burrows.
Many projects I worked on were highly interdisciplinary and involved animal work. I
want to thank everyone who provided support and help for the related works, including but not
limited to Dr. Dong Song, Dr. Huijing Xu, Wenxuan Jiang, Dr. Gordon McComb, Dr. Lynlee
Stevey, Dr. Bradley Ahrens, and Vivian La.
vi

Besides the name I mentioned above, many others have helped me with varied projects,
and I want to express my appreciation to all of you. And, of course, thanks to all the funding
agencies for supporting me through the graduate program and different research projects.
Lastly, I want to thank Aoyang for the unconditional love. For me, you are my senior,
friend, and family, and you are always on my side to accompany and support me whenever I am
happy or sad. Thank you for saying I am doing good and will keep doing good during the
moments I doubted life, work, and myself.

vii

TABLE OF CONTENTS
BACKGROUND………………………………………………………………………...1
PARYLENE-BASED 64-CHANNEL HIPPOCAMPAL PROBE…………………...…………..5
1.2.1 Probe Design and Fabrication …………………………………………………….8
1.2.2 Probe Insertion…………………………………………………………………10
1.2.3 Mechanical Evaluation of the Parylene Probe………………………………..…..13
1.2.4 In Vivo Study with Parylene Probe……………………………………………….21
IMPROVE PROBE LONG-TERM PERFORMANCE………………………………………..31
1.3.1 Debugging the Crosstalk Source……………………………………………….34
1.3.2 Reducing Crosstalk by Surface Treatment…………………………………...…..37
PROBE INSERTION STUDY…………………………….………………………….…..41
HIGH-DENSITY PROBE THROUGH THREE-DIMENSIONAL STACKING………………….50
1.5.1 Insertion Method……….………………………………………………………52
1.5.2 In Vivo Recording with the 3D Probe……….……………………………………66
1.6.1 Probe Design………………………………………………….………………..74
1.6.2 Insertion Method……………………………………………………………….75
SUMMARY…………………………………………………………………………...81
REFERENCES……………………………………………………………………….84
BACKGROUND…………………………………………………………………….....94
FLOW SENSOR FOR BLOOD FLOW DETECTION………...……………………………...98
2.2.1 Design Requirements…………………………………………………………….99
2.2.2 Flow Sensing Fundamentals……………………………………………………...99
CIRCUIT DESIGN……………………………………………………………………103
2.3.1 Input Impedance Buffer…………………………………………………………105
2.3.2 Demodulation Block……………………………………………………............108
2.3.3 Circuit Test with Single Pick-up Electrode…………….......................................113
2.3.4 Circuit Design and Measurement with Three Pick-up Electrodes……………….115
2.4 SUMMARY…………………………………………………………………………116
2.5 REFERENCES………………………………..………………………………………121
3.1 BACKGROUND…………………………………….………………………………..123
3.1.1 CSF Flow Monitoring.……………………………………………………….…124
3.1.2 ICP Monitoring.………..…………………………………………………….…125
3.1.3 Multimodal Sensing for Detecting Hydrocephalus Shunt Patency…………..…126

viii

3.2 FLOW AND TEMPERATURE SENSING………………………………………………126
3.2.1 Materials and Package……………………………………………………….…129
3.2.2 Benchtop Calibration……………………………………………………….…..129
3.2.3 Sensor Test in Swine………………………………………………………..…..134
3.3 PRESSURE SENSING……………………………………………………………...159
3.3.1 Sensor Design……………………...……………………….……………….…..160
3.3.2 Sensor Characterization…………………………………………………….…..163
3.3.3 Bubble Generation and Pressure Sensing………………………………….…..164
3.4 SUMMARY....………....………….………………………………………………...167
3.5 REFERENCES….……………………………………………………………….……169
APPENDICES………………………………………………………………………………….172

ix

LIST OF TABLES

Table 1-1. 1kHz impedance (mean ± SD) of difference devices [48]. © 2020 IEEE……………..25

Table 1-2. Summary of the mechanical study done with brain probes from other literature…....42

Table 1-3. Table showing the dimension and shape of Neuronexus probes versus Parylene
sham probe………………………………………………... ………………………………….....44

Table 1-4. Table showing measured insertion and shear force for probe (A, B, and C) with
different shank shapes (mean ± SD, N=3). A, B, and C devices are the devices shown in Fig.
1-47 from left to right. The probe with asymmetric and symmetric shanks has the same
spacing between shanks………………………………...……………..………………………….46

Table 1-5. Measured insertion and shear force collected with the probe with different shank
thicknesses (mean ± SD, N=3)……………………………………………...…………………....49

Table. 1-6. Table showing the neuron number captured by electrodes aligned on four probe
shanks………………………………………………………..…………………………………...68

Table. 2-1. The maximum achievable gain for different circuit blocks.…………………...…...116

Table 3-1. Diffusion coefficients of gas-liquid mixtures at 25 ℃ [27]. © 2022 IEEE ………...164

Table A-1. Bonding area of Parylene microchannels under varied bonding parameters……......175

Table B-1. DRIE parameters for deposition and etch steps……………………………………..180

Table. C-1. DRIE etch and deposition parameters for each loop …………… …………………..182

x

LIST OF FIGURES

Figure 1-1. Schematic showing the structure of a neuron. The communication between two
neurons (neuron A: blue; neuron B: orange) happens at the axon terminal through the synapse.......1

Figure 1-2. a) Cresyl violet stained coronal rat hippocampus slice showing the location of sub-
regions CA1, CA3, and DG (scale bar = 600 μm). (b) Schematic showing probe design with
electrode group layout to target CA1 and CA3 regions [48].………………………………..……..6

Figure 1-3. (a) Fabricated brain probe array compared to US dime (scale bar = 5 mm). (b)
Schematic showing probe design with electrode layout to target CA1 and CA3 regions. (c)
Schematic showing probe design with electrode layout to target CA1 and DG. (d) As-
fabricated electrodes are 50 μm in diameter, of which 30 μm is exposed after applying
surrounding insulation. The center-to-center distance between the electrodes is 70 μm. Probe
arrays measured 20 μm in thickness and consisted of a Parylene C-platinum-Parylene C
sandwich [48]. © 2020 IEEE……..………………………………………………………………..7

Figure 1-4. Schematic showing (a) probe buckled at the brain surface when Finsertion>Fbuckling;
(b) probe coupling to rigid shuttle and stiffener coating successfully inserted into the brain……..11

Figure 1-5. Fabrication process for the PEG brace. (a) An exploded view of the mold parts is
depicted and followed by (b) the assembled mold and method of PEG brace application. (c)
Parylene C array with PEG brace after releasing from the mold (scale bar = 2 mm) [48]. ©
2020 IEEE………………………………………......................…………………………………13

Figure 1-6. (a) Optical micrographs showing shank array for a sham Parylene device (left)
and a completed Parylene device with electrodes (right). Scale bar = 2 mm. (b) SEM
micrographs showing the tip region of the array (scale bar = 100 μm) and (c) a single probe
tip (scale bar = 10 μm) [48]. © 2020 IEEE ………...……………………………………………14

Figure 1-7. (a) The mechanical testing setup for insertion force measurement. (b) Close-up
view of the clamped probe array over an agarose gel block placed on the load cell. (c) The
mechanical testing setup for buckling force measurement. (d) Close-up view of the clamped
probe array over the Instron 5490 Series built-in load cell [48]. © 2020 IEEE……………………15

Figure 1-8. A representative buckling force curve obtained with a 2.8 mm 8-shank Parylene
C probe and smoothed with adjacent point average. The threshold force is taken with the
setup in Fig. 1-6 (c). The buckling threshold force is highlighted with a blue dashed line.
Inset shows the close-up view of a buckled 8-shank probe [48]. © 2020 IEEE………………...…16

Figure 1-9. A representative raw (grey) force-displacement data and smoothed curve (blue)
with adjacent averages for a 2.8 mm 4-shank Parylene C probe inserted into a block of 0.6%
agarose gel. After the probe contacts the gel surface, the force increases until it reaches a
threshold known as insertion force (labeled as an unfilled black star), which occurs when the
probe initially penetrates the gel. Prior to that point, the probe displaces gel resulting in
xi

dimpling. As the probe advances deeper, the force increases until reaching the shear force
(labeled as a solid blue star), at which point the maximum force is measured. This also
corresponds to the completion of insertion. Then the probe rests in the gel corresponding to
the relaxation phase [48]. © 2020 IEEE………………………………………………….……….17

Figure 1-10. (a) The buckling force thresholds (mean ± SD, N = 3; bars) and insertion force
(mean ± SD, N = 5; blue squares). Buckling force is shown for 5.5 (solid) and 2.8 mm
(hatched) shank lengths for single and multi-shank arrays (2, 4, and 8). (b) The buckling
force as a function of shank number for both 2.8- (grey) and 5.5- mm (black) shank lengths
(mean ± SD, N = 3) compared with the theoretical linear extrapolation calculated from (1)
[48]. © 2020 IEEE……….…………………………………………………………………….…18

Figure 1-11. (a) Braced Parylene arrays (2.8 mm exposed) successfully inserted into the
0.6% agarose across all conditions (1, 2, 4, and 8 shanks; scale bar = 1 mm). (b) A single
shank unbraced Parylene probe (5.5 mm exposed) buckled before penetrating agarose (scale
bar = 2 mm) [48]. © 2020 IEEE………………………………………..…………………………19

Figure 1-12. Characteristic features of the force-displacement curve compared across
different numbers of shanks: (a) insertion force and (b) shear force (mean ± SD, N = 5).
Each plot exhibits a high degree of linearity, as evidenced by the quality of the fit. As the
insertion force = 0.22 * shank number, R
2
=0.99; shear force = 0.37 * shank number, R
2
=
0.99 [48]. © 2020 IEEE……………………………………………………………………..……20

Figure 1-13. Photo of a Parylene probe sandwiched between two 0.5 mm thick PEG blocks……21

Figure 1-14. Histological slices after Parylene array implantation and removal highlighting
the hippocampus. Slices were stained with hematoxylin and eosin. (a) Coronal and (b & c)
transverse slices taken at 2.2 and 2.5 mm from the brain surface showing probe tracks (red
arrows). Scale bar for (a) is at 1 mm; scale bar for (b & c) is at 500 μm [48]. © 2020 IEEE……….22

Figure 1-15. (a) A representative CV curve of a single platinum electrode after the 2nd cycle
(orange) and 30th cycle cleaning (black) in the 0.05 M H2SO4. Characteristics peaks
corresponding to oxidation-reduction reactions between platinum and the ions in the solution
are labeled. (b) The electrochemical impedance spectroscopy (EIS) graph of the average
electrode impedance magnitude (mean ± SD, N =18 electrodes) before (orange) and after
(black) CV cleaning. (c) EIS phase curve taken before (orange) and after (black) CV cleaning
(mean ± SD, N = 18 electrodes). All electrodes are from a single device. The CV and EIS
measurements were taken after the array was thermally annealed [48]. © 2020 IEEE……………24

Figure 1-16. Illustration of the implanted array, electrical packaging, and recording setup for
the in vivo study [48]. © 2020 IEEE……………………………………………….…………...…26

Figure 1-17. (a) Fully packaged arrays using the PCB. (b) The dimensions of the PCB
designs. (c) A fully packaged array with the PCB shown in (a) and (b) highlighting the acrylic
backing and the built-in ground wires [48]. © 2020 IEEE……………………………………..…26

xii

Figure 1-18. Photo showing the surgical setup. The array is secured to the skull using dental
cement in preparation for recording (scale bar = 1 cm) [48]. © 2020 IEEE…………………..…27

Figure 1-19. Representative spike waveforms from multiple units were recorded from one
lightly anesthetized animal immediately after implantation [48]. © 2020 IEEE………..……...…28

Figure 1-20. Chronic recordings were obtained with a Parylene C multi-electrodes array. (a)
Average noise level (mean) and the spike amplitude (mean ± SD) of neural signals were
recorded from four animals over 5 to 12 weeks post-implantation. (b) Representative
recording from one animal at 12 weeks. The average spike amplitude of the units recorded
from a single neuron was 261.58 μV, and the 3-sigma of noise level was 37.45 μV. A group
of complex spikes is identified in the orange dashed box. (c) The top shows the overhead
frame capture of a rat running freely in an open field. The bottom two plots show the place
field of two units simultaneously recorded at 12 weeks post-implantation from the CA1 and
the CA3 sub-regions while the animal was running freely. The color bar represents the firing
rate of neurons (in Hz) [48]. © 2020 IEEE…………............................................................……..29

Figure 1-21. 48-week chronic recordings obtained with a Parylene probe showing the
average spike amplitude and average noise level over time (Courtesy of H. Xu).………..……..30

Figure 1-22. Neural recordings acquired from 8 channels located on a single probe shank
with crosstalk (Courtesy of W. Jiang).…………...…………………………..…………………...31

Figure 1-23. Benchtop setup used to obtain crosstalk measurement……………………………..32

Figure 1-24. (a) Result table showing a zero crosstalk between 8 selected traces. (b) Cross
section of a device illustrating channels with zero crosstalk and 50% crosstalk………………..…32

Figure 1-25. Schematic showing the probe-brain interfaces while acquiring neural recording
from a live rat……………………………………………………………………………….……33

Figure 1-26. Benchtop crosstalk measurement taken from probe recycled from surgery (top)
and after the omnetics connector and PCB were separated…………………………………….…34

Figure 1-27. (a) Setup for the soaking experiment. (b) Probe soaked in the liquid vail. (c)
Probes packaged with dental cement and marine epoxy………………………………….………35

Figure 1-28. Representative crosstalk results acquired with devices packaged with marine
epoxy and dental cement and soaked for two months…………………………………….………36

Figure 1-29. Schematic showing the cross-section of Parylene probe with crosstalk between
traces (left); without crosstalk between traces after surface treatment (right)…………………….37

Figure 1-30. Schematic showing the cross-section of the device with different surface
treatments………………………………………………………………………………………...38

xiii

Figure 1-31. Graph showing crosstalk percentage over time for the device with different
surface treatment (N=3, except for control device N=1)……………………………………..…...39

Figure 1-32. Schematic showing the resistivity measurement for probe with intact, broken,
and shorted traces…………………………….…………………………………………………..39

Figure 1-33. Resistivity measurement result for control and AdPro Plus treated device
(highlighted in green dashed box) soaked for up to 190 days……………………………………..40

Figure. 1-34. Photo showing the Neuronexus probe and Parylene C probe………………………43

Figure 1-35. Graph showing force and dimpling measured with Neuronexus and Parylene C
probe (mean ± SD, N=3)…………………………………………………………………………45

Figure 1-36. Schematic showing probe with different shank shape and shank thickness………...46

Figure 1-37. Measured insertion and shear force acquired with the probe with different shank
width and pitch (mean ± SD, N=3)……………………………………………………………….48

Figure 1-38. Photo showing a representative side view of an 8-shank probe inserted into
agarose phantom highlighting shank straightness of probe shanks with different thickness and
shank shape to a depth of 4 mm…………………………………………………………………..50

Figure 1-39. Schematic showing the method of increasing the recording volume of the probe
by 3D stacking. For the planner 8-shank probe shown on the left, each shank has 8 electrodes,
so a single planar probe has a total of 64 channels on 8 shanks. By stacking 8 of the probes
together (right), there is 64 x 8, a total of 512 channels………………………………………….51

Figure 1-40. Schematic showing the multiple-layer PDMS mold used to create 2x8 (16-
shank) 3D arrays…………………………………………………………………………………53

Figure 1-41. Photo showing the 6x8 3D 2.8 mm long Parylene sham device packaged with
double-sided tape………………………………………………………………………………...55

Figure 1-42. Photo showing the 8x8 3D 2.8 mm long Parylene sham device packaged with
double-sided tape………………………………………………………………………………...55

Figure 1-43. Graph showing the buckling and insertion force (Mean ± SD) for probe with
shank numbers of 1, 8, 32, 48, and 64 (n = 3 for 1, 8, 32, and 64 shanks; n = 2 for 48 shanks).........57

Figure 1-44. Graph showing the mean and standard deviation of dimpling for probe with
shank numbers of 8, 16, 24, 32, 48, and 64 (n=3)…………………………………………………58

Figure 1-45. Graph showing the mean and standard deviation of insertion force for probe
with shank numbers 8, 16, 24, 32, 48, and 64 (n=3)……………………………………………..59

xiv

Figure 1-46. Graph showing the mean and standard deviation of shear force for probe with
shank numbers of 8, 16, 24, 32, 48, and 64 (n=3)…………………………………………………60

Figure 1-47. Picture showing a 3x8 (24-shank) probe successfully packaged with the PEG
spacer. The spacer’s shape and dimension are labeled in the right image. The spacer with the
PEG cutout (right) corresponds to PDMS layer 1 in Fig. 1-40……………………………………61

Figure 1-48. Processed X-ray image showing the brain lobe where the 3x8 probes were
inserted. 2 out of 3 probes show the shank trace along the insertion path. The enclosed
window shows the measured depth of the probe inserted…………………………………………62

Figure 1-49. Picture showing a 4x8 (32-shank) probe successfully packaged with the 4mm
wide rectangular PEG spacer. The probes are attached to a rigid acrylic backing (highlighted
with a blue dashed line) with PEG. The shank tips are shown (highlighted in blue lines) in the
picture taken from the side of the device………………………………………………………….63

Figure 1-50. Pictures showing 4x8 (32-shank) probes stopped at an insertion depth of 4 mm
caused by the detachment of the probes from the acrylic support. The photos are taken from
different angles demonstrating the detached probe (highlighted in green) deviated from the
initial insertion path (highlighted in blue)………………………………………………………..64

Figure 1-51. (a) Schematic showing the dimension of the 1st PDMS layer (Fig. 2-1(2)). (b)
Bottom view of 3D matrix probe released from the PDMS mold with shank tips pointing
outward. (c) Top view of the 3D matrix probe. (d) Side view highlighting two Parylne probes
sandwiched between three 0.5 mm thick PEG spacers…………………………………………...65

Figure 1-52. Photo showing the mock insertion experiment setup with rat brain (left). A
histological slice was taken at a depth of 1.2 mm (right) (Courtesy of W. Jiang).………………66

Figure 1-53. (a) Schematic showing the dimension of the PCB design. (b) Fully packaged
2x8 arrays with PCBs. (c) Back view of the completed PCB package highlighting the acrylic
backing…………………………………………………………………………………………...67

Figure. 1-54. Graph showing a representative neural recording acquired from the cortex
region through a channel of the Parylene probe during surgery…………………………………68

Figure. 1-55. Schematic demonstrating the location of the 16 shanks inserted into the right
lobe of a rat (the figure is not drawn in actual scale). The bottom array is the one closed to
acrylic backing. 4 (highlighted in blue) out of 16 shanks successfully recorded neural signals.
The right figure shows the shanks with electrodes (highlighted in blue dots) that successfully
captured neural signals…………………………………………………………………………...69

Figure. 1-56. Schematic demonstrating the location of the 16 shanks inserted into the right
lobe of a rat for targeting the hippocampus. The figure is not drawn in actual scale. The
bottom array is the one closed to acrylic backing. Electrodes (highlighted in green dots)
located on four of 16 shanks successfully recorded neural signals………………………………..70

xv

Figure. 1-57. Graphs showing the EIS result of the same electrode (50 from a top array in
Fig. 1-56) before (top) and after (bottom) surgery. The impedance at 1 kHz increased from
643 to 733 kΩ (highlighted in yellow dashed line)……………………………………………….71

Figure. 1-58. Photo showing the histology of the brain tissue taken at a depth of 2.35 and
4.30 mm. The inserted shank left holes in the tissue (highlighted by red arrows) (Courtesy of
W. Jiang).………………………………………………………………………………………...71

Figure. 1-59. Photo showing the alignments of two stacked probes with electrodes facing the
same direction (left) and opposite (right)…………………………………………………………72

Figure 1-60. (a) Schematic showing the shanks and channels acquired neural signals from the
hippocampus. (b) Representative neural signals were obtained by four different groups of
electrodes from four channels…………………………………………………………………….73

Figure 1-61. Photo showing the geometry and size of the 10 mm and 15 m long probe. For a
10 mm long probe, section A has a width of 300 µm over a probe length of 8.5 mm. A 15 mm
long probe has a width of 300 µm over a probe length of 13.5 mm……………………………….75

Figure 1-62. Photo of sham Parylene probe with different lengths………………………………76

Figure 1-63. Photo of a 10 mm long Parylene probe shortened with PEG brace…………………76

Figure 1-64. Photo of a 10 mm long Parylene probe inserted in agarose phantom. Left: front
view: right: side view…………………………………………………………………………….77

Figure 1-65. (a) Photo of two 15 mm long Parylene probes stacked together with drops of
IPA. (b) The stacked probe probes, after being dipped coated in PEG solution and solidify
under room temperature………………………………………………………………………….79

Figure 1-66. (Left) Photo of probe insertion setup. (Middle) The side view of two stacked
probe probes inserted into 0.6% agarose phantom to a depth of 10mm. (Right) The side view
of the inserted probes……………………………………………………………………………..80

Figure 2-1. Schematic showing how a normal heart functions…………………………………...94

Figure 2-2. Schematic showing an HLHS heart…………………………………………….……95

Figure 2-3. Schematic showing the HLHS treatment through the Norwood procedure………….96

Figure 2-4. Schematic showing the mBTS failure caused by blood obstruction…………………97

Figure 2-5. Schematic showing the ring electrode attached to the external wall of mBTS for
blood flow monitoring…………………………………………………………………………....98

Figure 2-6. Schematic showing the ring electrode integrated with the mBTS designed based
xvi

on capacitively coupled contactless conductivity detection……………………………………..100

Figure 2-7. Photo showing the prototyped sensors with five ring electrodes by directly
painting silver epoxy on the mBTS shunt with an outer diameter of 3.5 mm. Multiple
sensors are made with electrodes having different widths and spacing………………………….101

Figure 2-8. Schematic showing the electrode layout and dimension……………………………101

Figure 2-9. Schematic showing the equivalent circuit for the mBTS sensors…………………...102

Figure 2-10. The circuit architecture for voltage read-out from a single pick-up electrode……103

Figure 2-11. The testing setup with the breadboard system………………………………….....104

Figure 2-12. Schematic for bootstrapping circuit……………………………………………….105

Figure 2-13. (a) Circuit with shunt model connected to the oscilloscope; (b) Measured
waveform after the bootstrapping circuit………………………………………………………..106

Figure 2-14. (a) Shunt model hooked to the bootstrapping circuit; (b) Measured waveform
with the bootstrapping circuit…………………………………………………………………...107

Figure 2-15. (a) Layout of the bootstrapping circuit with adjustable gain controlled by the
resistor's value highlighted in the red circle. (b) The measured waveform with bootstrapping
circuit shows an amplified signal………………………………………………………………..107

Figure 2-16. (a) Measured waveform of the input signal; (b) Modulated waveform output by
the mixer………………………………………………………………………………………...108

Figure 2-17. The circuit schematic for the low pass filter block………………………………...109

Figure 2-18. Simulation result for the low pass filter block……………………………………..110

Figure 2-19. Measured waveform for low pass filter with the input signal at different
frequencies. “Sin” labeled in the figure represents the sinusoidal wave of the input signal……...111

Figure 2-20. Measured waveform (left) and FFT (right) for mixer and low pass filter at
different frequencies. “Sin” labeled in the figure represents the sinusoidal wave of the input
signal……………………………………………………………………………………………112

Figure 2-21. Measurement setup used to test the circuit………………………………………...113

Figure 2-22. Measured waveform and FFT for the completed circuit with a single pick-up
electrode………………………………………………………………………………………...114

Figure 2-23. The block diagram for the circuit design for three pick-up electrodes…………….114
xvii

Figure 2-24. Circuit layout for the differential amplifier………………………………………..115

Figure 2-25. The schematic showing the layout of the PCB and its integration with the
mBTS shunt. The arrow indicates the flow direction of signals………………………………..117

Figure 2-26. Schematic showing the setup used mBTS sensor to transduce the flow signal……118

Figure 2-27. Graph showing the weight measurement (y-axis) along the time of infusion
(x-axis) for 100 – 280 mL.min. The weight measurement starts 5s after the pump is turned
on for consistency between measurements (Courtesy of L. Custer). ……………………………119

Figure 2-28. Graph showing the weight measurement (y-axis) along the time of infusion
(x-axis) for 300 and 369 mL/min. The weight measurement starts 5s after the pump is turned
on for consistency between measurements (Courtesy of L. Custer). ……………………………119

Figure 2-29. Schematic showing the Parylne sensor designed to acquire flow data from a
pig model……………………………………………………….……………………………….120

Figure 3-1. Schematic showing an implanted ventricular shunt for draining access CSF into
the abdomen. Zoomed-in picture on the left shows a normal shunt versus a clogged shunt……..123

Figure 3-2. Low profile, flexible sensor die packaged in a Luer module. Fabricated
Parylene impedimetric sensor after release from the silicon substrate. Inset shows a close-up
of the flow sensor elements (yellow box) [17]. Reprinted with the permission of the author
and the Transducer Research Foundation ………………………………………………………127

Figure 3-3. Impedance response of sensor over a heating cycle. Flow rate is transduced from
the normalized maximum rate of impedance (Z) change during the “Heater ON” phase
(10-20 s) [17]. Reprinted with the permission of the author and the Transducer Research
Foundation …………….………………………………………………………………………..128

Figure 3-4. Schematic of fully packaged impedimetric and LD20 sensors. Blue arrows
indicate the direction of CSF flow [17]. Reprinted with the permission of the author and the
Transducer Research Foundation ………………………………………………………………130

Figure 3-5. Benchtop testing setup for calibrating the sensor. Phosphate-buffered saline
(PBS) flow and an oven were used to mimic the body’s warm saline environment [17].
Reprinted with the permission of the author and the Transducer Research Foundation ………...131

Figure 3-6. Graph showing a representative calibration curve of an impedimetric flow
sensor over the flow range between 0 and 1000 µL/min………………………………………...132

Figure 3-7. Graph showing the separated calibration curves for high (top, linear, > 200
µL/min) and low (bottom, quadratic, < 200 µL/min) flow rates………………………………...133

Figure 3-8. Schematic showing the testing setup used in the swine study………………………135

xviii

Figure 3-9. Photo showing the assembly of the sensors and valves for priming purposes.
The devices listed from left to right are 3-way valve #1, impedimetric flow sensor #1, LD 20
#1, adjustable valve (Medtronic Strata II Valve Regular), impedimetric flow sensor #2,
LD 20 #2 and 3-way valve #2…………………………………………………………………...136

Figure 3-10. Photo showing detachment of the sensor assembly from the syringes used for
priming………………………………………………………………………………………….136

Figure 3-11. Photo demonstrating the connection and priming of the sensor assembly with
the newly added infusion and butterfly needle catheters………………………………………...140

Figure 3-12. Photo demonstrating the connection of the primed assembly of sensors and
catheters to the syringe pump used for aCSF infusion…………………………………………..141

Figure 3-13. Testing setup for the swine study. The ventricular catheter was implanted into
the ventricle with the reservoir under the scalp then the wound was sutured. A needle with a
catheter was connected to the reservoir (yellow dashed box) to access CSF via the shunt for
draining or infusing. Flow sensors were connected to the output to measure CSF's flow rate
and temperature [17]. Reprinted with the permission of the author and the Transducer
Research Foundation …………………………………………………………………………...142

Figure 3-14. Testing setup for the swine study highlighting the location and connection of
the manometer used to monitor the ICP. The liquid's height in the manometer corresponds to
the height of the buoyancy ball in the tube. The zoom-in box showed the valve position
when the manometer was off. An “ON” position corresponds to a 90° angle between the
direction of the valve and the tube……………………………………………………………...143

Figure 3-15. Schematic showing the setup for measuring the ICP through a manometer. The
red arrows indicate the flow direction of CSF. The 3-way valve #2 was turned off for the
infusion line and sensor paths…………………………………………………………………..143

Figure 3-16. Photo demonstrating the swine study setup highlighting the setup used for
gravimetric flow measurements………………………………………………………………...144

Figure 3-17. Photo showing the hand drill used to create the burr hole after the incision was
made on the scalp……………………………………………………………………………….146

Figure 3-18. Photo showing the burr holes made by the drill hole for targeting the left and
right ventricle…………………………………………………………………………………...147

Figure 3-19. Photo showing the ventricular reservoirs inserted into the ventricle……………..147

Figure 3-20. Photo (left) demonstrating the proper placement of the ventricular reservoir by
the observed CSF outflow from the reservoir sidearm. Image (right) with a yellow mark
showing the closed part of the reservoir…………………………………………………………148

xix

Figure 3-21. Photo (left) with blue mark demonstrating the location of the sutured added to
the reservoir sidearm for attachment to pericranium. Picture (right) with a blue arrow
showing the suture for skin incision closure…………………………………………………….148

Figure 3-22. Schematic (left) showing the reservoir tap procedure. Photo (right) shows two
needles tapped in the reservoirs and sensors connected to the needle outlet highlighted with
a yellow arrow…………………………………………………………………………………..151

Figure 3-23. Photo demonstrating the swine study setup highlighting the placement of the
sensor box and the detailed design of the sensor box……………………………………………151

Figure 3-24. Photo demonstrating the sensor testing setup under the bolus aCSF infusion
mode in time sequence………………………………………………………………………….152

Figure 3-25. Flow and temperature of pig CSF acquired by impedimetric Parylene and
LD20 sensors acquired after sequential aCSF bolus infusions at 80, 320, and 560 s.
Measurements were obtained after infusion…………………………………………………….153

Figure 3-26. Graph showing the comparison between gravimetric flow and flow
measurements acquired from impedimetric Parylene and LD20 sensors with three bolus
aCSF infusions………………………………………………………………………………….154

Figure 3-27. Photo demonstrating the sensor testing setup under the constant aCSF infusion
mode in time sequence………………………………………………………………………….155

Figure 3-28. Flow and temperature of pig CSF acquired by impedimetric Parylene and
LD20 sensors acquired under constant aCSF infusion. The infusion was stopped after 640 s,
and measurements were taken during infusion………………………………………………….155

Figure 3-29. The measured flow rate of the impedimetric sensor (y) was linearly correlated
to the LD20 sensor (y=0.95 x, R
2
=0.99) [17]. Reprinted with the permission of the author
and the Transducer Research Foundation ……………………………………………..………..156

Figure 3-30. The measured flow rates of the impedimetric sensors (y) were linearly
correlated to the LD20 sensors across different devices and animals as R
2
>0.9…………………156

Figure 3-31. Comparison of flow measurements over one week for the impedimetric and
LD20 sensors (implantation on Day 1) [17]. Reprinted with the permission of the author and
the Transducer Research Foundation ……………………………………………….…………..157

Figure. 3-32. (Top) Micrograph of nanobubble pressure sensor with counter electrodes on a
single die. The sensor is highlighted in the dotted box. (Bottom) Schematic showing sensor
with major components labeled. The working and counter electrodes were used to inject
current and generate the electrolytic bubble. Sensing electrodes tracked impedance across
the bubble microchannel to transduce pressure. The constriction valves trapped the bubble
in the sensing region of the microchannel [27]. © 2022 IEEE ………………………………...160
xx

Figure. 3-33. Photo showing the varying size for counter and working electrodes……………..161

Figure. 3-34. (Top) Randles equivalent circuit model of sensing electrode-electrolyte
interface. (Bottom) Impedance and phase of the sensing electrodes. The measurement
frequency (dashed line) is selected when the response is dominated by solution resistance
[27]. © 2022 IEEE…...........................................................................................................……162

Figure. 3-35. Photograph of testing setup with the sensor testing fixture, microscope, and
Fluigent pressure source. Inset shows a close-up of the sensor in the testing fixture
positioned under the microscope [27]. © 2022 IEEE…………………………………………..163

Figure. 3-36. Lifetime of electrolytically generated O2 (0.4 µA, 8s) and H2 (-0.4 µA, 8s)
bubbles in the microchannel after the current pulse was terminated (n=5, mean ± SD) [27].
© 2022 IEEE……………………………………………………………………………………164

Figure. 3-37. Schematic showing the sensor operation in time sequence………………………165

Figure. 3-38. (Top plot) Impedance response measured by sensing electrodes (black) and
(bottom plot) O2 bubble volume tracked by video feed from the microscope (blue) after the
applied current pulse (0.3 µA, 25s). Impedance magnitude decreased as the bubble dissolved
back into the solution. (Bottom panels) Frame captures of bubble detachment from channel
surfaces correspond to the dramatic impedance drops observed at ~20 and ~320 minutes
[27]. © 2022 IEEE ……………………………………………………………………………...166

Figure A-1. Schematic of thermos-bonding steps…………………………………..…….……173

Figure A-2. Schematic of the pressure applicator used for thermocompressive bonding………174

Figure A-3. Schematic of microfluidics channel flow characterization. (a) Bare Parylene
microfluidic channels; (b) channels filled with red dye.……………………………………..…176

Figure A-4. Picture showing the devices under different failure modes………………………...177

Figure A-5. Schematic of the flow test setup with a syringe pump……………………………...177

xxi

Biomedical microelectromechanical systems (BioMEMS) use micromachining techniques
to design miniaturized systems for biology and medicine. Conventional MEMS often use silicon,
but it is not preferred for medical device applications for many reasons. First, the silicon is not
biocompatible and often needs some biocompatible coating to prevent it from directly contacting
the tissue. It is also stiff and has a high Young’s modulus. The mismatch between tissue and
silicon’s stiffness will cause severe damage. The scar formation around the device limits the
device’s lifetime. It is also brittle, so silicon could not be made too thin; otherwise, have the risk
of fracture.
The flexible polymer has been explored to combine with MEMS techniques to make
medical devices. My work focuses on applying Parylene bioMEMS to make biomedical devices.
Parylene is selected because it is a transparent polymer with a low Young’s modulus that better
matches the tissue stiffness than silicon. It is also classified as a Class VI material by the United
States Pharmacopeia and is suitable for implants. More importantly, it is flexible and can be used
to make low-profile or thin devices.
Our body is operated by eleven physiological systems daily, including the urinary,
cardiovascular, endocrine, nervous, and many others. We lack tools that can be placed inside the
body to help acquire information from different systems to assist clinicians with disease diagnosis
and treatment. The motivation here is to use Parylene MEMS to develop tools that could be safely
placed in the body and acquire physiological signals to help better understand how different
systems work individually and collaborate with others.
My work includes developing brain probes to detect neural signals from the brain, which
will be discussed in chapter 1. In addition to the neural probes, sensors have also been developed
ABSTRACT
xxii

to monitor physiological flow for patients with cardiovascular diseases and neurological diseases.
The related work will be described in chapters 2 and 3, respectively.

