University of Southern California Dissertations and Theses

Next generation neural interfaces for neural modulation and prosthesis

(USC Thesis Other)

Next generation neural interfaces for neural modulation and prosthesis

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content NEXT GENERATION NEURAL INTERFACES FOR NEURAL MODULATION AND
PROSTHESIS
by
Sahar Elyahoodayan
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
BIOMEDICAL ENGINEERING
August 2020
Copyright 2020 Sahar Elyahoodayan
ii

ACKNOWLEDGMENTS
The thesis presented here is a record of my six years of research and study at the University
of Southern California. Here, I had an opportunity to meet, learn from and work with many
remarkable people. I would like to express my sincere gratitude to those whose support and
inspiration have enabled me to dedicate to my research. Thank you for your help making this
dissertation reality.
First and foremost, I would like to thank my thesis supervisor, Dr. Dong Song. Thanks for
him, I had this unique opportunity to gain a broad experience in hardware design and animal
studies. With his knowledge, support, and encouragement, I could concentrate on my goal,
focusing on academic research that has potential of being commercialized to help patients. I really
appreciate the time he invested in me.
I would like to thank my co-advisor, Dr. Theodore Berger, who has supported and inspired
me through my journey. I would like to thank my former advisor, Dr. James Weiland who advised
me during my first two years of PhD work. Thanks for him, I became interested in the field of
neuromodulation. His continuous support even after his move to Michigan University has helped
improve the quality of this thesis. I appreciate their patience, encouragement and support.
I would also like to thank my committee members, Dr. Ellis Meng and Dr. Hossein
Hashemi for their valuable feedback that have kept me focused on right questions. In addition to
mentorship, they are good role models as researchers and leaders.
I would like to thank my collaborators within Platinum Group Coatings LLC, Translational
Imaging Center, and Dr. Ellis Meng’s team. I appreciate all their input that has made this
dissertation possible. I am grateful to Christopher Larson, Jack Whalen, Curtis Lee, Artin
iii

Petrossians, Thai Truong, and Peter Luu, for their contributions in technical development. Their
continuous hard work and advice have made this research a reality.
I would like to express my appreciation to Wenxuan Jiang and Huijing Xu. Thank you for
your assistance in animal studies. I am grateful to Gene Yu, Clayton Bringham, and Aaron Curry
for their inputs and support.
I would like to thank my undergraduate students: Michael Kyzer, Betelhem Alemu, Tifanni
Szeto, Ryan Li and my graduate students: Amirhossein Forouzani and Siddharth Bhonge for their
assistance in various parts of the project. Their hard work truly helped propelled this project
forward.
I would like to express my appreciation to the generous funding from National Science
Foundation that have kept me able to dedicate my time to this work.
Lastly, I want to thank all my family and friends who have supported me throughout the
years. To my brother and good friend, for encouraging me and inspiring me through his own
success. To my cousin, Payam Eliahoo, for his unconditional support, advice and continuous
encouragement. To my best friends, Samira Niki, Amy Tran, and Natalie Partovi who always
supported me without hesitation. Most importantly, thank you to my parents, who have showed
me unconditional love. They taught me perseverance and discipline. It is the freedom they
provided me that enabled me to follow my heart in my education and career.

iv

TABLE OF CONTENTS
Acknowledgments ......................................................................................................................................... ii
List of Tables ............................................................................................................................................... vii
List of Figures ............................................................................................................................................. viii
Abstract .......................................................................................................................................................xiii
Chapter 1: Introduction ................................................................................................................................ 1
1.1 Motivation ..................................................................................................................................... 1
1.2 Outline........................................................................................................................................... 4
Chapter 2: Review of Neural Interface Technology Background .................................................................. 7
2.1 Introduction .................................................................................................................................. 7
2.2 Review of Neural Interface Technology Applications ................................................................... 7
2.3 Need Identification ....................................................................................................................... 9
2.4 Challenges ................................................................................................................................... 10
2.4.1 Large Scale Neural Code-Based Stimulation ....................................................................... 10
2.4.2 Stimulation and Recording from the Same Electrode ........................................................ 11
2.5 Review of Existing Technology .................................................................................................... 12
2.5.1 Large Scale Neural Code-Based Stimulation ....................................................................... 12
2.5.2 Stimulation and Recording from the Same Electrode ........................................................ 13
2.5.2.1 Hardware Electronic Circuit Technology ......................................................................... 13
2.5.2.2 Electrode Technology ...................................................................................................... 15
2.5.3 Peripheral Cuff Electrodes .................................................................................................. 15
2.6 Design Process Overview ............................................................................................................ 16
2.6.1 Hardware Electronics Circuit Design ................................................................................... 16
2.6.2 Electrode Design ................................................................................................................. 17
2.6.3 Design of Animal Studies .................................................................................................... 17
2.6.4 Integrated Microfluidic Channels ........................................................................................ 18
Chapter 3. Stimulation of Neural Tissue ..................................................................................................... 19
3.1 Physiology of Neural Tissue ........................................................................................................ 19
3.2 Electrode-Tissue Interface .......................................................................................................... 21
3.3 Stimulation Waveform ................................................................................................................ 22
3.4 Safety Requirements ................................................................................................................... 23
Chapter 4: System Design of Neural-Code Based Stimulation ................................................................... 25
4.1 Introduction ................................................................................................................................ 25
4.2 System Architecture .................................................................................................................... 25
4.3 Hardware Electronic Circuit Design ............................................................................................ 27
4.4 System Bench-top Characterization............................................................................................ 29
4.5 Results ......................................................................................................................................... 29
4.5.1 Asynchronous Arbitrary Pulse Pattern Generator .............................................................. 30
4.5.2 Limitations ........................................................................................................................... 31
4.5.3 Power Consumption............................................................................................................ 32
v

4.5.4 System Cost ......................................................................................................................... 32
4.6 Discussion .................................................................................................................................... 33
Chapter 5: Stimulation in Conjunction with Recording .............................................................................. 35
5.1 Introduction ................................................................................................................................ 35
5.2 Stimulus Artifact .......................................................................................................................... 36
5.3 Stimulus Artifact Suppression Technique ................................................................................... 36
5.4 Electrochemical Properties Consideration.................................................................................. 38
5.5 Methods ...................................................................................................................................... 42
5.5.1 System Bench-Top Testing .................................................................................................. 42
5.5.2 Preliminary Validation in vivo ............................................................................................. 43
5.6 Results ......................................................................................................................................... 45
5.6.1 Electrochemical Characterization ....................................................................................... 46
5.6.2 Verification in Phantom ...................................................................................................... 46
5.6.3 System Validation in Vivo .................................................................................................... 49
5.7 Discussion .................................................................................................................................... 52
Chapter 6: Microelectrode Design for Same Electrode Stimulation and Recording .................................. 54
6.1 Introduction ................................................................................................................................ 54
6.2 Electrode Design ......................................................................................................................... 58
6.3 Electrode Preparation ................................................................................................................. 58
6.4 Safety Requirements ................................................................................................................... 59
6.5 Discussion .................................................................................................................................... 61
Chapter 7. Proposal for In vivo Validation .................................................................................................. 63
7.1 Design of Electrophysiology Experiments for in vivo Evaluation ................................................ 63
7.1.1 Goal ..................................................................................................................................... 63
7.1.2 Surgical Procedure .............................................................................................................. 63
7.1.3 Data Acquisition .................................................................................................................. 65
7.1.4 Signal to Noise Ratio Calculation ........................................................................................ 65
7.1.5 Electrochemical Characterization ....................................................................................... 66
7.1.6 Voltage Transient Response ................................................................................................ 66
7.1.7 Neural Response to Stimulation ......................................................................................... 68
7.1.8 Spike sorting analysis .......................................................................................................... 69
7.2 Statistical Analysis to Demonstrate Functionality ...................................................................... 69
7.2.1 Electrochemical Measurements ......................................................................................... 69
7.2.2 Electrode Response to Stimulation Pulses .......................................................................... 70
7.2.3 Signal to Noise Ratio of Spontaneous Neural Activity ........................................................ 73
7.2.4 Short Latency Neural Response to Stimulation .................................................................. 76
7.2.5 Prolong Effect of stimulation .............................................................................................. 77
7.3 Discussion .................................................................................................................................... 80
Chapter 8: Integrated Microfluidic Channels .............................................................................................. 83
8.1 Introduction ................................................................................................................................ 83
8.2 Methods ...................................................................................................................................... 84
8.2.1 Electrode Design and Fabrication ....................................................................................... 84
8.2.2 Surgical Technique .............................................................................................................. 86
8.2.3 Recording Set-Up ................................................................................................................ 87
vi

8.2.4 Stimulation Set-Up .............................................................................................................. 88
8.2.5 Drug Delivery ....................................................................................................................... 88
8.3 Results ......................................................................................................................................... 89
8.3.1 Fabrication Results .............................................................................................................. 89
8.3.2 Acute Recording Test .......................................................................................................... 90
8.3.3 Acute Stimulation Test ........................................................................................................ 93
8.3.4 Drug Delivery ....................................................................................................................... 95
8.4 Discussion .................................................................................................................................... 98
Chapter 9: System Miniaturization ........................................................................................................... 101
9.1 System Description ................................................................................................................... 101
9.2 PCB Prototype ........................................................................................................................... 102
9.3 Miniaturized Design .................................................................................................................. 103
9.4 Chassis Design ........................................................................................................................... 104
9.5 Chassis Prototype Test in Rat .................................................................................................... 106
Chapter 10: Conclusion and Prospect ....................................................................................................... 109
References ................................................................................................................................................ 111

vii

LIST OF TABLES
Table 5-1 Electrochemical parameters of the microelectrode used in the experiments. ............................. 46
Table 6-2 List of stimulation test pulse parameters (pulse amplitude and total charge delivered). The same
pulse duration (200 µs per phase) was used for all tests. Test numbers 1-5 were applied to both electrode
types (coated and uncoated). Test numbers 6-10 (in grey) were only tested on the coated electrodes because
the parameters exceeded safety limits when used on the uncoated electrodes. .......................................... 61

viii

LIST OF FIGURES
Figure 1-1 Illustration of (A) a conventional open-loop DBS and (B) a configurable multi-channel
neurostimulator for implementing closed-loop processing from the region of stimulation. ......................... 4
Figure 3-1 Sketch of a neuron including cell body, axon with myelin and axon terminal. ........................ 20
Figure 3-2 Equivalent electrical circuit model of the electrode/electrolyte interface. C dl is the double layer
capacitor representing capacitive reactions, R p is the parallel resistor representing faradaic reactions, and
Rs represents solution resistance................................................................................................................. 22
Figure 3-3 Example of strength duration curve. Chronaxie is the time duration and 2x rheobase is the
amplitude at which stimulation is most efficient. ....................................................................................... 23
Figure 3-4 Voltage transient of an electrode (blue) in response to a biphasic constant current biphasic pulse
(red) (Reprinted from Cogan 2008). ........................................................................................................... 24
Figure 4-1 Block diagram and data stream of the multi-channel neurostimulator with stimulus artifact
suppression (SAS). The hardware is programmed by the embedded system to generate highly configurable
constant current monopolar biphasic-pulse stimulation. Each block in the embedded system represents an
algorithm to control the hardware. It consists of a MIMO non-linear dynamical model simulator, using a
random number generator. The data is inputted to a multiplexer control unit to set the appropriate select
lines of the multiplexer. The multiplexer control unit is synchronized with the SPI switch control unit to
control CMOS switches. ............................................................................................................................. 27
Figure 4-2 Design of a 32-channel configurable constant current biphasic neurostimulator. It includes (A)
a microcontroller (MSP430) which generates a PWM at a 40 kHz frequency and an adjustable duty cycle
to configure the stimulus amplitude. The duty cycle is also automatically adjusted as the battery supply
voltage drops, governed by equation 1, avoiding the need for a regulator. PWM then goes through (B) an
integrator to generate a constant -DCV, (C) an inverter to generate a +DCV, (D) a set of analog switches
dictating polarity and duration of the pulse controlled by the microcontroller, (E) a voltage to current
converter and a DC blocking capacitor to generate safe single channel constant current biphasic pulses. (F)
Lastly, a multiplexer is used to expand the design to 32 channels.............................................................. 28
Figure 4-3 Representative electrochemical impedance spectroscopy plots of impedance magnitude in
1XPBS for three trials (E1-E3) of a Tungsten electrode used in all experiments. ...................................... 30
Figure 4-4 Spatiotemporal pattern of cathodic or anodic first stimulation pulses with varying amplitudes
and intervals generated by the 32-channel neurostimulator. The stimuli were applied across 50 kΩ resistor
mimicking |z| electrode. The amplitudes of the pulses range from ±1 µA to ±60 µA, and the pulse intervals
range from 0.5 ms to 100 ms. ..................................................................................................................... 31
Figure 5-1 Stimulation artifact suppression set-up. (A) Stimulation is synchronized with a set of serially
controlled CMOS switches (S2) to block out the stimulus from the recording amplifier during stimulation
while connecting the amplifier input to a 100 Ω resistor via S3 to prevent ringing caused in the recording
system due to any extra charge coupled across S2. Charge coupling across S2 is due to (B) parasitic
components of analog switches. The PCB is connected to the recording amplifier and (C) the electrode
using a coaxial cable. The timing of the switches in (A) is determined by (C) the electrochemical properties
of the electrode, and (D) the voltage transient across a microelectrode (red) in response to a biphasic
cathodic first current pulse (blue). (C) RS represents the electrolyte resistance, RP represents faradaic
reaction, and Cdl represents capacitive reactions at the interface. .............................................................. 38
ix