1

1.1 Background
There are 86 billion neurons carrying information around the body to control how a person
thinks, talks, and performs life activities every day [1]. As the primary functioning cell in the brain,
a single neuron is connected to thousands of neurons. A neuron contains soma, dendrites, and axon.

Figure 1-1. Schematic showing the structure of a neuron. The communication between two neurons (neuron
A: blue; neuron B: orange) happens at the axon terminal through the synapse.

CHAPTER 1
FLEXIBLE PARYLENE-BASED
INTRACRANIAL PROBE FOR DEEP BRAIN
RECORDING
2

The soma (cell body) receives signals through dendrites and processes the signals through
the cell nucleus (Fig. 1-1). The signal is then transported to the neighboring neurons through the
axon. The neuron-neuron communications happen at the axon terminal through the synapse.
Neuron-neuron communication is not performed by a physical connection between neurons.
A synapse is a gap between neurons that can transfer either electrical or chemical signals. The
chemical synapse transfers signal through the release of neurotransmitters [2]. When a signal
reaches the axon terminal of neuron A, neurotransmitters are released into the synapse cleft and
binding to the receptors on the membrane of the second neuron cell. The binding between the
neurotransmitters released from neuron A and the receptors on neuron B triggers a neural response.
Electrical synapses transfer signals through gap junctions which are the channels located on
the axon terminal of neuron A and the cell membrane of neuron B [3]. Because those gap junctions
are closely aligned, the electrical signal can transfer directly. Brain-machine interfaces are
developed to better understand how neurons function and communicate with each other.
The brain-machine interface (BCI) is a system receiving, processing, and transferring signals
from the brain. It connects the neural system of humans with external equipment and machinery.
BCI can either capture signals from the brain or send signals to the brain. It is a powerful tool for
understanding the disease model and assisting clinicians in diagnosing and treating neurological
diseases. It can also be used in people with disability to restore or replace the part of the body that
did not work. For instance, in 2006, a microelectrode brain array was implanted into a paralyzed
human’s motor cortex, enabling the patient to operate a robotic arm and open and close a prosthetic
hand [4].
The signal acquisition unit of a BCI system is grouped into three major categories based on
its invasiveness [5, 6]. If looking at the cross-section of the human skull and meninges, the brain
3

is encapsulated by a shield consisting of pia mater (most inner layer), subarachnoid space,
arachnoid mater, subdural space, dura mater, skull bone, periosteum, and scalp skin (most outer
layer) [7].
Electroencephalogram (EEG) is noninvasive and involves electrodes directly placed on the
scalp to record an electrogram summarizing the electrical activity from the surficial brain layers.
EEG often consists of small metal discs connecting to the recording equipment. The metal discs
that captured brain activity are directly attached onto the patients’ scalp and connected to the
recording equipment. Due to its lower risk compared to other more invasive brain-machine
interfaces, it is a standard diagnostic tool helping clinicians to detect the abnormal brain activity
resulting from various diseases, including seizures, stroke, Alzheimer’s disease, and sleep disorder.
The limitation of EEG is its low spatial resolution; the neural signal gets attenuated when it passes
through the skull and meninges layer above the cortical brain layer.
Electrocorticography (ECoG) is semi-invasive, requires cranial surgery, and involves
electrodes placed in the epidural and subdural regions [8]. It captures signals directly from the
cortical surface, resulting in higher spatial resolution than EEG. It also recorded the neural signals
with higher amplitude and frequency (40 Hz-200 Hz), which is sufficient to help decode the
cortical signal related to motor and language function [9]. In 2010, ECoG was demonstrated on
patients with a disability to accurately type words on a computer screen using their minds [10].
ECoG records signals from groups of neurons located at the superficial layer of the cortex but does
not record the signal from single neurons.
To understand more complex neural functions, penetrating neural probes are developed to
be placed close to neurons and record signals from single neurons. Those probes include platform
array, microwire, and multisite probe. The Utah array is the commonly used platform array, which
4

only records from the cortex. There are some brain regions are deep and in charge of complicated
neural functions. To better understand the neural functionality of the deeper brain structure,
microwire has become the gold standard for decades. However, only one recording site is located
at the tip, limiting the number of neurons that can be accessed. Wire bundles have been made to
increase the channel count; however, it significantly increased the damage to the surrounding
tissue and limited the device for short-time recordings. To reduce the tissue damage to the
surrounding tissue, researchers use MEMS techniques to minimize the probe size and pack more
electrodes for recording from more neurons.
Penetrating microelectrode arrays produced using conventional silicon micromachining
techniques can support 10's to 100's recording sites on shanks having widths of ~100 microns and
measuring several millimeters in length; recent efforts are approaching 1000's sites [11-13]. These
tools offer high spatiotemporal resolution neural recording but often fail under chronic in vivo
conditions. Fundamental neuroscience research and neuroprosthetic technologies not only require
penetrating neural probes closed to neurons to capture high-resolution electrophysiological
recording but also to maintain stable, long-term performance.
The mismatch in stiffness of common MEMS materials, silicon (E ~ 150 GPa [14]), and
brain tissue (~ 6 kPa [15]) has been identified as a potential cause of chronic injury to the brain
and associated retaliatory immune response [16-19]. Instability at the device-tissue interface
arising from micromotion can result in neuronal death and glial scarring in the surrounding tissue.
This prevents the acquisition of neural recordings beyond weeks or months after implantation [16,
19-22].
To improve the recording lifetime of these devices, penetrating microelectrode arrays made
from softer materials in various form factors are being explored to minimize the chronic foreign
5

body response [17, 23, 24]. These materials include polyimide [25-27], SU-8 [28-31], poly-(para-
chloro-xylylene) (Parylene C) [32-34], thermoset shape memory polymer (SMP) such as thiol-
ene/acrylate [35-37], composite material containing elastomers such as polydimethylsiloxane
(PDMS) [38, 39], and soft-nanocomposites made from cellulose fiber scaffold [24, 40, 41]. The
consequence of using low elastic moduli materials is that the long, slender neural probe shanks are
susceptible to mechanical buckling during insertion into brain tissue [42, 43], preventing accurate
targeting and access to deep brain structures. In practice, penetrating polymer microelectrode
arrays are limited to superficial brain targets (< 3 mm deep).
To access and study deeper brain structures such as the hippocampus, thalamus, and basal
ganglia across species with brains of varying sizes [44-46], the flexible probe is often coupled to
a rigid support shuttle during insertion [47]. Most flexible probes designed for deep brain recording
contain fewer recording electrodes laid out on a single probe shank [47].
The project aims to create a high-density Parylene probe that overcomes the surgical
placement challenges to acquire recordings from deep brain structures such as the hippocampus
from different species. The long-term soaking experiment was conducted with probes for
mimicking the chronic use of the devices in vivo and characterizing the long-term performance of
the probes with different surface treatments. The first part of the chapter focuses on developing
neural probes for rodents. The later part focuses on scaling up the probe for larger animals and
developing a novel insertion method for inserting long probes (>10 mm).
1.2 Parylene-based 64-channel Hippocampal Probe
The hippocampus is a deep brain structure located under the cortex responsible for memory,
learning, and spatial navigation, including long-term memory formation and spatial information
6

Figure 1-2. a) Cresyl violet stained coronal rat hippocampus slice showing the location of sub-regions CA1,
CA3, and DG (scale bar = 600 μm). (b) Schematic showing probe design with electrode group layout to
target CA1 and CA3 regions [48].

processing. It is also a brain region commonly damaged in many neurological and psychiatric
diseases [49]. Studies have shown that the hippocampus is the earliest damaged brain structure
resulting from neurological disorders, including Alzheimer’s [50] and epilepsy.
7

Figure 1-3. (a) Fabricated brain probe array compared to US dime (scale bar = 5 mm). (b) Schematic
showing probe design with electrode layout to target CA1 and CA3 regions. (c) Schematic showing probe
design with electrode layout to target CA1 and DG. (d) As-fabricated electrodes are 50 μm in diameter, of
which 30 μm is exposed after applying surrounding insulation. The center-to-center distance between the
electrodes is 70 μm. Probe arrays measured 20 μm in thickness and consisted of a Parylene C-platinum-
Parylene C sandwich [48]. © 2020 IEEE

The neurons in the hippocampus communicate through a one-way trisynaptic neural circuit
located at three different layers, including cornu ammonis (CA)1, CA3, and dentate gyrus (DG).
8

Few polymer penetrating neural probes targeting the hippocampus in rats (> 3 mm deep) were
reported; these devices possessed a single probe shank with regularly spaced electrode layouts that
do not conform to the anatomy of hippocampal circuits [37, 51].
To address the need for high channel count probe arrays that can access multi-region of the
hippocampus, a 64-channel Parylene array with 5.5 mm long shanks was designed to span a 2000
μm length of the brain along the septal-temporal axis with eight probe shanks that are 150 μm wide
and spaced with a 250 μm center-to-center distance (Figs. 1-2 and 1-3). The 64 electrodes are
divided into electrode clusters positioned to match the laminar anatomy of the hippocampus layers
[48].
1.2.1 Probe Design and Fabrication
Each shank was constructed using a 20 μm thick Parylene C-platinum-Parylene C sandwich
and terminating in a pointed tip with a 45° angle. Numerous studies describe the fabrication and
in vivo performance of rigid and soft neural probes with different designs, including tip geometries
and probe dimensions [14, 52-56]. The array geometry and dimensions were chosen in the range
reported in the literature corresponding to relatively minimal tissue damage [20, 57]. An individual
shank's width expanded from the pointed tip to 110 μm and then slightly widened to a maximum
of 150 μm. Each shank had eight electrodes for 64 electrodes across the 8-shank array. The eight
electrodes were arranged in two or three linear clusters on each shank. Thus, two different arrays
were designed to target specific regions in the trisynaptic circuit of the hippocampus (Fig. 1-2).
All polymer devices are micromachined in a layer-by-layer process on a bare 4" silicon (Si)
wafer. The wafer was pre-baked at 110 °C (> 10 mins) before fabrication to dehydrate the surface.
A base layer of 10 μm thick Parylene C (Specialty Coating Systems, Indianapolis, IN) was
9

deposited by chemical vapor deposition (CVD) under vacuum. A 1.5 μm thick AZ5214-IR
(Integrated Micro Materials, Argyle, TX) layer was spin-coated on top of the base Parylene C layer
(step 1: 8 s, 500 rpm, step 2: 45 s, 2,000 rpm), and then patterned via photolithography as a lift-off
mask used to define the electrode sites (50 μm diameter) and metal traces (5 μm width with 5 μm
spacing). Before platinum deposition, the wafer surface was cleaned. The Parylene C was activated
using O2 plasma in a reactive ion etcher (RIE) (Technics, 800 Series Micro RIE System) at 100 W
and 100 mTorr for 1 minute. 2000 Å of platinum was deposited by electron-beam deposition
(Caltech Kavli Nanoscience Institute, Pasadena, CA). Excess metal was lifted off in acetone heated
to 50 °C with gentle brushing to dislodge excess metal between the traces and followed by
successive rinses in isopropyl alcohol (IPA) and deionized (DI) water for 5 minutes. Before the
deposition of a second Parylene C layer, wafers were cleaned. The Parylene C was activated in O2
plasma at 100 W and 100 mTorr for 1 minute and then dehydrated at 110 °C under vacuum (> 10
mins). A second 10 μm thick Parylene C insulation layer was deposited. AZ4620 (Integrated Micro
Materials, Argyle, TX) was spin-coated (step 1: 5 s, 500 rpm, step 2: 45 s, 1200 rpm) to form a 15
μm thick etch mask that defined the device and probe outline. The device outline was then etched
10 μm deep using a switched chemistry process in a deep reactive ion etcher (DRIE, Oxford Plasma
Lab System 100; 700 W ICP, 80 W RF power, 23 mTorr) [58]. The remaining photoresist mask
was removed in acetone, IPA, and DI water. A subsequent 30 μm thick AZ4620 mask (two spins
separated by soft bake at 90 °C for 12 minutes, both spins performed in two steps with step 1: 8 s,
500 rpm and step 2: 45 s, 2000 rpm) was used as a mask for the final switched chemistry C4F8/O2
etch step (same parameters as above) which exposed electrodes and contact pads and completed
the array outline. The electrode sites and contact pads were etched out at the end of the process to
avoid scum deposition from another micromachining process. The remaining photoresist was
10

stripped by rinsing in acetone, IPA, and DI water. To release devices from the wafer, devices were
soaked in DI water and gently peeled away from the native oxide layer on the silicon substrate.
Released Parylene C arrays were annealed to increase Parylene C chain entanglement and reduce
stress [59]. For this process, arrays were sandwiched between two Teflon sheets lining two glass
slides, and the assembly was held together with clips. Arrays were placed in an oven, vacuum
purged three times with N2, and then annealed under vacuum for 48 hours at 200 °C [60]. The
intermediate Teflon layer prevented Parylene C from irreversibly adhering to the glass slides [48].
1.2.2 Probe Insertion
A single neural probe can be modeled as a thin, uniform beam, with one end fixed and the
other end pinned against the brain's surface at the moment of implantation. Euler's buckling
formula (Eq. 1) approximates the force required to buckle a probe of length L, width w, thickness
t, Young's modulus E, and an effective length factor k [25, 52, 61-63].

𝐹 𝐸𝑢𝑙𝑒 𝑟 =
𝜋 2
𝐸𝑤 𝑡 3
12 ( 𝑘𝐿 )
2
(1)

The column effective length factor, k, captures the degree to which each end of the column
is constrained against movement. During insertion into brain tissue, the probe’s base is clamped
to the insertion tool and fixed (allowing for no translation or rotation). The probe’s tip is free to
move laterally as soon as it contacts brain tissue where k is 2. As the tip moves and further deforms
the brain surface without penetration, the boundary condition changes to be pinned in the x-y plane
(only allows for rotation, not translation). Accordingly, the commonly accepted value of k is 0.7,
11

validated experimentally in [42]. Once the probe has penetrated the tissue, both ends of the probe
are now effectively fixed, and k drops to 0.5, yielding a higher buckling force threshold. This
predicts that probes inserted in tissue can withstand more stress without buckling when inserted
deeper into the brain, according to Equation (1). However, Equation (1) also indicates that polymer
probes having low Young's modulus and long shanks may have low buckling force.

Figure 1-4. Schematic showing (a) probe buckled at the brain surface when F insertion>F buckling; (b) probe
coupling to rigid shuttle and stiffener coating successfully inserted into the brain.

To insert a long flexible probe without buckling, the force taken to insert the probe needs to
be smaller than the force taken to buckle the probe. A common approach is to temporarily stiffen
the probe shank to avoid buckling by increasing the buckling force to exceed the force required to
penetrate brain tissue (Fig. 1-4). Examples include insertion shuttles [32, 64-69] which are
temporarily coupled to the probe and then retracted following implantation, and water-soluble
coatings such as silk [70, 71], polyethylene glycol (PEG) [72-74], tyrosine-derived polymers [75,
76], gelatin [54, 77], saccharose [78, 79], maltose [61], and carboxymethylcellulose (CMC) [80]
12

coated over the probes, which increases the effective Young's modulus of the assembly. However,
these approaches significantly increase the cross-sectional area of the probe shank. This increases
acute injury to the brain, further contributing to an increase in gliosis and isolation of recording
sites [81]. The impact of introducing a high concentration of coating material at the tissue-device
interface on neuronal health, recording quality, and local brain chemistry is unknown and requires
further investigation. Alternatively, it can increase the buckling force by decreasing length (L). To
reach deep brain structures beyond the first few millimeters, this decrease in length needs to be
temporary. This motivates the approach here to use a dissolvable brace that temporarily shortens
probe length and increases buckling force during implantation. Removal of the brace returns
probes to their native mechanical properties.
We introduced a novel method for deep brain implantation of micromachined Parylene C-
based probe arrays that preserved the original as-fabricated shank cross-section to minimize acute
insertion trauma [82, 83]. The probe’s back end was braced in a dissolvable polyethylene glycol
(PEG) slab to temporarily shorten the probe length and increase its stiffness during initial
implantation [84]. PEG was selected as the dissolvable material for its availability in a range of
molecular weights (MW), allowing the dissolution rate to be tuned. In addition, it is easy to prepare
and apply and has anti-immunogenic and antigenic properties [85].
The process of applying the PEG brace onto the array is depicted in Fig. 1-5. Molten PEG
(Sigma Aldrich, Darmstadt, Germany; MW 3350) was injected into a three-layer
polydimethylsiloxane (PDMS) mold that supported the array while the brace was applied in a 60
°C oven. A 1.5 mm thick acrylic backing was temporally attached to the probe's cable portion
(shown in Fig. 1-3b) through a PEG block. PDMS Layer 1 (0.5 mm, labeled 2 in Fig. 1-5)
contained a pocket defined by a vinyl cutter (Graphtec® cutting plotter CE6000-40) as a receptacle
13

to receive the acrylic backing and define the brace. The top PDMS layer (0.5 mm thick) also
contained a precisely cut slot and was used to complete the brace's top half. After filling, the entire
assembly was cooled to room temperature to allow the PEG to solidify. The braced array was
released by carefully peeling away the mold starting from the contact pad end to the probe tips
[48].

Figure 1-5. Fabrication process for the PEG brace. (a) An exploded view of the mold parts is depicted and
followed by (b) the assembled mold and method of PEG brace application. (c) Parylene C array with PEG
brace after releasing from the mold (scale bar = 2 mm) [48]. © 2020 IEEE
1.2.3 Mechanical Evaluation of the Parylene Probe
To surgically insert these soft polymer probes, their mechanical performance must be
considered. Sham probe arrays having identical geometry and dimensions but without any metal,
features were prepared for the mechanical test (Fig. 1-6 (a), (b), and (c)). A 20 μm thick layer of
14

Parylene C was deposited on a 4" dehydrated prime silicon wafer via two consecutive depositions
of 10 μm each. AZ 4620 photoresist was spin-coated to serve as an etch mask (30 μm thick) and
patterned via photolithography to define the probe array outline. Parylene C was etched via
switched chemistry C4F8/O2 DRIE (230 loops, 700 W ICP, 20 W RF Power, 23 mTorr). The
photoresist mask was then stripped in sequential baths of acetone, IPA, and DI water. Devices
were individually released by gently peeling under DI water immersion [48].

Figure 1-6. (a) Optical micrographs showing shank array for a sham Parylene device (left) and a completed
Parylene device with electrodes (right). Scale bar = 2 mm. (b) SEM micrographs showing the tip region of
the array (scale bar = 100 μm) and (c) a single probe tip (scale bar = 10 μm) [48]. © 2020 IEEE

According to the critical buckling equation (k = 2, E = 2.76 GPa [86], 20 μm thick and 110
μm wide probe shank), a Parylene C probe with an effective length of 2.8 mm can theoretically
withstand > 0.5 – 1.0 mN of force without buckling, the commonly reported insertion force for
penetrating the cortical tissue without dura mater [52, 53, 87]. For in vitro tests, probes were
shorted to 2.8 mm long and inserted into flat agarose (Sigma-Aldrich, Darmstadt, Germany) mold
as a representative mechanical brain phantom.
We measured the insertion and buckling forces for Parylene C probe arrays having different
shank numbers (1, 2, 4, 6, or 8; N = 5 for the insertion test; N = 3 for the buckling test) with
15

different probe lengths (5.5 mm and 2.8 mm). Insertion force was measured only for shorter probes
as it is not dependent on length. The shorter probes were obtained by covering half of the shank
with PEG leaving 2.8 mm exposed. Differing shank numbers were obtained simply by cutting off
unwanted shanks. A Bose model 3100 motorized bench coupled with a 50 g Honeywell load cell

Figure 1-7. (a) The mechanical testing setup for insertion force measurement. (b) Close-up view of the
clamped probe array over an agarose gel block placed on the load cell. (c) The mechanical testing setup for
buckling force measurement. (d) Close-up view of the clamped probe array over the Instron 5490 Series
built-in load cell [48]. © 2020 IEEE

(Sensotec, Model 31, Columbus, OH) (Fig. 1-7a) was used to measure the insertion force. Probes
were mounted on the Bose motorized clamp and driven vertically into a brain phantom (0.6%
agarose) on the load cell at a constant speed of 0.01 mm/s to determine insertion force (Fig. 1-7b).
16

Figure 1-8. A representative buckling force curve obtained with a 2.8 mm 8-shank Parylene C probe and
smoothed with adjacent point average. The threshold force is taken with the setup in Fig. 1-6 (c). The
buckling threshold force is highlighted with a blue dashed line. Inset shows the close-up view of a buckled
8-shank probe [48]. © 2020 IEEE

The brain phantom composition was selected for its similarity to brain tissue's bulk
mechanical properties [88]. The insertion speed of 0.01 mm/s was chosen to minimize the shear
force on the surrounding brain tissue, based on prior reports [89]. The agarose gel was prepared in
transparent plastic boxes less than 48 hours before insertion trials and sealed in Parafilm to prevent
water evaporation and maintain consistent properties across gel batches. However, the force
resolution of the Bose system was inadequate to resolve the single shank buckling force. Therefore,
the buckling force was obtained using an Instron 5940 Series Single Column Tabletop (Norwood,
MA) for the different probe lengths (2.8 or 5.5 mm, Fig. 1-7c). With a setup similar to that shown
in Fig. 1-7a, the probe was driven against a metal plate at 0.01 mm/s. Force was recorded as a
function of stage displacement. Data was analyzed in Origin software using the adjacent averaging
filter. Fig. 1-8 showed a representative mechanical measurement of a Parylene C probe array's
17

buckling force. The buckling force threshold was defined as the applied force to the entire array
when all shanks buckled. Fig. 1-9 showed representative results from insertion tests of Parylene C
probes into gel agarose. Here we defined insertion force as the force required to initially pierce the
brain phantom and the shear force as the maximum force recorded as a function of the insertion
depth [48, 90].

Figure 1-9. A representative raw (grey) force-displacement data and smoothed curve (blue) with adjacent
averages for a 2.8 mm 4-shank Parylene C probe inserted into a block of 0.6% agarose gel. After the probe
contacts the gel surface, the force increases until it reaches a threshold known as insertion force (labeled as
an unfilled black star), which occurs when the probe initially penetrates the gel. Prior to that point, the probe
displaces gel resulting in dimpling. As the probe advances deeper, the force increases until reaching the
shear force (labeled as a solid blue star), at which point the maximum force is measured. This also
corresponds to the completion of insertion. Then the probe rests in the gel corresponding to the relaxation
phase [48]. © 2020 IEEE

18

Figure 1-10. (a) The buckling force thresholds (mean ± SD, N = 3; bars) and insertion force (mean ± SD,
N = 5; blue squares). Buckling force is shown for 5.5 (solid) and 2.8 mm (hatched) shank lengths for single
and multi-shank arrays (2, 4, and 8). (b) The buckling force as a function of shank number for both 2.8-
(grey) and 5.5- mm (black) shank lengths (mean ± SD, N = 3) compared with the theoretical linear
extrapolation calculated from (1) [48]. © 2020 IEEE

The buckling force for probe arrays of up to 8 shanks of full or shortened length was
compared to the measured insertion force in Fig. 1-10a. Notably, in all cases, the 5.5 mm probes
buckled at forces below that required for insertion, confirming prior observations that long, thin,
19

polymer probes cannot be implanted into brain tissue successfully (Fig. 1-11). As expected, shorter
probes had a higher threshold for buckling, but for larger arrays, the buckling force was within a
standard deviation of the insertion force, prompting the need to shorten arrays even further for in
vivo implantation as a measure of safety [48].

Figure 1-11. (a) Braced Parylene arrays (2.8 mm exposed) successfully inserted into the 0.6% agarose
across all conditions (1, 2, 4, and 8 shanks; scale bar = 1 mm). (b) A single shank unbraced Parylene probe
(5.5 mm exposed) buckled before penetrating agarose (scale bar = 2 mm) [48]. © 2020 IEEE

Fig. 1-10b compared the experimentally determined buckling force as a function of array
size to the theoretical buckling force for a single probe as calculated with (1) (k = 0.7, E = 2.76
GPa, w = 110 μm, t = 20 μm). A single shank Parylene C probe (L = 5.5 mm) was expected to
buckle at an applied axial force of 0.13 mN and a shortened probe (L = 2.8 mm) at 0.52 mN. These
calculations fall within the range of experimentally determined values: 0.13 ± 0.02 mN and 0.45
± 0.21 mN, respectively. The buckling force for a Parylene C array probe did not scale linearly
20

Figure 1-12. Characteristic features of the force-displacement curve compared across different numbers of
shanks: (a) insertion force and (b) shear force (mean ± SD, N = 5). Each plot exhibits a high degree of
linearity, as evidenced by the quality of the fit. As the insertion force = 0.22 * shank number, R
2
=0.99;
shear force = 0.37 * shank number, R
2
=0.99 [48]. © 2020 IEEE

with the shank number. The 8-shank array buckled at 1.22 ± 0.12 mN for the full length and 2.09
± 0.06 mN at half-length, a nine- and fivefold increase over single-shank measurements,
respectively. Fig. 1-12 shows the linear regression of insertion force and maximum shear force on
arrays measured with 0.6% agarose gel as a function of array size. To determine the shear force,
21

all arrays were inserted to a depth of 2.5 mm. The shear force for a single shank is measured as
0.43 ± 0.17 mN, comparable to values reported in the literature taken with the same insertion speed
[89]. Both insertion and shear force scaled linearly with the number of shanks (R
2
> 0.99, N = 5).
As the number of shanks changed from 1 to 8, the insertion force increased six-fold (0.24 ± 0.09
mN changed to 1.50 ± 0.60 mN); the shear force increased seven-fold (0.43 ± 0.17 mN changed
to 2.84 ± 0.68 mN) [48].

Figure 1-13. Photo of a Parylene probe sandwiched between two 0.5 mm thick PEG blocks.
1.2.4 In Vivo Study with Parylene Probe
Initial feasibility studies on mechanical insertion were performed using probe shams
having no metal features. The sham device was packaged with a PEG brace with a 1mm exposed
length (Fig. 1-13). The exposed probe tips of braced arrays were inserted into the brain until the
PEG brace reached the brain's surface. The brace was incrementally dissolved in saline, and the
newly exposed bare probe length was advanced in increments at a speed of 10 μm/s down to 4 - 5
22

Figure 1-14. Histological slices after Parylene array implantation and removal highlighting the
hippocampus. Slices were stained with hematoxylin and eosin. (a) Coronal and (b & c) transverse slices
taken at 2.2 and 2.5 mm from the brain surface showing probe tracks (red arrows). Scale bar for (a) is at 1
mm; scale bar for (b & c) is at 500 μm [48]. © 2020 IEEE

mm depth (measured from probe tip contact with the brain surface). The animal was euthanized
and perfused with paraformaldehyde right after the insertion. After the array was removed from
23

the brain, the brain tissue was dissected from the cranium. The tissue was fixed in formalin
overnight. The tissue was then dehydrated with 18% sucrose solution and sliced for histological
staining. Histological brain sections reveal that the stab wounds matched probe cross-sectional
dimensions indicating minimal damage to surrounding tissues (Fig. 1-14) [48].
1.2.4.1 Electrochemical Characterization of Electrode Sites
Before applying the PEG brace in preparation for implantation, all 64 electrodes in each
array were electrochemically evaluated using cyclic voltammetry (CV) and electrochemical
impedance spectroscopy (EIS) using a 3-electrode setup on a Reference 600 potentiostat (Gamry
Instrument, Warminster, PA) [91, 92]. A 1 cm
2
platinum plate was used as the counter electrode,
and the reference electrode was Ag/AgCl (3M NaCl, BASi, MF-2052, West Lafayette, IN). The
setup was contained in a Faraday cage to minimize external noise. CV was performed at room
temperature in 0.05 M H2SO4 (30 cycles from -0.2 to 1.2 V versus Ag/AgCl (3M NaCl), scan rate
of 250 mV/s), and in the process, provided electrochemical cleaning of electrode surfaces [91, 93].
EIS was performed in 1× PBS at room temperature (25 mVRMS, 1-10
6
Hz). Electrode impedance
was analyzed at 1 kHz, reported as the frequency where neurons fire action potentials [94].
Electrodes exhibiting at 1 kHz an impedance > 2 MΩ or with uncharacteristic phase behavior were
considered open circuits and discarded during analysis.
The comparison of the CV curve taken after the 2
nd
and 30
th
cycles suggested an increase
in the electroactive area following the CV cleaning process (Fig. 1-15a). As the cycles progress,
the current response broadened at different voltages, indicating increased cleanliness of the
electrode surface area, allowing for more surface reactions and current flow. At cycle 30, the CV
was comparable to a typical representative CV curve for Pt having characteristic peaks of Pt
reaction in H2SO4 solution, indicating successful electrode cleaning. This improvement was also
24

Figure 1-15. (a) A representative CV curve of a single platinum electrode after the 2nd cycle (orange) and
30th cycle cleaning (black) in the 0.05 M H2SO4.Characteristics peaks corresponding to oxidation-
reduction reactions between platinum and the ions in the solution are labeled. (b) The electrochemical
impedance spectroscopy (EIS) graph of the average electrode impedance magnitude (mean ± SD, N =18
electrodes) before (orange) and after (black) CV cleaning. (c) EIS phase curve taken before (orange) and
after (black) CV cleaning (mean ± SD, N = 18 electrodes). All electrodes are from a single device. The CV
and EIS measurements were taken after the array was thermally annealed [48]. © 2020 IEEE
25

evident across the EIS data on electrodes obtained before and after the CV cleaning; impedance
magnitude decreased slightly from 715 to 513 kΩ at 1 kHz (Fig. 1-15b).
The mean impedance at 1 kHz across seven different devices (N = 420 electrodes) was 691
± 257 kΩ with variation between individual devices reported in Table 1-1. After CV cleaning, the
phase angle at higher frequency was more resistive (closer to 0°) and roughly constant at the lower
frequency range (Fig. 1-15c) [48].
1.2.4.2 Electrical Package
The implanted neural interface incorporated a flexible ribbon cable (5 μm wide platinum
traces with 5 μm edge-to-edge spacing) that fanned out and terminated in a contact pad array. The
backside of this portion was supported by a 0.05 mm thick polyetheretherketone (PEEK) tape with
a 0.06 mm thick acrylic adhesive (CS Hyde Co., Lake Villa, IL) to build up the thickness of the
probe for mating to a 71 pin zero insertion force (ZIF) connector (Hirose Electric Co., Japan).

Table 1-1. 1kHz Impedance (mean ± SD) of difference devices [48]. © 2020 IEEE

Device
ID
Electrode
Count
1 kHz
Impedance (kΩ)
A 53 734 ± 238
B 63 795 ± 364
C 62 718 ± 147
D 59 695 ± 122
E 60 883 ± 141
F 61 556 ± 132
G 62 468 ± 279

26

Figure 1-16. Illustration of the implanted array, electrical packaging, and recording setup for the in vivo
study [48]. © 2020 IEEE

Figure 1-17. (a) Fully packaged arrays using the PCB. (b) The dimensions of the PCB designs. (c) A fully
packaged array with the PCB shown in (a) and (b) highlighting the acrylic backing and the built-in ground
wires [48]. © 2020 IEEE

27

This ZIF connector bridged the recording electrodes via a printed circuit board (PCB; Gold
Phoenix PCB, China) to two Omnetics connectors (Omnetics Connector Corporation, Minneapolis,
MN) for direct connection to the electrophysiological recording system (Fig. 1-16). The acrylic
backing attached to the probe was then matched to the ZIF height and attached to the PCB's
backside through a double-side tape (Fig. 1-17). A single-PCB approach was developed to
miniaturize the size of the head-mounted electrical packaging for the long-term in vivo study. One
Omnetics connector was angled at 70° from the horizontal to minimize overall PCB size (Fig. 1-
17a and b). Built-in ground traces were added to decrease external noise sources from the ground
wires’ movement (Fig. 1-17c) [48].

Figure 1-18. Photo showing the surgical setup. The array is secured to the skull using dental cement in
preparation for recording (scale bar = 1 cm) [48]. © 2020 IEEE

28

1.2.4.3 In Vivo Recording from Rat Hippocampus
Probe arrays were implanted according to the guidance of the Institutional Animal Care and
Use Committee (IACUC) and the Department of Animal Resources of the University of Southern
California (USC). The surgical procedure used [34] is summarized here. Animals were
anesthetized with isoflurane during the surgery after a pre-implantation injection of ketamine and
xylazine. A 2 × 4 mm cranial window was made above the hippocampus of a Sprague-Dawley rat,
and the dura was carefully removed with forceps. Five screws were anchored into the rat skull
around the surgical window. The PCB was coupled to a stereotaxic with Parafilm for insertion.
Surgical implantation lasted approximately 2 hours after alignment and includes time for
electrophysiological monitoring of sites for characteristic compound spikes indicative of proper
placement in the hippocampus. Dental cement was applied to the insertion site up to the PCB to
secure the array (Fig. 1-18) [48].

Figure 1-19. Representative spike waveforms from multiple units were recorded from one lightly
anesthetized animal immediately after implantation [48]. © 2020 IEEE
29

During implantation of the array, neural signals from a single electrode were monitored with
an oscilloscope to appropriately locate the target by recording characteristic compound action
potential recordings. For chronic recordings, each subject was given 7-16 days post-implantation

Figure 1-20. Chronic recordings were obtained with a Parylene C multi-electrodes array. (a) Average noise
level (mean) and the spike amplitude (mean ± SD) of neural signals were recorded from four animals over
5 to 12 weeks post-implantation. (b) Representative recording from one animal at 12 weeks. The average
spike amplitude of the units recorded from a single neuron was 261.58 μV, and the 3-sigma of noise level
was 37.45 μV. A group of complex spikes is identified in the orange dashed box. (c) The top shows the
overhead frame capture of a rat running freely in an open field. The bottom two plots show the place field
of two units simultaneously recorded at 12 weeks post-implantation from the CA1 and the CA3 sub-regions
while the animal was running freely. The color bar represents the firing rate of neurons (in Hz) [48]. © 2020
IEEE

30

to recover, after which neural activities were recorded on a 64-channel data acquisition system
(Plexon Inc., Dallas, TX) while the animal freely roamed in an open field. Simultaneous recordings
from all 64 electrodes were captured at 40 kHz along with video recording (CinePlex, Plexon Inc.,
Dallas, TX) from the top of the open field to track the movement path of the animal [95]. Spike
sorting (Offline Sorter, Plexon Inc., Dallas, TX) was applied to the recorded data offline.

Figure 1-21. 48-week chronic recordings obtained with a Parylene probe showing the average spike
amplitude and average noise level over time (Courtesy of H. Xu).

The representative templates of spike waveforms of multiple units recorded from one
animal are depicted in Fig. 1-19; 5 out 8 shanks recorded from both the CA1 and the CA3 region
simultaneously, while the other three shanks recorded from either the CA1 or CA3 region due to
the curved structure of the hippocampus. Over 80 neuronal units were recorded in this experiment.
The results of long-term recordings are shown in Fig. 1-20a, with 75% (3 out of 4 rats) yielding
recordings with stable noise levels over five weeks. The probe acquired recordings from the rat
hippocampus for a maximum of 48 weeks with a steady average spike and noise level over time
(Fig. 1-21). Fig. 20b showed a representative recording through one channel from the CA1 and
31

CA3 regions of an animal at 12 weeks. The maximum spike recorded signal amplitude of 330.10
μV and a 3-sigma noise level of 37.45. The probe with shanks able to capture the signal from CA1
and CA3 regions was further used in the behavioral study with free-moving rats (Fig. 1-20c) [48].

Figure 1-22. Neural recordings acquired from 8 channels located on a single probe shank with crosstalk
(Courtesy of W. Jiang).

1.3 Improve Long-term Probe Performance
One factor influencing the probe performance in the long-term neural recording is the crosstalk.
Crosstalk is defined as the electrical signal leakage between channels caused by moisture intrusion.
As shown in Fig. 1-22, 8 channels on a single probe shank detected the same neural action potential
spike at a similar time point, reflecting a leakage in the system. Benchtop crosstalk measurements
with devices recycled from surgery and devices soaked in phosphate-buffered saline (PBS) were
32

taken to debug the crosstalk source and suggest solutions for reducing crosstalk over time.
Crosstalk was measured as the voltage leakage between channels. Before measurement, the tested
probe was removed from tissue or PBS buffer and dried up in the air. The probe was then connected
to an 8-channel data acquisition box with eight randomly selected consecutive traces (Fig. 1-23).
A 1 kHz, 0.5 V sine wave was sent to channel 1. The voltage was measured from channels 1 to 8,
one channel at a time, controlled by the automatic switch. The crosstalk for each channel was
calculated as the percentage of dividing the feedback voltage by input voltage.

Figure 1-23. Benchtop setup used to obtain crosstalk measurement.

Figure 1-24. (a) Result table showing a zero crosstalk between 8 selected traces. (b) Cross section of a
device illustrating channels with zero crosstalk and 50% crosstalk.

33

cross-talkn (%)=
V r ea d , n
V s en d
× 100 (%) (where n is the channel number)

For the ideal case with zero crosstalk, the channels were electrically isolated. Besides the
sending channel, no signal should be detected on other channels (Fig. 1-24 (a)). As shown in Fig.
1-24 (b), diagonal cells' crosstalks were acquired by measuring from the sending channel.
Therefore, they were 100%. All other cells should read 0 as no crosstalk was presented. Due to the
measurement resolution and embedded noise in the recording system, readings below 10% were
considered "zero" crosstalk.

Figure 1-25. Schematic showing the probe-brain interfaces while acquiring neural recording from a live
rat.
34

1.3.1 Debugging the Crosstalk Source
Fig. 1-25 illustrated a polymer probe inserted into the brain with a closed wound. The PCB
was fixed to the skull with dental cement. The possible crosstalk sources could be categorized into
three groups: the parts exposed in the air, covered by dental cement, and inserted in the brain.
To debug region (1), a probe observed crosstalk was recycled from surgery, and a cut was
made along the top yellow line shown in Fig. 1-25. The crosstalk was measured with the benchtop
testing setup before and after the cutting. As shown in Fig. 1-26, 7 consecutive traces showed
crosstalk right after the probe was recycled from surgery. The isolated omnetics connector and
PCB did not exhibit crosstalk, demonstrating proper channel isolation in the region (1).

Figure 1-26. Benchtop crosstalk measurement taken from probe recycled from surgery (top) and after the
omnetics connector and PCB were separated.
35

Figure 1-27. (a) Setup for the soaking experiment. (b) Probe soaked in the liquid vail. (c) Probes packaged
with dental cement and marine epoxy.

To determine if crosstalk was caused by moisture trapped in the region (2), probes
packaged with different materials were soaked in PBS. Crosstalk was monitored over time through
the benchtop testing setup (Fig. 1-27(a)). Waterproof LOCTITE
®
marine epoxy was used as the
control compared to the dental cement used in the animal study. There is less study on how well
the dental cement prevented water from entering the package. Parylene probes (n=4) were
connected with PCB and glued to a vial cap through dental cement to mimic the dental cement cap
Fig. 1-18, 1-27(b)).
In comparison, three Parylene probes were packaged with marine epoxy as the control. For
each probe, eight consecutive traces were selected and connected to the jumper cable through the
PCB. The vial was filled with 1X phosphate buffer saline (PBS) to maintain a moisture
environment for the electrodes. The vials were kept in a water bath heated to 37°C to mimic the
36

body temperature (Fig. 1-27(a)). The crosstalk was measured once daily for the first week and
weekly for later weeks.

Figure 1-28. Representative crosstalk results acquired with devices packaged with marine epoxy and dental
cement and soaked for two months.

37

Fig. 1-28 showed representative crosstalk data for devices packaged with dental cement
and marine epoxy soaked for up to 2 months. The marine epoxy device started with a short between
channels 1 and 3 (possibly caused by packaging) before being soaked, so the crosstalk shown on
these channels was not counted. Starting on day 50, the device packaged with marine epoxy started
to show crosstalk between channel 2 and channel 1, channel 2 and channel 3. In comparison,
devices packaged with dental cement did not show crosstalk until 54 days. In summary, the dental
cement showed similar performance to marine epoxy in protection from water intrusion. So far,
our benchtop crosstalk measurements suggested crosstalk did not occur by water intrusion in
Omnetics connector, PCB, or dental cement. Crosstalk is likely to occur by water intrusion in the
Parylene probe.

Figure 1-29. Schematic showing the cross-section of Parylene probe with crosstalk between traces (left);
without crosstalk between traces after surface treatment (right).

1.3.2 Reducing Crosstalk by Surface Treatment
Crosstalk is often caused by liquid trapped between Parylene-metal layers. To reduce
crosstalk over time, the approach taken here decreased the chance of forming water cavities
38

between layers by improving layer cleanliness and adhesion (Fig. 1-29). Parylene probes with
different surface treatments were soaked in 1X PBS at 37°C to mimic the in vivo environment (Fig.
1-30). 1X PBS was used because it contained similar chemical composition as the cerebrospinal
fluid. AdPro Plus
®
was an adhesion promotor that helped increase the bonding force between
Parylene-metal-Parylene, according to research conducted by J. Ortigoza-Diaz et al. in 2018 [60].
Hydrofluoric acid (HF) was commonly used to remove organic debris left from the
micromachining processes. After deposition of the base Parylene insulation and the metal layer,
the wafer was dipped in an HF bath to wash away the residue left on the wafer surface after the
metal lift-off process to reduce the chance of voids on the material surface. For the last group of
the probe, dehydration was done to remove moisture from the base Parylene layer to avoid

Figure 1-30. Schematic showing the cross-section of the device with different surface treatments.
39

moisture trapped in layers early. Devices without any of the surface treatments mentioned were
soaked as the control. All devices were soaked for up to 200 days, and crosstalk was monitored
regularly over time.

Figure 1-31. Graph showing crosstalk percentage over time for the device with different surface treatment
(N=3, except for control device N=1).

Figure 1-32. Schematic showing the resistivity measurement for probe with intact, broken, and shorted
traces.

40

In Fig. 1-31, any readings lower than 10% were considered in the green zone with
"neglectable crosstalk" and it is counted as the noise embedded in the measurement system. The
control device observed crosstalk the earliest on day 43. Next, the HF-treated devices started to
show crosstalk on days 50, 64, and 170. The dehydrated devices started to see crosstalk on days
84, 149, and 170. AdPro Plus treated device did not see crosstalk over the 198-day soaking period.

Figure 1-33. Resistivity measurement result for control and AdPro Plus treated device (highlighted in green
dashed box) soaked for up to 190 days.

To verify the neglectable crosstalk demonstrated by the AdPro Plus treated devices was
not caused by an open circuit, the resistance between the traces was measured with a DC
multimeter. As shown in Fig. 1-32, the circuit is equivalent to resistors connected in series for two
intact channels without crosstalk. The resistance for two intact traces was measured in a range of
12-14 MΩ. If two traces were no longer connected, the resistance was measured as infinitely large,
41

and the readout showed overload (OL). If there was crosstalk, two traces were partially shorted,
and the resulted resistance was measured as a value lower than 12 MΩ.
The resistivity measurement result is shown in Fig. 1-33. Trace 1 was selected as the
reference electrode for the control device, and the resistance was measured between trace 1 and 7
other traces. The Table showed the decreased resistance between channels 1, 2, 5, and 6 on day
44, which agrees with the data shown in Fig. 1-31, where crosstalk was first detected on day 43.
For AdPro Plus devices, the resistance between most channels showed measurement in the normal
range. However, starting on day 90, some open circuits between channels were observed. All other
traces were still connected and showed zero crosstalk agreeing with the data shown in Fig. 1-31.
Since the benchtop measurements required aggressive handling for repetitive crosstalk
measurement, the carefulness during those handling processes may cause some traces to break
after a 3-month soaking. In summary, the AdPro Plus treated device showed decreased crosstalk
over time compared to devices without or with other surface treatments.
1.4 Probe Insertion Study
A thin polymer probe often has the problem of penetrating through the brain surface without
buckling and inserting deep due to the low mechanical compliance. Designing probes to target
different brain structures with different mechanical stiffness requires understanding how probe
shape and dimension influence the insertion performance. Hence study on optimizing the design
for the intracortical probe is needed. Numerous research on the mechanical performance of
inserting neural probes was done to investigate different materials and probe designs, but most of
the study was done with the rigid probes having varied tip angles, shank dimensions, and pitch, as
42

Table 1-2. Summary of the mechanical study done with brain probes from other literature.

Probe
Material
Tip Angle Shank Dimension
(width x thickness)
Shank
Number/Shank
Pitch
Insertion
Depth
C. S.
Bjornsson et
al., 2006 [96]
Silicon 5-150° 100 µm x 60 µm 1 <2 mm
N. H. Hosseini
et al., 2007
[97]
Silicon 17° 120 µm x 100 µm Up to 10/ [550 –
1100 µm]
3-6 mm
S. Singh et al.,
2016 [63]
SU-8 Unknown 320 µm x [5 – 20
µm]
1 0-5 mm
Tyrosine-
based polymer
coated
Parylene C
Unknown [50 - 350 µm] x [50
– 100 µm]
1 0-5 mm
W. Jensen et
al., 2006 [53]
Silicon 4° [38 - 200 µm] x
[25µm]
1-8 shanks/ [200
– 600 µm]
<2 mm
Tungsten 3-10° [50 – 150] µm Ø 1 <2 mm
A. Andrei et
al., 2012 [89]
Silicon 10-50° [200 - 400] µm x
[50 - 150] µm
1 6 mm
Our Work Parylene C Single tip 45°,
double tip 45°
[100 - 150 µm] x [20
– 40 µm]
8/ [250 - 300 µm] 0 - 5 mm
43

Figure. 1-34. Photo showing the Neuronexus probe and Parylene C probe.

shown in Table 1-2. Most works with flexible probes only demonstrated successful neural
recording from animals by coupling the probe to a rigid insertion wire. Due to the method used to
insert the probe, mechanical study for the probe insertion is limited to the rigid probe or rigid
insertion aid coupled to the flexible probe. To minimize the tissue damage, a new method was
developed for insetting flexible probes without relying on a rigid insertion shuttle, including the
dissolvable PEG brace method described in previous sessions. With those novel methods, the
probes can be inserted in their “soft” condition. To ensure the flexible probe has enough
mechanical robustness to be inserted into the brain, work on investigating the mechanical
performance of flexible probes with different design parameters is needed.
The motivation here is to conduct the mechanical study with bare Parylene probes with
varied designs and develop a guideline for the community to optimize the design for probes made
from thin film to fulfill the surgical requirement for animal study for the case when an insertion
shuttle is not used. Work has been done to study the mechanical performance of silicon probes
with different design parameters including shank number, the distance between shanks, tip angle,
shank dimension, and shank shape [48, 53].
44

For the studies reported here, we first compared the insertion performance of probes made
from silicon and Parylene, as silicon has become a common material used to make neural probes
since 1985 [98]. Studies have shown that polymer probes result in less tissue damage than silicon
probes. Less study is conducted to directly compare the mechanical performance of inserting
silicon and bare polymer probe due to the limitation of the used insertion method described in the
previous paragraphs.
Besides the material of the probe build, the effect of the shank shape including tip angle
and shank dimension on the insertion force of the Parylene probe was also studied. Last, we
investigated which design parameters influenced the probe straightness during insertion as the
shank straightness is a vital parameter to investigate for checking the reliability and consistency of
probe insertion.

Table 1-3. Table showing the dimension and shape of Neuronexus probes versus Parylene sham probe.

The insertion, shear force, and dimpling were measured with the probe made from silicon
and Parylene C using the setup described in previous sections (Fig. 1-34). NeuroNexus silicon
probe is selected for study comparison with Parylene probe because it has been used extensively
45

in research labs for neural recording and stimulating from various brain structures [99-101]. Table
1-4 shows that the commercial silicon probe from Neuronexus had four shanks; each shank had a
thickness of 15 μm, a width of 95 μm, and a length of 2.8 mm. Parylene sham probes with a similar
dimension and shank shape were used and compared with silicon probes. However, there was a
slight difference in shank thickness and width, where the Parylene device width tapered from 100
to 60 μm and had a thickness of 20 μm. Both types of the probe had a tip angle of 45°.
As shown in Figure 1-35, the insertion force and dimpling for the Neuronexus probe were
higher than the force and compression acquired by the Parylene sham probe. The average insertion
force for Neuronexus was higher than the insertion force of the Parylene probe (no statistical
difference). Compared to insertion force and dimpling, the shear force showed a smaller difference
for probes made from different materials. The results demonstrated that the silicon probe
potentially caused more stress and compression to the brain surface when penetrating the surface.
To minimize the stress and compression to the brain surface during insertion, polymer probes with
different designs were further studied.

Figure 1-35. Graph showing force and dimpling measured with Neuronexus and Parylene C probe (mean
± SD, N=3).
46

Figure 1-36. Schematic showing probe with different shank shape and shank thickness.

Table 1-4. Table showing measured insertion and shear force for probe (A, B, and C) with different shank
shapes (mean ± SD, N=3). A, B, and C devices are the devices shown in Fig. 1-47 from left to right. The
probe with asymmetric and symmetric shanks has the same spacing between shanks.

Device ID Shank Shape Insertion Force
(mN)
Shear Force
Slope (mN/mm)
A Asymmetric 1.20±0.25 2.12±0.13
B Symmetric 1.31±0.27 2.14±0.16
C 2- 45° tip 2.04±0.34 2.38±0.24
47

The hippocampal probe (Device A in Table 1-4) mentioned in the previous chapters was
selected as the control group as the specific design has been demonstrated successfully in rodent
implantation presented in the previous work [34]. Besides the control group, Parylene probes with
shank designs differed from the hippocampal probe were studies. As shown in Fig. 1-36, shank
shapes, including asymmetric, symmetric, single-tip, and double-tip were investigated. To
determine the shank dimension's influence on insertion performance, shank thickness of 20, 25,
and 40 μm was studied. Also, the effect of shank width and pitch on insertion performance was
investigated.
Other groups have researched how the needle tip symmetricity influences needles'
mechanical performance while inserted into tissues [102]. The research showed that the needle
with asymmetric tips received more force on one side and was easier to bend. Since no similar
study was done with the brain probe, we were interested in seeing if the probe shank's symmetricity
would influence the brain probe's insertion performance. As shown in Fig. 1-36, both samples A
(asymmetric) and B (symmetric) had the same shank width of 110-150 µm and the same tip of 45°.
As shown in Table 1-4, by comparing Device A with B, the symmetry did not influence the
insertion force and shear force. By comparing Device B and C, the double-tip probe increased the
insertion force by 56 % but was still lower than the theoretical buckling force, so the probe could
penetrate the brain surface without buckling. The shear force was increased by 11% when the
probe tip doubled.
The relationship between shank width, pitch, and insertion performance was also studied.
As shown in Fig. 1-37, the force was measured to insert a probe with different shank widths and
spacing. More specifically, probes having a shank width of 60-100 µm, 110-150 µm, and shank
spacing of 250 µm and 300 µm were studied. Device D, compared with Device A, the width
48

increased from 60-100 µm to 110-150 µm, and the insertion force and shear force were
approximately the same. In other words, the 50 µm width increase did not influence the insertion
and shear force. By comparing Device A and Device E with the same shank width but different
shank pitch, the insertion, and shear force was roughly equaled. In summary, neither the 50 µm
shank width nor pitch increase influenced the insertion and shear force.

Figure 1-37. Measured insertion and shear force acquired with the probe with different shank width and
pitch (mean ± SD, N=3).

Probes with different shank thicknesses were also studied. Probes with shank thicknesses of
20 µm and 25 µm were fabricated and used for the mechanical study. The 40 µm thick device was
also tested by attaching two 20 µm thick devices together with drops of IPA. The probe was dried
and inserted into the agarose brain phantom to acquire the insertion force.
49

Table 1-5. Measured insertion and shear force collected with the probe with different shank thicknesses
(mean ± SD, N=3).

In Table 1-5, as the probe thickness increased from 20 µm (Device A) to 25 µm (Device F),
the insertion force stayed roughly the same, and the shear force increased by 28%. As the thickness
increased from 20 µm to 40 µm, the insertion force remained approximately the same; the shear
force increased by 33%. However, the buckling force significantly increased as the critical
buckling force is proportional to the cube of the probe thickness. According to Euler's buckling
equation, when the thickness increases by 2-fold, the buckling force theoretically increases by 8-
fold. As a result, the probe with increased thickness is more likely to penetrate the brain surface
without buckling.
As shown in Fig. 1-38, shank thickness helped to improve the shank straightness inserted
into the gel, as the 40 µm thick probe deviated less from the probe midline (79 µm) compared to
the 20 µm thick probe (134 µm) at an insertion depth of 4 mm. The double-tip shank geometry
also helped improve the shank straightness as its shanks deviate less (59 µm) from the midline
than the single-tip probe (134 µm). It was surgically acceptable to target our interesting brain
regions when the probe deviated within 150 µm from the probe midline. The probe design resulted
50

in less splitting. Straighter shanks will be more likely to benefit from the deeper insertion, as the
shank deviated from the midline more as inserted deeper.

Figure 1-38. Photo showing a representative side view of an 8-shank probe inserted into agarose phantom
highlighting shank straightness of probe shanks with different thickness and shank shape to a depth of 4
mm.

1.5 High-density Probe Through Three-dimensional Stacking
Three-dimensional (3D) or stacked neural probe arrays feature higher channel counts and
the ability to target larger anatomical areas than planar or two-dimensional arrays. Three-
dimensional arrays can be built as a single device, such as the Utah array [103, 104], or by aligning
several individual neural probes (Fig. 1-39); the latter approach has been used for years using
traditional microwire and silicon-based probes [6, 105-107]. Flexible and soft-neural probes built
from polymer substrates are suggested to reduce acute injury during implantation in the brain and
chronic irritation following implantation [17, 23, 24]. However, accurate implantation of polymer-
51

based neural probes is difficult due to their tendency to bend or buckle during implantation. Many
groups inserted flexible probes by coupling the probe to the rigid support [26]. The rigid support
is usually designed to match the specific geometry and dimension of the polymer probe. Extensive
work is required to create the insertion support and align the support with the polymer probe [66].
Also, stacking causes more damage to tissue as more rigid shuttles are inserted along with polymer
probes. As a result, scaling up the recording volume for the polymer probe is limited. Other groups

Figure 1-39. Schematic showing the method of increasing the recording volume of the probe by 3D
stacking. For the planner 8-shank probe shown on the left, each shank has 8 electrodes, so a single planar
probe has a total of 64 channels on 8 shanks. By stacking 8 of the probes together (right), there is 64 x 8, a
total of 512 channels.

are designing automatic robots to quickly insert small and flexible probes multiple times to target
more recording regions [108]. However, those insertions are limited by inserting only a single-
52

shank probe coupled to a rigid insertion guide wire at a time. Inserting multiple probes increased
surgery time. These methods are not simple or broadly adaptable for most researchers. Using
silicon/metal insertion shuttles and robotic implantation can increase acute injury. The custom
backbones and supporting materials are not easily transferable to different designs. In summary,
three-dimensional polymer probe arrays face challenges both in package and implantation. There
is no well-developed technique for creating a 3D polymer neural probe with varied shank shapes.
This project aims to develop a universal method of inserting multiple polymer probes in parallel
to enable multi-region brain recording without using the rigid shuttle.
1.5.1 Insertion Method
The method of inserting three-dimensional stacked probes is developed based on the method
used to insert single planar probes described in the previous chapter. The probes are stacked with
PEG spacer where the PEG brace acts as the adhesive layer for stacking and serves as the insertion
aid. Based on the same mechanical fundamentals, the PEG brace temporarily decreases the probe
length to increase the buckling force. The shorted probe ends up with a buckling force greater than
the insertion force. This will result in successful probe insertion without buckling.
1.5.1.1 Packaging 3D Probes with PEG Spacer
PEG spacers are created in a molding process using machine-cut PDMS layers with a
cavity featuring the PEG spacer's shape (Fig. 1-40). PDMS sheets (0.5 mm thick) are prepared
with a vinyl cutter to create the desired molds. The first PDMS layer creates a PEG brace attaching
the first Parylene C array to an acrylic backer. The second PDMS layer defines a PEG spacer
between the two arrays (0.5 mm thick), and the third PDMS layer encases the 2
nd
array in PEG,
53

ensuring the entire structure is symmetric. The tips of the Parylene C shanks are left exposed.
Liquid-phase PEG is poured into the mold within a 60 ºC oven and solidified after cooling to room
temperature. The process is performed sequentially, with each layer allowed to cool before the
subsequent layer is deposited.

Figure 1-40. Schematic showing the multiple-layer PDMS mold used to create 2x8 (16-shank) 3D arrays.

Scaling up the probe from a 2D planer to a 3D matrix increases shank count, increasing the
amount of force applied to the brain, risking injury. Dense spacing of shanks can increase the risk
of increased dimpling on the brain surface. Polymer probes are at risk of buckling or curving during
implantation, making accurate anatomical targeting impossible. Few studies on the mechanics of
flexible probe insertion, and few studies focus on flexible probes stacked into 3D arrays.
For a probe inserted into the brain without bucking, the force required to penetrate the brain
surface needs to be higher than the buckling force [48]. According to Euler's equation, the buckling
54

force can be increased by temporarily decreasing the length. The method of inserting a 3D probe
was inspired by the insertion method used to insert the planar 8-shank probe mentioned in the
previous chapter. Polyethylene glycol (PEG) was used to brace the Parylene C arrays to help insert
and space separate arrays at a fixed distance to create a 3D array.
1.5.1.2 Mechanical Evaluation of 3D Probe
PEG brace increased the buckling force by temporally decreasing the shank length during
insertion. This method was successfully demonstrated with a 2D planar probe in the previous
chapter. As the probe scales from 2D to 3D, the shank number increases, and both the buckling
and insertion force should increase theoretically, with the buckling force being higher than the
insertion force. To test the PEG method's feasibility in inserting a 3D stacked probe, buckling and
insertion force with a probe having different shank numbers were measured. Sham probe arrays
with identical geometry of the hippocampal probes described in the previous chapter, but without
any metal features were prepared. The focus of the work presented below is to investigate the
insertion method for 3D stacked probe, and no neural recording is performed.
The Parylene planar probe consists of 8 shanks with a tip-to-tip spacing of 250 µm. Each
shank is 5.5 millimeters long, 150 µm wide, and terminated with a 45° angle. The shank number,
width, and spacing are designed to span a 2000 µm distance along the septotemporal axis. Multiple
single-row planar probes are physically stacked together to increase the spanning distance along
the axis perpendicular to the septotemporal axis (Fig. 1-39). For instance, a 6x8 and 8x8 probe
created by stacking eight 8-shank probes with a device-to-device spacing of 500 µm span a 2000
µm distance along the septotemporal axis and 2620 µm (Fig. 1-41) and 3660 µm (Fig. 1-42)
distance along the other axis perspectively.