Figure 5-2 Polarization voltage (VP) across the electrode with respect to a reference of the same material
after termination of the stimulation pulse with an initial value of VP-i. The voltage must drop below VT -
before recording is resumed, to avoid amplifier saturation, which takes a time duration of multiple factors
of τ represented as xτ. ................................................................................................................................. 40
Figure 5-3 The state of the switches (1 = closed; 0 = open) shown in Figure 3a with respect to the stimulation
pulse. In all cases when the stimulator is not pulsing, it is connected to ground. States 1 to 6 are as follows:
1. Recording: 1. Recording: The electrode is connected to the recording amplifier while the stimulator and
the resistor are disconnected. 2. Grounding: 200 µs before stimulation, the stimulator is connected to the
electrode and the resistor is connected to the input of the amplifier. 3. Stimulation: The amplifier is
disconnected from the electrode. The stimulation pulse is then applied to the electrode. 4. Discharge period:
The electrode stays disconnected from the amplifier until the polarization voltage on the electrode falls
below V T. 5. Discharge of residual charge (200 µs): The electrode is disconnected from the stimulator and
connected to the recording system. The resistor remains connected to the input of the amplifier to absorb
any charge injection due to residual offset on the electrode. 6. Recording resumed: The termination resistor
is disconnected, and recording is resumed. ................................................................................................. 41
Figure 5-4 Test bench setup for evaluating the stimulus artifact with (Vout2) and without (Vout1) the SAS
technique. The setup consists of diluted PBS mimicking neural tissue and a sinusoidal signal (1 kHz, 10
mV peak to peak) applied to a neighboring electrode mimicking neural activity. ..................................... 43
Figure 5-5 Acute in vivo setup for system evaluation. (A) PCB design of the system as shown in Figure 6
connected to (B) an implanted microelectrode in rat CA1 region of the hippocampus. (C) Stimulus artifacts
and evoked neural responses are amplified, digitized, and saved with a bench-top neural recording system.
.................................................................................................................................................................... 45
Figure 5-6 Stimulus artifact (A) without and (B) with the SAS in response to a current pulse (cathodic first,
60 µA, 200 µs). Recording is resumed more than (A) 60 ms as opposed to (B) 2.3 ms after the onset of
stimulation. (C) 15 ms time window from (A) demonstrating amplifier saturation and signal reflection. (D)
15ms time window from (B) demonstrating the results of states 1 to 6 described in Figure 5 and summarized
as follows: 1. Recording phase, 2. switching artifact (200 µs), 3. charge coupling across S2 during
stimulation (400 µs), 4. discharge period of the electrode (1.5 ms), 5. switching artifact plus residual charge
across the electrode (200 µs), 6. resume recording. .................................................................................... 48
Figure 5-7 Recording of evoked neural responses to increasing micro-stimulation amplitudes in the CA1
region of the hippocampus of an anesthetized rat using the SAS component. All pulses are constant-current,
cathodic-first and 200 µs in duration. Stimulus artifacts are shown in red. Evoked potentials are shown in
black. As stimulation amplitudes increase, the evoked potentials show increased amplitudes and more
complex waveforms. ................................................................................................................................... 50
Figure 5-8 Stimulus artifacts and evoked potentials to a stimulation with 40 µA amplitude. (A) Three trials
of recordings with the SAS. (B) Recordings with (red) and without (black) the SAS are overlaid for
comparison. (C) and (D) Verification of the evoked potentials by comparing the signals before (red) and
after (black) the rat was euthanized. Neural signals are retrieved within (C) ~2 ms after the stimulation onset
when SAS is used, and (D) ~60 ms when SAS is not used......................................................................... 51
Figure 5-9 Frequency analysis of evoked potentials. 7-30 Hz band is apparent in all cases which increases
in magnitude with the stimulus amplitude. Evoked potentials to higher stimulus amplitudes (40 µA, 50 µA,
and 60 µA) also contain frequency components in the 40–100 Hz band. 60 Hz noise is shown in grey. .. 52
x

Figure 6-1 a) Schematic showing relative locations of coated and uncoated contacts. This arrangement
allowed a side-by-side comparison of contacts with and without Pt-Ir coating. b) Optical micrograph of
Microprobes microelectrode array. SEM micrograph of c) Pt-Ir coated and, d) uncoated microelectrode tips.
c) The fractal nodules on the coated electrode dramatically increase the electrochemical surface area while
the geometric surface area remains similar to d) the uncoated electrode. ................................................... 59
Figure 6-2 Representative cyclic voltammograms for coated (black) and uncoated (red) electrodes in room
temperature PBS. Artifacts from silver leakage caused during the coating process are greyed out. The
calculated CSCC is an average of 0.3±0.025 mC/cm
2
for uncoated (n =6) and 12.5+0.75 mC/cm
2
for the
coated (n =10) electrodes. ........................................................................................................................... 60
Figure 7-1 Quantification of electrode polarization from current pulsing. (a) Simplified electrical equivalent
circuit model of the electrode-electrolyte interface. Electrode/electrolyte interface is modeled by a capacitor
(C dl) and a resistor in parallel (R p). The resistance of the electrolyte is modeled using a simple resistor (R s).
b) Voltage response of constant biphasic pulse (with current I) sent through a single small (5 kΩ) resistor.
There is a linear response that maintains the square wave of the biphasic pulse. c) Voltage response to
constant biphasic pulse sent through a single 50 kΩ resistor meant to approximate the Rs of tissue. Because
of the size of the resistor, the voltage response is no longer linear likely due to the large time constant at the
output of the stimulator. d) The voltage response to the 50 kΩ resistor superimposed onto the voltage
response of the coated (blue) and uncoated (red) electrodes. The polarization voltage for the uncoated
electrode (Ep 1) and coated electrode (Ep 2) is the subtraction between the access polarization estimated in
(c) and the voltage responses of the uncoated (red) and coated electrode (blue), respectively. ................. 67
Figure 7-2 Bode plots of impedance magnitude |Z| vs. frequency recorded for Pt-Ir coated (black traces,
n=5) and uncoated (red traces, n=3) in vitro (left plot), and in rat hippocampus (right plot). 60 Hz noise
picked up by high impedance electrodes observed in vitro are greyed out. Impedance magnitude at 1 kHz
demonstrated an average 9.3× in vitro and 7.4× in vivo reduction in impedance magnitude. R s (highlighted
in yellow) approaches 5 kΩ±3 (n=8) in vitro and 50±9 kΩ (n = 8) in vivo. ............................................... 70
Figure 7-3 In vivo voltage transient of Pt-Ir coated vs. uncoated microelectrodes, in response to biphasic
current pulsing in rat hippocampus. a) Representative voltage response traces recorded from Pt-Ir coated
(black) and uncoated (red) microelectrodes, in response to biphasic current pulses (pulse duration of 200
µs, amplitudes of 1 µA to 5 µA). b) Average electrode polarization as a function of stimulus current pulses
across the coated (black) and uncoated (red) electrodes. The uncoated microelectrode surface potential
crosses the cathodic potential safety limit (U = -700 mV) at ~5 µA, whereas the coated electrode reaches
the same polarization in response to a 10× larger current pulse (i = ~50 µA). On average there is an 83%
reduction in electrode polarization for all stimuli. c) Calculated average power consumption associated with
driving the electrodes with biphasic stimulation pulses, plotted as a function of stimulus current magnitudes.
Data comparing the Pt-Ir coated (grey) and uncoated (red) power consumption are shown on the left. Higher
stimulation pulses applied to the coated electrode are shown on the right. Error bars indicate standard error
calculated for coated (n =15) and uncoated (n = 9) electrodes. There is a statistically significant
improvement in power consumption using the Pt-Ir coated electrodes (p < 0.001). On average there is 64%
reduction in power across all stimuli. ......................................................................................................... 72
Figure 7-4 Comparison of spontaneous recordings made from uncoated (top 3) and coated (bottom 5)
electrodes in the CA1 region of the hippocampus (all recordings are plotted on the same scale).
Superimposed grey traces are recordings made from the same electrodes after the animal has been
confirmed. Zoomed insets (right) are examples of complex spikes indicating proper placement of electrodes
in the rat CA1 region of hippocampus (euthanized recordings are omitted). ............................................. 73
xi

Figure 7-5 Example of acute single unit recordings from each electrode. Red, blue and green traces represent
different single units. White and light grey background indicate coated and uncoated electrodes
respectively. ................................................................................................................................................ 74
Figure 7-6 SNR of spontaneous activity. a) PSD from 1-5 kHz of spontaneous neural recordings in the CA1
region of rat hippocampus (PSD anesthetized) and activity after the animal was euthanized (PSD euthanized) using
either a 1 Hz HPF (left) or 300 Hz HPF (right). b) The SNR from 1-5 kHz for uncoated (red) and coated
(black) electrodes for recordings made with a 1 Hz HPF (left) or a 300 Hz HPF (right). SNR approaches 1
dB for frequencies above ~3 kHz for uncoated electrodes when using a 1 Hz HPF and above 4 kHz when
using a 300 Hz HPF. All results presented as mean ± SE for uncoated (n = 6) and coated (n = 10) electrodes.
c) Mean SNR calculated from neural recording data (300Hz-5kHz filtered) for uncoated (n=12) and coated
(n=20) electrodes. The effect of the coating was statistically significant in the mixed linear model
(p<0.0002), but the effect of the filter was not (p=0.81). ........................................................................... 75
Figure 8-1 Peripheral nerve interfaces can be used for a) closed-loop control of prostheses or limbs after
spinal cord injury, and b) modulation of various organs. ........................................................................... 84
Figure 8-2 Fabricated Parylene C cuff electrode with integrated microfluidic channels. A red photoresist
layer is shown in the channels for visibility. ............................................................................................... 85
Figure 8-3 Magnified view of the microfluidic outlets and the electrode arrangement (left) and illustrated
cross section of one channel and its electrodes (right). Electrodes E1, E3, E5, and E7 are partially occluded
by the microfluidic channels with the outlets directly above the electrodes. Electrodes E2, E4, E6, and E8
are fully occluded by the channel (not used in experiments described in this paper). The surface electrode
(ES) is fully exposed. .................................................................................................................................. 85
Figure 8-4 a) LACE threaded underneath sciatic nerve. b) LACE locked around the sciatic nerve. ......... 86
Figure 8-5 Acute in vivo recording setup is shown on the right of each panel, and the corresponding
recordings from LACE are shown on the left. a) LACE was used to record neural activities from the nerve.
E7 was used as the working electrode and SE was used as the reference electrode. Stimulation pulses were
applied to the nerve 20 mm from the recording site using needle electrodes. The stimulation pulses were
charge-balanced anodic-first biphasic pulses with a fixed duration of 200 µs, and pulse amplitudes ranging
from 100 µA to 700 µA in steps of 50 µA. The recording amplifier was set to a gain of 80 dB, with 10 Hz-
10 kHz band pass filter. b) Lidocaine was applied to the nerve distal from the stimulation electrodes and (a)
was repeated. c) Lidocaine was again applied to the nerve, the leg muscle was stimulated at 700 µA, and
activity was recorded from the nerve using LACE. d) CAPs elicited with stimulations of varying amplitude
were recorded using LACE. e) and f) are verification of recorded nerve response. e) Application of lidocaine
to the nerve abolished neural responses while stimulation artifacts were unchanged. f) EMG artifacts are
absent during direct muscle stimulation. ..................................................................................................... 91
Figure 8-6 Recruitment curve of sciatic nerve fibers. Error bars are standard error mean values from three
rats. a) Rectified and integrated values of each CAP activity following stimulation. The integrated CAPs
are normalized by the maxima and plotted versus the stimulus intensity. b) Latency to onset versus the
stimulus intensity. ....................................................................................................................................... 93
Figure 8-7 Acute in vivo stimulation setup is shown on the right panel, and corresponding recorded evoked
EMG shown on the left panel. a) E5 was used as the working stimulation electrode and SE was used as the
reference. Anodic-first stimulation pulses set to a fixed pulse duration of 200 µs and amplitudes ranging
from 40 µA to 200 µA were applied. EMG was recorded bipolarly using needle electrodes in the extensor
digitorum longus muscle in the rat’s right hind leg. The recording amplifier was set to a gain of 60 dB with
xii