55

Figure 1-41. Photo showing the 6x8 3D 2.8 mm long Parylene sham device packaged with double-sided
tape.

Figure 1-42. Photo showing the 8x8 3D 2.8 mm long Parylene sham device packaged with double-sided
tape.
56

2D and 3D arrays of 20 µm thick Parylene sham device with various shank numbers (1, 8,
16, 24, 32, 48, and 64) were driven into the load cell, and 0.6% agarose gel was placed on the
acrylic plate to measure buckling force, dimpling, insertion force, and shear force. For force
measurement, probes were shortened with tape from a length of 5.5 mm to 2.8 mm. The 3D probe
was made by stacking multiple 2D 8- shank planar arrays with the 250-µm thick double-side tape
to leave an effective length of 2.8 mm. The 250-µm spacing between individual 2D planar arrays
was selected to match the tip-to-tip spacing between shanks from a single 2D array. A motor drove
the probes into the load cell or agarose gel to measure the buckling and insertion force at a speed
of 0.01 mm/s. The setup was installed on a pressurized vibration isolation table to reduce the
background noise. A dino camera (Microscope, USB, portable, Dino-Lite) was used to record the
insertion to investigate when the probe shank penetrates the gel surface.
2.8 mm long 3D Parylene probes were inserted into 0.6% agarose without buckling. As
probes scale from 2D with 1 shank to 8 shanks to 3D with 32, 48, and 64 shanks, the buckling
force consistently exceeded the insertion force (Fig. 1-43). The difference between the buckling
force and insertion force for 32, 48, and 64 shanks was more significant than the difference for 1
and 8 shanks. It reflected the method of temporarily shortening the probe shank length with a PEG
brace was promising to insert a 3D probe with more probe shanks.
As the probe scales from 2D planar to 3D, the dimpling and insertion force are linearly
correlated to the shank number with an R
2
value greater than 0.99. Similar to other research work,
the dimpling and insertion force show closed correlation, and these two parameters follow a similar
linear trend as the probe shank dimension changes (Fig. 1-44, 1-45) [109]. Similar to dimpling and
insertion force, the shear force increased roughly linearly (R
2
≥ 0.99) as the shank number
57

increased from 8 to 16, 24, 32, 48, and 64 (1-46). The results reflected no extra stress being added
to the brain while inserting the 3D probe than the 2D probe.

Figure 1-43. Graph showing the buckling and insertion force (Mean ± SD) for probe with shank numbers
of 1, 8, 32, 48, and 64 (n = 3 for 1, 8, 32, and 64 shanks; n = 2 for 48 shanks).

58

Figure 1-44. Graph showing the mean and standard deviation of dimpling for probe with shank numbers
of 8, 16, 24, 32, 48, and 64 (n=3).

59

Figure 1-45. Graph showing the mean and standard deviation of insertion force for probe with shank
numbers 8, 16, 24, 32, 48, and 64 (n=3).

60

Figure 1-46. Graph showing the mean and standard deviation of shear force for probe with shank numbers
of 8, 16, 24, 32, 48, and 64 (n=3).

1.5.1.3 Insertion Test of 3D Probe with Fresh Rat Brain
The force measurement from previous sections demonstrated the feasibility of inserting a
3D probe with dissolvable PEG braces. However, as the multiple probes are stacked with PEG
spacers, the PEG melting profile is unknown as the volume increases, which may cause problems
during probe insertion. The PEG spacers with different shapes and molecular weights were tested
in fresh rat brains.
61

Figure 1-47. Picture showing a 3x8 (24-shank) probe successfully packaged with the PEG spacer. The
spacer’s shape and dimension are labeled in the right image. The spacer with the PEG cutout (right)
corresponds to PDMS layer 1 in Fig. 1-40.

First, a stair-shape PEG mold is used to pack a 3x8 (24 shanks) probe matrix with PEG (MW:
3.35 kDa) (Figure 1-47). The probe shanks are covered with the PEG braces to leave a 1 mm
exposed length. The probe was then inserted into the rat’s brain following the procedure described
in the previous chapter about inserting a 2D planar probe. The animals were anesthetized, and a
cranium window was made to expose both left and right lobes. The first exposed 1 mm length was
inserted into either right or left lobe. Saline is added to dissolve the PEG brace to expose more
probe length and inserted deeper.
While inserting the probe, an uneven PEG melting profile was observed with a slower
melting rate for the PEG located in the center of the blocks. The uneven molten PEG caused an
earlier exposure (or more exposed length) for the probe shanks close to the PEG block's edge.
Those over-exposed probe shanks had a high risk of buckling on the brain surface. Moreover,
62

during the insertion of this specific 3x8 probe, significant dimpling on the brain surface was
observed.

Figure 1-48. Processed X-ray image showing the brain lobe where the 3x8 probes were inserted. 2 out of
3 probes show the shank trace along the insertion path. The enclosed window shows the measured depth of
the probe inserted.

After the probe was inserted to the target depth, the probe shank was cut and left inside the
brain. The brain tissue is extracted and prepared for micro X-Ray imaging to inspect the location
of the probe shanks inside the brain. The brain tissue was soaked in potassium triiodide (I3K)
63

overnight to prepare the brain for imaging. On the imaging day, the half lobe containing the probe
shanks is placed into an imaging tube on a rotational stage. Four X-Rays scans were acquired for
every 0.7° rotation. Over 300 2D X-rays were taken and stacked to form a final image showing
the probe shank traces in the brain tissue (Fig. 1-48). Two out of three inserted shanks are visible
in the X-Ray image; the inserted depth was measured at 2.6 mm and 3.0 mm from the brain surface.
This is equivalent to a final insertion depth of ~3.6 mm and ~4 mm for these two visible probes
counting the shrinkage of the brain after being soaked in I3K overnight. This insertion test approves
that the PEG method is feasible for inserting multiple probes in parallel to the target rat’s
hippocampus, commonly reported at a depth of 4-5 mm [34].

Figure 1-49. Picture showing a 4x8 (32-shank) probe successfully packaged with the 4mm wide rectangular
PEG spacer. The probes are attached to a rigid acrylic backing (highlighted with a blue dashed line) with
PEG. The shank tips are shown (highlighted in blue lines) in the picture taken from the side of the device.
64

To reduce the effect of uneven PEG molten profile to insertion, the shape of the PEG brace
was modified. A 4x8 (32-shank) probe was packaged with a 4 mm wide rectangular PEG spacer
(MW: 2 kDa) with a 1 mm length exposed (Figure 1-49). The 32-shank probe was successfully
inserted into the rat’s brain to a depth of 4 mm with relatively lower dimpling on the brain surface.
However, the assembled probes started to detach from the acrylic backing and deviated from the
initial insertion path when the probe reached a depth of 4 mm (Figure 1-50). This detachment
between the shank and acrylic back is caused by fast PEG dissolvement around the acrylic support,
stopping the probe from going deeper.

Figure 1-50. Pictures showing 4x8 (32-shank) probes stopped at an insertion depth of 4 mm caused by the
detachment of the probes from the acrylic support. The photos are taken from different angles
demonstrating the detached probe (highlighted in green) deviated from the initial insertion path (highlighted
in blue).

The PEG mold is further optimized to add more PEG volume around the acrylic backing to
address the above problem. The complete assembly with the optimized PEG mold is depicted in
Fig. 1-51 showed a full 2x8, 16 shank probe assembly successfully packaged with the modified
PMDS mold. The method is expected to work with bigger stacks up to 64 shanks.
65

Figure 1-51. (a) Schematic showing the dimension of the 1st PDMS layer (Fig. 2-1(2)). (b) Bottom view
of 3D matrix probe released from the PDMS mold with shank tips pointing outward. (c) Top view of the
3D matrix probe. (d) Side view highlighting two Parylne probes sandwiched between three 0.5 mm thick
PEG spacers.

Next, the insertion of a 4x8 (32-shank) probe (packaged with the new PEG mold) into a fresh
rat brain was demonstrated. The animals were anesthetized and sacrificed. The brain was removed
after the rat was sacrificed and placed in an aluminum beaker filled with saline. The brain was
fixed to a foam board glued inside the beaker. A 4x8 probe was sandwiched between PEG braces
and left with a 1 mm effective length for initial insertion. The packaged probe was attached to the
stereotaxic frame for insertion (Fig. 1-52). After the insertion, the brain was dehydrated with 18%
66

sucrose solution and sliced for histological staining. The histology showed the insertion wound
matched the probe footprint at a depth of 1.2 mm, located in the cortex region (Fig. 1-52).

Figure 1-52. Photo showing the mock insertion experiment setup with rat brain (left). A histological slice
was taken at a depth of 1.2 mm (right) (Courtesy of W. Jiang).

1.5.2 In Vivo Recording with the 3D Probe
After a 4x8 probe was successfully inserted into a fresh rat brain and reached the cortex
region, a 3D probe array (2x8) was prepared for acute recording using the PEG spacer method
described above. Before the test in vivo, each of the 128 electrodes from the 2x8 3D matrix probe
was characterized through cyclic voltammetry (CV) and electrochemical impedance spectroscopy
(EIS) with a Reference 600 potentiostat (Gamry Instrument, Warminster, PA) in a Faraday cage.
67

For CV, the current was measured between the individual electrode and a reference Ag/AgCl
electrode (3M NaCl, BASi, MF-2052, West Lafayette, IN) in 0.05 M H2SO4 solution at the
frequency of 1 – 10
6
Hz (30 cycles, -0.2 to 1.2 V, 250 mV/s ). A 1 cm
2
platinum electrode was
used as the counter electrode. The EIS measured individual electrodes' impedance in 1X PBS at
1-10
6
Hz (25 mVRMS). Only the electrodes with less than 2 MΩ impedance at 1 kHz were used
in the animal study.

Figure 1-53. (a) Schematic showing the dimension of the PCB design. (b) Fully packaged 2x8 arrays with
PCBs. (c) Back view of the completed PCB package highlighting the acrylic backing.

Electrical connections to fully functional probes were established using zero-insertion force
connectors on custom-made PCBs. The PCB routed analog signals from each probe array to a pair
of Omnetics connectors. For the 3D array, two identical PCBs were adhered together using epoxy.
The acrylic backer was taped to the PCB's underside to support the PEG-braced polymer probes
(Fig. 1-53).
68

Figure. 1-54. Graph showing a representative neural recording acquired from the cortex region through a
channel of the Parylene probe during surgery.

The probe was implanted to a depth of 2.3 mm counting from the brain surface and acquired
neural recording from the primary somatosensory cortex. Figure 1-54 showed a representative
neural signal captured during the surgery, with an amplitude of 430 μV and a noise level of 50 μV.
Not all 16 shanks successfully captured neural signals; 8 electrodes from 4 probe shanks recorded
neural signals (Fig. 1-55). Neural signals were recorded from 17 neurons through electrodes
arranged on four shanks (Table 1-6).

Table. 1-6. Table showing the neuron number captured by electrodes aligned on four probe shanks.
Electrode # 5 6 7 8 13 15 17 32
Neuron # 1 1 3 3 1 3 3 2
69

Figure. 1-55. Schematic demonstrating the location of the 16 shanks inserted into the right lobe of a rat
(the figure is not drawn in actual scale). The bottom array is the one closed to acrylic backing. 4 (highlighted
in blue) out of 16 shanks successfully recorded neural signals. The right figure shows the shanks with
electrodes (highlighted in blue dots) that successfully captured neural signals.

After successfully acquiring neural signals from the cortex, another 2x8 probe was inserted
into the rat to obtain recordings from a deeper brain structure, the hippocampus. The probe was
successfully inserted into the right lobe of the rat. The insertion site was posterior to the bregma
and 2.4 mm lateral to the midline at a depth of 4.4 mm. The probe was aligned to the midline with
a 30-degree angle to match the hippocampus's anatomical location.
A 2x8 (16-shank) probe was successfully inserted into the brain to a depth of 4.1 mm without
buckling. The neural recording spikes were acquired from the hippocampus through two electrode
groups on the front and back probe arrays. Neural signals from 21 neurons were recorded from 8
channels, where the most recorded signals are captured from the bottom clusters of the electrode
(Fig. 1-56). No signals were recorded from the top groups of electrodes.

70

Figure. 1-56. Schematic demonstrating the location of the 16 shanks inserted into the right lobe of a rat for
targeting the hippocampus. The figure is not drawn in actual scale. The bottom array is the one closed to
acrylic backing. Electrodes (highlighted in green dots) located on four of 16 shanks successfully recorded
neural signals.

The electrode arrays were recycled from the animal after acute recording. EIS was taken on
a randomly selected electrode picked from those top electrode clusters, which did not detect neural
signals. The result was compared with the EIS taken on the same electrode before being implanted.
As shown in Fig. 1-57, electrode 50 from the top array (Fig. 1-56) had an impedance in the
operating range but higher (733 kΩ vs. 643 kΩ) at 1 kHz after being explanted from an animal.
This approved the electrodes are functional during the implantation. This increased impedance
may be caused by the left-over tissue attached to the recording sites.
Histology was taken to inspect the probe shunt’s location in the brain. Figure 1-58 showed
that the front and back probes were relatively parallel at a depth of 2.35 mm. 15 of 16 shanks were
visible in the histological slice. When it reached a depth of 4.30 mm, the distance between the two
probes decreased, and the side of the probe started to touch with others. At this depth, 10 of 16

71

Figure. 1-57. Graphs showing the EIS result of the same electrode (50 from a top array in Fig. 1-56) before
(top) and after (bottom) surgery. The impedance at 1 kHz increased from 643 to 733 kΩ (highlighted in
yellow dashed line).

Figure. 1-58. Photo showing the histology of the brain tissue taken at a depth of 2.35 and 4.30 mm. The
inserted shank left holes in the tissue (highlighted by red arrows) (Courtesy of W. Jiang).
72

shanks were visible, demonstrating that some probe shanks began deviating from the original path
when inserted deeper. This resulted in the probe tips reaching different target depths. The deviation
of shanks toward others may also cause the blockage of the recording electrodes, which stops those
covered electrodes from reading neural signals.

Figure. 1-59. Photo showing the alignments of two stacked probes with electrodes facing the same direction
(left) and opposite (right).

To address this problem, another 2x8 (16-shank) probe with the recording sites facing
outward was prepared and packaged (Fig. 1-59). The neural recording spikes were successfully
acquired from hippocampus subregion CA1 and CA3 through all four electrode groups (Fig. 1-60
(a)). There were 12 out of 16 shanks that successfully captured neural signals. Fig. 1-60 (b) showed
the representative neural spike acquired from CA1 and CA3 through each electrode group. The
amplitude and noise levels were in a typical range compared to the neural signal obtained by
microwires.
73

Figure 1-60. (a) Schematic showing the shanks and channels acquired neural signals from the
hippocampus. (b) Representative neural signals were obtained by four different groups of electrodes from
four channels.

1.6 Scaling Up the Probe for Application in Larger Animal
Most work on intracranial probe development is demonstrated in rats with probe shank
length < 10 mm. To translate the recording device to human application, the probe is often needed
to be first tested in larger animals, including pigs and sheep, or even non-human primates. To
target the deep brain structure such as the hippocampus in those large animals, the probe length is
often needed to be many times longer than the probe used for rodents, in tens of mm long. For
74

instance, a rat’s hippocampus is located ~4 mm below the skull surface. In comparison, the sheep’s
hippocampus could locate at a depth of ~15 mm. Most existing intracranial probes designed for
application in larger animals focus on shallow cortical regions due to the limitation of inserting
the probe deep. Few probes target large animals' deep brain regions, such as the hippocampus. The
major challenge with scaling up the probe construction’s flexibility and size for limiting the
damage caused to the surrounding tissue. Meantime, it is challenging to insert long slender probes
deep. A reliable insertion method capable of inserting long and flexible probes is needed to target
those deep brain regions in the large animal while minimizing the tissue damage for the long-term
application. This work aims to design a flexible brain probe that overcomes the surgical placement
challenges to acquire recording from deep brain structures in larger animals such as sheep.
1.6.1 Probe Design
To target the sheep’s hippocampus, the intracranial probe needs to have a minimum length
of 15 mm. Since the focus is to investigate a reliable method for inserting a flexible probe with
varied lengths, only a sham device with a single shank without metal traces is fabricated and tested
in a brain phantom made from agarose. Flexible Parylene probes with 10 mm and 15 mm lengths
are made, which are ~2X and ~3X long compared to the rodent hippocampal probes described in
the previous chapter. Similar to the brain probe shape to the rat’s probe described in the earlier
sessions, the probe has a tip angle of 45 ° and a tapered shape with different widths (300 µm, 140
µm, and 100 µm) along the shank. The probe has a similar thickness of 25 µm for 10 mm long and
20 µm for 15 mm long as the rodent hippocampal probe (20 µm) (Figure 1-61). The probe with
different widths is also fabricated for testing different insertion methods.
75

Figure 1-61. Photo showing the geometry and size of the 10 mm and 15 m long probe. For a 10 mm long
probe, section A has a width of 300 µm over a probe length of 8.5 mm. A 15 mm long probe has a width
of 300 µm over a probe length of 13.5 mm.

1.6.2 Insertion Method
Sham Parylene probes without a metal trace are fabricated for investigating different
insertion methods (Figure 1-62). Similar to the rat probe, insertion with a dissolvable PEG brace
was first tested. According to Euler’s buckling equation mentioned in the previous sessions, a 25
µm thick and 110 µm wide probe with 2.8 mm effective length can withstand ~ 0.5-1 mN of force
without buckling, the commonly reported insertion force for penetrating the cortical tissue with
dura removal [48, 52, 53]. A 10 mm probe was packed between two 250 µm thick PEG braces
(molecular weight: 3350) with a 1 mm exposed length (Figure 1-63). Referring to the previous
chapters, the probe designed for rodents was usually left with a 1 mm length for in vivo experiments
instead of 2.5 mm accommodating the challenge of insertion with the uneven surface of the brain
[48].
To closer mimic the insertion condition with animals, the insertion test with agarose phantom
is performed with a probe with an effective length of 1mm. After the probe is packaged with the
76

molding method described in previous sessions, the first 1 mm exposed probe length is inserted.
Saline is added to dissolve the PEG brace to expose more probe length. The new exposed length
is inserted. The steps were repeated until the probe reached the target depth.

Figure 1-62. Photo of sham Parylene probe with different lengths.

Figure 1-63. Photo of a 10 mm long Parylene probe shortened with PEG brace.
77

Figure 1-64. Photo of a 10 mm long Parylene probe inserted in agarose phantom. Left: front view: right:
side view.

The problem with inserting a long probe (> 5 mm) with the PEG method is that the probe
tends to deviate from the original insertion path when inserted deeper (3 out of 3 insertions). A 10
mm was inserted into a 0.6% agarose phantom with significant curvature (Figure 1-64). With the
curvature, a 10 mm long probe reached a maximum depth of ~ 8.5 mm. The considerable
displacement of the probe tip from the original insertion path makes it challenging to target the
specific brain region. Other insertion methods were investigated to address the problem, including
insertion with metal wire and a “self-support” soft shuttle.
A research group has demonstrated inserting a polymer probe in the depth of 7 mm in rats
by attaching the probe to a rigid tungsten shuttle [47]. Using a similar approach, a steel wire is
used to insert our probe. A 10 mm long probe with a shank width of 200 µm is dip coated in PEG
(molecular weight:3350) solution for once. A stainless-steel wire with a 250 µm diameter is dip
coated in the same PEG solution for once. Two were then aligned under a microscope. Drops of
78

IPA are added to fuse the PEG-coated probe with PEG-coated stainless-steel wire. The assembly
was then dipped and coated in PEG solution for one more time. The insertion result showed the
probe could not be inserted at a 10 mm depth because the probe separated from the insertion shuttle
at a depth of around 8 mm. The bonding of the polymer probe and metal insertion shuttle must be
further optimized to achieve deep insertion to avoid the probe being detached from the shuttle
before reaching the target depth. Research has shown adding a sleeve at the probe tip helps to bond
the polymer probe to the rigid guide through insertion. The sleeve can be a three-dimensional cone
[110] or a hole fabricated to the end of the probe shank [28]. Further testing with the rigid shuttle
method required a customized designed probe with the structure to secure the insertion aid.
Another method without the need to modify the probe design is investigated and discussed below.
The insertion method involved a “self-supported” soft shuttle that used the probe itself as
the support for insertion to double the overall thickness during insertion by increasing the cross-
section of the probe; no additional modification on fabrication is needed. According to Euler’s
equation described in previous sessions, as the thickness doubled, the critical buckling force
increased by 8x when other design parameters stayed the same. A 50 µm thick (formed by stacking
two 25 µm thick devices) and 110 µm wide probe with 10 mm effective length can withstand ~
0.5-1 mN of force without buckling, the commonly reported insertion force for penetrating the
cortical tissue with dura removal [50, 52, 53].
Table 1-5 shows the insertion force of the probe with different thicknesses. As the probe
shank thickness doubled from 20 µm to 40 µm, the insertion force stayed around the same. As a
result, more significant differences between the buckling and insertion force caused by stacking
two probes decreased the change of probe buckling while penetrating through the brain surface.
79

To test if two stacked probes can be inserted into agarose phantom as suggested by the
mechanical test result, two identical probes are stacked together with drops of isopropanol or
deionized water under a microscope and coated with a thin layer of PEG (molecular weight of
8000) for one time (Figure 1-65). The PEG was added to ensure the probes stayed together through
the insertion. The result showed this method successfully inserted two stacked 10 mm long probes
(Figure 1-66) and two stacked 15 mm probes.

Figure 1-65. (a) Photo of two 15 mm long Parylene probes stacked together with drops of IPA. (b) The
stacked probe probes, after being dipped coated in PEG solution and solidify under room temperature.

80

Figure 1-66. (Left) Photo of probe insertion setup. (Middle) The side view of two stacked probe probes
inserted into 0.6% agarose phantom to a depth of 10mm. (Right) The side view of the inserted probes.

Unlike the conventional way of inserting deep requires coupling the polymer probe to a rigid
metal shuttle which creates severe damage during insertion, the soft-shuttle method described here
does not increase Young’s modulus of the probe and limits the footprint caused by the insertion.
The PEG coat can be father optimized for accommodating different insertion depths. With the
minimized footprint resulting from insertion, this insertion method can potentially be used to
increase the recording channels or achieve stimulating and recording simultaneously. The second
probe used as the insertion guide can have active electrodes for recording for stimulating. When
81

two probes are stacked together with the electrode side facing externally, the number of electrodes
will be doubled.
1.7 Summary
A flexible, penetrating Parylene C-based neural probe array whose electrodes conformed to
hippocampal anatomy was designed, fabricated, characterized, and implemented in vivo. In
particular, probe arrays' mechanical properties and performance were investigated, and the
polymer material motivated the use of a novel bracing method to overcome premature mechanical
buckling. The PEG bracing strategy overcame challenges of minimally disruptive surgical
implantation of soft polymer probes and could be generalized to other designs, whether single or
multi-shank. Arrays were successfully implanted in agarose phantoms and rat brains using this
bracing approach; bare shanks could be placed > 4 mm below the brain surface in a deep brain
structure. The PEG brace temporarily shortened the probe length to allow implantation and access
to hippocampal structures. Simultaneous recordings from multiple electrode sites captured
complex spikes that are characteristic of pyramidal cells in both the CA1 and CA3 regions of the
hippocampus for a maximum of 48 weeks. Individual shanks inflicted minimal tissue disruption
and left tracks within the tissue matching the bare probe cross-section.
The long-term performance of the probe soaked in phosphate buffer saline was investigated
to correlate with the long-term recording acquired with the probe from the rats. Channel leakage
between channels was detected and defined as crosstalk. The crosstalk is defined here as when
neighbor channels capture similar neural signals due to the moisture intrusion between the
Parylene-metal-Parylene layer, causing a short between channels. Parylene probes with different
surface treatments were investigated. The result suggests applying surface adhesion promotors
82

such as Adpro Plus improved the adhesion between layers and further reduced the crosstalk over
time.
To expand the neural probe’s access for use in targeting different brain structures or
different species, the probe length needs to be longer. The insertion of those slender flexible probe
could be challenging. 10- and 15 mm long sham Parylene probes are made to investigate the new
methods that could be used to insert probes longer than 10 mm. A novel method of inserting the
probe with the “soft shuttle” is demonstrated. It is achieved by physically stacking two flexible
probes together and insert. This method does not require modification on fabrication and can be
easily adapted for probes with different designs. Future work will include further optimizing the
“soft shuttle” insertion method for accommodating probes with different lengths and testing the
probe insertion with animal brains.
A simple method for stacking 2D polymer neural probe arrays to create 3D arrays was
successfully demonstrated to insert into the agarose brain phantom. Mechanical data of probes
with different shank numbers showed the possibility of using such a method to insert a probe with
a maximum of 64 shanks. Archiving bigger stacks with more probe arrays will increase the overall
volume of the added PEG which will take more time to dissolve. Additional work needs to be done
to optimize the minimal amount of PEG required for each individual spacer. If PEG volume
required for the spacer decreased, it is possible to extend the capability to more than 64 shanks.
Preliminary animal data showed a 2x8 (16-shank, 128-electrode) probe could be successfully
inserted into the brain to different depths up to 4.4 mm. The probe acquired neural signals from
the live animal from multiple brain regions, including the cortex and hippocampus. The PEG mold
needed further optimized to scale up the probe shanks with more shank numbers for animal study.
The surgery time depended on the PEG block's melting time, and thicker blocks would increase
83

the surgery time. The optimization can be done by changing the PEG block's size, thickness, and
molecular weight to fulfill the surgical requirement. Future work includes further optimizing the
PEG probe for stacking different numbers of 2D planar arrays and accommodating the different
insertion depths. The electrical package of the 3D stacked project also needs to be further
miniaturized for lighter weight for chronic use.
Additional future work can focus on increasing the probe’s functionality beyond neural
signal recording, a microfluidics channels could be integrated to the existing Parylene probe
through thermoforming (Appendix A-1). Unlike the conventional fabrication processes involving
the long dissolving time of photoresists, the preliminary result demonstrated that microscale
channels could be potentially made by thermobonding prepatterned Parylene layers.

84

1.8 References
[1] B. Voytek. "Are There Really as Many Neurons in the Human Brain as Stars in the Milky
Way?" https://www.nature.com/scitable/blog/brain-metrics/are_there_really_as_many/
(accessed.

[2] CK-12. "17.12 Communication Between Neurons - Advanced."
https://www.ck12.org/book/ck-12-biology-advanced-concepts/section/17.12/ (accessed.

[3] A. E. Pereda, "Electrical synapses and their functional interactions with chemical
synapses," Nature Reviews Neuroscience, vol. 15, no. 4, pp. 250-263, 2014/04/01 2014,
doi: 10.1038/nrn3708.

[4] L. R. Hochberg et al., "Neuronal ensemble control of prosthetic devices by a human with
tetraplegia," (in eng), Nature, vol. 442, no. 7099, pp. 164-71, 2006, doi:
10.1038/nature04970.

[5] P. Fattahi, G. Yang, G. Kim, and M. R. Abidian, "A Review of Organic and Inorganic
Biomaterials for Neural Interfaces," Advanced Materials,
https://doi.org/10.1002/adma.201304496 vol. 26, no. 12, pp. 1846-1885, 2014/03/01 2014,
doi: https://doi.org/10.1002/adma.201304496.

[6] G. Rios, E. V. Lubenov, D. Chi, M. L. Roukes, and A. G. Siapas, "Nanofabricated Neural
Probes for Dense 3-D Recordings of Brain Activity," Nano Letters, vol. 16, no. 11, pp.
6857-6862, 2016/11/09 2016, doi: 10.1021/acs.nanolett.6b02673.

[7] C. Clinic. "Meninges." https://my.clevelandclinic.org/health/articles/22266-meninges
(accessed.

[8] S. K. Mudgal, S. K. Sharma, J. Chaturvedi, and A. Sharma, "Brain computer interface
advancement in neurosciences: Applications and issues," Interdisciplinary Neurosurgery,
vol. 20, p. 100694, 2020/06/01/ 2020, doi: https://doi.org/10.1016/j.inat.2020.100694.

[9] J. J. Shih, D. J. Krusienski, and J. R. Wolpaw, "Brain-Computer Interfaces in Medicine,"
Mayo Clinic Proceedings, vol. 87, no. 3, pp. 268-279, 2012/03/01/ 2012, doi:
https://doi.org/10.1016/j.mayocp.2011.12.008.

[10] D. J. Krusienski and J. J. Shih, "Control of a Visual Keyboard Using an
Electrocorticographic Brain–Computer Interface," Neurorehabilitation and Neural Repair,
vol. 25, no. 4, pp. 323-331, 2011/05/01 2010, doi: 10.1177/1545968310382425.

[11] J. J. Jun et al., "Fully integrated silicon probes for high-density recording of neural
activity," Nature, vol. 551, no. 7679, pp. 232-236, 2017/11/01 2017, doi:
10.1038/nature24636.
85

[12] C. M. Lopez et al., "A Neural Probe With Up to 966 Electrodes and Up to 384 Configurable
Channels in 0.13 \mum SOI CMOS," IEEE Transactions on Biomedical Circuits and
Systems, vol. 11, no. 3, pp. 510-522, 2017, doi: 10.1109/TBCAS.2016.2646901.

[13] P.-M. Wang et al., "Challenges in the Design of Large-Scale, High-Density, Wireless
Stimulation and Recording Interface," in Interfacing Bioelectronics and Biomedical
Sensing, H. Cao, T. Coleman, T. K. Hsiai, and A. Khademhosseini Eds. Cham: Springer
International Publishing, 2020, pp. 1-28.

[14] K. Scholten and E. Meng, "Materials for microfabricated implantable devices: a review,"
Lab on a Chip, vol. 15, no. 22, pp. 4256-4272, 2015, doi: 10.1039/c5lc00809c.

[15] A. K. Ommaya, "Mechanical properties of tissues of the nervous system," Journal of
Biomechanics, vol. 1, no. 2, pp. 127-138, 1968/07/01/ 1968, doi:
https://doi.org/10.1016/0021-9290(68)90015-8.

[16] T. D. Y. Kozai, A. S. Jaquins-Gerstl, A. L. Vazquez, A. C. Michael, and X. T. Cui, "Brain
tissue responses to neural implants impact signal sensitivity and intervention strategies,"
ACS Chemical Neuroscience, vol. 6, no. 1, pp. 48-67, 2015, doi: 10.1021/cn500256e.

[17] J. Subbaroyan, D. C. Martin, and D. R. Kipke, "A finite-element model of the mechanical
effects of implantable microelectrodes in the cerebral cortex," Journal of Neural
Engineering, vol. 2, no. 4, pp. 103-113, 2005, doi: 10.1088/1741-2560/2/4/006.