10 Hz-10 kHz band pass filter. b) Lidocaine was applied proximal from the neuromuscular junction and (a)
was repeated. c) EMG activity recorded following nerve stimulation. d) Verification of the recorded muscle
response. Application of lidocaine to the nerve abolished EMG response while the stimulus artifact was
unchanged. .................................................................................................................................................. 94
Figure 9-1 Illustration of a miniaturized multi-channel neurostimulation and recording module with
wireless data telemetry housed and mounted on the animal’s skull. The power source is two-coin batteries
mounted on the animal’s back and routed to the PCB. ............................................................................. 102
Figure 9-2 PCB layout (designed in Altium Designer) integrating a 16 channel neurostimulation and
recording module chip (Intan) with a wireless microcontroller (ESP32) for data telemetry. The PCB also
includes an antenna chip for wireless data telemetry, a flash to store the programmed data, and FTDI to
bridge USB to UART commands, two regulators for battery powering, and an on-board charger to recharge
batteries. .................................................................................................................................................... 103
Figure 9-3 Rendering of rigid-flex PCB consisting of double-sided 4-layer PCB designed in Altium
Designer. ................................................................................................................................................... 104
Figure 9-4 SolidWorks rendering of PCB housing. a) Top piece with protruding cantilevers. b) Bottom
piece with undercuts. The PCB with the Omnetics connectors first gets secured in (b) using four 1 mm
screws. c) Jointed mating components (width: 20 mm; length: 20 mm; height: 17 mm). Wall thickness is 1
mm. The openings are for connectors and a switch for power control as labeled. ................................... 105
Figure 9-5 Rendering of assembled folded PCB inside the housing......................................................... 106
Figure 9-6 Mock-up implantation process of the housing on rat’s skull. a) Five anchoring screws on the left
hemisphere of the skull and a marked craniotomy region on the contralateral hemisphere. b) 3D printed
Omnetics connector over the craniotomy region secured in place with dental cement. c) Omnetics connector
glued to the housing is glued to the mating connector on the skull. d) Batteries attached to a rodent jacket
via Velcro and worn on the back of the animal. ....................................................................................... 107

xiii

ABSTRACT
New technologies have emerged in recent years to treat various neurological disorders via
electrical neuromodulation or replace a lost function via neural prosthesis. For example, deep brain
stimulation has provided therapy to patients with Parkinson’s disease, epilepsy, severe depression,
and dystonia. Cochlear implants and retinal prothesis have successfully provided hearing and some
visual acuity to patients with sensory impairment by electrical stimulation of the cochlea or retina
using appropriate patterns of pulses. Hippocampal memory prothesis has also shown success in
restoring memory by reinstating the input-out properties of neurons, which requires delivery of
neural-code-like stimulation patterns to the brain tissue with multiple electrodes.
Current technology to support such systems lack feedback control and the ability to
precisely modulate and monitor neural activities. These needs may be met with high density
recording and modulation (electrical and chemical) systems with high resolution, stimulation and
recording from the same electrode for feedback control, ability to deliver complex pattern of
stimulation, and power efficient interfaces. Furthermore, miniaturized wireless systems is essential
for chronic free-roaming applications.
To better support neuromodulation and neural prosthesis, we present the following neural
interface systems: 1) a highly configurable multichannel asynchronous neural-code-based
stimulator with ability to stimulate and record from the same electrode near simultaneously, 2)
implantable electrode design with high signal-to-noise-ratio, power efficiency, and spatial
resolution for both recording and stimulation, 3) design and performance of flexible electrodes
with microfluidic channels to inject drug focally with controlled doses for chemical modulation of
the electrode-tissue interface, 4) system miniaturization for mounting on rat’s skull for future free-
roaming small animal neuroscience research. The body of work presented int this thesis starts with
xiv

background review of the need for neural interface technology, to technology design, prototyping,
and testing on benchtop and then in vivo rats, concluding with the plan for next generation design
and validation. We expect the final device to be a valuable tool for studying neurobiological basis
of cognitive functions and a step closer to meet the unmet needs of next generation neural interface
technologies.
1

CHAPTER 1: INTRODUCTION
1.1 Motivation
Neural interface technology has made much progress in recent years aiming to provide
therapies via neuromodulation to patients with neurological disorders or replace a lost function via
neural prosthesis. In the clinic, neuromodulation using deep brain stimulation (DBS) has provided
treatments to various neurological disorders such as epilepsy (Lee et al. 2015), depression
(Schläpfer and Kayser 2014), Parkinson’s disease (Voges, Koulousakis, and Sturm 2007),
memory loss (Hamani et al. 2008), and Tourette’s syndrome (Ackermans et al. 2006). Researchers
have also been using neuromodulation as a tool to record and manipulate neural circuits to study
neural correlates of sensory, motor and cognitive functions (Nimmagada and Weiland 2018;
Brecht et al. 2004; Salas et al. 2018).
In neural prosthesis applications, neural interfaces have been used to convert sensory input
signals to neural stimulations as in cochlea prostheses (Holden et al. 2013) and retinal prostheses
(Weiland, Cho, and Humayun 2011), or decode motor cortical output signals into movements as
in motor prostheses (Velliste et al. 2008). A novel form of neural prosthesis is a hippocampal
memory prosthesis, which aims to restore cognitive functions lost in injures or diseases due to
destruction of neurons and their connections in a specific region of the brain (Theodore W. Berger
et al. 2011). It relies on a computational model that mimics the input-output properties of the neural
circuit to be replaced (D Song et al. 2007; 2018). To implement a hippocampal memory prosthesis,
it is essential to be able to deliver temporally and spatially distributed neural code-based
stimulation patterns to the brain tissue with multiple electrodes.
Next generation neural interfaces for neuromodulation and neural prosthesis require
feedback control and high special resolution for neural-code-based precise recording and
2

modulation (both chemical and electrical) of neural activities. Existing technologies for brain
implantation, as in DBS, utilize fixed interval trains of pulses, with a single or small number of
stimulation electrodes. However, to enable neural code-based stimulation, a system for real-time
precise delivery of large-scale spatiotemporal patterns of electrical pulses must be designed and
implemented. In addition, the ability to record from single neurons and stimulate to same
microelectrodes in parallel is highly needed for feedback control (Figure 1.1). In
electrophysiological studies, same electrode recording and stimulation would enable stimulus-
response experiments at single neuron or small neuronal population level (Shepherd, Hardie and
Baxi, 2001; Houweling and Brecht, 2008; Krause et al., 2019). In DBS such technique would
allow delicate micro-manipulation of complex neural circuits and monitoring feedback neural
signals with high spatial resolution (Vesper et al., 2002; Little et al., 2013; Priori et al., 2013;
Salam, Velazquez and Genov, 2016; Swan et al., 2018). In hippocampal memory prosthesis,
stimulating and recording from the same single neurons becomes vital for successful
implementation of the single neuron-level, multi-input, multi-output model-based
microstimulation (Deadwyler, 2018; Song et al., 2018).
The main challenge of generating large-scale stimulation pulses is hardware efficiency.
Power and area are crucial parameters of an implantable and apparent solution of separate pulse
generators for each electrode is inefficient for brain neuromodulation and prosthesis. This is
because the natural rate of neural activity is slow, so each pulse generator would need to spend a
major amount of time in quiescent mode, which stills consumes power. Whereas, time
multiplexing a single pulse generator into multiple channels enables combining several sources
into a single high rate pulse generator.
3

One challenge of recording in conjunction with stimulation is stimulus artifacts, which
masks neural activities of interest. Minimizing such recording contaminants is vital for better
recording evoked neural activities and controlling neural interface devices. The second challenge
is that the geometric area of a recording electrode should be comparable to the size of a single
neuron to record unitary activities, but at the same time, stimulation electrodes require relatively
large surface area to obtain low electrochemical impedance that allows safe charge injection to
evoke desired neural responses. It is therefore highly desirable to minimize electrochemical
impedance of the electrode while keeping the electrode area small enough for single neuron
recording.
To better support neuromodulation and neural prosthesis, I present a neural interface
technology with high signal-to-noise-ratio, power efficiency, and spatial resolution. The system
includes a highly configurable asynchronous multi-channel neurostimulator. Stimulation pulses
can be generated in real-time driven by an external source and feedback from neural response to
stimulation from the region of stimulation. Each stimulation channel is equipped with an artifact
suppression technique to reduce recovery period from the artifact. The system also includes design
of microelectrodes with enhanced coatings to enable same electrode stimulation and recording
single neuron activities in parallel. The electrodes were rigorously characterized and demonstrated
statistically significant electrochemical and electrophysiological improvement over traditional
electrodes in terms of impedance, range of stimulation parameters, and precision. Separately, I
discuss design and perform evaluation of polymer-based electrodes with microfluidic channels to
inject drug focally with controlled doses for chemical modulation of the electrode-tissue interface.
Finally, I demonstrate system miniaturization for mounting on rat’s skull for future free-roaming
expedients.
4

The body of work presented here starts with background review of the need, to technology
design, prototyping, and testing on benchtop and then in vivo rats, concluding with the plan for
next generation design and validation.

Figure 1-1 Illustration of (A) a conventional open-loop DBS and (B) a configurable multi-channel
neurostimulator for implementing closed-loop processing from the region of stimulation.
1.2 Outline
Chapter 2: Review of Neural Interface Technology Background
This chapter is a background review of neuromodulation systems, commercial products available
for rat electrophysiology experiments, need to improve state of the art technology, and challenges
to meet such needs are identified. It also presents the advantages of stimulation, recording and
drug delivery from the same interface.
Chapter 3: Stimulation of Neural Tissue
This chapter reviews the underlying neural tissue physiology and the underlying mechanism in
neural stimulation and recording using electrodes. It addresses concerns about how to produce a
practical electrical stimulation of neural tissue with safety and efficacy.

5

Chapter 4: System Design of Neural-Code Based Stimulation
This chapter reviews the underlying brain firing pattern, which enable an alternative method for
electronic circuit design of a multi-channel neurostimulator that mimics neural activities. It
addresses the design requirements to develop a multi-channel highly configurable electrical
stimulation pulse pattern generator. It also presents design implementation and system
characterization on bench-top set-up.
Chapter 5: Stimulation in Conjunction with Recording
This chapter presents an inexpensive technique to allow stimulation and recording from the same
electrode. The objective of this design is to recover short latency neural response to stimulation
from stimulation artifact. Technical design requirements and integration with the neurostimulator
is described. It provides details of engineering work to bring the device into realization. Bench top
tests to evaluate the design in a phantom are also presented to set the stage for in vivo validation
using coated microelectrodes.
Chapter 6: Microelectrode Design for Same Electrode Stimulation and Recording
This chapter details limitations of current electrode technology for same electrode focal
stimulation and high-resolution recording. It provides design and characterization of an
implantable microelectrode array with electrodeposited Platinum Iridium Coating to overcome this
challenge. Preliminary data from in vitro are presented to set the stage for in vivo experimental
direction.
Chapter 7: Proposal for in vivo Validation
This chapter delves into questions related to in vivo validation and verification of the design for
bidirectional stimulation and recording. It presents the collected data and results of the animal
6

studies. Quantitative analysis of device assessment in generating and recording neural response
are provided. The discussion presents plan for future studies of neurophysiological mechanisms
using this technology.
Chapter 8: Integrated Microfluidic Channels
This chapter describes integration of microfluidic channels with electrodes. Its application to
deliver drug focally with controlled boluses is demonstrated in vivo.
Chapter 9: System Miniaturization
This chapter presents the required strategy for system miniaturization. It demonstrates rendering
of PCB design and housing to be mounted on rat’s skull along with in vivo evaluation of a mock-
up for chronic implantation. The goal of this is to provide guidance for technological translation
for free roaming small animal neuromodulation systems.
Chapter 10: Conclusion and Prospect
This chapter summarizes the results from this body of work and proposes adaption of current
technology to address neuroscience problems. It discusses an outlook on the future of neural
interface technology.

7

Chapter 2: Review of Neural Interface Technology Background
2.1 Introduction
Neural interface technology provides a way of communication between the nervous system
and electronic circuitry. It has been deployed in basic neuroscience research to study input-output
properties of neural tissue and in the clinic as deep brain stimulation (DBS) to provide therapy to
patients with neurological disorders. The interface between the electronic circuits and the excitable
tissue is an electrode, which must be meticulously designed and characterized per application. An
electrode senses neural activity and may also be driven by electronic circuitry to stimulate and
thereby manipulate neural activities. Current technology to support such systems lacks precision
and feedback control from the stimulated tissue. What is needed is large-scale neural code-based
stimulation and the ability to stimulate and record from the same electrode. In addition, focal
chemical manipulation of neural tissue is essential in basic neuroscience research. We propose the
design and in vivo evaluation of a highly configurable 32-channel neurostimulator with capability
to record short latency neural response to stimulation from the stimulation electrode (Figure 1.1).
Separately, we demonstrate the capability of an integrated microfluidic channel with polymer-
based microelectrodes to deliver drug focally and with highly controlled boluses.
2.2 Review of Neural Interface Technology Applications
Researchers have been using neural interface technology as a tool to record and manipulate
neural circuits to study neural correlates of sensory, motor and cognitive functions. In the clinic,
DBS has provided treatments to various neurological disorders such as epilepsy (Lee et al., 2015),
depression (Schläpfer and Kayser, 2014), Parkinson’s disease (Voges et al., 2007), memory loss
(Hamani et al., 2008), and Tourette’s syndrome (Ackermans et al., 2006). In neural prosthesis
applications, neural interface has been used to convert sensory input signals to neural stimulations
8

as in cochlea prostheses (Holden et al., 2013) and retinal prostheses (Weiland et al., 2011), or
decode motor cortical output signals into movements as in motor prostheses (Velliste et al., 2008).
Hippocampal memory prosthesis, is a novel form of neural prosthesis that aims to restore
cognitive functions lost in injures or diseases due to destruction of neurons and their connections
in a specific region of the brain (Theodore W. Berger et al., 2011). It relies on a computational
model that mimics the nonlinear dynamical multi-input, multi-output (MIMO) properties of the
neural circuit to be replaced (Song et al., 2018, 2007). The MIMO model enables the prosthesis to
stimulate a downstream brain region with appropriate output spatiotemporal patterns of neural
codes predicted from input spatiotemporal patterns of neural activities recorded from an upstream
brain region (Berger et al., 2010; Song et al., 2009). By reinstating the neural code processing and
transmission, the damaged brain region is bypassed, and the cognitive function is thus restored
(Theodore W Berger et al., 2011; Deadwyler, 2018; Hampson et al., 2013, 2012). To implement
a hippocampal memory prosthesis, it is essential to be able to deliver temporally and spatially
distributed neural code-based stimulation patterns to the brain tissue with multiple electrodes.
Peripheral nerve interfaces also aim to provide therapies for various diseases or restoration
of lost functions through their interactions with the peripheral nervous system. For example,
electrical stimulation of the peripheral nerve i.e. sciatic, ulnar, and occipital nerve have been used
to treat chronic pain (Picaza et al., 1977; Sun and Morrell, 2014). Therapeutic modulation of organ
functions through vagus nerve stimulation (VNS) has demonstrated success in controlling blood
pressure (Plachta et al., 2014), reducing body fat (Banni et al., 2012), suppressing epileptic seizures
(Zabara, 1985), and treating arthritis (Giagka and Serdijn, 2018; Koopman et al., 2016).
Furthermore, bidirectional recording and stimulation from peripheral nerves has become an
important approach for achieving closed-loop control of neural prostheses such as robotic hands
9

(Grill and Mortimer, 1996; Schultz and Kuiken, 2011; Tan et al., 2015) or a limb after spinal cord
injury (del Valle and Navarro, 2013; Russell et al., 2019).
2.3 Need Identification
Recording and stimulation from same microelectrodes is highly desirable for both basic
neuroscience research and neural engineering applications. In electrophysiological studies, same
electrode recording and stimulation would enable stimulus-response experiments at single neuron
level or small neuronal population level (Houweling and Brecht, 2008; Krause et al., 2019;
Shepherd et al., 2001). DBS, such technique would allow delicate micro-manipulation of complex
neural circuits and monitoring feedback neural signals with high spatial resolution (Little et al.,
2013; Priori et al., 2013; Salam et al., 2016; Brandon D. Swan et al., 2018; Vesper et al., 2002). In
cortical prostheses such as the hippocampal memory prosthesis, stimulating and recording from
the same single neurons becomes vital for successful implementation of the single neuron-level,
multi-input, multi-output model-based microstimulation (Deadwyler, 2018).
Furthermore, to enable neural code-based stimulation, a system for real-time precise
delivery of large-scale spatiotemporal patterns of electrical pulses must be designed and
implemented. For example, multi-channel microstimulation can provide more focal and effective
modulations to the brain compared with single-channel DBS (Brandon D Swan et al., 2018; Vesper
et al., 2002). DBS parameters such as the pulse width, waveform, frequency, amplitude and
duration can be more effectively tuned based on feedback from neural activities (Little et al., 2013;
Priori et al., 2013; Salam et al., 2016). Thus, a configurable multi-channel neurostimulator with
feedback recording capability may potentially enable safer, more precise and efficient DBS
treatments, as well.
10

Lastly, in all applications of peripheral nerve interfaces, selective recording and stimulation
of nerve fibers within a nerve bundle is highly desired so that targeted neural effects can be
achieved while undesired off-target effects are avoided. For example, motor control requires
recording and translating neural commands from efferent fibers within the nerve, while stimulating
afferent fibers to send sensory feedback to the user. Furthermore, organ modulation via nerve
stimulation, as in bioelectronic medicine, requires selective targeting of the specific fibers within
a nerve trunk that innervate the organ of interest. Focal chemical manipulation of the nerve tissue
may provide a way to improve selectivity.
Summary of needs:
• Precise delivery of large scale spatio-temporal patterns of stimulation pulses
• Stimulation and recording from the same electrode with
o High spatial resolution
o High signal-to-noise ratio
o Power efficiency
o Feedback signals recorded from the manipulated tissue
• Focal chemical manipulation
2.4 Challenges
2.4.1 Large Scale Neural Code-Based Stimulation
The main challenge of generating large-scale stimulation pulses is hardware efficiency.
Power and area are crucial parameters of an implantable and apparent solution of separate pulse
generators for each electrode is inefficient for neural code-based stimulation. This is because the
natural rate of neural activity is slow, so each pulse generator would need to spend a major amount
of time in quiescent mode, which stills consumes power.
11

2.4.2 Stimulation and Recording from the Same Electrode
One challenge of recording in conjunction with stimulation is due to the prolonged
saturation of the recording amplifier caused by stimulus artifacts, which masks neural activities of
interest. Stimulus artifact may last for tens of milliseconds or even hundreds of milliseconds
depending on the amplifier, electrode property, tissue property, and stimulus parameters (Rolston
et al., 2009a). Minimizing such recording contaminants is vital for better recording evoked neural
activities and controlling neural interface devices.
Also, current electrode technology to support stimulation and recording from the same
electrode lacks precision, and power efficiency. For recording electrodes, precision is important
for differentiating single neuron activity from background noise including noise at electrode-
electrolyte interface. For stimulation electrodes, precision is essential for focal delivery of charge
to the target neural tissue. Furthermore, feedback control with precision from the stimulated tissue
enables proper adjustment of the stimulation parameters over time. This is especially crucial in
chronic applications where glial cell encapsulation can weaken the electrode-tissue interaction and
cause reduction in the stimulation effect over time (Vadim S Polikov et al., 2005). Neural plasticity
may also alter response to stimulation and require feedback control to optimize stimulation
parameters (Kerr et al., 2011; Månsson et al., 2016a). Lastly, battery operated free-roaming
experiments and implantable neuromodulation devices, require low power systems for long-term
use of the device.
The aforementioned needs may be addressed with low-impedance microelectrodes that
allow both stimulation and recording. The issue with such system, however, is that the geometric
area of a recording electrode should be comparable to the size of a single neuron to record unitary
activities. On the other hand, traditional stimulation electrodes require relatively large surface area
12

for obtaining low electrochemical impedance that allows safe charge injection while evoking
neural responses. The challenge with reducing the electrode dimension for high precision
stimulation and recording is that it would result in an increase in the electrochemical impedance
of the electrode-tissue interface (Stuart F Cogan, 2008). For recording electrodes this would result
in higher thermal noise (Suner et al., 2005). For stimulation electrodes, where the same amount of
charge must be delivered across a smaller interface, the increased impedance results in increased
electrode polarization, power consumption, and limits maximum electrochemically reversible
stimulation pulses (Kelly et al., 2014). It is therefore desirable to minimize electrochemical
impedance of the electrode while keeping the electrode dimension small enough for single neuron
recording.
2.5 Review of Existing Technology
2.5.1 Large Scale Neural Code-Based Stimulation
Existing neural interface technologies for brain implantation, as in DBS, utilize fixed
interval trains of pulses, with a single or small number of stimulation electrodes. However, to
enable neural code-based stimulation, a system for real-time precise delivery of large-scale
spatiotemporal patterns of electrical pulses must be designed and implemented. Systems such as
the ones developed by Lo et al. (Lo et al., 2017) is a fully integrated closed-loop wireless
neuromodulation system with 40 individual current sources for spinal cord stimulation. However,
it does not include the function of generating arbitrary pulse patterns independently for each
channel. The system developed by O’Leary et. al (O’Leary et al., 2018) is an arbitrary single
channel waveform generator for adaptive seizure control, which suffers from low channel count.
Tran et al. (Tran et al., 2014) offers a CMOS 256 channel neurostimulator for retinal prosthesis,
13

which is expensive and inefficient for neural code-based stimulation in small animal neuroscience
research.
2.5.2 Stimulation and Recording from the Same Electrode
2.5.2.1 Hardware Electronic Circuit Technology
The challenge in realizing a system that allows stimulation and recording from the same
electrode is imposed by direct connection of the stimulator and amplifier through a recording
electrode. If a stimulation pulse causes amplifier saturation, the input signal would be clipped at
the amplifier’s maximum input range. Consequently, neural response would be completely masked
with this artifact and cannot be recovered. On the other hand, if recording is from a neighboring
electrode to the stimulation electrode, or if the recording is from the stimulation electrode but the
applied stimulus magnitude is small enough, the amplifier may not get saturated. In this case, often
back-end signal processing may be used to separate artifact from neural response (Zhou et al.,
2018).
A fundamental limitation of back-end signal processing approach is that it relies on
unsaturated recordings of neural signals and stimulus artifacts, which are often unavailable due to
the commonly encountered saturation of recording amplifiers. Even back-end signal processing of
unsaturated recordings such as the ones proposed by Wagenaar et al. (Wagenaar and Potter, 2002)
and by Wichmann et al. (Wichmann, 2000) face difficulties in separating neural activity from
stimulus artifact due to their overlap in both time and frequency domains. More recently,
Limnuson et al. (Limnuson et al., 2015) developed a more sophisticated real time artifact
suppression technique based on template subtraction, which requires a VLSI chip.
14