[18] V. S. Polikov, P. A. Tresco, and W. M. Reichert, "Response of brain tissue to chronically
implanted neural electrodes," Journal of Neuroscience Methods, vol. 148, no. 1, pp. 1-18,
2005, doi: 10.1016/j.jneumeth.2005.08.015.

[19] R. Biran, D. C. Martin, and P. A. Tresco, "Neuronal cell loss accompanies the brain tissue
response to chronically implanted silicon microelectrode arrays," Experimental Neurology,
vol. 195, no. 1, pp. 115-126, 2005, doi: 10.1016/j.expneurol.2005.04.020.

[20] T. D. Y. Kozai, A. L. Vazquez, C. L. Weaver, S.-G. Kim, and X. T. Cui, "in vivo two-
photon microscopy reveals immediate microglial reaction to implantation of
microelectrode through extension of processes," Journal of Neural Engineering, vol. 9, no.
6, p. 066001, 2012, doi: 10.1088/1741-2560/9/6/066001.

[21] R. Chen, A. Canales, and P. Anikeeva, "Neural Recording and Modulation Technologies,"
(in eng), Nat Rev Mater, vol. 2, no. 2, p. 16093, 2017, doi: 10.1038/natrevmats.2016.93.

[22] J. C. Williams, R. L. Rennaker, and D. R. Kipke, "Long-term neural recording
characteristics of wire microelectrode arrays implanted in cerebral cortex," Brain Research
Protocols, vol. 4, no. 3, pp. 303-313, 1999, doi: 10.1016/S1385-299X(99)00034-3.

[23] A. Weltman, J. Yoo, and E. Meng, "Flexible, penetrating brain probes enabled by advances
in polymer microfabrication," vol. 7, ed. Micromachines: MDPI AG, 2016, p. 180.
86

[24] J. K. Nguyen et al., "Mechanically-compliant intracortical implants reduce the
neuroinflammatory response," Journal of Neural Engineering, vol. 11, no. 5, p. 056014,
2014/08/15 2014, doi: 10.1088/1741-2560/11/5/056014.

[25] P. J. Rousche, D. S. Pellinen, D. P. Pivin, J. C. Williams, R. J. Vetter, and D. R. Kipke,
"Flexible polyimide-based intracortical electrode arrays with bioactive capability," IEEE
Transactions on Biomedical Engineering, vol. 48, no. 3, pp. 361-371, 2001, doi:
10.1109/10.914800.

[26] J. E. Chung et al., "High-density, long-lasting, and multi-region electrophysiological
recordings using polymer electrode arrays," Neuron, vol. 101, no. 1, pp. 21-31.e5, 2019,
doi: 10.1016/j.neuron.2018.11.002.

[27] R. Fiáth et al., "Long-term recording performance and biocompatibility of chronically
implanted cylindrically-shaped, polymer-based neural interfaces," Biomedical
Engineering/Biomedizinische Technik, vol. 63, no. 3, pp. 301-315, 2018.

[28] X. Wei et al., "Nanofabricated Ultraflexible Electrode Arrays for High‐Density
Intracortical Recording," Advanced Science, vol. 5, no. 6, pp. n/a-n/a, 2018, doi:
10.1002/advs.201700625.

[29] A. Altuna et al., "SU-8 based microprobes for simultaneous neural depth recording and
drug delivery in the brain," Lab on a Chip, vol. 13, no. 7, pp. 1422-1430, 2013, doi:
10.1039/c3lc41364k.

[30] S.-H. Huang, S.-P. Lin, and J.-J. J. Chen, "In vitro and in vivo characterization of SU-8
flexible neuroprobe: From mechanical properties to electrophysiological recording,"
Sensors and Actuators A: Physical, vol. 216, pp. 257-265, 2014/09/01/ 2014, doi:
https://doi.org/10.1016/j.sna.2014.06.005.

[31] G. Márton et al., "The neural tissue around SU-8 implants: A quantitative in vivo
biocompatibility study," Materials Science and Engineering: C, p. 110870, 2020/03/20/
2020, doi: https://doi.org/10.1016/j.msec.2020.110870.

[32] B. J. Kim et al., "3D parylene sheath neural probe for chronic recordings," Journal of
Neural Engineering, vol. 10, no. 4, p. 045002, 2013, doi: 10.1088/1741-2560/10/4/045002.

[33] J. W. Reddy, I. Kimukin, E. Towe, and M. Chamanzar, "Flexible, Monolithic, High-
Density µLED Neural Probes for Simultaneous Optogenetics Stimulation and
Recording*," in 2019 9th International IEEE/EMBS Conference on Neural Engineering
(NER), 20-23 March 2019 2019, pp. 831-834, doi: 10.1109/NER.2019.8717116.

[34] H. Xu, A. W. Hirschberg, K. Scholten, T. W. Berger, D. Song, and E. Meng, "Acute in
vivo testing of a conformal polymer microelectrode array for multi-region hippocampal
recordings," Journal of Neural Engineering, vol. 15, no. 1, p. 016017, 2018, doi:
10.1088/1741-2552/aa9451.
87

[35] T. Ware et al., "Fabrication of responsive, softening neural interfaces," Advanced
Functional Materials, vol. 22, no. 16, pp. 3470-3479, 2012, doi: 10.1002/adfm.201200200.

[36] A. Zátonyi et al., "A softening laminar electrode for recording single unit activity from the
rat hippocampus," Scientific Reports, vol. 9, no. 1, p. 2321, 2019.

[37] M. S. Allison et al., "Chronic intracortical recording and electrochemical stability of thiol-
ene/acrylate shape memory polymer electrode arrays," Micromachines, vol. 9, no. 10, p.
500, 2018, doi: 10.3390/mi9100500.

[38] C. L. Kolarcik et al., "Elastomeric and soft conducting microwires for implantable neural
interfaces," Soft Matter, vol. 11, no. 24, pp. 4847-4861, 2015, doi: 10.1039/c5sm00174a.

[39] M. McClain, I. Clements, R. Shafer, R. Bellamkonda, M. LaPlaca, and M. Allen, "Highly-
compliant, microcable neuroelectrodes fabricated from thin-film gold and PDMS,"
Biomedical Microdevices, vol. 13, no. 2, pp. 361-373, 2011, doi: 10.1007/s10544-010-
9505-3.

[40] J. P. Harris et al., "In vivo deployment of mechanically adaptive nanocomposites for
intracortical microelectrodes," Journal of Neural Engineering, vol. 8, no. 4, p. 046010,
2011, doi: 10.1088/1741-2560/8/4/046010.

[41] J. P. Harris et al., "Mechanically adaptive intracortical implants improve the proximity of
neuronal cell bodies," Journal of Neural Engineering, vol. 8, no. 6, p. 066011, 2011, doi:
10.1088/1741-2560/8/6/066011.

[42] S. P. Marshall, P. R. Wei-Chen Lin, A. J. Patel, C. A. Shih, and S. P. Chestek, "Effects of
geometry and material on the insertion of very small neural electrode," Conference
proceedings : International Conference of the IEEE Engineering in Medicine and Biology
Society., vol. 2016, pp. 2784-2788, 2016, doi: 10.1109/EMBC.2016.7591308.

[43] A. Lecomte, E. Descamps, and C. Bergaud, "A review on mechanical considerations for
chronically-implanted neural probes," Journal of Neural Engineering, vol. 15, no. 3, p.
031001, 2018, doi: 10.1088/1741-2552/aa8b4f.

[44] M. Jorfi, J. L. Skousen, C. Weder, and J. R. Capadona, "Progress towards biocompatible
intracortical microelectrodes for neural interfacing applications," vol. 12, ed. Journal of
neural engineering, 2015, p. 011001.

[45] M. A. Lebedev and M. A. L. Nicolelis, "Brain–machine interfaces: past, present and
future," Trends in Neurosciences, vol. 29, no. 9, pp. 536-546, 2006, doi:
10.1016/j.tins.2006.07.004.

[46] W. M. Grill, S. E. Norman, and R. V. Bellamkonda, "Implanted neural interfaces:
biochallenges and engineered solutions," Annual Review of Biomedical Engineering, vol.
11, pp. 1-24, 2009, doi: 10.1146/annurev-bioeng-061008-124927.
88

[47] J. H. Kim et al., "Flexible deep brain neural probe for localized stimulation and detection
with metal guide," Biosensors and Bioelectronics, vol. 117, pp. 436-443, 2018/10/15/ 2018,
doi: https://doi.org/10.1016/j.bios.2018.06.035.

[48] X. Wang et al., "A Parylene Neural Probe Array for Multi-Region Deep Brain Recordings,"
Journal of Microelectromechanical Systems, vol. 29, no. 4, pp. 499-513, 2020, doi:
10.1109/JMEMS.2020.3000235.

[49] K. S. Anand and V. Dhikav, "Hippocampus in health and disease: An overview," (in eng),
Ann Indian Acad Neurol, vol. 15, no. 4, pp. 239-246, 2012.

[50] G. B. Frisoni, C. R. Fox Nc Fau - Jack, Jr., P. Jack Cr Jr Fau - Scheltens, P. M. Scheltens
P Fau - Thompson, and P. M. Thompson, "The clinical use of structural MRI in Alzheimer
disease," (in eng), Nat Rev Neurol, vol. 6, no. 2, pp. 67-77, 2010.

[51] V. Castagnola et al., Parylene-based flexible neural probes with PEDOT coated surface
for brain stimulation and recording. Biosensors and Bioelectronics, 2015.

[52] B. A. Wester, R. H. Lee, and M. C. LaPlaca, "Development and characterization of in vivo
flexible electrodes compatible with large tissue displacements," Journal of Neural
Engineering, vol. 6, no. 2, p. 024002, 2009, doi: 10.1088/1741-2560/6/2/024002.

[53] W. Jensen, K. Yoshida, and U. G. Hofmann, "In-vivo implant mechanics of flexible,
silicon-based ACREO microelectrode arrays in rat cerebral cortex," IEEE Transactions on
Biomedical Engineering, vol. 53, no. 5, pp. 934-940, 2006, doi:
10.1109/TBME.2006.872824.

[54] J. Agorelius, F. Tsanakalis, A. Friberg, P. T. Thorbergsson, L. M. E. Pettersson, and J.
Schouenborg, "An array of highly flexible electrodes with a tailored configuration locked
by gelatin during implantation-initial evaluation in cortex cerebri of awake rats," Frontiers
in Neuroscience, vol. 9, 2015, doi: 10.3389/fnins.2015.00331.

[55] D. J. Edell, V. V. Toi, V. M. McNeil, and L. D. Clark, "Factors influencing the
biocompatibility of insertable silicon microshafts in cerebral cortex," IEEE Transactions
on Biomedical Engineering, vol. 39, no. 6, pp. 635-643, 1992, doi: 10.1109/10.141202.

[56] A. Sridharan, J. K. Nguyen, J. R. Capadona, and J. Muthuswamy, "Compliant intracortical
implants reduce strains and strain rates in brain tissue in vivo," Journal of Neural
Engineering, vol. 12, no. 3, p. 036002, 2015, doi: 10.1088/1741-2560/12/3/036002.

[57] R. Chen, A. Canales, and P. Anikeeva, "Neural recording and modulation technologies,"
Nature Reviews Materials, vol. 2, no. 2, p. 16093, 2017, doi: 10.1038/natrevmats.2016.93.

[58] E. Meng, P.-Y. Li, and Y.-C. Tai, "Plasma removal of parylene c," Journal of
Micromechanics and Microengineering, vol. 18, no. 4, p. 045004, 2008, doi:
10.1088/0960-1317/18/4/045004.
89

[59] S. Dabral et al., "Stress in thermally annealed parylene films," Journal of Electronic
Materials, vol. 21, no. 10, pp. 989-994, 1992, doi: 10.1007/BF02684208.

[60] J. Ortigoza-Diaz, K. Scholten, and E. Meng, "Characterization and modification of
adhesion in dry and wet environments in thin-film parylene systems," Journal of
Microelectromechanical Systems, vol. 27, no. 5, pp. 874-885, 2018, doi:
10.1109/JMEMS.2018.2854636.

[61] Z. Xiang et al., "Ultra-thin flexible polyimide neural probe embedded in a dissolvable
maltose-coated microneedle," Journal of Micromechanics and Microengineering, vol. 24,
no. 6, p. 065015, 2014, doi: 10.1088/0960-1317/24/6/065015.

[62] K. Najafi and J. F. Hetke, "Strength characterization of silicon microprobes in
neurophysiological tissues," IEEE Transactions on Biomedical Engineering, vol. 37, no. 5,
pp. 474-481, 1990, doi: 10.1109/10.55638.

[63] S. Singh et al., "Modeling the insertion mechanics of flexible neural probes coated with
sacrificial polymers for optimizing probe design," Sensors, vol. 16, no. 3, p. 330, 2016,
doi: 10.3390/s16030330.

[64] S. Felix et al., "Removable silicon insertion stiffeners for neural probes using polyethylene
glycol as a biodissolvable adhesive," Conference proceedings : International Conference
of the IEEE Engineering in Medicine and Biology Society., pp. 871-874, 2012.

[65] T. D. Y. Kozai and D. R. Kipke, "Insertion shuttle with carboxyl terminated self-assembled
monolayer coatings for implanting flexible polymer neural probes in the brain," Journal of
Neuroscience Methods, vol. 184, no. 2, pp. 199-205, 2009/11/15/ 2009, doi:
https://doi.org/10.1016/j.jneumeth.2009.08.002.

[66] Z. Zhao, X. Li, F. He, X. Wei, S. Lin, and C. Xie, "Parallel, minimally-invasive
implantation of ultra-flexible neural electrode arrays," Journal of Neural Engineering, vol.
16, no. 3, p. 035001, 2019, doi: 10.1088/1741-2552/ab05b6.

[67] J. Shi and Y. Fang, "Flexible and Implantable Microelectrodes for Chronically Stable
Neural Interfaces," Advanced Materials, vol. 31, no. 45, pp. n/a-n/a, 2019, doi:
10.1002/adma.201804895.

[68] J. E. Chung et al., "Chronic Implantation of Multiple Flexible Polymer Electrode Arrays,"
Journal of Visualized Experiments, no. 152, 2019, doi: 10.3791/59957.

[69] H. R. Joo et al., "A microfabricated, 3D-sharpened silicon shuttle for insertion of flexible
electrode arrays through dura mater into brain," Journal of Neural Engineering, vol. 16,
no. 6, p. 066021, 2019, doi: 10.1088/1741-2552/ab2b2e.

[70] L. W. Tien, F. Wu, M. D. Tang‐Schomer, E. Yoon, F. G. Omenetto, and D. L. Kaplan,
"Silk as a multifunctional biomaterial substrate for reduced glial scarring around brain‐
90

penetrating electrodes," Advanced Functional Materials, vol. 23, no. 25, pp. 3185-3193,
2013, doi: 10.1002/adfm.201203716.

[71] M. Cinzia and A. T. Barry, "Silk coating as a novel delivery system and reversible adhesive
for stiffening and shaping flexible probes," Journal of Biological Methods, vol. 2, no. 1,
pp. e13-e13, 2015, doi: 10.14440/jbm.2015.41.

[72] A. Lecomte et al., "Silk and peg as means to stiffen a parylene probe for insertion in the
brain: toward a double time-scale tool for local drug delivery," Journal of Micromechanics
and Microengineering, vol. 25, no. 12, p. 125003, 2015, doi: 10.1088/0960-
1317/25/12/125003.

[73] T. Suzuki, K. Mabuchi, and S. Takeuchi, "A 3D flexible parylene probe array for
multichannel neural recording," in First International IEEE EMBS Conference on Neural
Engineering, 2003. Conference Proceedings., 2003: IEEE, pp. 154-156.

[74] K. J. Seo et al., "Transparent, Flexible, Penetrating Microelectrode Arrays with
Capabilities of Single‐Unit Electrophysiology," Advanced Biosystems, vol. 3, no. 3, p.
1800276, 2019.

[75] M.-c. Lo et al., "Coating flexible probes with an ultra fast degrading polymer to aid in
tissue insertion," Biomedical Microdevices, vol. 17, no. 2, p. 34, 2015.

[76] D. Lewitus, K. L. Smith, W. Shain, and J. Kohn, "Ultrafast resorbing polymers for use as
carriers for cortical neural probes," Acta Biomaterialia, vol. 7, no. 6, pp. 2483-2491, 2011,
doi: 10.1016/j.actbio.2011.02.027.

[77] G. Lind, C. E. Linsmeier, J. Thelin, and J. Schouenborg, "Gelatine-embedded electrodes—
a novel biocompatible vehicle allowing implantation of highly flexible microelectrodes,"
Journal of Neural Engineering, vol. 7, no. 4, p. 046005, 2010.

[78] A. Singh, Z. Haixin, and H. Jiping, "Improving mechanical stiffness of coated
benzocyclobutene (BCB) based neural implant," in The 26th Annual International
Conference of the IEEE Engineering in Medicine and Biology Society, 1-5 Sept. 2004 2004,
vol. 2, pp. 4298-4301, doi: 10.1109/IEMBS.2004.1404197.

[79] D. Kil et al., "Dextran as a resorbable coating material for flexible neural probes,"
Micromachines, vol. 10, no. 1, 2019, doi: 10.3390/mi10010061.

[80] P. J. Gilgunn et al., "An ultra-compliant, scalable neural probe with molded biodissolvable
delivery vehicle," ed. 2012 IEEE 25th International Conference on Micro Electro
Mechanical Systems (MEMS): IEEE, 2012, pp. 56-59.

[81] F. Pei and B. Tian, "Nanoelectronics for Minimally Invasive Cellular Recordings,"
Advanced Functional Materials, p. 1906210, 2019, doi: 10.1002/adfm.201906210.
91

[82] A. W. Hirschberg, H. Xu, K. Scholien, T. W. Berger, D. Song, and E. Meng,
"Development of an anatomically conformal parylene neural probe array for multi-region
hippocampal recordings," in 2017 IEEE 30th International Conference on Micro Electro
Mechanical Systems, 22-26 Jan. 2017 2017, pp. 129-132, doi:
10.1109/MEMSYS.2017.7863357.

[83] A. Weltman, H. Xu, K. Scholten, T. W. Berger, D. Song, and E. Meng, Deep brain
targeting strategy for bare parylene neural probe arrays. Proceedings of the Hilton Head
Workshop, 2016, pp. 302-305.

[84] S. Yagi et al., "Dissolvable Base Scaffolds Allow Tissue Penetration of High‐Aspect‐Ratio
Flexible Microneedles," Advanced Healthcare Materials, vol. 4, no. 13, pp. 1949-1955,
2015, doi: 10.1002/adhm.201500305.

[85] N. A. Alcantar, E. S. Aydil, and J. N. Israelachvili, "Polyethylene glycol-coated
biocompatible surfaces," Journal of Biomedical Materials Research, vol. 51, no. 3, p. 343,
2000, doi: 10.1002/1097-4636(20000905)51:33.0.CO;2-D.

[86] H. M. Gensler and E. Meng, "Rapid fabrication and characterization of MEMS Parylene C
bellows for large deflection applications," Journal of Micromechanics and
Microengineering, vol. 22, no. 11, p. 115031, 2012, doi: 10.1088/0960-1317/22/11/115031.

[87] A. Sharp, A. Ortega, D. Restrepo, D. Curran-Everett, and K. Gall, "In Vivo Penetration
Mechanics and Mechanical Properties of Mouse Brain Tissue at Micrometer Scales," IEEE
transactions on bio-medical engineering, vol. 56, pp. 45-53, 02/01 2009, doi:
10.1109/TBME.2008.2003261.

[88] Z.-J. Chen et al., "A realistic brain tissue phantom for intraparenchymal infusion studies,"
Journal of Neurosurgery, vol. 101, no. 2, pp. 314-322, 2004, doi:
10.3171/jns.2004.101.2.0314.

[89] A. Andrei, M. Welkenhuysen, B. Nuttin, and W. Eberle, "A response surface model
predicting the in vivo insertion behavior of micromachined neural implants," Journal of
Neural Engineering, vol. 9, no. 1, 2012, doi: 10.1088/1741-2560/9/1/016005.

[90] M. Welkenhuysen, A. Andrei, L. Ameye, W. Eberle, and B. Nuttin, "Effect of insertion
speed on tissue response and insertion mechanics of a chronically implanted silicon-based
neural probe," IEEE Transactions on Biomedical Engineering, vol. 58, no. 11, pp. 3250-
3259, 2011, doi: 10.1109/TBME.2011.2166963.

[91] S. A. Hara, B. J. Kim, J. T. W. Kuo, and E. Meng, "An electrochemical investigation of
the impact of microfabrication techniques on polymer-based microelectrode neural
interfaces," Journal of Microelectromechanical Systems, vol. 24, no. 4, pp. 801-809, 2015,
doi: 10.1109/JMEMS.2015.2434827.

92

[92] N. A. B. Ismail, F. Abd-Wahab, and W. W. A. W. Salim, "Cyclic voltammetry and
electrochemical impedance spectroscopy of partially reduced graphene oxide -
PEDOT:PSS transducer for biochemical sensing," in 2018 IEEE-EMBS Conference on
Biomedical Engineering and Sciences, 3-6 Dec. 2018 2018, pp. 330-335, doi:
10.1109/IECBES.2018.8626618.

[93] J.-M. Hsu et al., Characterization of parylene-c film as an encapsulation material for
neural interface devices. Conference on Multi-Material Micro Manufacture, 2007.

[94] S. F. Cogan, "Neural Stimulation and Recording Electrodes," Annual Review of Biomedical
Engineering, vol. 10, no. 1, pp. 275-309, 2008, doi:
10.1146/annurev.bioeng.10.061807.160518.

[95] H. Xu, A. W. Hirschberg, K. Scholten, E. Meng, T. W. Berger, and D. Song, "Application
of Parylene-Based Flexible Multi-Electrode Array for Recording From Subcortical Brain
Regions From Behaving Rats," in 2018 40th Annual International Conference of the IEEE
Engineering in Medicine and Biology Society (EMBC), 18-21 July 2018 2018, pp. 4599-
4602, doi: 10.1109/EMBC.2018.8513202.

[96] C. S. Bjornsson et al., "Effects of insertion conditions on tissue strain and vascular damage
during neuroprosthetic device insertion," Journal of Neural Engineering, vol. 3, no. 3, pp.
196-207, 2006, doi: 10.1088/1741-2560/3/3/002.

[97] N. H. Hosseini, R. Hoffmann, S. Kisban, T. Stieglitz, O. Paul, and P. Ruther, "Comparative
study on the insertion behavior of cerebral microprobes," Conference proceedings : ...
Annual International Conference of the IEEE Engineering in Medicine and Biology Society.
IEEE Engineering in Medicine and Biology Society. Annual Conference, vol. 2007, p. 4711,
2007.

[98] B. Thielen and E. Meng, "A comparison of insertion methods for surgical placement of
penetrating neural interfaces," Journal of Neural Engineering, vol. 18, no. 4, p. 041003,
2021/04/26 2021, doi: 10.1088/1741-2552/abf6f2.

[99] J. E. Gilbert, T. Zhang, R. Esteller, and W. M. Grill, "Evaluating optimized temporal
patterns of spinal cord stimulation (SCS)," Brain Stimulation: Basic, Translational, and
Clinical Research in Neuromodulation, vol. 15, no. 5, pp. 1051-1062, 2022, doi:
10.1016/j.brs.2022.07.046.

[100] J. Sohn et al., "Presynaptic supervision of cortical spine dynamics in motor learning,"
Science Advances, vol. 8, no. 30, p. eabm0531, doi: 10.1126/sciadv.abm0531.

[101] K. Kajikawa, B. K. Hulse, A. G. Siapas, and E. V. Lubenov, "UP-DOWN states and ripples
differentially modulate membrane potential dynamics across DG, CA3, and CA1 in awake
mice," eLife, vol. 11, p. e69596, 2022/07/12 2022, doi: 10.7554/eLife.69596.
93

[102] A. M. Okamura, C. Simone, and M. D. O'Leary, "Force Modeling for Needle Insertion Into
Soft Tissue," IEEE Transactions on Biomedical Engineering, vol. 51, no. 10, pp. 1707-
1716, 2004, doi: 10.1109/tbme.2004.831542.

[103] E. M. Maynard, C. T. Nordhausen, and R. A. Normann, "The Utah Intracortical Electrode
Array: A recording structure for potential brain-computer interfaces,"
Electroencephalography and Clinical Neurophysiology, vol. 102, no. 3, pp. 228-239,
1997/03/01/ 1997, doi: https://doi.org/10.1016/S0013-4694(96)95176-0.

[104] S.-J. Kim, S. Manyam, D. Warren, and R. Normann, "Electrophysiological Mapping of Cat
Primary Auditory Cortex with Multielectrode Arrays," Annals of biomedical engineering,
vol. 34, pp. 300-9, 03/01 2006, doi: 10.1007/s10439-005-9037-9.

[105] C. W. Chang and J. C. Chiou, "A Wireless and Batteryless Microsystem with Implantable
Grid Electrode/3-Dimensional Probe Array lor ECoG and Extracellular Neural Recording
in Rats," Sensors, vol. 13, no. 4, pp. 4624-4639, 2013, doi: 10.3390/s130404624.

[106] C. W. Chang and J. C. Chiou, "Development of a Three Dimensional Neural Sensing
Device by a Stacking Method (vol 10, pg 4238, 2010)," Sensors, vol. 10, no. 5, pp. 4716-
4716, 2010, doi: 10.3390/s100504716.

[107] J. Du, M. L. Roukes, and S. C. Masmanidis, "Dual-side and three-dimensional
microelectrode arrays fabricated from ultra-thin silicon substrates," Journal of
Micromechanics and Microengineering, vol. 19, no. 7, p. 075008, 2009, doi:
10.1088/0960-1317/19/7/075008.

[108] E. Musk, "An integrated brain-machine interface platform with thousands of channels,"
bioRxiv, p. 703801, 2019, doi: 10.1101/703801.

[109] L. Chen, J. P. Hartner, T. K. Dong, A. D. R. Li, B. O. Watson, and A. J. Shih, "Flexible
High-Resolution Force and Dimpling Measurement System for Pia and Dura Penetration
During In Vivo Microelectrode Insertion Into Rat Brain," IEEE Transactions on
Biomedical Engineering, vol. 68, no. 8, pp. 2602-2612, 2021, doi:
10.1109/TBME.2021.3070781.

[110] B. J. Kim, B. Chen, M. Gupta, and E. Meng, "Formation of three-dimensional Parylene C
structures via thermoforming," Journal of Micromechanics and Microengineering, vol. 24,
no. 6, p. 065003, 2014/04/25 2014, doi: 10.1088/0960-1317/24/6/065003.

94

Background
Hypoplastic left heart syndrome (HLHS) is a rare disease that happens in 1 out of every
3841 babies born in the United States each year [1]. The mortality rate of HLHS is over 90% within
30 days after birth without surgical treatment [2]. The surgery decreases the mortality rate to 25%

Figure 2-1. Schematic showing how a normal heart functions.
CHAPTER TWO
A NON-CONTACT FLOW SENSOR FOR
V ASCULAR SHUNT
95

[3, 4]. HLHS infants' symptoms include but are not limited to poor feeding, abnormal breathing
rate, weak heart pulse, and cyanotic condition [5].
A normal heart contains four chambers: the right atrium, right ventricle, left atrium and left
ventricle. The heart's left side takes the oxygen-rich blood from the lung through the pulmonary
vein to the left atrium, then to the left ventricle. The left ventricle pumps oxygen-rich blood to the
body through the aorta. After the tissues and organs absorb the oxygen, the oxygen-poor blood is
recycled back to the heart's right atrium. The right atrium pumps the oxygen-poor blood to the
right ventricle and back to the lung through the pulmonary artery (Fig. 2-1). Oxygen translation
between the lung and heart is called the pulmonary circuit. In parallel, oxygen translation between
the heart and the rest of the body is the systemic circuit.

Figure 2-2. Schematic showing an HLHS heart.
96

HLHS patients are often born with significant failure on the heart's left side (Fig. 2-2). The
most common defect is the underdeveloped left ventricle which is typically abnormally small.
Other defects include narrowed aorta, patent ductus arteriosus, and atrial septal defect. The patent
ductus arteriosus is the opening between the aorta and pulmonary artery, and the atrial septal defect
is the opening between the left and right atrium [6]. Overall, the left heart defects result in the aorta
not sending blood containing enough oxygen to support organs and tissues adequately.

Figure 2-3. Schematic showing the HLHS treatment through the Norwood procedure.

As shown in Fig. 2-3, the early-stage treatment of HLHS usually starts with the Norwood
procedure [7]. The pulmonary artery is first disconnected from the right ventricle. Next, the aorta
97

is widened and connected to the right ventricle. After that, the patent ductus arteriosus is closed,
and the atrial septal defect is cleaned to connect the right with the left atrium. The oxygen-rich and
oxygen-poor blood is mixed in the right atrium and enters the right ventricle. The mixed blood is
pumped from the right ventricle to the body through the newly built aorta. In the end, a modified

Figure 2-4. Schematic showing the mBTS failure caused by blood obstruction.

Blalock-Tausig shunt (mBTS) is sutured between the aorta and pulmonary artery to ensure the
pulmonary artery can recycle blood to the lung for oxygen recharging [8]. The outcome of the
Norwood procedure is to make the right side of the heart pump blood to the body for oxygen supply
and recycle blood to the lung for oxygen intake.
The mBTS frequently fails due to thrombin clotting in the shunt. There is no specified
symptom reflecting the shunt obstruction (Fig. 2-4). The shunt obstruction causes insufficient
pulmonary circulation resulting in hypoxia or even death if found in the late stage. There is no
98

existing imaging modality to detect the shunt obstruction due to the shunt's small size for pediatric
patients. There is a need to develop a tool to monitor the shunt obstruction in real time.
Flow Sensor for Blood Flow Detection
There are flow sensors on the market designed for human patients, including the Cook-
Swartz Doppler probe. The Doppler probe is a one-time use sensor. Due to the measurement
system's bulk size, the Doppler probe was only used during the Free Flaps surgery to monitor the
blood flow through vessels [9]. Abbott-CardioMEMS HF systems are designed for chronic use,
even though it does not measure blood flow directly, it measures blood pressure which can be used
to diagnose heart failure. However the size of the Abbott-CardioMEMS HF system is too large to
fit into the neonatal shunt. So far, there are no sensors designed for pediatric vascular shunts [10].
Our goal is to develop a sensor integrated onto pediatric mBTS for real-time monitoring of the
blood flow change for detecting shunt obstruction (Fig. 2-5).

Figure 2-5. Schematic showing the ring electrode attached to the external wall of mBTS for blood flow
monitoring.
99

Design Requirements
To design a flow sensor for a mBTS, the sensor needs to be on the exterior wall of the shunt
and should not be in contact with blood. It also needs to fit the mBTS with an outer diameter of
3.5 mm and a length of 2 cm. A 0.25 cm length on each side is reserved for suturing purposes. The
whole sensor should not be longer than 1.5 cm. The sensor should transduce the flow rate in the
range of 0-1000 mL/min with a resolution of 100 mL/min.
Flow Sensing Fundamentals
Different flow sensing methods have been used for detecting the flow rate, including
electromagnetic, ultrasonic, thermal sensing, and capacitively coupled contactless conductivity
detection (C
4
D). Most of these techniques were well developed for industrial use but not for
medical usage. We chose C
4
D because it does not require direct contact of sensor elements with
the blood.
The C
4
D technique was initially proposed by Zemann et al. in 1998 [11] for capillary
electrophoresis. It was later adapted by Qi et al. in 2013 [12] to measure the flow rate of multi-
phase liquid in 0.5 – 3.9 mm inner diameter tubes with wide ring electrodes (10 mm wide). The
electrode size is too big to fit onto mBTS. The effectiveness and resolution of detecting flow rate
are limited by the liquid's conductivity, flow rate, and electrode size. The work reported here
focuses on optimizing the sensor and measurement circuitry design to detect flow rates with small
electrodes that fit onto mBTS. The next step is to test the feasibility of implementing the C
4
D for
real-time cardiovascular shunt blood flow monitoring.
Referring to Fig. 2-6, an AC source is connected to the excitation electrode to send out an
input signal, then the pick-up electrodes read out the voltage. The voltage difference between two
100

neighboring electrodes, V1-2 and V2-3, reflects the liquid's conductive signal between the pick-up
electrodes. The flow rate (Q) measurements can be converted by using the following equation,

Figure 2-6. Schematic showing the ring electrode integrated with the mBTS designed based on capacitively
coupled contactless conductivity detection.

𝑄 =
𝐾𝐴𝐿 τ

where A is the cross-section area of the shunt; L is the distance between the center of two
electrodes; τ is the time delay between two differential voltage signals, V1-2 and V2-3, where the
maximum cross-correlation coefficient was [12], and K is the calibration coefficient for the sensor.
101

Figure 2-7. Photo showing the prototyped sensors with five ring electrodes by directly painting silver epoxy
on the mBTS shunt with an outer diameter of 3.5 mm. Multiple sensors are made with electrodes having
different widths and spacing.

Figure 2-8. Schematic showing the electrode layout and dimension.

The mBTS sensors have five-ring electrodes, as shown in Fig. 2-7. Each electrode is 1 mm
wide and 2.5 mm away from neighboring electrodes (Fig. 2-8). The equivalent circuit for the shunt
102

sensors is shown in Fig. 2-9. The capacitance of the ring electrode is roughly 1 pF at 100 kHz,
calculated by the plate capacitance equation,

Figure 2-9. Schematic showing the equivalent circuit for the mBTS sensors.

𝐶 =
𝜀𝐴
𝑑

C is capacitance; ɛ is the permittivity; A is the plate area, which equals the ring electrode's surface
area; d is the distance between plates equivalent to the ring electrodes' radius. The blood resistance
between electrodes was calculated as 200 Ω through the following equation:

𝑅 =
𝜌𝐿
𝐴
R is resistance; ρ is blood resistivity and roughly equals 100 Ω cm [13]; L is the shunt length
between two electrodes, and A is the cross-section of the shunt. The challenge is that the voltage
103

signal between the pick-up electrodes is small as voltage equals the product of current and
resistance. The current is low due to the electrodes' small capacitance, and the resistance of the
blood is relatively low compared to the impedance of the capacitance. Therefore, the circuitry was
developed to read a small voltage signal from a small capacitor.

Figure 2-10. The circuit architecture for voltage read-out from a single pick-up electrode.

Circuitry Design
The circuit design started with designing a voltmeter circuit to read a small voltage signal
from a single ring electrode. As shown in Fig. 2-10, the circuitry consists of three blocks, starting
with a high impedance buffer used to boost the signal swings collected by the pick-up electrodes.
Afterward, a mixer combined with a low pass filter block served as the demodulation block to
104

down-convert the signals and filter out the high-frequency signal and noise. The circuit was built
and tested on the breadboard, as shown in Fig. 2-11. Each individual block was tested in sequence
for performance.

Figure 2-11. The testing setup with the breadboard system.
105

Figure 2-12. Schematic for bootstrapping circuit.

Input Impedance Buffer
The input impedance buffer was a bootstrapped circuit driven by an operational amplifier
AD8045 from Analog Devices, as shown in Fig. 2-12. When a high-frequency AC signal was sent
at the input port, the capacitor between b and c was equivalent to a short. As a result, the voltage
at points b and c are equal to V- of the operational amplifier. For an ideal operational amplifier,
V- equals V+, so V+ equals Vb. As V+ equals Va, Vb equals Va. Therefore, there is no current passing
through paths a-b. In other words, the bootstrapping circuit resulted in an infinite large impedance
at the point after the pick-up electrode. As a result, it would divide most of the voltage signal.
Therefore, the measured voltage amplitude increased.
106

Figure 2-13. (a) Circuit with shunt model connected to the oscilloscope; (b) Measured waveform after the
bootstrapping circuit.

The shunt model containing a single pick-up electrode without input impedance buffer was
first connected to an oscilloscope (Tektronix TDS 2024B,4-channel digital storage oscilloscope)
(Fig. 2-13). A single-channel arbitrary function generator (Tektronix AFG 3021) was connected
to the excitation electrode and sent a sine wave with Vp-p of 2V at 100 kHz. The oscilloscope only
measured a voltage with 27.2 mV amplitude, which equaled 1.36% of the sending signal's
amplitude. The signal attenuation was caused by the oscilloscope's built-in capacitance (27 μF),
which is relatively high compared to the pick-up electrode's capacitance, 1 pF. In other words, at
100 kHz, the pick-up electrode had an impedance much higher than the oscilloscope's built-in
impedance. The pick-up electrode became the voltage divider, resulting in the oscilloscope picking
a small voltage swing. After the bootstrapping circuit was connected to the pick-up electrode, the
107

measured voltage increased to 278 mV (Fig. 2-14), reflecting the input buffer’s ability to increase
the detected signal swing effectively.

Figure 2-14. (a) Shunt model hooked to the bootstrapping circuit; (b) Measured waveform with the
bootstrapping circuit.

Figure 2-15. (a) Layout of the bootstrapping circuit with adjustable gain controlled by the resistor's value
highlighted in the red circle. (b) The measured waveform with bootstrapping circuit shows an amplified
signal.
108

The voltage read-out can be increased by adjusting the gain of the bootstrapping circuit.
The gain was controllable by changing the resistance of the feedback resistor. If the resistance
changed from 4.87 kΩ to 10 kΩ (Fig. 2-15), the measured voltage increased to 360 mV.

Figure 2-16. (a) Measured waveform of the input signal; (b) Modulated waveform output by the mixer.

Demodulation Block
The demodulation block consisted of a mixer and a low-pass filter block. To understand how
the blocks work with the shunt system, assuming a sine wave with a peak-to-peak voltage of 2V
at 100 kHz was sent to the excitation electrode as the input signal after liquid passed through the
shunt, the pickup electrode would detect a modulated wave that equals the input wave multiplied
by the wave reflecting the liquid's stochastic fluctuation. If the liquid fluctuation were at 20 Hz,
the output signal's FFT would show two spikes, one at 100.02 kHz; the other at 99.98 kHz. If a
carrier wave at 100 kHz were mixed with the signal, the output signal after the mixer would show
three spikes at 200.02 kHz, 199.98 kHz, and 20 Hz. Since the 20-Hz spike reflected the information
109

of liquid fluctuation, the spikes at 200.02 kHz and 199.98 kHz were filtered out by a low-pass
filter with the cut-off frequency greater or equal to 20 Hz and much smaller than 199.98 kHz.
Specifically, the cut-off frequency was selected to be 10 kHz. The mixer and low pass-filter block
were tested individually for functionality and overall performance.
A sine wave with 2V Vp-p at 10 kHz and a sine wave with 2V Vp-p at 100 kHz were mixed
to test the mixer. As shown in Fig. 2-16, the 100 kHz sine wave was successfully modulated with
a wave at a lower frequency of 10 kHz, demonstrating the mixer block was functioning properly.

Figure 2-17. The circuit schematic for the low pass filter block.

After the signal was mixed, the higher-frequency wave was filtered out through the low-pass
filter block. Two low pass filters operated by operational amplifier AD 8541(Analog Devices)
were cascaded together with a cut-off frequency at 10 kHz and a gain of 1 (Fig. 2-17).
A circuit simulation was built to predict the waveform’s voltage at different frequencies after
passing through the low pass filter block. As shown in Fig. 2-18, at 1 kHz, the gain read 1. The
gain of 0.707 (-3dB), 0.316 (-10dB) and 0.001 (-60dB) were estimated at 10 kHz, 15 kHz and 100
110

kHz respectively according to the simulation. To test the low-pass filter performance, a sine wave
was given as the input signal with a Vp-p of 1 V at 1 kHz. The measured result showed a sine
wave with a Vp-p of 1.06 V at 998 Hz (Fig. 2-16(a)). It resulted in a gain roughly equal to 1.1,
which agreed with the simulation (1). As the input signal frequency increased from 1 kHz to 10
kHz, the measured voltage decreased to 700 mV, roughly equaling the predicted attenuation factor
of 0.707 (Fig. 2-16(b)). As the input signal frequency increased from 10 kHz to 15 kHz, the
amplitude decreased to 272 mV, closed to 316 mV predicted by the simulation (Fig. 2-16(c)). As
the input signal frequency increased from 15 kHz to 100 kHz, the measured results showed no AC
signal meaning all signals were filtered out (Fig. 2-16(d)). In summary, the measurement
demonstrated that the low-pass filter block worked properly as the measured result agreed with the
simulated results at all measured frequencies.

Figure 2-18. Simulation result for the low pass filter block.
111

Figure 2-19. Measured waveform for low pass filter with the input signal at different frequencies. “Sin”
labeled in the figure represents the sinusoidal wave of the input signal.

The mixer and low pass filter were connected in series and tested for their ability to
demodulate signals with different frequencies. The result in the top row of Fig. 2-19 demonstrated
that the circuit successfully mixed and demodulated the sine waves at 100 kHz and 105 kHz. The
FFT of the output signal showed spikes at 0 and 5 kHz. The 5 kHz spike reflected the demodulated
low-frequency signal after mixing the 100 kHz and 105 kHz signals. The spike at 0 Hz reflected
the DC bias added to the input signal in the beginning. As mentioned in the prior sections, the
mBTS sensor was required to detect flow with a 100 mL/min resolution, equaling 70 Hz. The
demodulation block was tested to demodulate signals with lower frequency. The result (Fig. 2-20)
showed the circuit successfully demodulated the signals and output a waveform with a frequency
112

as low as 30 Hz. The lowest frequency the circuit was able to detect was 5 Hz. This meets the
sensing requirement for the mBTS sensors.

Figure 2-20. Measured waveform (left) and FFT (right) for mixer and low pass filter at different frequencies.
“Sin” labeled in the figure represents the sinusoidal wave of the input signal.
113

Figure 2-21. Measurement setup used to test the circuit.

Circuit Test with Single Pick-up Electrode
The circuit blocks were connected and tested for performance (Fig. 2-21). A single-channel
function generator (Tektronix AFG 3021) was used to send an input signal at the excitation
electrode. The input signal was a sine wave with a Vp-p of 2 V at 100 kHz. A second function
generation (NI VirtualBench) was used to send the carrier wave to the mixer block, a sine wave
with a Vp-p of 3.1 V at 105 kHz. A DC supply (Agilent) was used to power the operational amplifier
in the input buffer and mixer block. A second DC supply (NI VirtualBench) was used to power
the low-pass filter block. As shown in Fig. 2-22, the measured output signal of 160 mV at 5 kHz
agreed with the spike at 5 kHz shown in the FFT. The FFT also showed a spike at 0 Hz, a DC bias
added to the initial input signal.
114

Figure 2-22. Measured waveform and FFT for the completed circuit with a single pick-up electrode.

Figure 2-23. The block diagram for the circuit design for three pick-up electrodes.
115

Circuit Design and Measurement with Three Pick-up Electrodes
The sensors designed for mBTS contained three pick-up electrodes. A differential amplifier
was added between the input buffer and mixer block to amplify the voltage difference between the
two pick-up electrodes (Fig. 2-23). The differential amplifier contained an operational amplifier
AD 8045 (Analog Device), and the gain was adjustable by modifying the ratio of R1 and R2 (Fig.
2-24). The differential amplifier was tested for its ability to amplify the voltage signal with
different gains. When the voltage difference between two pick-up electrodes was 10 mV, R1 was
487 Ω, and R2 was 140 kΩ, the differential amplifier's max output was measured at 984 mV,
equivalent to a gain of 98.4.

Figure 2-24. Circuit layout for the differential amplifier.

Besides the differential amplifier, other blocks were also designed to achieve various gains
for handling signals with varied amplitude ranges. Table 3-1 shows that the input buffer block
116

could reach a maximum gain of 18. The mixer could achieve a maximum gain of 2. The low-pass
filter was designed with unity gain because the major amplification is designed to be done at the
prior blocks to reduce overall system noise.

Table 2-1. The maximum achievable gain for different circuit blocks.

Summary
A first non-contact sensor based on C
4
D and cross-correlation was designed to be
integrated with mBTS for monitoring the blood flow in real-time. A circuit containing three blocks
was designed to detect small voltage signals from one or multiple electrodes with small
capacitance. The blocks were tested and demonstrated to increase the input impedance, amplify
the voltage difference between two pick-up electrodes, and down-convert the signals to output the
low-frequency signal reflecting the conductivity signals of fluid flowing through the shunt. The
circuit was also successfully tuned to reach different gains by changing the resistor values for
handling signals with varied amplitude and frequency ranges.
117

The circuitry designed to read voltage from the pick-up electrode was built and tested with
a breadboard system. To decrease the overall system noise, the breadboard model was converted
into a PCB model (Fig. 2-24 ). The PCB is under testing, and more work is underway to tune the
PCB to accommodate the sensing of different flow rates.

Figure 2-25. The schematic showing the layout of the PCB and its integration with the mBTS shunt. The
arrow indicates the flow direction of signals.

After the PCB is properly tuned, before using the PCB to sense the flow, steps will be taken
first to convert the electrical signal (voltage) to the conductivity signal of fluid. PBS solution with
different conductivity will pass through the shunt, and the electrical signals will be obtained
through the ring electrodes. The measurement will show if the sensor can respond to the interested
conductivity range of physiological fluids (ex. blood).
118

Next, the sensor will be tested for its ability to convert the electrical signal to a flow signal.
A sensor testing setup with multi-phase liquids and syringe pumps (PHD 2000 Series, Harvard
Apparatus) was built for this test (Fig. 2-26). The accuracy of how the syringe-pump setup injected
fluid was investigated by comparing the pump set injection volume and gravimetric measurements
of pumped fluid (Fig. 2-27).

Figure 2-26. Schematic showing the setup used mBTS sensor to transduce the flow signal.

First, only one syringe pump was activated, and the weight of the injected fluid was
measured every 10 s for at least 50 s at different flow rates, including 100 mL/min, 200 mL/min,
280 mL/min (Fig. 2-27), 300 mL/min and 368 mL/min (Fig. 2-28). The results showed the pump
fluid consistently as the linear regression of the graphs of the liquid weight versus time has an R
2

> 0.99. The accuracy of the syringe pump in percentage is calculated by dividing the measured
flow rate by the infusion rate with which the pump was set up. For the flow rates of 100 mL/min,
200 mL/min, 280 mL/min, 300 mL/min and 368 mL/min, the accuracy was 99.3%, 99.1%, 97.6%,
97.4% and 98.9%. The highest flow rate with a single syringe pump is 368 mL/min. To test the
119

high flow rate, two syringe pumps were connected in parallel through a three-way valve and can
reach a max flow of 738 mL/min. The measured accuracy for the flow injection at a rate of 560
mL/min with two pumps is 99.8%. Due to the capacity of the syringe loaded to the pump (300
mL), the setup is good for measurement of less than 1 minute. For a longer measurement duration,
the peristaltic pump was recommended to replace the syringe pump, which is feasible to
continuously pump fluid for a longer time.

Figure 2-27. Graph showing the weight measurement (y-axis) along the time of infusion (x-axis) for 100 –
280 mL.min. The weight measurement starts 5s after the pump is turned on for consistency between
measurements (Courtesy of L. Custer).

Figure 2-28. Graph showing the weight measurement (y-axis) along the time of infusion (x-axis) for 300
and 369 mL/min. The weight measurement starts 5s after the pump is turned on for consistency between
measurements (Courtesy of L. Custer).
120

To mimic the blood flow in the mBTS shunt, two liquids (water and milk) will be injected
through two syringe pumps into a mixer. The add-in of milk acts as the tracer to mimic the
stochastic fluctuation that blood provides. The mixed liquid is then passed through the shunt. The
sensors will generate several voltage plots through the voltage reading circuit. Then cross-
correlation is applied to those voltage plots for transducing the flow information.
Since this project's end goal is to implant the sensor into a swine model and acquire acute
blood flow data. We collaborated with physicians at Seattle Children's Hospital with extensive
experience treating patients with HLHS and operating mBTS placement procedures. After the
sensor is successfully characterized on the benchtop, the sensor and electronics will be further
miniaturized to fit the surgical need. The optimization may require us to decrease our pick-up
electrode's size further, which is no longer achievable by manually painting silver epoxy on the
shunt. If that is the case, the sheet Parylene electrode would be fabricated with the stripe electrodes
with holes etched out for suture (Fig. 2-29). The sheet electrode will be wrapped and secured to
the shunt by suturing through the holes or applying medical-grade adhesive. The electronics may
also be further miniaturized depending on the surgical setup.

Figure 2-29. Schematic showing the Parylne sensor designed to acquire flow data from a pig model.

121

References
[1] C. Mai et al., "National population‐based estimates for major birth defects, 2010–2014,"
Birth Defects Research, vol. 111, 10/01 2019, doi: 10.1002/bdr2.1589.

[2] D. S. Fruitman, "Hypoplastic left heart syndrome: Prognosis and management options,"
Paediatrics & Child Health, vol. 5, no. 4, pp. 219-225, 2000, doi: 10.1093/pch/5.4.219.

[3] T. Mahle William, L. Spray Thomas, G. Wernovsky, J. W. Gaynor, and J. Clark Bernard,
"Survival After Reconstructive Surgery for Hypoplastic Left Heart Syndrome,"
Circulation, vol. 102, no. suppl_3, pp. Iii-136-Iii-141, 2000/11/07 2000, doi:
10.1161/circ.102.suppl_3.III-136.

[4] B. Alsoufi et al., "Results of Palliation With an Initial Modified Blalock-Taussig Shunt in
Neonates With Single Ventricle Anomalies Associated With Restrictive Pulmonary Blood
Flow," The Annals of Thoracic Surgery, vol. 99, no. 5, pp. 1639-1647, 2015, doi:
10.1016/j.athoracsur.2014.12.082.

[5] D. J. Barron, M. D. Kilby, B. Davies, J. G. C. Wright, T. J. Jones, and W. J. Brawn,
"Hypoplastic left heart syndrome," The Lancet, vol. 374, no. 9689, pp. 551-564, 2009, doi:
10.1016/S0140-6736(09)60563-8.

[6] D. M. Cohen and H. D. Allen, "New developments in the treatment of hypoplastic left heart
syndrome," Current Opinion in Cardiology, vol. 12, no. 1, 1997. [Online]. Available:
https://journals.lww.com/co-
cardiology/Fulltext/1997/01000/New_developments_in_the_treatment_of_hypoplastic.8.
aspx.

[7] K. Bando et al., "Surgical management of hypoplastic left heart syndrome," The Annals of
Thoracic Surgery, vol. 62, no. 1, pp. 70-77, 1996/07/01/ 1996, doi:
https://doi.org/10.1016/0003-4975(96)00251-2.

[8] W. I. Norwood, "Hypoplastic left heart syndrome," The Annals of Thoracic Surgery, vol.
52, no. 3, pp. 688-695, 1991/09/01/ 1991, doi: https://doi.org/10.1016/0003-
4975(91)90978-Y.

[9] F. J. Paprottka, D. Klimas, N. Krezdorn, D. Schlarb, A. E. J. Trevatt, and D. Hebebrand,
"Cook-Swartz Doppler Probe Surveillance for Free Flaps-Defining Pros and Cons," (in
eng), Surg J (N Y), vol. 6, no. 1, pp. e42-e46, 2020, doi: 10.1055/s-0040-1702922.

[10] B. Adamson Philip et al., "Wireless Pulmonary Artery Pressure Monitoring Guides
Management to Reduce Decompensation in Heart Failure With Preserved Ejection
Fraction," Circulation: Heart Failure, vol. 7, no. 6, pp. 935-944, 2014/11/01 2014, doi:
10.1161/CIRCHEARTFAILURE.113.001229.
122

[11] A. J. Zemann, E. Schnell, D. Volgger, and G. K. Bonn, "Contactless Conductivity
Detection for Capillary Electrophoresis," Analytical Chemistry, vol. 70, no. 3, pp. 563-567,
1998/02/01 1998, doi: 10.1021/ac9707592.

[12] H. Ji, X. Gao, B. Wang, Z. Huang, and H. Li, "A new method for flow rate measurement
in millimeter-scale pipes," (in eng), Sensors (Basel), vol. 13, no. 2, pp. 1563-1577, 2013,
doi: 10.3390/s130201563.

[13] P. Schwan Herman and F. Kay Calvin, "Specific Resistance of Body Tissues," Circulation
Research, vol. 4, no. 6, pp. 664-670, 1956/11/01 1956, doi: 10.1161/01.RES.4.6.664.

123

Background
Hydrocephalus is a disease with elevated intracranial pressure (ICP) resulting in an enlarged
brain ventricle. It is commonly caused by imbalanced cerebrospinal fluid (CSF) production,
absorption, and circulation. Hydrocephalus can be congenital or developed after birth. Congenital
hydrocephalus happens in two out of every thousand new births in the United States [1]. The
standard treatment involves implanting a drainage shunt in the brain ventricle to safely absorb the
excess fluid in other body parts such as the abdomen (Fig. 3-1).

Figure 3-1. Schematic showing an implanted ventricular shunt for draining access CSF into the abdomen.
Zoomed-in picture on the left shows a normal shunt versus a clogged shunt.
CHAPTER THREE
A MULTIMODAL SENSING SYSTEM FOR
DETECTING HYDROCEPHALUS SHUNT
PATENCY
124

Hydrocephalus is a life-long condition that has no cure; patients need to live with the shunt
for their whole life to regulate brain pressure. Approximately 125000 patients live with the
ventricular shunt, and 33000 new shunts are implanted annually in the United States [1]. However,
shunts often fail due to tissue growth which clogs the drainage holes (Fig. 3-1). About 18-32% of
shunts failed within five years after implantation [2]. The number increases to 50% for pediatric
patients as the shunt size is usually smaller. Shunt failure is challenging to detect from symptoms
including headache, nausea, and vomiting; these are general, nonspecific, and may indicate other
maladies [2]. A suspected shunt malfunction can involve a hospital visit for diagnostic imaging,
including magnetic resonance imaging, computed tomography, and plain X-rays. The additional
imaging is costly and can expose patients to radiation.
CSF Flow Monitoring
Sensors have been developed to monitor CSF flow to detect hydrocephalus shunt
malfunction. The ShuntCheck, developed by NeuroDx Development, is an FDA-approved device
used to measure CSF flow rate, which is a non-invasive thermal flow sensor positioned on the skin.
There are many challenges while working with this device. The sensor reading depends on the
accuracy of identifying the implanted shunt’s location underneath the skin and the proper
alignment of the sensors with the CSF flow. An ice pack is set upstream, cooling the flowing CSF.
The flow is then transduced through the induced temperature gradient by downstream thermistors.
This approach can cause discomfort but, more importantly, has limited resolution and accuracy. It
also provides single discrete measurements, which do not provide sufficient information for
detecting progressive shunt failure caused by gradual tissue growth or accumulation.
For early detection of shunt malfunction, a sensor capable of periodically monitoring
physiological flow with high resolution is needed. Sensors have been developed for continuous in
125

vivo measurement of CSF flow in research labs, including ultrasonic sensors [3]. However, the
operation of the sensor requires external, expensive ultrasound equipment. Flow measurement
relies on access to an ultrasonic imager and requires extensive training. It measures volumetric
flow in shunts transdermally, has low resolution, and is impractical to wear. Another research
group took a different approach and built a thermistor-integrated shunt to measure CSF flow from
a pig model in 2019 [4]. The sensor is bulky and wired, required the testing subject to lay still on
the bed, and was feasible for monitoring flow in acute studies. Other thermal flow sensors can be
implanted [5], but most have a silicon construction subject to corrosion, fracture, and exposure to
the body environment while operating in vivo [6].
ICP Monitoring
Physiological pressure monitoring is a standard of care in treating many health conditions
and diseases, including hydrocephalus. Pressures such as ICP that span a relatively small range (7-
15 mmHg [7]) require sensors that can resolve changes down to 1 mmHg. Those pressures are
difficult to acquire using non-invasive methods and are most accurately tracked with implanted
pressure sensors with real-time monitoring capability. Continuous measurements of ICP with high
resolution are currently possible in a hospital setting. There are commercial pressure sensors with
catheter-tip transducers, such as the Camino® fiber optic parenchymal bolted catheter and Codman
MicroSensor®. The limitations with those traditional membrane-based sensors are susceptible to
biofouling, fluid ingress, and ion permeation [8], all of which can affect performance when used
in vivo.
126

Multimodal Sensing for Detecting Hydrocephalus Shunt Patency
Since hydrocephalus is a lifelong condition with no cure, the sensor needs to work
discretely over a long time. An ideal sensor would be implanted in line with the shunt and be able
to continuously measure multiple parameters, including CSF flow and ICP. Most existing sensors
only detect one parameter, CSF flow or ICP, but not both. CSF circulation is dynamic, and it is
driven by pressure. A multi-modal sensor will provide more information for diagnosing shunt
malfunction.
The work here highlights the development of a smart shunt incorporating multiple sensors
to provide early, definitive warning of shunt malfunction and reduce hospitalization. The sensor is
made from biocompatible and non-corroding Parylene C, and it does not rely on movable parts
such as the membrane built in many conventional sensing devices. The sensor is designed to
measure flow, temperature, and pressure in real time through electrochemical impedance
transduction of biological saline solutions [9-15]. The previous work has demonstrated the sensing
fundamentals and the sensor’s functionality through benchtop experiments. The focus here is to
show the device, package modifications, and calibrations for in vivo investigation. The first part is
about testing the impedimetric flow and temperature sensor in a survival swine model for the first
time and the sensor’s performance compared to benchmark commercial flow sensors. The second
part will talk about the current version of the nanobubble-based pressure sensor modified for in
vivo testing.
Flow and Temperature Sensor
The flow and temperature sensor is made from Parylene and platinum thin films that easily
fit into the shunt system without blocking the fluid path (Fig. 3-4). The sensor is designed to work
in the body’s saline environment, using the physiological fluid as the sensing element. The sensing
127

fundamental is based on the concept that the impedance of body fluid changes with temperature.
The impedimetric sensor consists of an upstream, resistive, serpentine platinum (Pt) heater (~400
Ω) and downstream sensing electrode pairs (exposed Pt) on a Parylene substrate (Fig. 3-2) [16].

Figure 3-2. Low profile, flexible sensor die packaged in a Luer module. Fabricated Parylene impedimetric
sensor after release from the silicon substrate. Inset shows a close-up of the flow sensor elements (yellow
box) [17]. Reprinted with the permission of the author and the Transducer Research Foundation

The flow was detected using a time-of-flight approach: the upstream heater was activated by
a voltage pulse, heating the body fluid. The heat transfer downstream was then detected through
the electrolyte’s impedance response, which is sensitive to thermal changes. Specifically, the flow
is transduced from the maximum rate of impedance change – defined here as the heating slope
(HS, %/s) – during the 10-second heater activation was used to transduce the flow rate (Fig. 3-3)
[14, 15]. Measurements were acquired every 80 s (including 60 s for cooling between
128

measurements) to minimize the overheating temperature (~2 °C). Impedance was obtained at 50
kHz to maximize the contribution from solution resistance.

Figure 3-3. Impedance response of sensor over a heating cycle. Flow rate is transduced from the normalized
maximum rate of impedance (Z) change during the “Heater ON” phase (10-20 s) [17]. Reprinted with the
permission of the author and the Transducer Research Foundation

Body temperature usually remains at 37 ℃ but can change if a patient has a fever. A
temperature sensing mode was added to the sensor to track body temperature changes for in vivo
tests. The upstream platinum heater was used as a resistance temperature detector (Fig. 3-2). The
temperature coefficient of resistance for platinum is 0.392 %/℃. In other words, the resistance of
platinum changes 0.392% for 1 ℃ of temperature change. The temperature reading is transferred
from the measured electrical current of the heater [15].
129

Materials and Package
Multilayer micromachining on a silicon support wafer was used to fabricate the sensor [15].
The fabrication process is summarized below. A 10 µm thick Parylene insulation layer was
deposited through chemical vapor deposition (CVD). To define the metal features, a 2 µm thick
AZ 5214-IR photoresist layer was spin-coated, and Pt (2000 Å) was deposited through e-beam
deposition followed by lift-off in heated acetone and rinses in isopropanol alcohol (IPA) and
deionized (DI) water. Another 10 µm thick Parylene layer was then deposited. 15 µm thick AZ
4620 photoresist was spin-coated to serve as the sacrificial layer. The metal pad, electrode sites,
and device shape were etched out through deep reactive ion etching. Photoresist was removed by
sequential acetone, IPA, and DI water baths, and the individual sensor was released using a razor
blade to cut along the etched shape. The device was thermoformed in a vacuum oven at 200 ℃ for
48 hours to enhance the adhesion between Parylene and metal layers.
The contact pad region was attached to a PEEK adhesive backing and mated to an FFC cable
through a zero-insertion force (ZIF) connector. This assembly was packaged with the sensor
positioned in the lumen of a luer module and affixed with epoxy (EpoTek 353 ND-T) for
integration into the EVD system (Fig. 3-6). The sensor was connected to a printed circuit board
(PCB) with a commercial flow sensor LD20 (Sensirion) via the FFC (Fig. 3-6). A cable interfaced
the PCB to the power source and recording system, allowing both the impedimetric and LD20
sensors to be activated to acquire simultaneous flow measurements.
Benchtop Calibration
The impedimetric sensor was calibrated at the benchtop using phosphate-buffered saline
(PBS) and compared against a commercial sensor LD20 over the flow range of 0 to 1000 µL/min
(Fig. 3-7). The lightweight commercial flow sensor LD20 from Sensirion was used here to directly
130

measure the CSF flow rate simultaneously with the impedimetric sensor to serve as a benchmark
flow measurement. Even though the LD20 sensor is not designed to be implanted, it can measure
flow rates up to 1000 mL/hr, which is sufficient for monitoring the CSF flow.