To avoid amplifier saturation, front-end artifact reduction has been implemented, which
typically involves increasing the dynamic range of the amplifier to withstand larger voltages. This
approach sacrifices power efficiency by using a higher voltage supply (Rolston et al., 2009b) and
still requires back-end signal processing to reduce the stimulus artifact. Another front-end
approach is to subtract the artifact at the negative input of the amplifier based on a model that
replicate the electrode-tissue properties (Nag et al., 2015). This technique holds promise but
involves relatively complex computation and is to be validated in biological preparations.
Blanking techniques have also been used to reduce the artifact recorded from non-
stimulation electrodes, where residual charge left on the electrode after termination of the
stimulation pulse does not need to be accounted for (Cheng et al., 2017; Venkatraman et al., 2009).
The system that accounts for the discharge period of the stimulation electrode for same electrode
stimulating and recording is by DeMichele et al. (DeMichele and Troyk, 2003). This system does
not demonstrate data from bench-top system or neural tissue. Howttoway et al. (Hottowy et al.,
2012) designed a system with capability to generate spatio-temporal pattern of stimulation and
stimulus artifact reduction. The range of stimuli used in this system to test the artifact rejection
technique in vitro are low (0.43µA, 100µs), which did not cause amplifier saturation.
Lastly, state of the art neurostimulation integrated circuit has been realized in CMOS by
Intan Technologies and utilized by many research groups (Ewing et al., 2013; Sessolo et al., 2013).
The RHS2000 is a 16-channel stimulator/amplifier chip, which employs a fast recovery settling
time scheme to minimize artifacts from stimulation. The technique Intan chip uses is “Low-
frequency Cutoff shifting” to reduce recovery time from stimulation. This means that the user can
switch the cut-off frequency of the amplifier high pass filter to a higher value during stimulation,
effectively reducing the time constant and therefore the recovery period. However, this technique
15

is mainly for stimulation and recording from neighboring electrodes. If it is implemented for same
electrode stimulation and recording, the potential problem would be that some of the charge from
the amplifier may couple into the amplifier and damage it over time. Plus, the amount of intended
charge delivery to the electrode would be different than calculated.
2.5.2.2 Electrode Technology
One of the most common ways to decrease the impedance of microelectrodes is applying
activated iridium oxide (IrOx) to the surface of the electrode. Iridium electrodes benefit from high
charge injection capacities per unit of geometric area and reduced electrochemical impedance due
to its high number of oxidative states and increase in the effective area. Three approaches have
been used to make IrOx coated microelectrodes: Activated iridium oxide film where an iridium
electrode is oxidized by cycling it through positive and negative voltages (Beebe and Rose, 1988),
sputtered iridium oxide films where an iridium target is used in the presence of oxygen (S F Cogan
et al., 2004), and electrodeposited iridium oxide(Lu et al., 2009).
2.5.3 Peripheral Cuff Electrodes
Extraneural cuff electrodes provide a minimally invasive method to interface with the
peripheral nerve as they wrap gently around the nerve and stimulate or record from multiple nerve
fibers with non-penetrating surface electrodes. They enclose the circumference of the nerve with
an insulative material to restrict ionic currents. Metal electrodes on the inner wall of the cuff enable
stimulation and recording of the encircled nerve fibers (Grill and Mortimer, 1996; Marks and Loeb,
1976). More recent cuff electrodes take advantage of microfabrication techniques and are made
from thin-film polymers and metals. They consist of insulated thin-film metal traces, exposed
electrode sites and integrated interconnects that allow external connection (Caravaca et al., 2017;
Rodríguez et al., 2000). Ideally, the cuff diameter can be adjustable to fit nerves of varying sizes,
16

thereby allowing surgical flexibility and better integration of the cuff with the nerve (Loeb and
Peck, 1996).
Cuff electrodes induce minimal immune response when properly installed (Navarro et al.,
2005). However, electrodes on the surface of the cuff are separated from nerve fibers by the
epineurium and perineurium and therefore suffer from lack of selectivity and sensitivity (Hoffer
and Kallesøe, 2001; Kang et al., 2015; Loeb and Peck, 1996). A variety of strategies have been
investigated to improve selectivity and sensitivity (Larson and Meng, 2019), including extraneural
cuffs which reshape the nerve (Leventhal and Durand, 2003; Schiefer et al., 2010), (Caparso et al.,
2009; Leventhal et al., 2006) and intrafascicular electrodes which penetrate the nerve trunk (Badia
et al., 2011; Lawrence et al., 2002). Such approaches have varying degrees of success in improving
device functionality while maintaining long-term performance and nerve health. The latter
penetrating approach is highly invasive and breaches the integrity of the nerve. To our knowledge,
non-surgical methods to focally remove epineurium and perineurium have not been explored in
conjunction with cuff electrodes.
2.6 Design Process Overview
2.6.1 Hardware Electronics Circuit Design
Time multiplexing a single pulse generator into multiple channels enables combining
several current sources into a single high rate pulse generator. Furthermore, neural spiking activity
is sparse in nature and simultaneous pulse generation from multiple electrodes would not result in
sparse neural code. On the other hand, asynchronous low magnitude stimulation pulsing has the
potential of generating sparse neural code. Accordingly, a highly configurable and asynchronous
neurostimulator would enable assessment of whether sparse low amplitude pulses would generate
sparse neural code.
17

We present a highly configurable asynchronous multi-channel neurostimulator that can be
driven by a MIMO model-based computational unit to continuously generate neural code-like
spatiotemporal patterns of stimulation pulses with adjustable pulse parameters. Stimulation pulses
may be generated in real-time driven by an external source (i.e., the output of the MIMO model)
and feedback from neural response to stimulation from the region of stimulation. Each stimulation
channel is equipped with a switching mechanism designed to reduce recovery period from the
artifact from tens to hundreds of milliseconds to ~2 ms. We designed, fabricated and characterized
the system first in phantom preparations and then the hippocampi of behaving rats.
2.6.2 Electrode Design
Electrodeposition has the advantage of being cheaper as it does not require a cleanroom
and can be applied to biomedical electrodes made from almost any electrically conductive
material. The addition of Pt to electrodeposited IrOx to form an Electrodeposited Pt-Ir Coating
(EPIC) maintains the advantages of electrodeposited iridium, with the added benefit of containing
Pt which makes the coating more robust and less prone to delaminate compared to IrOx (Stuart F
Cogan et al., 2004; Dalrymple et al., 2019; Petrossians et al., 2011).
2.6.3 Design of Animal Studies
We evaluated the in vivo performance of the designed neurostimulator with artefact
suppression connected to EPIC coatings on microelectrodes by acutely implanting 8-channel Pt-Ir
microelectrodes, with approximately every other electrode coated and the other half uncoated. The
arrays were implanted in the CA1 cell body layer of rat’s hippocampi. Performance of coated
electrodes was quantitatively compared against uncoated electrodes within the same device and
across different animals. The electrodes were tested for signal to noise ratio, electrode polarization,
18

power efficiency, charge storage capacity, and capability to stimulate and record short latency and
prolong neural response to various electrochemically reversable stimulation parameters.
2.6.4 Integrated Microfluidic Channels
To provide a means to evaluate our approach towards achieving fascicular selectivity and
sensitivity, we developed a novel hybrid, multi-functional cuff electrode (Cobo et al., 2019). The
design consists of a microfabricated thin-film polymer cuff with electrodes for recording and
stimulation, and microfluidic channels to deliver drugs focally. The microfluidic channels
terminate in a well-defined outlet and allow focal drug delivery of a chemical lysing agent to non-
surgically remove connective tissue and thereby minimize the distance between the electrodes and
nerve fibers without introducing mechanical trauma to the nerve. The resulting “window” provides
greater access to the underlying nerve fibers for electrodes on the cuff. This could possibly be
enhanced further by delivery of neurotrophic factors in order to attract axonal sprouting towards
the electrodes within the channels. Reflective of this design concept, the device has been termed
the lyse-and-attract cuff electrode (LACE) (Cobo et al., 2019).
The engineering design, fabrication, packaging, electrical characterization and fluidic
performance of LACE on a benchtop set-up were demonstrated in a previous study (Cobo et al.,
2019). In this work, the implantation procedure and a systematic evaluation of the LACE’s
capability to acutely record, stimulate, and deliver lysing agent to rat sciatic nerves in vivo is
described.
19

CHAPTER 3. STIMULATION OF NEURAL TISSUE
3.1 Physiology of Neural Tissue
Our brains contain about 85 billion neurons. The neuron is the primary functional unit of
the nervous system. Its structure consists of dendrites, cell body and axons (Figure 3.1). Dendrites
are the area where neurons receive their information through receptors that are designed to pick
up chemical signals called neurotransmitters from other neurons. The signals picked up by
dendrites cause electrical changes in a neuron interpreted in the cell body, or soma. The soma
contains the nucleus, which contains the DNA of the cell. If the signal coming from dendrites is
strong enough, then the signal is sent to the axon. At this point the signal is called an action
potential. An action potential is a nerve impulse that causes movement of ions across the cell
membrane of a neuron.
The cell membrane of a neuron contains thousands of ion channels allowing either Na
+
or
k
+
to pass through. At resting state, the channels on the neuron are closed, and the charge inside of
the cell membrane is more negative than the outside (~-70mV). In this state, the cell membrane is
polarized because of this electrical difference across the membrane. An action potential starts when
an input disturbs the neuron plasma membrane to the point that the potential difference reaches a
threshold voltage of about -55 mV. This causes hundreds of Na
+
to flow inside the cell membrane
causing the membrane to temporarily become more positive than the outside, which in turn causes
the potential difference to rise. This is called the depolarized state because the net charge inside
the cell changed from negative to positive as Na
+
enters in. As the impulse passes, the k
+
channels
begin to open allowing positively charged k
+
to flow out, which causes inside of the cell to resume
a net negative charge at this point. The voltage is now said to be repolarized with more negative
ions inside than outside. After an action potential, some voltage gated k
+
channels remain open
20

resulting in further movement of k
+
out of the cell causing a voltage drop below the resting state.
This is called the refractory period of the neuron. At this stage, the neuron is unable to conduct an
action potential and it is said to be in refractory period during which time the neuron is returning
to resting potential.
The action potential then travels down the axon. The axon may be covered with myelin, an
insulatory material that helps prevent the signal degrading, with gaps known as nodes of ranvier
which are unmyelinated. The wave of depolarization can only occur at the node of ranvier. Thus,
an action potential appears to jump from one node to another when traveling down the axon. This
phenomenon is known as saltatory conduction. The last step of the action potential is the axon
terminal, also called the synapse, where the signal reaches the axon terminal. It can cause release
of neurotransmitters, which can interact with receptors on the dendrites of the next neuron, and the
process repeats.

Figure 3-1Sketch of a neuron including cell body, axon with myelin and axon terminal.
21

3.2 Electrode-Tissue Interface
Increasing the potential difference across a cell body or an axon above its threshold to
generate an action potential is possible with artificial stimulation using electrical impulses.
Similarly, changes in electrical activity of a neuron or population of neurons may be recorded
using probes connected to an amplifier. The interface between neural tissue and electronic
hardware is an electrode which serves as a metal to transfer charge to or from the extracellular
environment. When two electrodes come in contact with the neural tissue, it forms a closed path
for current.
At the electrode tissue interface, there is an interaction between electrons on the metal
electrode and ions on the extracellular environment. There are two mechanisms during charge
transfer at the interface: reversable reactions and irreversible reactions. Reversable reactions
involve physical absorption and desorption of charged particles in the extracellular environment.
This reaction may sustain using alternating current and is modeled as a capacitor (Cdl) in the
electrical equivalent circuit model of electrode/electrolyte interface (Figure 3.2). If too much
charge is applied in one phase of the waveform, irreversible faradaic reactions may occur causing
formation of new species. Such reactions are represented by a resistor (Rp) as they involve local
donation of electrons. Irreversible reactions are undesirable because they cause corrosion and
tissue damage. Lastly, the ion conductivity of the electrolyte is represented by a resistor (Rs).
22

Figure 3-2 Equivalent electrical circuit model of the electrode/electrolyte interface. C dl is the double layer
capacitor representing capacitive reactions, R p is the parallel resistor representing faradaic reactions, and
Rs represents solution resistance.
3.3 Stimulation Waveform
Cathodic stimulation pulses are typically more effective and efficient than anodic
stimulation. A cathodic (negative) pulse attracts positive ions away from the cell and causes an
increase in the potential difference across it. When this rise in voltage reaches above a threshold
of voltage dependent Na+ channels, an action potential occurs.
The minimum pulse duration and magnitude of the cathodic pulse to generate an action
potential is dependent on the proximity of the electrode and neural tissue, stimulation waveform,
and the state of the neural tissue. The strength duration curve relationship is the relationship
between the applied amplitude and the pulse duration (Figure 3.3). As the pulse duration increase,
the stimulation amplitude threshold decreases. However, as the pulse duration approaches zero,
the amplitude approaches infinity asymptotically. This is due to the RC time constant of a cell
required to approach a given voltage before eliciting an action potential. Also, as the pulse duration
approaches infinity, pulse amplitude asymptotically approaches zero. This is because neurons act
as a high pass filter and filter DC like waveforms. The minimum current to elicit excitation is
23

called rheobase when the pulse duration is very long. A current amplitude of twice the rheobase
corresponds to the time called chronaxie. It is generally most efficient to provide a pulse duration
of chronaxie.