Figure 3-4. Schematic of fully packaged impedimetric and LD20 sensors. Blue arrows indicate the
direction of CSF flow [17]. Reprinted with the permission of the author and the Transducer Research
Foundation

A Fluigent pressure unit was connected to a PBS source to pump fluid at different flow rates
to mimic the brain ventricle with different ICP. To simulate in vivo conditions, the testing setup
was placed in a temperature-controlled chamber (Fig. 3-5). A temperature probe was used to
actively monitor the temperature of the reference PBS (with a similar fluid volume as the source
PBS) to ensure the liquid flowing through the sensor reached the expected temperature. The fluid
temperature was varied around body temperature (37 °C), 20, 30, and 40 °C specifically to
investigate the impact of temperature on sensitivity.
131

Figure 3-5. Benchtop testing setup for calibrating the sensor. Phosphate-buffered saline (PBS) flow and an
oven were used to mimic the body’s warm saline environment [17]. Reprinted with the permission of the
author and the Transducer Research Foundation

Fig. 3-6 showed a representative calibration curve of an impedimetric sensor over the flow
range of 0 to 1000 µL/min. A non-linear response with flow rate was recorded, requiring two
separate calibrations. In a 2016 clinical study conducted with hydrocephalus patients with
implanted drainage shunts, the measured CSF flow rate had a mean of 133- 163 µL/min [3]. While
calibrating the impedimetric sensor under the low flow range of 0 to 200 uL/min, one calibration
measurement was taken every 20 μL/min. The measurement was taken every 250 µL/min for the
higher flow rates.

132

Figure 3-6. Graph showing a representative calibration curve of an impedimetric flow sensor over the flow
range between 0 and 1000 µL/min.

For flow > 200 μL/min, HS was linearly correlated with flow rate (R
2
=0.94) but quadratic
for lower flow rates (R
2
=0.95) (Fig. 3-7). Considering different regression curves by flow range,
the sensor accuracy (% error) improved over sixfold compared to previous work for the low flow
rate domain, which corresponds to the typical CSF flow range observed in patients with an EVD
[3]. Longer response times are expected at low flow, further supporting two separate calibrations
[18].
133

Figure 3-7. Graph showing the separated calibration curves for high (top, linear, > 200 µL/min) and low
(bottom, quadratic, < 200 µL/min) flow rates.

134

Sensor Test in Swine
To pave the way for fully implanted devices in patients, a pathway was developed to first
validate devices in a simplified model with an external ventricular drain (EVD). This system can
be placed acutely to relieve high intracranial pressure. A proximal catheter is inserted into the brain,
and the distal end drains into an external bag. The sensors can be connected to the external drainage
tube with limited disturbance to existing therapeutic procedures in clinical settings. This setup
exposes the sensor to CSF with realistic flow dynamics but does not have to be fully implanted for
these initial testing and validation phases.
The sensor was first tested in a pig model, which is compatible with an EVD system, and
measured the CSF flow drained from the brain ventricle. Healthy pigs were used, and the CSF
drainage rate was directly controlled by modifying the ICP. The animal procedure is developed
based on an acute study conducted by Qin et al. in 2019 [4] in which a thermistor measures CSF
flow drained from the pig’s ventricle under healthy and hydrocephalus conditions induced by aCSF
infusion. It is a non-survival study, aCSF was infused at various flow rates to test the thermistor’s
response to detecting varied flow ranges. The animal was sacrificed after the measurements on
surgery day. Since hydrocephalus is a lifelong condition, to better mimic the chronic condition, a
modified model for a survival pig study is developed to enable the continuous test of sensors over
weeks for evaluating the sensor’s performance and stability over a long period. To induce chronic
hydrocephalus condition, the common practice is to use Kaolin to induce the hydrocephalus
condition chronically such that the pigs will live with hydrocephalus during their life. To decrease
the discomfort of pigs over time, here we use aCSF infusion to induce hydrocephalus temporarily
to allow us to evaluate the devices. After the infusion stops, the ICP should recover to a healthy
135

range, the hydrocephalus symptoms are expected to go away, and the pig should recover to a
healthy state.
Specifically, a temporary hydrocephalus condition is introduced by infusing artificial CSF
(aCSF) into a pig’s ventricle through a syringe pump, causing the ICP and outflow to increase (Fig.
3-8). Flow measurements are acquired by an impedimetric sensor and a commercial LD20 sensor
connected to an EVD. The goals of this swine study are to verify CSF flow sensing in response to
the changing ICP and to compare the performance against the benchmark LD20 sensor. The
preparation process of the sensor and setting up the testing system will be described first. Next,
surgical procedures for animals will be presented in sequence to highlight the refinement process
of the animal model based on the observations and lessons learned from each surgery. In the end,
the sensor measurements acquired from those in vivo studies will be presented along with future
work.

Figure 3-8. Schematic showing the testing setup used in the swine study.

136

Sensor Preparation for Surgery
To minimize trapped bubbles in the system during testing, the sensors were connected,
primed, and stored one day before the recording session. The steps for assembling and priming are
summarized here. The 3-way valve #1, impedimetric flow sensor #1, LD 20 #1, adjustable valve
(Medtronic Strata II Valve Regular), impedimetric flow sensor #2, LD 20 #2, and 3-way valve #2
were assembled together through Luer connectors (Fig. 3-9).
The assembly was then flushed with sterile aCSF with syringes (the left one was filled with
aCSF; the right one was left empty) in one direction to remove any air bubbles and fill the line
with sterile aCSF. If air bubbles remained after a few rounds of aCSF flushing, holding the
assembly perpendicular and flushing aCSF through the assembly upward helped to remove the
leftover bubbles.

Figure 3-9. Photo showing the assembly of the sensors and valves for priming purposes. The devices listed
from left to right are 3-way valve #1, impedimetric flow sensor #1, LD 20 #1, adjustable valve (Medtronic
Strata II Valve Regular), impedimetric flow sensor #2, LD 20 #2 and 3-way valve #2.

Figure 3-10. Photo showing detachment of the sensor assembly from the syringes used for priming.
137

After the assembly was primed, the 3-way valves at two ends were turned off (rotate 90°
clockwise), and the syringes were detached from the valves (Fig. 3-10). An LCR meter was then
connected to the primed impedimetric sensor through a PCB for pretesting. The resistance of the
heater was measured; the working heater should end up with a tested resistance in the range of
~300-350 Ω. The impedance (50 kHz, 2V) between two sensing electrodes is expected to be 4-7
kΩ.
Cadaver Surgery
The purpose of this cadaver surgery is for the surgeon to practice the surgical procedure of
implanting the drainage shunt into the ventricle. Sensors were also connected to the drainage tube
to acquire flow measurement of CSF to familiarize with the measurement acquiring process. The
surgical procedure and sensor testing were performed according to the guidance of both the
Institutional Animal Care and Use Committee (IACUC) and the Department of Animal Resources
(DAR) of the University of Southern California (USC). All the catheter placement surgical
procedures were performed by CHLA physician Dr. McComb with USC DAR’s assistance.
30-36 kg pigs were selected for this study because of their similarities in size and anatomy
with humans, especially in the cerebroventricular architecture. In addition, pigs are readily
available and easy to work with compared to other similar animal models (i.e., dogs/non-human
primates). Pigs are a common animal model for studying hydrocephalus [4, 19]. In addition,
similar procedures of placing a ventricular shunt were well evaluated in the pig model [4, 19]. In
addition, pig CSF physiology has been broadly studied in the literature [20, 21].
A 32 kg pig was sacrificed from another study and prepared for this cadaver surgery. The
surgical on the scalp was prepared by shaving the hair and sterilizing afterward. A 5-10 cm long
linear incision was made, and the scalp was detached from the skull around the incision. A hand
138

drill was used to drill a burr hole with a diameter of 1.5 cm on the coronal suture with a 1-1.5 cm
distance from the midline. The lamboid suture was not exposed. A drainage shunt terminated with
a ventricular reservoir was inserted into the frontal horn of the right lateral ventricle. The outflow
of CSF confirmed the proper placement. The sensors were hooked up to the drainage tube to
acquire flow measurements.
Several observations were made from this cadaver surgery. First, after the placement of the
drainage shunt, not much CSF flowed out. Even after some aCSF was infused into the ventricle
through the drainage tube, no CSF flowed out. One possible reason is that the pig was dead, and
no more CSF was produced or left in the ventricle, as most of the CSF got drained out during the
surgical placement of the shunt. Another possibility is that the drainage tube was too narrow, with
an inner diameter of 1.02 mm. The narrow tube posted high pressure and stopped the CSF from
flowing out from the ventricle. The clinical ventricular shunt used for draining the CSF has an
inner diameter of 1.3 mm, which is wider than the one used in cadaver surgery. The drainage tube
was switched to a wider tube with an inner diameter of 1.3 mm for future study.
A manometer with an inner diameter of 6 cm was used to measure the ICP of this pig. It
turned out the manometer size was inappropriate and took up a large volume of CSF to fill the tube
for acquiring pressure readings. It also resulted in a long waiting time for the liquid level to
stabilize and get the ICP reading. A manometer with a smaller inner diameter (3 cm) was used for
later studies. The cadaver surgery also confirms the surgical site by properly placing the ventricular
catheter. The proper placement is confirmed by the outflow of CSF.
First Round of Survival Study
Yucatan pigs were used for survival study instead of domestic Yorkshire pigs because of the
slower growth rate for chronic study. The surgical procedure and sensor testing were performed
139

according to the guidance of both the Institutional Animal Care and Use Committee (IACUC) and
the Department of Animal Resources (DAR) of the University of Southern California (USC). The
animal arrived at the facility one week before the study for acclimation and was fasted for at least
12 hours before surgery.
Before the surgical procedure, the 36 kg Yucatan female pig was given anesthesia, antibiotics,
and analgesics. After anesthesia, the animal was placed on a surgery table in a prone position. The
head area was clipped, cleaned, and prepared with a chlorhexidine (or iodine) scrub followed by
70% isopropyl alcohol (IPA) 3 times. Similar to cadaver surgery, an incision was made, and a burr
hole was drilled on the coronal sure. The ventricular reservoir was inserted into the lateral ventricle
through the burr hole. Differing from cadaver surgery, an 4-5 cm long arc-shaped incision was
made (shorter than the incision made in cadaver surgery) on this pig. A shorter incision was
attempted for this first survival study to help the post-surgical recovery of the pig.
CSF's outflow from the silicone reservoir's sidearm confirmed proper ventricular catheter
placement. After the placement was confirmed, a plug was used to seal the reservoir sidearm. The
reservoir was then bent and placed under the scalp. The wound is closed with 2 layers of suture.
Intraoperative crystalloid fluid was given to the animal through intravenous therapy (IV) at
a rate of 3-5 mL/kg/hr throughout the surgery. A monitor machine was wired to the animal,
documenting its health state every 5-10 minutes during the surgical procedure and recording
sessions. The parameters recorded include heart rate, respiratory rate, blood pressure, and body
temperature. A warming device such as a Bair Hugger
TM
was used to maintain thermal homeostasis.
A rectal probe was used to monitor the body temperature.
140

After the animal was prepared for recording, the sterile manometer, infusion line (to the
syringe pump), and butterfly needle line were assembled to the primed sensor assembly with
syringes (Fig. 3-11). To maintain sterility, special attention was taken to prevent the open ports of
the valves from touching the surface of the testing table. Any uncapped ports are later capped with
a Luer plug to avoid any openings to the sterile system that may be contaminated and later cause
infections in the animal. To prime the newly added catheter, a syringe filled with sterile aCSF was
flushed toward the end connecting to the butterfly needle. Flushes of aCSF under the same
direction were repeated a few times until all the visible air bubbles were removed (Fig. 3-11). After
the catheters were primed, the manometer was prefilled to a 10 cm H2O mark with priming

Figure 3-11. Photo demonstrating the connection and priming of the sensor assembly with the newly added
infusion and butterfly needle catheters.
141

Figure 3-12. Photo demonstrating the connection of the primed assembly of sensors and catheters to the
syringe pump used for aCSF infusion.

syringes filled with aCSF. The priming syringe at the infusion side was then connected to the
syringe pump (Fig. 3-12). The syringe pump was set up by inputting the diameter of the syringe
(23 mm for a 30 mL needle; 14 mm for a 10 mL needle).
Similar to what is done in clinical settings [22], the 25-gauge butterfly needle with primed
catheters was used to puncture the reservoir from outside the skin to get access to CSF (Fig. 3-13).
The manometer monitors the intracranial pressure (ICP) during the CSF infusion and drainage
activities. The manometer was prefilled with sterile aCSF to the level of the pig's normal ICP. The
ICP is then taken by opening the valve of the manometer (Fig. 3-14), allowing the CSF to drain
into the manometer (Fig. 3-15). The liquid level in the manometer rises or reduces depending on
if the pressure increases or decreases. The fluid level in the manometer stabilized ~5 minutes after
142

the valve was left open, and the pressure was taken as the height (marked by a red buoyancy ball)
of the liquid in the manometer with a unit of mmH2O (Fig. 3-14). According to the literature, the
density of the CSF is 1.00059 ± 0.00020 (mean ± standard deviation), which is close to water [23].
The gravimetric flow was acquired and used as a reference to compare with the sensor data. In the
clinical setting, the gravimetric flow was acquired hourly by reading the drainage volume from the
EVD drainage bag, as shown in Fig. 3-15. For the sensor test with swine, the gravimetric
measurements were acquired every three flow measurements (240s) and compared with the flow
rates taken through the sensors.

Figure 3-13. Testing setup for the swine study. The ventricular catheter was implanted into the ventricle
with the reservoir under the scalp then the wound was sutured. A needle with a catheter was connected to
the reservoir (yellow dashed box) to access CSF via the shunt for draining or infusing. Flow sensors were
connected to the output to measure CSF's flow rate and temperature [17]. Reprinted with the permission of
the author and the Transducer Research Foundation

143

Figure 3-14. Testing setup for the swine study highlighting the location and connection of the manometer
used to monitor the ICP. The liquid's height in the manometer corresponds to the height of the buoyancy
ball in the tube. The zoom-in box showed the valve position when the manometer was off. An “ON” position
corresponds to a 90° angle between the direction of the valve and the tube.

Figure 3-15. Schematic showing the setup for measuring the ICP through a manometer. The red arrows
indicate the flow direction of CSF. The 3-way valve #2 was turned off for the infusion line and sensor paths.
144

Several observations were made from this specific pig. First, after the reservoir was tapped,
no CSF flowed out. ACSF was infused at a rate of 1 mL/min into the ventricle for 5 mins, and the
liquid level in the manometer rose demonstrating that the aCSF was successfully induced and
caused the elevated ICP to push the CSF out. Then the EVD bag height was lowered to create
some pressure difference to help CSF flow. The adjustment was stopped until the CSF started to
drip into the collection bag. It turned out that simply lowering the CSF bag did not control the CSF
flow well. Oftentimes, after the CSF collection bag was lower, the CSF was over-drained, and the
primed tube was emptied. In addition, the mark on the drainage bag did not give enough resolution
to read the drained CSF volume collected in 4 mins.

Figure 3-16. Photo demonstrating the swine study setup highlighting the setup used for gravimetric flow
measurements.

To address the problem of low resolution and a poorly controlled CSF drainage rate, a
drainage bottle placed on a weight scale was used to replace the clinical EVD bag (Fig. 3-16). A
145

customized tube with decreasing diameter was used to add resistance to prevent the CSF from
draining too fast. The drainage bottle was also placed in the marked location on the weight scale
between measurements to limit any variation associated with scale measurement.
After the surgical and recording sessions, the pig spent 4 days in post-surgical recovery. This
animal did not recover well from surgery and exhibited neurologic symptoms, including head
pressing, slow movement, dull mentation, and poor appetite. Due to the poor recovery conditions,
the animal was sacrificed one week after surgery.
Tracing back to the animal records, many reasons could cause poor recovery from the
surgery, including surgical placement of the shunt and anesthesia overdosing. For this specific pig,
it took abnormally longer (~ 3 hours) to sedate the pig on the surgery day, and multiple injections
of anesthetic drugs were given to the pig. The pigs woke up in the middle of the recording sessions,
and more anesthetic drug was administered. These actions might cause overdosing.
During the aCSF infusing and draining sessions, a strawberry-colored liquid (looked like
CSF) discharged from the pig’s right nose (the same side where the ventricular reservoir was
placed). Considering some papers showed that CSF could be collected from the nose in
hydrocephalus dog models [24]. The nasal discharge to this pig was possibly caused by overdosing
on aCSF infusion. Therefore, the CSF-infused rate and volume are carefully controlled for the next
pig.
One last critical observation from this pig experiment is that the priming of the 25-gauge
needle used for tapping the reservoir is essential for initiating the CSF flow from the ventricle to
the drainage line. It is often challenging to ensure that the reservoir tapping procedure does not
introduce air bubbles. If a bubble is induced during the tapping process, it is difficult to remove as
the system is sealed once the needles are tapped into the reservoir. To help with the priming, an
146

additional burr hole was made on the other side for the next pig. One ventricular reservoir was
inserted into each burr hole, and one was used as the priming line to remove the bubble induced
during the tapping process.

Figure 3-17. Photo showing the hand drill used to create the burr hole after the incision was made on the
scalp.

Second Round of Survival Study
A 1.5 by 4.5 cm incision was made on the pig’s scalp (Fig. 3-17). Two ~1.5 cm diameter
burr holes were made with a hand drill (the commercial hand drill used for human patients at the
hospital for the surgical procedure of inserting an external ventricular drain) on the coronal suture,
with a 1-1.5 cm distance from the midline on each side of the skull for inserting two ventricular
catheters to target the pig's left and right lateral ventricle (Fig. 3-18).
147

Metal guide wires were used to insert the proximal ends of two ventricular reservoirs
(ventricular shunt terminated with a reservoir) through the burr hole to a depth of 3-8 cm for
targeting the frontal horn of the right and left ventricles of the brain (Fig. 3-19).

Figure 3-18. Photo showing the burr holes made by the drill hole for targeting the left and right ventricles.

Figure 3-19. Photo showing the ventricular reservoirs inserted into the ventricle.
148

Figure 3-20. Photo (left) demonstrating the proper placement of the ventricular reservoir by the observed
CSF outflow from the reservoir sidearm. Image (right) with a yellow mark showing the closed part of the
reservoir.

Proper ventricular catheter placement was confirmed by the outflow of CSF from the
sidearm of the silicone reservoir (Fig. 3-20). After the placement was confirmed, a plug was used
to seal the reservoir sidearm (Fig. 3-20). The reservoir was then bent and placed under the skin.
The reservoir wings were sutured to the pericranium with a dissolvable suture (Fig. 3-21). Finally,
the skin is sutured with a non-dissolvable suture (Fig. 3-21), and the suture is removed two weeks
after surgery.

Figure 3-21. Photo (left) with blue mark demonstrating the location of the sutured added to the reservoir
sidearm for attachment to pericranium. Picture (right) with a blue arrow showing the suture for skin incision
closure.
149

After the wound was sutured up, both of the reservoirs were tapped. As mentioned
previously, one of the reservoirs was used for priming purposes to remove any air bubbles
introduced during the tapping process. However, after 5 mL aCSF infusion through one side,
nothing came out from the other side.
The anatomical structure of the ventricle is a possible reason causing this. Specifically, the
infused aCSF in the right lateral ventricle may not always go into the left lateral ventricle as it also
goes into the 3
rd
and 4
th
ventricles and the subarachnoid space. Another possibility can be the
placement of the shunt in the ventricle and the blockage of the shunt.
To get a better idea of the surgical placement of the ventricular reservoir, CT and MRI
imaging will be added to future animals to inspect the location of the implanted shunts in the
ventricle. Brain extraction and fixation will also be planned for the next animal for the final
investigation of the shunt’s location in the ventricle after the animal is euthanized. The details
regarding imaging and tissue collection will be described in detail in the later sessions.
Since the last pig did not recover well, this pig was given a more extended recovery period
of 1 week (compared to 4 days for the previous pig). The pig recovered quickly from the surgery
this time; no neurological symptoms showed up. However after the first recording session one
week after the surgery, the pig did not recover well and started to move slowly. In addition, the
pig was lethargic and had a poor appetite. However, no neurological symptoms like the ones seen
in the previous pigs showed up. The pig was euthanized one week after the symptoms showed for
the first time. This supports the need to add imaging to the future pig, which will help provide
additional information on the pig’s recovery from surgery and recording sessions, and the
placement of the catheter. To understand the difference between the two recording sessions and
how it correlated to the different recovery conditions of the pig. Tracing back to the animal records,
150

the longer recording length and larger volume of aCSF infusion during the recording sessions may
cause poor recovery. Therefore, the recording session for future pigs is modified to a maximum of
4 hours.
The other observation from this experiment is that the Yucatan pig’s cranium thickness was
not always correlated to age and weight. The previous pig was ~9 months old and weighed ~36 kg
around the surgery. The pig in this experiment is ~6.5 months old and weighs ~33 kg. However,
this younger pig has a thicker cranium (1-2 cm thicker) than the last pig, making it challenging to
secure the current size of the ventricular reservoir to the cranium. Due to the varied cranium
thickness of the pigs, future experiments are planned with lighter pigs weighing ~30 kg.
Sensor Measurement
To prepare for sensor testing, the pig is sedated. Similar to what is done in clinical settings
[22], a 25-gauge butterfly needle was used to puncture the reservoir from outside the skin to get
access to CSF (Fig. 3-22). The sensors, pressure transducers/manometer, and aCSF infusion unit
(syringe pump) were then connected to the butterfly needle through Luer connectors.
Impedimetric and commercial LD 20 sensors are connected to the drainage line through 3-
way valves. Multiple 3-way valves were added to the liquid path for priming purposes and to
remove visible air bubbles or clogs with aCSF flushes. The shunt, catheter, connector, and sensor
parts that have direct contact with the infusing aCSF were sterilized by hydrogen peroxide
sterilization processes/ethylene oxide sterilization (EtO) before usage.
The sensor system used for the swine study was packaged in a 3D-printed casing (Fig. 3-23).
The case contains a transparent window to detect air bubbles during the pre-measurement priming
stage. Each sensor box includes a pair of Parylene impedimetric and LD20 sensors. Both sensors
are connected to a recording box through a Molex connector. The boxes contain a recording PCB
151

with an SD card slot for storing the data. A lithium polymer battery was used to power up the PCB
and sensors.

Figure 3-22. Schematic (left) showing the reservoir tap procedure [25]. Photo (right) shows two needles
tapped in the reservoirs and sensors connected to the needle outlet highlighted with a yellow arrow.

Figure 3-23. Photo demonstrating the swine study setup highlighting the placement of the sensor box and
the detailed design of the sensor box.
152

Two infusion modes were tested. First, the sensor was investigated under bolus infusion. A
single dose of aCSF was infused through a syringe pump into the pig’s ventricle to bring the
intracranial pressure up to mimic the hydrocephalus condition (Fig. 3-24). The pressurized CSF
will flow out while the sensors are taking measurements. Fig. 3-25 shows the measured flow and
temperature of drained CSF acquired by the impedimetric (black curve) and LD20 sensors (blue
curve). Sequential aCSF boluses were infused at 80, 320, and 560 s. Measurements were obtained
after infusion. The impedimetric sensors successfully tracked temperature and flow rate as the CSF
drained. The CSF temperature consistently reads at around room temperature because the CSF
quickly cools while flowing out through the external catheter.

Figure 3-24. Photo demonstrating the sensor testing setup under the bolus aCSF infusion mode in time
sequence.

153

Figure 3-25. Flow and temperature of pig CSF acquired by impedimetric Parylene and LD20 sensors
acquired after sequential aCSF bolus infusions at 80, 320, and 560 s. Measurements were obtained after
infusion.

Figure 3-26 shows the gravimetric flow taken once every 240s. The result showed that the
gravimetric flow is in the range of the measured flow from sensors but higher than the average
measured flow rate. The variation is caused by the over drainage of CSF, as the drainage valve
was manually closed after the sensor measurements were completed. The CSF continued to drain
into the collection bottle for certain times after the sensors finished acquiring the data.
The sensors were then investigated under constant flow. The aCSF was constantly infused
until 640 s, and measurements were taken during infusion (Fig. 3-27). The result showed that the
impedimetric sensors successfully tracked temperature and flow rate under the case in which the
flow was relatively constant (Fig. 3-28).

154

Figure 3-26. Graph showing the comparison between gravimetric flow and flow measurements acquired
from impedimetric Parylene and LD20 sensors with three bolus aCSF infusions.

Figure 3-29 shows the response of the Parylene impedimetric sensor and the LD20. The
sensors’ responses showed good agreement, as the measured flow rates were linearly correlated
with an R square value of 0.99. The data taken from different pairs of impedimetric and LD20
sensors in different tested pigs are reported in Fig. 3-30. Impedimetric and LD20 sensors showed
good agreement across other devices tested in pigs.

155

Figure 3-27. Photo demonstrating the sensor testing setup under the constant aCSF infusion mode in time
sequence.

Figure 3-28. Flow and temperature of pig CSF acquired by impedimetric Parylene and LD20 sensors
acquired under constant aCSF infusion. The infusion was stopped after 640 s, and measurements were taken
during infusion.
156

Figure 3-29. The measured flow rate of the impedimetric sensor (y) was linearly correlated to the LD20
sensor (y=0.95 x, R
2
=0.99) [17]. Reprinted with the permission of the author and the Transducer Research
Foundation

Figure 3-30. The measured flow rates of the impedimetric sensors (y) were linearly correlated to the LD20
sensors across different devices and animals as R
2
>0.9.
157

After the measurement, the needle was removed, and the hole in the reservoir would self-
heal. The multimodal sensor, pressure transducer (manometer), and infusion unit were detached
from the needle at the 3-way valve. The needle was disposed, and the sensors were rinsed with DI
water and stored in dry condition before the next recording sessions. The animal was weaned from
the anesthesia machine. If necessary, the animal would be on the ventilator until it started breathing
spontaneously. Then all IV lines and patient monitoring equipment would be disconnected.

Figure 3-31. Comparison of flow measurements over one week for the impedimetric and LD20 sensors
(implantation on Day 1) [17]. Reprinted with the permission of the author and the Transducer Research
Foundation

158

Once the animal is breathing on its own, it will be removed from the surgery table and
carefully placed into a recovery cage in the recovery room. After fully recovering from anesthesia,
the animal will be returned to its cage/run/pen. The animal will heal for 2-7 days, depending on
the health state. The sutures will be removed 10 - 14 days after surgery. 1
st
chronic measurement
was taken as early as two days after the surgery based on the animal's health state.
Figure 3-31 compares flow measurements over one week for the impedimetric and LD20
sensors. The impedimetric sensor reliably tracked flow at different time points over one week: day
1 (surgery day), 6, and 7 (Fig. 3-34). A stable RMS error of 21.39, 36.62, and 32.25 μL/min was
recorded.
Planninng for Future Pig Experiment
CT or MRI imaging will be conducted on the animal's head before and after surgery. CT will
be attempted first as it provides better resolution for our purpose of inspecting the implanted shunt's
location in the brain ventricle [26]. MRI imaging will be used as the backup when CT is not
available. The details regarding the imaging procedure and frequency are described below.
Before the imaging session, the pig will be anesthetized and transferred to the imaging
facility. During imaging, animals will be monitored for responsiveness or spontaneous movement.
Special attention will be paid to respiratory rate, eye movement, eye position, temperature, and
heart rate/rhythm.
The 1st imaging will be conducted before surgery for surgery planning. The 2nd imaging
will be planned for 2-10 days after surgery, depending on the animal’s recovery from surgery. This
imaging is used to inspect the recovery condition of the tissue around the implants in the brain
ventricle and shunt’s location. Another imaging will be conducted after at least one post-surgical
recording to inspect the tissue condition around the implants and look for any shunt displacement.
159

Besides the imaging, the animals will also be euthanized with 1 mL/10 lbs euthasol IV under
general anesthesia. After the pig is euthanized, the brain tissue will be collected and fixed in PFA
overnight in a hood. After the brain is fixed, the brain will be cut along the catheter to inspect the
catheter's location in the brain ventricle.
Pressure Sensor
The pressure sensor is designed to be on the same die as the flow and temperature sensor
and integrable to the implantable hydrocephalus shunt, which is designed to be compact and
operates in the body’s saline environment, leveraging the near-instantaneous response of a gas
bubble to pressure changes. As pressure increases, the size of the bubble decreases. In prior work,
the sensing concept was introduced and demonstrated [9-11]. A single microbubble is generated
through faradaic gas evolution in a microchannel. The pressure is transduced by tracking the
bubble size through electrochemical impedance measurement. If pressure increases, the bubble
shrinks. This decreases the electrolyte conductive path and reduces the measured impedance.
However, the bubble did not last long and was not precisely controlled, limiting the
sensor’s resolution and sensitivity for low-pressure ranges. Later work has shown that the size of
the electrode pairs used to generate the bubble influences the control of bubble generation [12, 13].
The work here highlighted a new version of the sensor with the implementation of unevenly
sized electrodes on the die for precise picoliter (pL) - volume bubble generation and bubble
composition control enabling high-resolution pressure tracking suitable for measuring
physiological pressure within 0-20 mmHg in vivo.
160

Sensor Design
Sensors were fabricated using Parylene surface micromachining and included four exposed
platinum electrodes within a bubble confining microchannel and an exterior exposed counter
electrode (Fig. 3-32) [12, 13]. A single working electrode (230 μm
2
) and the pair of counter
electrodes (length: 1500 μm; width: 180 μm) shorted together formed an asymmetric electrode

Figure. 3-32. (Top) Micrograph of nanobubble pressure sensor with counter electrodes on a single die. The
sensor is highlighted in the dotted box. (Bottom) Schematic showing sensor with major components labeled.
The working and counter electrodes were used to inject current and generate the electrolytic bubble. Sensing
electrodes tracked impedance across the bubble microchannel to transduce pressure. The constriction valves
trapped the bubble in the sensing region of the microchannel [27]. © 2022 IEEE
161

pair, allowing a single pL-volume O2 or H2 bubble to be reliably generated by controlling the
amplitude and pulse duration of current passing through the electrode pair [13]. A key feature of
this version’s nanobubble-based pressure sensor has the working and counter electrodes with
different sizes on the same die to generate nanobubble rather than using the same sized electrodes
(Fig. 3-33). Previous work demonstrated that using a larger counter electrode could control the
bubble size and composition [13]. By limiting electrode polarization, power consumption is
reduced. Also, a hydrogen or oxygen gas bubble generation could be selected by controlling the
applied current polarity. This also allows tuning of the bubble dissolution rate because of the
different diffusion coefficients of gas. All of these factors contribute to higher resolution bubble-
based pressure sensing.

Figure. 3-33. Photo showing the varying size for counter and working electrodes.

This sensor contains constriction valves located at the ends of the bubble microchannel
(length: 400 μm; width: 50 μm; height: 16 μm) and is designed to trap the formed bubble within
the sensing region. Localizing the bubble to this region allowed the electrochemical impedance to
162

be monitored by the sensing electrode pair located on either end of the bubble microchannel
regardless of sensor orientation .