Figure 3-3 Example of strength duration curve. Chronaxie is the time duration and 2x rheobase is the
amplitude at which stimulation is most efficient.
3.4 Safety Requirements
A continuous application of a cathodic pulse to the electrode will cause it to polarize. If the
polarization goes beyond water window, electrolysis will occur and cause irreversible reactions.
Thus, a stimulation pulse of opposite polarity must be applied to the electrode following the
cathodic phase to reverse the polarization voltage across the electrode and prevent buildup of
charge at the electrode/electrolyte interface. This is called biphasic pulsing.
Voltage transient measurements are typically used to estimate the maximum charge that
can be safety injected to an electrode in one phase of the biphasic pulse when using constant current
stimulation. An example of a constant current biphasic pulse and its resultants voltage transient is
shown in Figure 3.4. The voltage transient consists of several elements including the ohmic voltage
drop (Va) across Rs and a polarization voltage across the electrode/electrolyte interface (Ep). Ep is
24

then compared with established water oxidation and reduction potentials for an electrode material
to determine its potential limits. Multiplying the current amplitude and duration in one phase will
provide the charge injection limit of the target electrode.

Figure 3-4 Voltage transient of an electrode (blue) in response to a biphasic constant current biphasic pulse
(red) (Reprinted from Cogan 2008).

25

CHAPTER 4: SYSTEM DESIGN OF NEURAL-CODE BASED STIMULATION
4.1 Introduction
Mimicking the neural code using stimulation pulses is desirable in a variety of neural
interface applications, particularly a cognitive prosthesis aiming to restore cognitive functions by
reinstating neural code transmissions in the brain. Neural code-based stimulation requires a real-
time precise delivery of large-scale spatiotemporal patterns of electrical pulses. Because neural
spiking activity is sparse and slow in nature and simultaneous pulse generation from multiple
electrodes is seldom, asynchronous low magnitude stimulation pulsing has the potential of
generating sparse neural code.
In this chapter, I present a highly configurable asynchronous multi-channel
neurostimulator that can be driven by an external source to continuously generate neural code-like
spatiotemporal patterns of stimulation pulses with adjustable pulse parameters in real time. The
system is implemented with low-power and compact packaged microchips to constitute an
effective, cost-efficient and miniaturized neurostimulator. Benchtop results demonstrate the
capability of the stimulator to generate arbitrary spatio-temporal pattern of stimulation pulses
dictated by random number generators to control magnitude and timing between each individual
biphasic pulse.
4.2 System Architecture
The principle elements of the design include a stimulation pattern generator, a multiplexer,
a micro-processor-based controller, and a set of serially controlled CMOS switches for stimulus
artifact suppression (SAS).
26

Figure 4.1 illustrates the block diagram and data stream of the neurostimulator with the
stimulus artifact suppression technique (descried in more depth in the next chapter). An ultra-low
power microcontroller (MSP430G2553 from Texas Instruments) is selected and programmed to
generate command signals including the shape of the stimulation pulse, timing of the pulses, and
the electrode to be pulsed. The microcontroller runs at a clock frequency of 10 MHz and is
programmed with 4 blocks of control units: multiple input multiple output (MIMO) model
simulator, pulse pattern generation control unit, multiplexer control unit, and switch control unit.
The MIMO model simulator simulates inputs from an external source such as the output of the
MIMO model using a random number generator (RNG), which are generated by the
microcontroller through recording noise from a floating general-purpose pin. It also uses another
set of random numbers to vary the magnitude of the stimulus simulating feedback from neural
response to stimulation. The timing and magnitude information across 32 channels are then
collapsed into a single array. In case two channels need to be stimulated simultaneously (which is
a rare event), one may be delayed by the duration of the biphasic pulse. The pulse pattern
generation unit then translates the information into commands controlling the single channel
neurostimulator to generate biphasic pulses with varying timing and magnitude. Next, the
multiplexer control unit activates proper select lines of the multiplexer to send each stimulus to
the target channel.
The switch control unit, which is synched with the MIMO model simulator, controls timing
and state of each switch used to suppress the stimulus artifact. The switches for each channel may
be controlled individually, but because the stimulus artifact will contaminate all channels within
the same media, the stimulation should be blocked from all recording channels (described in the
next chapter).
27

Figure 4-1 Block diagram and data stream of the multi-channel neurostimulator with stimulus artifact
suppression (SAS). The hardware is programmed by the embedded system to generate highly configurable
constant current monopolar biphasic-pulse stimulation. Each block in the embedded system represents an
algorithm to control the hardware. It consists of a MIMO non-linear dynamical model simulator, using a
random number generator. The data is inputted to a multiplexer control unit to set the appropriate select
lines of the multiplexer. The multiplexer control unit is synchronized with the SPI switch control unit to
control CMOS switches.
4.3 Hardware Electronic Circuit Design
The stimulator consists of a configurable constant current biphasic and monopolar
waveform generator and a pattern generator independently specifying spatiotemporal timings and
magnitudes of pulses across 32 stimulating electrodes. First, a single-channel configurable current
source capable of providing charge-balanced biphasic pulses is designed and tested as follows.
The microcontroller is programmed to generate a 40 kHz pulse width modulator (PWM). An op-
amp integrator with a cut-off frequency of 4 kHz is designed and used to average the PWM to
output a negative DC voltage with minimal ripples, which is then inverted to output a positive DC
voltage using an inverting amplifier. Three other signals from the microcontroller are generated to
drive analog switches (TS5A22362) dictating polarity and duration of each pulse or the inter-pulse
28

intervals. An op-amp-based current source is designed to convert the output voltage biphasic
pulses to constant current biphasic pulses with a 1 µA resolution. The output of the current pulse
is then fed into a multiplexer to expand a single channel to 32 channels (Figure 4.2). A DC blocking
capacitor is placed at the input of the multiplexer to block input off-set. Furthermore, since the
application of the design requires sparse and low pulse rate, charge build-up is not expected as the
electrode is shorted to ground after each stimulation pulse.

Figure 4-2 Design of a 32-channel configurable constant current biphasic neurostimulator. It includes (A)
a microcontroller (MSP430) which generates a PWM at a 40 kHz frequency and an adjustable duty cycle
to configure the stimulus amplitude. The duty cycle is also automatically adjusted as the battery supply
voltage drops, governed by equation 1, avoiding the need for a regulator. PWM then goes through (B) an
integrator to generate a constant -DCV, (C) an inverter to generate a +DCV, (D) a set of analog switches
dictating polarity and duration of the pulse controlled by the microcontroller, (E) a voltage to current
converter and a DC blocking capacitor to generate safe single channel constant current biphasic pulses.
(F) Lastly, a multiplexer is used to expand the design to 32 channels.
This circuit is powered by two 3.7V coin batteries connected in series to obtain ±3.7V. The
absolute maximum supply rating is determined by the microcontroller which is +4.1V. Other chips
have a maximum voltage rating of ±5V. To minimize hardware design for future miniaturization
of the PCB, voltage regulators are eliminated since all chips can operate at a minimum voltage of
±3V. To ensure the output DC voltage from the PWM stays constant even with voltage supply
29

drop, the microcontroller is programmed to sample the supply voltage and adjust the PWM duty
cycle according to Equation 4.1:
𝑫 𝒏𝒆𝒘
= 𝑫 𝐛𝐚𝐬𝐞 ∗ (
𝟑 .𝟕 𝑽 𝐛𝐚𝐭𝐭𝐞𝐫𝐲 ), (4.1)
where Dnew is the adjusted duty cycle; Dbase is the target duty cycle for when the battery voltage is
3.7V, and Vbattery is the voltage of the battery at a time point.
4.4 System Bench-top Characterization
The microcontroller was programmed to generate random numbers dictating both the
timing and amplitude of pulses across each of the 32 channels independently. The range of
amplitudes (Imax) can be selected based on the user need and is limited to the supply voltage
(Vsupply) and the total impedance of the electrode-tissue interface (|z|electrode)
𝐼 𝑚𝑎𝑥
=
𝑉 supply
|𝑧 |
electrode
(4.2)
|z|electrode includes the electrolyte resistance plus the polarization impedance across the electrode-
electrolyte interface, which is frequency dependent. An electrochemical impedance spectroscopy
(EIS) of a chosen electrode can be used to determine |z|electrode at a frequency equal to the inverse
of the pulse duration. This frequency is a reasonable approximation for non-sinusoidal pulses.
4.5 Results
We have successfully designed, fabricated and tested a multiplexed 32-channel
microstimulator that can generate arbitrary spaciotemporal pattern of pulses driven by an external
source. System characterization including examination of asynchronous arbitrary pulse pattern
generation are presented below.
30

4.5.1 Asynchronous Arbitrary Pulse Pattern Generator
The EIS plot of an electrode sued to run bench-top experiments is presented in Figure 4.3.
Here, a pulse duration of 200 µs (5 kHz) corresponds to a |z|electrode of 50 kΩ. Thus, the stimuli
were applied across 50 kΩ resistors, mimicking |z|electrode, with amplitudes ranging from ±1 µA to
±60 µA, and the pulse intervals ranging from 0.5 ms to 100 ms. The system can be programmed
to generate cathodic or anodic first stimulation pulses across individual channels. This choice is
dependent on the region of the brain being stimulated as one waveform may manifest a lower
threshold than the other.

Figure 4-3 Representative electrochemical impedance spectroscopy plots of impedance magnitude in
1XPBS for three trials (E1-E3) of a Tungsten electrode used in all experiments.
The result of arbitrary stimulation pulse generator across 32 channels is shown in Figure
4.4 demonstrating the capability of the system to generate neural code-based stimulation with each
channel generating either anodic first or cathodic first pulse. Small positive spikes at the beginning
of the biphasic pulses are charge injection when switching from one channel of the mux to another.
This is a value of maximum 5 pC.
31

Figure 4-4 Spatiotemporal pattern of cathodic or anodic first stimulation pulses with varying amplitudes
and intervals generated by the 32-channel neurostimulator. The stimuli were applied across 50 kΩ resistor
mimicking |z| electrode. The amplitudes of the pulses range from ±1 µA to ±60 µA, and the pulse intervals
range from 0.5 ms to 100 ms.
4.5.2 Limitations
The limitation of this system is imposed by the multiplexer. Multiplexing prevents two or
more electrodes from being stimulated simultaneously. Instead, one stimulation pulse needs to be
delayed by the duration of the pulse. A typical pulse duration to evoke neural response is 100-200
µs in small animals such as rats (Hambrecht, 1995; McCreery et al., 1998). If the pulse duration
per phase is set to be 200 µs and two electrodes need to stimulate simultaneously, one pulse would
be delayed by 400 µs. This is in fact rarely needed due to the sparse nature of neural spiking
activities. Furthermore, our design specifications were based on hippocampal memory prosthesis
application, which uses weak stimulation pulses to activate small and localized population of
neurons. However, synchronized stimulation has also been shown effective in current steering and
focusing (Chaturvedi et al., 2012; McIntyre et al., 2015; Steigerwald et al., 2016). Our system is a
32

32n channel neurostimulator, where n is the number of individual current sources. Therefore, the
design may easily be expanded to larger channel counts with individual current sources to allow
synchronous stimulation through multiple channels.
4.5.3 Power Consumption
The total power dissipation is dependent on the load current and the quiescent current. The
output load current is dependent on the driving load (the electrode) and the stimulus waveform
(pulse amplitude, pulse duration, and pulse rate), which is a variable defined by the user. Quiescent
power consumption is the product of current drawn by the supply (Icc) and supply voltage (Vcc).
Here, the quiescent power dissipation from all active parts except the microcontroller and
multiplexer is 815 µW. The multiplexer consumes 60 µW. Thus, without multiplexing the system
power consumption would be 815µW*32=26mW, whereas multiplexing reduces this number to
875 µW. The microcontroller power consumption is dependent on usage of general-purpose input
output pins. To generate arbitrary pulse patterns in real-time through a single channel, 4 pins are
required by the pulse pattern generation control unit (Figure 4.1). Without multiplexing, the
number of required pins would be 32*4=128 for 32 channels. Whereas, with multiplexing, this
number would be reduced to 9. Thus, multiplexing greatly reduces power consumption and real-
estate usage.
4.5.4 System Cost
The cost to fabricate and assemble the PCB is approximately $100. The minimum
requirements are a personal computer, a TI MSP430 launchpad, and the Code Composer Studio to
upload the code.
33

4.6 Discussion
We have designed, fabricated, and tested a versatile and cost-efficient neurostimulator that
can deliver precise spatiotemporal patterns of stimulation pulses with arbitrary magnitudes and
intervals through 32 channels continuously and in real time. This stimulator can be controlled by
an external source such as a MIMO nonlinear dynamical model to achieve localized and patterned
micro-stimulation to the brain. We have systematically tested this stimulator in a phantom
preparation.
Highly configurable neuro-stimulators are highly desirable. For example, variations in
neural response may occur weeks or months after implantation due to inflammation. Inflammation
causes glial cell encapsulation around the electrodes and thus weakens the neuron-electrode
interaction, e.g., reduction of the stimulation effect and recorded signals (Vadim S. Polikov et al.,
2005). Furthermore, neural plasticity may also contribute to variations of neural responses as the
underlying neural circuits are constantly altered by behaviors (Kerr et al., 2011; Månsson et al.,
2016b). These variations can be compensated by adjusting stimulation parameters based on
behavior or the feedback signals provided by the recording electrodes.
Our neurostimulator is particularly suitable for building hippocampal memory prosthesis.
In a hippocampal memory prosthesis system, spatiotemporal patterns of stimulation to a
downstream brain region are calculated based on the ongoing spatiotemporal patterns of neural
activities in an upstream brain region using a predictive MIMO nonlinear dynamical model. The
stimulation patterns mimic the endogenous neural signals, which intrinsically are sparse,
asynchronous and involve multiple channels. The neurostimulator provides away for delivering
such patterns when connected to the output of a computational unit that contains the MIMO model.
34

A key feature of this design is the use of a multiplexer to save power and real estate to
handle large numbers of electrodes. Higher channel counts are achievable with simple hardware
and software modifications, with a complexity that scales sub-linearly with the channel counts. As
such, compared with commercially available neurostimulators such as Tucker Davis Technologies,
Ripple Neuro, and Black Microsystems our system is a low-cost (~$100) and reproducible design
for use as a neuroscience tool, which may be easily miniaturized for mounting on rat’s skull.