Figure. 3-34. (Top) Randles equivalent circuit model of sensing electrode-electrolyte interface. (Bottom)
Impedance and phase of the sensing electrodes. The measurement frequency (dashed line) is selected when
the response is dominated by solution resistance [27]. © 2022 IEEE
163

Sensor Characterization
The sensors were loaded into an acrylic testing fixture filled with 1× phosphate-buffered
saline (PBS, to mimic the composition of physiological fluids). This system was primed with
pressurized nitrogen gas (~200 torrs) overnight to fill the microfluidic channels with 1× PBS.
Sensing electrode pairs were first characterized using electrochemical impedance spectroscope
(EIS) measurements taken with a potentiostat. The sensor electrode-electrolyte interface can be
modeled by the Randles circuit (Fig. 3-34). The measurement frequency of the sensor (10 kHz)
was selected when the electrochemical impedance response was dominated by the solution
resistance (phase ≈ -20 °; Fig. 3-37) [11].

Figure. 3-35. Photograph of testing setup with the sensor testing fixture, microscope, and Fluigent pressure
source. Inset shows a close-up of the sensor in the testing fixture positioned under the microscope [27]. ©
2022 IEEE
164

Bubble Generation and Pressure Sensing
High-resolution and real-time pressure sensing require careful control of bubble life and size.
Bubbles were generated using current pulses to conserve power for in vivo applications. Bubble
size and shape were monitored using a microscope and Matlab script described in previous work
(Fig. 3-35) [12, 13]. At the same time, an LCR meter acquired impedance (at 10 kHz).
A pL volume bubble was reliably generated using the asymmetric electrode pair and by
controlling the magnitude and duration of the current pulse. A bubble of either H2 or O2 could be
selected by applying a negative or positive current at a single working electrode. O2 bubbles were
used for pressure sensing as they lasted over three times longer, which is attributed to the lower
diffusion coefficient at room temperature (Table 3-1; Fig. 3-36) [11, 28].

Table 3-1. Diffusion coefficients of gas-liquid mixtures at 25 ℃ [27]. © 2022 IEEE
Figure. 3-36. Lifetime of electrolytically generated O2 (0.4 µA, 8s) and H2 (-0.4 µA, 8s) bubbles in the
microchannel after the current pulse was terminated (n=5, mean ± SD) [27]. © 2022 IEEE
Solute-solvent mixture Diffusion coefficient (cm
2
/s)
H2-H2O 4.5 x 10
-5

O2-H2O 2.1 x 10
-5

165

Figure. 3-37. Schematic showing the sensor operation in time sequence.

Pressure sensing experiments utilized a Fluigent pressure source capable of imposing precise
one mmHg steps over the 0-20 mmHg physiologically relevant range. Sensor operation for
pressure sensing is illustrated in Fig. 3-37. First, a bubble is generated using a current pulse applied
to one of the working electrodes. The bubble fills the channel as it grows. Once the current pulse
terminates, the bubble starts to dissolve into solution. The bubble experiences a quasi-stable state
during which pressure measurements can be made.
Figure 3-38 shows representative impedance and volume data for a single oxygen bubble
with no applied pressure. As the positive current pulse was applied, the bubble volume increased,
reaching peak volume before the current pulse terminated. Then the bubble volume decreased as
it dissolved back into the solution. Concurrently, measured impedance followed a similar trend but
with significant drops corresponding to bubble detachment from the channel wall (at 22.51 and
166

321.87 minutes in Fig. 3-38). The highlighted quasi-stable regime was selected for sensing and is
bounded by two distinct phenomena [29]. The first rapid drop in impedance corresponds to the

Figure. 3-38. (Top plot) Impedance response measured by sensing electrodes (black) and (bottom plot) O2
bubble volume tracked by video feed from the microscope (blue) after the applied current pulse (0.3 µA,
25s). Impedance magnitude decreased as the bubble dissolved back into the solution. (Bottom panels)
Frame captures of bubble detachment from channel surfaces correspond to the dramatic impedance drops
observed at ~20 and ~320 minutes [27]. © 2022 IEEE
167

detachment of the bubble from the top wall of the microchannel. The second drop in impedance
corresponds to when the bubble further detaches from the side wall of the smaller bubble initiation
chamber. Real-time pressure sensing at one mmHg resolution was achieved over 0 to 20 mmHg.
This represents a 20× improvement in resolution over prior work [10, 11] and meets requirements
for in vivo clinical pressure sensing.
Summary
A microfabricated electrochemical impedance-based flow sensor for inline monitoring of
physiological fluid flow was demonstrated to track cerebrospinal fluid (CSF) flow changes in a
week-long, live porcine study. This impedimetric sensor was benchmarked against a Sensirion
LD20 thermal liquid flow sensor at the benchtop, followed by a CSF drainage study using an
external ventricular drain (EVD) in the pig. Sensor resolution was improved ~2× to 26.2 μL/min
at low flow versus previous work, and operation expanded to 0-1000 μL/min. Using separate
calibrations for low and high flow ranges, accuracy at low flow rates (0-200 μL/min) was improved
more than six-fold compared to prior work. High sensitivity, resolution, accuracy, and stable RMS
error were demonstrated in the porcine study. The performance of the impedimetric sensor was
comparable to the LD20 benchmark showing the sensor’s capability to monitor flow dynamics in
implanted hydrocephalus shunts.
Besides continuously monitoring the CSF flow in the shunt, a nanobubble-based sensor
working for the body’s saline environment with one mmHg resolution for sensing physiological
pressure in a range of 0-20 mmHg was also successfully designed. The improved sensor
performance was achieved using different-sized electrodes, which enabled precise control of
bubble composition, lifetime, and size in nL volumes. The bubble life increased 25 times compared
to the previous work by selectively generating nanobubbles with oxygen gas.
168

However, all the benchtop characterization and pressure sensing experiments were done
under static conditions where one end of the sensor outlet was closed, and no fluid flowed through.
The future work involves combing the impedimetric flow sensor and the nanobubble pressure
sensor onto the same die integrated into the implanted ventricular shunt for clinical use with EVD
and testing with natural cerebrospinal fluid. This requires the sensor to maintain reliable
functionality and performance when the CSF flow is presented. To prepare the pressure sensor for
animal study, testing the sensor under dynamic conditions where a flow is present is required. The
sensor’s performance under static and dynamic conditions will be compared, and the sensor will
be further tuned if needed for pressure sensing in the case of no CSF flowing and CSF flowing
through. In the longer term, for achieving a fully implantable sensor system, the work on the
miniaturization of the recording electronics and wireless transmission electronics is also underway.
169

References
[1] U. Health, "Pediatric Neurosurgery Hydrocephalus FAQs," 2022. [Online]. Available:
https://www.uclahealth.org/medical-services/pediatric-neurosurgery/conditions-
treatment/pediatric-hydrocephalus-program/hydrocephalus-
faqs#:~:text=How%20common%20is%20hydrocephalus%3F,annually%20in%20the%20
United%20States.

[2] J. M. Spirig, M. N. Frank, L. Regli, and L. H. Stieglitz, "Shunt age-related complications
in adult patients with suspected shunt dysfunction. A recommended diagnostic workup,"
Acta Neurochirurgica, vol. 159, no. 8, pp. 1421-1428, 2017/08/01 2017, doi:
10.1007/s00701-017-3237-6.

[3] 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,"
(in English), Journal of Neurosurgery: Pediatrics PED, vol. 17, no. 3, pp. 270-277, 01 Mar.
2016 2016, doi: 10.3171/2015.7.PEDS1577.

[4] C. Qin, A. H. Olivencia-Yurvati, A. G. Williams, D. Eskildsen, R. T. Mallet, and P. K.
Dasgupta, "Inline flow sensor for ventriculoperitoneal shunts: Experimental evaluation in
swine," Medical Engineering & Physics, vol. 67, pp. 66-72, 2019/05/01/ 2019, doi:
https://doi.org/10.1016/j.medengphy.2019.03.010.

[5] T. Bork et al., "Development and in-vitro characterization of an implantable flow sensing
transducer for hydrocephalus," Biomedical Microdevices, vol. 12, no. 4, pp. 607-618,
2010/08/01 2010, doi: 10.1007/s10544-010-9413-6.

[6] K. Scholten and E. Meng, "Materials for microfabricated implantable devices: a review,"
Lab on a Chip, vol. 15, no. 22, pp. 4256-4272, 2015, doi: 10.1039/c5lc00809c.

[7] M. Czosnyka and J. D. Pickard, "Monitoring and interpretation of intracranial pressure,"
Journal of Neurology, Neurosurgery & Psychiatry, vol. 75, no. 6, p. 813, 2004, doi:
10.1136/jnnp.2003.033126.

[8] L. Yu, B. J. Kim, and E. Meng, "Chronically implanted pressure sensors: challenges and
state of the field," (in eng), Sensors (Basel), vol. 14, no. 11, pp. 20620-20644, 2014, doi:
10.3390/s141120620.

[9] C. A. Gutierrez and E. Meng, "A subnanowatt microbubble pressure sensor based on
electrochemical impedance transduction in a flexible all-parylene package," in 2011 IEEE
24th International Conference on Micro Electro Mechanical Systems, 2011: IEEE, pp.
549-552.

[10] L. Yu and E. Meng, "A microbubble pressure transducer with bubble nucleation core," in
2014 IEEE 27th International Conference on Micro Electro Mechanical Systems (MEMS),
26-30 Jan. 2014 2014, pp. 104-107, doi: 10.1109/MEMSYS.2014.6765584.
170

[11] L. Yu, C. A. Gutierrez, and E. Meng, "An Electrochemical Microbubble-Based MEMS
Pressure Sensor," Journal of Microelectromechanical Systems, vol. 25, no. 1, pp. 144-152,
2016, doi: 10.1109/JMEMS.2015.2499167.

[12] E. Yoon and E. Meng, "Electrolytic Generation of Trapped Nanobubbles via Nucleation
Core with Picoliter Precision," in 2020 IEEE 33rd International Conference on Micro
Electro Mechanical Systems (MEMS), 18-22 Jan. 2020 2020, pp. 1064-1067, doi:
10.1109/MEMS46641.2020.9056455.

[13] E. Yoon and E. Meng, "Asymmetric Microelectrodes for Nanoliter Bubble Generation via
Electrolysis," Journal of Microelectromechanical Systems, vol. 31, no. 1, pp. 106-115,
2022, doi: 10.1109/JMEMS.2021.3126789.

[14] T. Hudson, A. Baldwin, E. Christian, J. G. McComb, and E. Meng, "Testing a Multi-
Sensor System For Hydrocephalus Monitoring in External Ventricular Drains," in 2019
20th International Conference on Solid-State Sensors, Actuators and Microsystems &
Eurosensors XXXIII (TRANSDUCERS & EUROSENSORS XXXIII), 23-27 June 2019 2019,
pp. 310-313, doi: 10.1109/TRANSDUCERS.2019.8808653.

[15] T. Q. Hudson, A. Baldwin, A. Samiei, P. Lee, J. G. McComb, and E. Meng, "A portable
multi-sensor module for monitoring external ventricular drains," Biomedical Microdevices,
vol. 23, no. 4, p. 45, 2021/09/20 2021, doi: 10.1007/s10544-021-00579-8.

[16] A. Baldwin, L. Yu, and E. Meng, "An Electrochemical Impedance-Based Thermal Flow
Sensor for Physiological Fluids," Journal of Microelectromechanical Systems, vol. 25, no.
6, pp. 1015-1024, 2016, doi: 10.1109/JMEMS.2016.2614664.

[17] X. Wang, T. Hudson, K. Scholten, E. Myong, J. G. McComb, and E. Meng, "Monitoring
of Physiological Flow with a Microfabricated Electrochemical Parylene Flow Sensor,"
Hilton Head Workshop 2022: A Solid-State Sensors, Actuators and Microsystems
Workshop, pp. 89-92, 2022.

[18] C. Sosna, T. Walter, and W. Lang, "Response time of thermal flow sensors," Procedia
Engineering, vol. 5, pp. 524-527, 2010/01/01/ 2010, doi:
https://doi.org/10.1016/j.proeng.2010.09.162.

[19] V. T. Grigorean, A. M. Sandu, M. Popescu, and V. Strambu, "Ventriculoportal Shunt, a
New Transomphalic Extraperitoneal Surgical Technique in Treatment of Hydrocephalus,"
Surgical Innovation, vol. 24, no. 3, pp. 223-232, 2017/06/01 2017, doi:
10.1177/1553350617692637.

[20] C. J. Morgan et al., "Bilirubin as a cerebrospinal fluid marker of sentinel subarachnoid
hemorrhage: a preliminary report in pigs," (in English), Journal of Neurosurgery, vol. 101,
no. 6, pp. 1026-1029, 01 Jan. 2004 2004, doi: 10.3171/jns.2004.101.6.1026.
171

[21] Christopher M. Bernards, "Cerebrospinal Fluid and Spinal Cord Distribution of Baclofen
and Bupivacaine during Slow Intrathecal Infusion in Pigs," Anesthesiology, vol. 105, no.
1, pp. 169-178, 2006, doi: 10.1097/00000542-200607000-00027.

[22] K. Kormanik, J. Praca, H. J. L. Garton, and S. Sarkar, "Repeated tapping of ventricular
reservoir in preterm infants with post-hemorrhagic ventricular dilatation does not increase
the risk of reservoir infection," Journal of Perinatology, vol. 30, no. 3, pp. 218-221,
2010/03/01 2010, doi: 10.1038/jp.2009.154.

[23] A. C. P. Lui, T. Z. Polis, and N. J. Cicutti, "Densities of cerebrospinal fluid and spinal
anaesthetic solutions in surgical patients at body temperature," Canadian Journal of
Anaesthesia, vol. 45, no. 4, p. 297, 1998/04/01 1998, doi: 10.1007/BF03012018.

[24] K. Kinoshita et al., "Case Report: Detection of Transferrin in a Dog Suspected of Having
Cerebrospinal Fluid Rhinorrhea," Front Vet Sci, vol. 9, p. 845809, 2022, doi:
10.3389/fvets.2022.845809.

[25] M. Butler. "Bleeds and Blockages: Post-Haemorrhagic Hydrocephalus."
https://www.paediatricfoam.com/2017/05/post-haemorrhagic-hydrocephalus/ (accessed
7/31, 2022).

[26] Q. Yang et al., "Application of magnetically actuated self-clearing catheter for rapid in situ
blood clot clearance in hemorrhagic stroke treatment," Nature Communications, vol. 13,
no. 1, p. 520, 2022/01/26 2022, doi: 10.1038/s41467-022-28101-5.

[27] X. Wang, E. Yoon, and E. Meng, "High Resolution Nanobubble-Based Pressure Sensor
for in Vivo Monitoring," in 2022 IEEE 35th International Conference on Micro Electro
Mechanical Systems Conference (MEMS), 9-13 Jan. 2022 2022, pp. 47-50, doi:
10.1109/MEMS51670.2022.9699477.

[28] P. S. Epstein and M. S. Plesset, "On the Stability of Gas Bubbles in Liquid‐Gas Solutions,"
The Journal of Chemical Physics, vol. 18, no. 11, pp. 1505-1509, 1950/11/01 1950, doi:
10.1063/1.1747520.

[29] E. J. Yoon, "PARYLENE C BIOMEMS FOR IMPLANTABLE DEVICES WITH
ELECTROCHEMICAL INTERFACES," DOCTOR OF PHILOSOPHY, BIOMEDICAL
ENGINEERING, UNIVERSITY OF SOUTHERN CALIFORNIA, 2021.

172

A PP E NDIC E S
Appendix A: Parylene C Thermoforming Experiments for Microfluidics Integration

Integrating microfluidic channels onto multi-shank polymer-based interface arrays enables
the delivery of pharmacological agents to focally inhibit or modify the neurotransmitter and further
control the neural activity. Furthermore, the integration of microfluidics will help to dissect
complex neural circuits and help with the drug screening for treating neurological diseases.
To deliver pharmacological agents into the brain, the conventional way is to use metal
cannulas. A cannula with a metal needle tip is screwed to the skull and connected to a syringe
pump. Those metal cannulas are made from rigid materials with young’s modulus much higher
than the brain tissue. Its bulk size and rigidness limit its use to short-term applications. Since it is
not integrable with the existing brain probe, a separated burr hole or incision is needed for
implanting the intracranial probe to monitor the neural response introduced by the delivery of
pharmacological agents.
The end goal of this project is to develop a flexible Parylene high-density brain interface with
microfluidic channels. The work reported here focuses on optimizing a Parylene-parylene bonding
method to create microchannels integrative into the Parylene neural probes array.
The conventional layer-by-layer fabrication relies on the photoresist to define the channels. It
is challenging to dissolve the photoresist encapsulated in the microscale channel, especially when
the channels are made to be long and narrow on the intracranial brain probe. The method described
here does not involve dissolving the photoresist left in an enclosed channel. It uses thermos-
compressive to fuse two polymer layers through heat and pressure. It is adhesion-free and
173

compatible with Parylene micromachining. Changing the thickness of the Parylene layer deposited
on a pre-etched silicon mold can adjust the channel dimension.

Figure A-1. Schematic of thermos-bonding steps.

The fabrication process of Parylene microfluidic channels for drug delivery is reported here
(Fig. A-1). Two silicon wafers were prepared, one flat and one with channels etched using a Bosch
process through deep reactive ion etching, to form a 2-part mold for bonding microchannels. A
layer of detergent serving as a release layer was deposited onto both wafers (polyethylene glycol
174

(PEG) or Micro 90), followed by chemical vapor deposition of 8 μm Parylene. The Parylene
deposition conformed to the silicon wafers, creating a series of polymer channels and flat film.
The two Parylene surfaces were joined by compression bonding in a fixture between a Teflon sheet
backing and two stainless steel plates. The joining force was set by torque screws (Fig. A-2). The
entire assembly was heated in a vacuum oven. After cooling, the detergent was removed by
soaking in DI water for 1 hour, and the free-standing channels were released.

Figure A-2. Schematic of the pressure applicator used for thermocompressive bonding.

To examine the effect of different bonding parameters on the adhesion strength between the
Parylene layers, samples were processed using different bonding temperatures (160 °C and 200 °C),
time (30 min and 60 min), and pressures (0 MPa, 2 MPa, and 5 MPa). In addition, we examined
different surface cleaning protocols using N2 and/or solvent rinses (acetone, isopropanol, and
175

deionized water) before bonding. The bonding pressure was determined by the torque applied to
the screws, as shown in the following equation (Fig. A-2);

M =
𝑃𝐴 ( 𝑟 +
𝑑 2
) 𝑢 4

where M is torque; P is pressure; r is the radius of screw heads; d is the diameter of the screw; u is
the friction cons, and A is the area of the bonding wafer.

Bonding
Temperature
(ºC)
Bonding
Time (mins)
Bonding
Pressure
(MPa)
Cleaning Method
Bonding
Area (%)
160 60 2
N
2

0
160 30 5
N
2

75
200 60 5
N
2

25
200 60 5
A ce to n e + IP A + DIH
2
O+N
2

70
200 60 5
A ce to n e + IP A + DIH
2
O+N
2

85

Table A-1. Bonding area of Parylene microchannels under varied bonding parameters.

Table A-1 summarizes results for a subset of the experiments showing the fraction of the
successfully bonded interface. Bonding of the channels only occurred with an applied bonding
pressure of 5 MPa. For experiments with less pressure, there was little or only partial bonding
176

along the sides of the channels. Parylene surfaces cleaned with solvents exhibited improved
adhesion (85% and 70% compared to 25% for 200 °C, 60 min). At 160 °C, bonding for 30 min
resulted in partial bonding of channels, but no bonding was observed at 60 min due to low applied
bonding pressure. Other trails with little bonding are not shown in the table.

Figure A-3. Schematic of microfluidics channel flow characterization. (a) Bare Parylene microfluidic
channels; (b) channels filled with red dye.

To look at the integrity of the channel and characterize the flow of water inside the channels,
the device was soaked in water with red food dye. Under the microscope, water bubbles were
observed flowing through the channels while it was physically compressed and released by
tweezers. The bubble movement in liquid-filled channels proved the channels are intact and can
capsulate liquid and bubble (Fig. A-3). Successful flow indicated channels are robust for use in
microfluidic delivery.
We observed that functional Parylene microchannels could be created through this method.
The cleanliness of the Parylene surface proved critical. Higher temperature yielded more consistent
177

adhesion for the same bonding time, whereas shorter bonding time resulted in better adhesion
under the same bonding temperature. Higher bonding pressure proved critical. However, some
silicon wafers began cracking during the bonding experiments under 5 MPa pressure (Fig. A-4).
More investigation is required to determine if the cracking is caused by the uneven pressure applied
during thermobonding.

Figure A-4. Picture showing the devices under different failure modes.

Figure A-5. Schematic of the flow test setup with a syringe pump.

Future studies are recommended to optimize the process parameters to achieve successful
bonding close to 100% over the target bonded region. To move forward, examining the channels'
178

cross sections is recommended to characterize the inner wall surface properties. Before further
testing, a quantitative measurement of bonding strength between the Parylene layers is also needed.
To determine the maximum flow rate and pressure supported by bonded channels, a syringe pump
could be connected to the microchannels through a catheter and apply varied pressure and flow
rate until the channel reaches the failure mode (Fig. A-5).

179

Appendix B: Fabrication of Sham Arrays

1. Bake clean 4” silicon wafer to remove moisture 110 °C, > 10 mins

2. Deposit Parylene (10 µm)

3. Pattern AZ 4620 etch mask (~ 8 µm thick) (Mask 1 – Sacrificial photoresist pockets)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 3500 rpm
Softbake 90 °C, 5 minutes
Hydration 45 minutes
Exposure 400 mJ/cm
2
(25 mW/cm
2
, 16 sec)
Development 80 seconds/ 20 seconds (two baths)
Hard bake 1 90 °C, 5 minutes, hotplate
Hard bake 2 110 °C, 15 minutes, under vacuum

4. Deposit Parylene ( 10 µm)

5. Pattern AZ 4620 etch mask (> 10 µm thick) (Mask 2 –Pocket inlet; cutout and ellipse)
Pre spin 5 sec, 500 rpm
Spin 40 sec, 1200 rpm
Softbake 90 °C, 5 minutes
Hydration 30 minutes
Align and expose 550 mJ/cm
2
(25 mW/cm
2
, 22 sec)
Development 90 seconds
Hard bake 1 90 °C, 5 minutes, hotplate
Hard bake 2 90 °C, 15 minutes, under vacuum

6. Deep Reactive Ion Etching (Oxygen plasma) 125 loops, rotate wafer 90° every 25 loops

180

Table B- 1: DRIE parameters for deposition and etch steps
Parameter Deposition Etch
ICP Power (W)
700 700
RF Power (W)
80 80
O2 (ccm)
1 60
C4F8 (ccm)
35 1
Ar (ccm)
40 40
SF6 (ccm)
0 0
Pressure (mTorr)
23 23
Time (s)
3 10

7. Strip remaining photoresist mask with Acetone, IPA, and DI water

8. Pattern AZ 4620 etch mask (> 10 µm thick) (Mask 3 –Cutout and ellipse completion)
Pre spin 5 sec, 500 rpm
Spin 40 sec, 1200 rpm
Softbake 90 °C, 5 minutes
Hydration 30 minutes
Align and expose 550 mJ/cm
2
(25 mW/cm
2
, 22 sec)
Development 90 seconds
Hard bake 1 90 °C, 5 minutes, hotplate
Hard bake 2 90 °C, 15 minutes, under vacuum

9. Deep Reactive Ion Etching (Oxygen plasma) 125 loops, rotate wafer 90° every 25 loops

10. Release Clean surface with acetone and IPA
Peel carefully while immersed in water

11. Strip any remaining photoresist mask with 5 min soaks in acetone, IPA, and DI water

181

Appendix C: Fabrication of Complete Hippocampal Arrays

1. Bake clean 4” silicon wafer to remove moisture 110 °C, > 10 mins

2. Bake clean 4” silicon wafer to remove moisture 110 °C, > 10 mins

3. Deposit Parylene (10 µm)

4. Pattern AZ 5214-IR for lift-off (2 µm thick) (Mask 1 - Metal)
Pre spin 8 sec, 500 rpm
Spin 45 sec, 1800 rpm
Softbake 90 °C, 70 seconds
Exposure 37.5 mJ/cm
2
(25 mW/cm
2
, 1.5 sec)
IR bake 110 °C , 55 sec
Hydration 3 minutes
Global exposure 1000 mJ/cm
2
(25 mW/cm
2
, 40 seconds)
Development (AZ 351 1:4 dilution) 18 seconds

5. Descum, O2 plasma 100 W, 100 mTorr, 1 min

6. Metal deposition (Pt) 2000 Å (in 4 runs of 500 Å )

7. Lift-off in acetone (gentle scrub if necessary) In warm acetone 50 °C

8. Descum, O2 plasma 100 W, 100 mTorr, 1 min

9. Deposit Parylene (10 µm)

10. Pattern AZ 4620 etch mask (12.8 – 13.5 µm thick) (Mask 2 – Insulation Cutout)
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1000 rpm
Softbake 90 °C, 5 minutes
Hydration 45 minutes
Exposure 550 mJ/cm
2
(20 mW/cm
2
, 27.5 sec)
Development 1.5 minutes
Hard bake 1 90 °C, 15 minutes, hotplate
Hard bake 2 90 °C, 15 minutes, under vacuum

182

11. Deep Reactive Ion Etching (Oxygen plasma) 100 loops, rotate wafer 90° every 25
loops

Table. C-1. DRIE etch and deposition parameters for each loop
Parameter Deposition Etch
ICP Power (W)
700 700
RF Power (W)
80 80
O2 (ccm)
1 60
C4F8 (ccm)
35 1
Ar (ccm)
40 40
SF6 (ccm)
0 0
Pressure (mTorr)
23 23
Time (s)
3 10

12. Strip remaining photoresist mas with Acetone, IPA, and DI water

13. Pattern AZ 4620 (double layer) sacrificial channel mold (20.1 µm thick) (Mask 3 – Sac PR)
LAYER 1
Pre spin 5 sec, 500 rpm
Spin 45sec, 1700 rpm
Softbake 90 °C, 6 minutes
LAYER 2
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1900 rpm
Softbake 90 °C, 7 minutes
Hydration 60 minutes
Exposure 600 mJ/cm
2
(20 mW/cm
2
, 30 sec)
Development 3 minutes using two baths
Hard bake 1 90 °C, 15 minutes, hotplate
Hard bake 2 90 °C, 15 minutes, under vacuum

14. Descum, O2 plasma 100 W, 100 mTorr, 1 min

15. Deposit Parylene (8 µm)

16. Pattern AZ 4620 (double layer) etch mask (32 µm thick) (Mask 4 – Pads Etch)
183

LAYER 1
Pre spin 5 sec, 500 rpm
Spin 45sec, 1000 rpm
Softbake 90 °C, 11 minutes
LAYER 2
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1000 rpm
Softbake 90 °C, 14 minutes
Hydration 60 minutes
Exposure 750 mJ/cm
2
(20 mW/cm
2
, 37.5 sec)
Development 3.5 minutes using two baths
Hard bake 1 100 °C, 15 minutes, hotplate
Hard bake 2 100 °C, 15 minutes, under vacuum

17. Deep Reactive Ion Etching (Oxygen plasma) 140 loops, rotate wafer 90° every 25
loops

18. Strip remaining photoresist mas with Acetone, IPA, and DI water

19. Pattern AZ 4620 (double layer) etch mask (32 µm thick) (Mask 5 – Ports Etch)
LAYER 1
Pre spin 5 sec, 500 rpm
Spin 45sec, 1000 rpm
Softbake 90 °C, 11 minutes
LAYER 2
Pre spin 5 sec, 500 rpm
Spin 45 sec, 1000 rpm
Softbake 90 °C, 14 minutes
Hydration 60 minutes
Exposure 750 mJ/cm
2
(20 mW/cm
2
, 37.5 sec)
Development 3.5 minutes using two baths
Hard bake 1 100 °C, 15 minutes, hotplate
Hard bake 2 100 °C, 15 minutes, under vacuum

20. Deep Reactive Ion Etching (Oxygen plasma) 140 loops, rotate wafer 90° every 25 loops

Release Clean surface with acetone and IPA.
Peel carefully while immersed in water

Abstract (if available)

Abstract Biomedical microelectromechanical systems (BioMEMS) use micromachining techniques to design miniaturized systems for biology and medicine. Conventional MEMS often use silicon, but it is not preferred for medical device applications for many reasons. First, the silicon is not biocompatible and often needs some biocompatible coating to prevent it from directly contacting the tissue. It is also stiff and has a high Young’s modulus. The mismatch between tissue and silicon’s stiffness will cause severe damage. The scar formation around the device limits the device’s lifetime. It is also brittle, so silicon could not be made too thin; otherwise, have the risk of fracture. The flexible polymer has been explored to combine with MEMS techniques to make medical devices. My work focuses on applying Parylene bioMEMS to make biomedical devices. Parylene is selected because it is a transparent polymer with a low Young’s modulus that better matches the tissue stiffness than silicon. It is also classified as a Class VI material by the United States Pharmacopeia and is suitable for implants. More importantly, it is flexible and can be used to make low-profile or thin devices.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Parylene-based implantable interfaces for biomedical applications

PDF

Parylene C bioMEMS for implantable devices with electrochemical interfaces

PDF

Development of implantable Parylene-based MEMS technologies for cortical applications

PDF

High-resolution data acquisition with neural and dermal interfaces

PDF

Development of fabrication technologies for robust Parylene medical implants

PDF

Development of micromachined technologies for neural interfaces

PDF

Penetrating parylene neural probe array for dense, in vivo, chronic recordings

PDF

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

PDF

Fabrication and packaging of three-dimensional Parylene C neural interfaces

PDF

Ultrasonic microelectromechanical system for microfluidics, cancer therapeutics and sensing applications

PDF

The electrochemical evaluation of Parylene-based electrodes for neural applications

PDF

Design and application of a C-shaped miniaturized coil for transcranial magnetic stimulation in rodents

PDF

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

PDF

Investigation of the electrode-tissue interface of retinal prostheses

PDF

A wireless implantable MEMS micropump system for site-specific anti-cancer drug delivery

PDF

Bidirectional neural interfaces for neuroprosthetics

PDF

Integrated wireless piezoelectric ultrasonic transducer system for biomedical applications

PDF

High frequency ultrasonic phased array system and its applications

PDF

CMOS mixed-signal charge-metering stimulus amplifier for biomimetic microelectronic systems

PDF

Next generation neural interfaces for neural modulation and prosthesis

Asset Metadata

Creator Wang, Xuechun (author)

Core Title Parylene-based biomems sensors for multiple physiological systems

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 11/16/2022

Defense Date 09/08/2022

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

Tag biomedical sensors,brain probes,flexible implantable sensors,flow sensors,neural interfaces,OAI-PMH Harvest,pressure sensors,thin-film sensors

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Meng, Ellis (committee chair), Kim, Eun Sok (committee member), Mousavi, Maral (committee member), Song, Dong (committee member), Zhou, Qifa (committee member)

Creator Email xuechun.wang@yahoo.com,xuechunw@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC112485548

Unique identifier UC112485548

Identifier etd-WangXuechu-11318.pdf (filename)

Legacy Identifier etd-WangXuechu-11318

Document Type Dissertation

Format theses (aat)

Rights Wang, Xuechun

Internet Media Type application/pdf

Type texts

Source 20221121-usctheses-batch-992 (batch), 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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

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

Repository Email cisadmin@lib.usc.edu