35

CHAPTER 5: STIMULATION IN CONJUNCTION WITH RECORDING
5.1 Introduction
The ability to stimulate and record from the same electrode would:
1. Maximize the number of electrodes for recording and stimulation
2. Provide feedback from stimulated tissue for validating stimulation effects and optimizing
stimulation parameters
3. Potentially enable safer, more precise and efficient treatment
4. Enable building single neuron-level MIMO model
There are two challenges with stimulation in conjunction with recording: 1. Electrode
design including material and sizing, which is described in detail in chapter 6 and 7, and 2.
Hardware electronic circuit design, descried in this chapter.
The challenge in realizing this feature is imposed by direct connection of the stimulator
and amplifier through a recording electrode. If a stimulation pulse causes amplifier saturation, the
input signal would be clipped at the amplifier’s maximum input range. Consequently, neural
response would be completely masked with this artifact and cannot be recovered. Stimulus artifact
may last for tens of milliseconds or even hundreds of milliseconds depending on the amplifier,
electrode property, tissue property, and stimulus parameters (Rolston et al., 2009a). Minimizing
such recording contaminants is vital for better recording of evoked neural activities and controlling
neural interface devices.
In this chapter, I have described the design of a stimulus artifact suppression technique
using switching mechanism designed to reduce recovery period from the artifact from tens to
hundreds of milliseconds to ~2 ms. I have designed, fabricated and characterized the system first
in phantom preparations and then the hippocampi of an anesthetized rat. Preliminary in vivo
recordings shown in this chapter demonstrate recovery of early onset compound potentials.
Recording of spike activity is also possible with this system based on results from phantom
36

recordings. However, since the same electrode is used for stimulation and recording, the surface
area of the electrode must be large for a charge storage capacity that allows delivering enough
charge to the tissue to evoke a response. Thus, here the geometric area of the electrode size limits
recording of spikes. I will explore the capability of recording single and multiple neuron activities
from the stimulation electrode in chapters 7.
5.2 Stimulus Artifact
Neural recording systems consist of small-signal (typically 10 µV-10 mV) voltage
amplifiers with adjustable gain and bandpass filters. Beyond power supply voltage after
amplification, the amplifier cannot produce amplification of the input signal. Not only does a large
signal causes signal distortion and loss of neural signal, the excess power transfer may damage the
amplifier over time. A stimulation pulse applied to an electrode is a large signal and when used in
conjunction with a neural recording system, it produces high amplitude artifacts with long recovery
period caused by amplifier saturation and filter ringing.
5.3 Stimulus Artifact Suppression Technique
To prevent blockade of neural data by stimulus artifact, CMOS switches (ADG714) are
used to block the stimulation current from transmitting to the recording system. Each switch is
connected between the electrode and the recording system and is synchronized with the stimulator
to be triggered a short time before and after the stimulation pulse. During this time, the switches
connecting the electrodes to the recording module (S2) will be kept open and the switches
connecting the stimulator to the same electrodes (S1) will be closed (Figure 5.1).
One challenge to using CMOS analog switches to block the stimulus from the recording
system is they contain parasitic components that affect the AC performance of the device (Figure
37

5.1b). This means that part of the stimulus may couple from the source to the drain of S2 during
stimulation. S2 acts to minimize charge coupling into the amplifier but since the amplifier is
typically set to a gain of at least 60 dB, even small voltages may generate long contaminated
signals.
To dissipate the charge coupled across the switch during stimulation, a 100 Ω resistor is
used at the input of the amplifier during stimulation to suppress ringing. This resistor is also used
to absorb any instantaneous charge coupled from the electrode to the amplifier when the electrode
is reconnected to the amplifier. The resistor is disconnected from the amplifier when recording is
resumed using another CMOS switch (S3) (Figure 5.1a). The timing of the switches depends on
the shape of the voltage transient waveform across the electrode in response to a given stimulation
pulse and is determined by the electrochemical properties of the interface as described in the next
section.
38

Figure 5-1 Stimulation artifact suppression set-up. (A) Stimulation is synchronized with a set of serially
controlled CMOS switches (S2) to block out the stimulus from the recording amplifier during stimulation
while connecting the amplifier input to a 100 Ω resistor via S3 to prevent ringing caused in the recording
system due to any extra charge coupled across S2. Charge coupling across S2 is due to (B) parasitic
components of analog switches. The PCB is connected to the recording amplifier and (C) the electrode
using a coaxial cable. The timing of the switches in (A) is determined by (C) the electrochemical properties
of the electrode, and (D) the voltage transient across a microelectrode (red) in response to a biphasic
cathodic first current pulse (blue). (C) RS represents the electrolyte resistance, RP represents faradaic
reaction, and Cdl represents capacitive reactions at the interface.
5.4 Electrochemical Properties Consideration
The shape of the voltage transient across a microelectrode is a factor of the electrochemical
processes at the interface which can be estimated with an electrical equivalent circuit model. The
model consists of an electrolyte resistance (Rs) in series with the parallel combination of a double
layer capacitance (Cdl) and an impedance of faradaic reactions (RP) (Figure 5.1c) (Conway, 1991).
39

Equation 5.1 describes the relationship between an applied constant current pulse (I) and the
resulting voltage (V) across the interface:
𝑉 (𝑡 ) = 𝐼 𝑅 s
+ 𝐼 𝑅 p
(1 − 𝑒 −
𝑡 𝑅 P
𝐶 dl
) (5.1)
An example of the voltage transient in response to a stimulation current pulse of 60 µA, 200 µs is
shown in Figure 5.1d.
In this model, capacitive charge injection represented by Cdl involves physical absorption
and desorption of ions in an electrolyte. Faradaic reaction represented by R P involves local
donation of electrons through oxidation and reduction reactions, which is less desirable than a
capacitive process since it involves formation of new species as discussed in chapter 3 (Stuart F.
Cogan, 2008a). Both charge injection mechanisms may be involved during stimulation
(irreversible faradaic reactions are to be avoided) (Merrill et al., 2005). Thus, the time constant of
the electrode is governed by:
𝜏 = 𝑅 p
∗ 𝐶 dl
(5.2)
After the termination of the stimulation pulse, the electrode is left with an initial
polarization voltage (Vp-i) and takes several τ’s to reach to a value close to its initial bias level (Vb)
(Figure 4). Vp-i is defined as:
𝑉 p_i
= 𝑉 b
+ (𝛥𝑉 − 𝑉 a
) (5.3)
where Va is the instantaneous voltage drop after the termination of the current pulse and ΔV is the
maximum voltage the electrode reaches at the end of the pulse (Bard AJ, 2001). Vb is typically a
few millivolts with respect to a large return electrode of the same material, as is the case of
electrophysiology experiments. It is often difficult to identify Va in a voltage transient plot, thus
40

Vp_i cannot be accurately determined. Alternatively, VP_max, the maximum polarization voltage
across an electrode to avoid potential exertion beyond the water window, may be used as Vp-i. This
is a worst-case scenario. This value is experimentally determined and is electrode material
dependent (Stuart F. Cogan, 2008b).

Figure 5-2 Polarization voltage (VP) across the electrode with respect to a reference of the same material
after termination of the stimulation pulse with an initial value of VP-i. The voltage must drop below VT -
before recording is resumed, to avoid amplifier saturation, which takes a time duration of multiple factors
of τ represented as xτ.
It is important that the polarization voltage on the working electrode (Vp) with respect to a
reference of the same material, drops to below a threshold before recording is resumed to avoid
saturation of the amplifier. This threshold is governed by the settings on the recording amplifier
such that:
V
p
<
𝑉 max−amp
Gain
= V
T
(5.4)
where Vmax-amp is the maximum output voltage from the recording amplifier, and Gain is the gain
of the amplifier.
After the termination of the stimulation pulse, the electrode will take several time constants
(xτ) to approach to below VT, which dictates the time duration the electrode must stay disconnected
from the amplifier. The timing and state of the switches are shown in Figure 5.3. 1)The stimulator
41

and the resistor at the input of the amplifier are disconnected while the electrode is connected to
the amplifier. This is the recording phase. 2) 200 µs before the initiation of the stimulation pulse,
S1 and S3 close. 3) S2 then opens to disconnect the recording system and the stimulation pulse is
applied to the electrode. 4) The switches stay in that state for the duration of the pulse plus a
predefined xτ. 5) S1 then disconnects the electrode from the stimulator and S2 reconnects the
electrode to the amplifier while the input of the amplifier is still connected to ground through the
100 Ω resistor to absorb any instantaneous charge injected from the electrode. 6) 200 µs later, the
resistor is then disconnected by S3 and recording is resumed.
It is important to note that the electrochemical properties of an electrode must be
characterized for different media to determine different time constants of the electrode-electrolyte.
Factors that affect the value of the time constant include electrode material, geometric and effective
area of exposed region, and the electrolyte impedance. Thus, an electrode should be separately
characterized for in vivo or in vitro experiments.

Figure 5-3 The state of the switches (1 = closed; 0 = open) shown in Figure 3a with respect to the stimulation
pulse. In all cases when the stimulator is not pulsing, it is connected to ground. States 1 to 6 are as follows:
1. Recording: 1. Recording: The electrode is connected to the recording amplifier while the stimulator and
the resistor are disconnected. 2. Grounding: 200 µs before stimulation, the stimulator is connected to the
42

electrode and the resistor is connected to the input of the amplifier. 3. Stimulation: The amplifier is
disconnected from the electrode. The stimulation pulse is then applied to the electrode. 4. Discharge period:
The electrode stays disconnected from the amplifier until the polarization voltage on the electrode falls
below V T. 5. Discharge of residual charge (200 µs): The electrode is disconnected from the stimulator and
connected to the recording system. The resistor remains connected to the input of the amplifier to absorb
any charge injection due to residual offset on the electrode. 6. Recording resumed: The termination resistor
is disconnected, and recording is resumed.
5.5 Methods
5.5.1 System Bench-Top Testing
A Tungsten electrode was used to initially evaluate the stimulus artifact suppression (SAS)
technique. The electrode was first characterized by its electrochemical properties, namely the
magnitude of each element in the equivalent circuit model shown in Figure 5.1c, its charge storage
capacity and its τ using Cyclic voltammetry (CV) and EIS performed by Gamry Reference 600
potentiostat (Gamry Instruments, Warminster, PA). The return electrode was also made of
Tungsten which was many times larger in area than the working electrode (WE).
Stimulation and recording experiments were performed in a phantom to compare the
artifact with and without the SAS component. The phantom was 1/6 diluted PBS mimicking brain
tissue impedance of approximately 0.25S/m (Kandadai et al., 2012). To mimic neural activity, a
known input of 1 kHz sinusoidal signal was applied to another electrode in the same solution. 1
kHz was chosen because it is within the spectral range of single unit activity. Also, this frequency
over the frequency of evoked potentials provides better visualization for earliest point at which
biological signals may be recovered from the stimulus artifact. The amplitude of the input sinusoid
is 10 mV peak to peak (Figure 5.4). The electric field from this electrode to the recording electrode
would be attenuated due to distance and electrolyte impedance. This is also the case for in vivo
recording as the source of an action potential is in millivolts and distance and tissue impedance
from the neuron to the recording electrode results in recordings of few hundred microvolts.
43

Figure 5-4 Test bench setup for evaluating the stimulus artifact with (Vout2) and without (Vout1) the SAS
technique. The setup consists of diluted PBS mimicking neural tissue and a sinusoidal signal (1 kHz, 10
mV peak to peak) applied to a neighboring electrode mimicking neural activity.
The time duration when no sinusoidal signal can be recorded due to the artifact is measured
and compared. The recording amplifier used (A-M systems, model 1700) was set to 60 dB gain
and a 10 Hz – 10 kHz band pass filter. The maximum output voltage of this amplifier is 10 V, thus
the voltage at the input of the amplifier must be less than 10 mV to avoid saturation. All signals
were digitized and recorded by a recording system (Digidata 1322A, Molecular Devices) and data
were saved by pClamps9 (Molecular Devices) software using a 100 kHz sampling frequency.
5.5.2 Preliminary Validation in vivo
To demonstrate the system’s functionality including the stimulator and the SAS technique
in vivo, electrode recordings were conducted in dorsal hippocampus of one male Sprague-Dawley
rat (12 weeks old, 350g) using our designed and fabricated PCB (Figure 5.5a). All procedures were
performed in accordance with protocols approved by the Institutional Animal Care and Use
Committee of the University of Southern California. The anesthetic induction was carried out in a
vaporizer-controlled induction chamber with a mixture of 4% isoflurane and O2. The rat was then
anesthetized with a mixture of Ketamine and Xylazine. Once the animal was deeply anesthetized,
44

it was placed on the surgery table. During the surgery, anesthesia was maintained with an
inhalation of isoflurane (1%~2% in pure oxygen) administered with a nose cone from isoflurane
machine with a scavenging cartridge attached. The status of anesthesia was checked periodically
(every 15 minutes) by pinching the toe of the hind paw. If a positive "toe pinch" response was
elicited, the doses of gaseous anesthesia would be increased. In addition, before, during and after
the surgical procedure, the respiratory rate, mucous membrane colour and body temperature of the
rat were monitored.
Ear bars on a stereotaxic frame were used to hold the rat’s head in place. Craniotomy of
2x4 mm was made over the right dorsal hippocampus and the dura was incised. The electrode was
inserted at 2.80 mm posterior to the bregma and 2.50 mm lateral to the midline at a depth of 2.65
mm, perpendicular to the brain surface. A micro-manipulator was used to insert the electrode
(Figure 5.5b). Two reference electrodes were inserted far away from the working electrode in the
hindbrain for the stimulator and recording system.
45

Figure 5-5 Acute in vivo setup for system evaluation. (A) PCB design of the system as shown in
Figure 6 connected to (B) an implanted microelectrode in rat CA1 region of the hippocampus. (C)
Stimulus artifacts and evoked neural responses are amplified, digitized, and saved with a bench-top
neural recording system.
Three sets of experiments were performed in vivo: (1) stimulate and record from the same
electrode without the proposed SAS to measure the artifact, (2) stimulate and record from the same
electrode with the SAS, and (3) repeat (1) and (2) after the animal is euthanized to separate neural
responses from the stimulus artifacts.
5.6 Results
We have successfully designed, fabricated and tested a stimulus artifact suppression
technique for recording from the same electrodes for feedback control of stimulation parameters.
System evaluation of stimulus artifact suppression in a phantom preparation are presented below.
46

5.6.1 Electrochemical Characterization
CV and EIS were performed to assess the electrochemical properties of the microelectrodes
used in the experiments. The measurements were analyzed with Echem Analyst (potentiostats and
electrochemical instrument software by Gamry Instruments) to generate the values listed in Table
5.1. The cathodic charge storage capacitance of the electrode measured at a scan rate of 100 mV/s
is 20 nC. The maximum charge applied to this electrode was 12 nC(60µA*200µs). This is below
the cathodic charge storage capacity of the electrode, which is an approximation of the charge
injection capacity(Stuart F. Cogan, 2008a).
The Vp on the electrode after termination of the maximum stimulation pulse of 60 µA, 200
µs is 0.5V (Figure 5.1d). After 5τ, Vp reaches to 0.5e-5=3.3mV with respect to a large Tungsten
return electrode. 5τ is chosen because the voltage across the electrode must reach to a value less
than the input range of the amplifier to avoid saturation, which here is 10 mV. The τ of this
electrode is 300 µs (Cdl*Rp), and 5τ corresponds to 1.5ms. Thus, a duration of 1.5 ms was spent
after the termination of the stimulation before the electrode is reconnected to the amplifier.
Table 5-1 Electrochemical parameters of the microelectrode used in the experiments.
Cdl RP RS Charge Storage
Capacity
Time constant (τ)
0.3nF 1MΩ 10KΩ 20nC 300µs

5.6.2 Verification in Phantom
Figure 5.6 demonstrates the resulting stimulus artifact in a phantom preparation with and
without the SAS component. The 10 mV sinusoidal input signal is attenuated and recorded at less
than 1 mV by the recording electrode. The input signal takes >60 ms to recover from the stimulus
artifact when the electrode is directly connected to the recording amplifier (Figure 5.6a). By
47

contrast, when the same signal is applied with the SAS component between the electrode and the
amplifier, the recovery period is reduced to 2.3 ms (Figure 5.6b). Based on the SNR of the recorded
signal, lower amplitudes may also be detected. A 15 ms time window of the two signals is shown
in Figure 5.6c and 5.6d with respect to the timing of the stimulation pulse for better visualization.
Figure 5.6c shows that the stimulation pulse saturates the recording amplifier, causes
ringing and finally settles according to the time constant of the electrode and the recording system.
Figure 5.6d illustrates the resulting signal with respect to the 6 states of S1-S3 shown in figure 5.3.
State 1 is the recording phase. The artifact seen in state 2 is due to switching artifact caused by
charge injection during closing of S1 and S2. State 3 shows the extra charge coupled from the
source to the drain of S2 during stimulation, which still saturates the amplifier. The extra charge
coupled is absorbed by the resistor connected at the input of the amplifier. The discharge period
of the electrode is represented in state 4. The artifact in state 5 is due to switching artifact and
possibly any residual charge left on the electrode when the electrode is reconnected to the
amplifier. In state 6, recording is resumed but the sinusoid takes some time to recover to the
baseline.
48

Figure 5-6 Stimulus artifact (A) without and (B) with the SAS in response to a current pulse (cathodic first,
60 µA, 200 µs). Recording is resumed more than (A) 60 ms as opposed to (B) 2.3 ms after the onset of
stimulation. (C) 15 ms time window from (A) demonstrating amplifier saturation and signal reflection. (D)
15ms time window from (B) demonstrating the results of states 1 to 6 described in Figure 5 and summarized
as follows: 1. Recording phase, 2. switching artifact (200 µs), 3. charge coupling across S2 during
stimulation (400 µs), 4. discharge period of the electrode (1.5 ms), 5. switching artifact plus residual charge
across the electrode (200 µs), 6. resume recording.
The control to compare recordings without the SAS technique is dependent on the amplifier
and its settings. For example, a lower gain and higher high pass cut-off frequency will result in
faster recovery period. Another factor affecting the recovery period is the input impedance of the
amplifier. A lower input impedance would mean that the charge from stimulation would be divided
between the electrode and the amplifier, which is undesirable. However, the settling time will be
49

shorter and vice versa. The SAS technique is designed to be used with a variety of amplifiers. It
minimizes charge coupling between the stimulator and the recording system and suppresses
ringing due to amplifier saturation while using a wide-band filter.
It is apparent that our system is not capable of stimulation and recording perfectly
simultaneously. However, it is important to note that neural tissue does not instantaneously
respond to the stimulation pulse. Instead, it requires a minimum amount of time termed latency to
generate the evoked response. The duration of latency depends on the properties of the target
excitable tissue (IRNICH, 1980).
5.6.3 System Validation in Vivo
We stimulated the CA1 region of the hippocampus in anesthetized rats using increasing
stimuli ranging in amplitude from 10 µA to 60 µA in increments of 10 µA separated in time by 1
second. Neural response was recorded with and without the SAS component. Each trial was
repeated 3 times with a 5-minute recovery period between trials.
Figure 5.7 shows neural responses following stimulation using the SAS component.
Evoked potentials are apparent at stimulus amplitudes of 10 µA, 20 µA, and 30 µA. The amplitudes
and durations of the evoked potentials increase with the stimulus amplitude. At and above 40 µA,
complex waveforms with increased magnitudes are observed, possibly because more neurons are
recruited, and more complex neural dynamic is elicited. The complex waveforms are
characteristics of population spikes in the hippocampus as there is an initial depolarization of the
nearby tissue followed by hyperpolarization deflections (Joëls and Fernhout, 1993).
50

Figure 5-7 Recording of evoked neural responses to increasing micro-stimulation amplitudes in the CA1
region of the hippocampus of an anesthetized rat using the SAS component. All pulses are constant-current,
cathodic-first and 200 µs in duration. Stimulus artifacts are shown in red. Evoked potentials are shown in
black. As stimulation amplitudes increase, the evoked potentials show increased amplitudes and more
complex waveforms.
Figure 5.8a shows results with the SAS component. Responses from 3 trials at 40 µA
stimulus are overlaid to demonstrate the repeatability and variations of the responses. Comparison
of the recorded signals with and without the SAS component is demonstrated in Figure 5.8b.
Notably, short latency neural response is obscured when the SAS component is not used. To verify
the recorded signals were indeed from neural tissue and clearly distinguish the artifact from neural
activity, the experiment was repeated after the animal was euthanized as a control. Overlaying the
signals recorded before and after euthanasia demonstrate clearly the effects of artifacts on the
stimulating electrode with and without usage of SAS (Figure 5.8 c, d).
51

Figure 5-8 Stimulus artifacts and evoked potentials to a stimulation with 40 µA amplitude. (A) Three trials
of recordings with the SAS. (B) Recordings with (red) and without (black) the SAS are overlaid for
comparison. (C) and (D) Verification of the evoked potentials by comparing the signals before (red) and
after (black) the rat was euthanized. Neural signals are retrieved within (C) ~2 ms after the stimulation onset
when SAS is used, and (D) ~60 ms when SAS is not used.
The frequency spectrum of the evoked potentials to stimuli of 10 µA, 20 µA, and 30 µA
contains frequency components between 7 Hz–30 Hz. Furthermore, at higher stimulus amplitudes
of 40 µA, 50 µA, and 60 µA, the neural responses also contain higher frequency components in
the range of 40 Hz-100 Hz associated with Gamma oscillations previously studied in the CA1
region of the hippocampus (Bragin et al., 1995; Lega et al., 2016) (Figure 5.9).
52

Figure 5-9 Frequency analysis of evoked potentials. 7-30 Hz band is apparent in all cases which increases
in magnitude with the stimulus amplitude. Evoked potentials to higher stimulus amplitudes (40 µA, 50 µA,
and 60 µA) also contain frequency components in the 40–100 Hz band. 60 Hz noise is shown in grey.
5.7 Discussion
The described SAS technique described in this chapter can be integrated with commercial
amplifiers. Using an array of CMOS switches, electrodes are disconnected from recording
amplifiers during stimulation, while the input of the recording system is shorted to ground through
another CMOS switch to suppress ringing in the recording system. The timing of the switches
used to block and suppress the stimulus artifact are crucial and are determined by the
electrochemical properties of the electrode.
The ability to recording neural responses after stimulation is critical for implementing
closed-loop neuromodulation or hippocampal memory prosthesis where consistent neural
responses are desirable but often difficult to maintain. The SAS technique allows stimulation and
recording from the same electrodes to monitor neural response to stimulation with short latencies
53

from the region of stimulation for achieving feedback control of neural stimulation. The in vivo
results showed that evoked potentials elicited by the neurostimulator can be recorded ~2ms after
the termination of stimulus pulses from the same electrodes where stimulation pulses are delivered,
whereas commercial amplifiers without such an artifact suppression typically result in tens to
hundreds of milliseconds recovery period.

54

CHAPTER 6: MICROELECTRODE DESIGN FOR SAME ELECTRODE
STIMULATION AND RECORDING
6.1 Introduction
Recording from single neurons and stimulation to same microelectrodes near
simultaneously is highly desirable for both basic neuroscience research and neural engineering
applications. In electrophysiological studies, same electrode recording and stimulation would
enable stimulus-response experiments at single neuron or small neuronal population level
(Houweling and Brecht, 2008; Krause et al., 2019; Shepherd et al., 2001). In DBS, such technique
would allow delicate micro-manipulation of complex neural circuits and monitoring feedback
neural signals with high spatial resolution (Little et al., 2013; Priori et al., 2013; Salam et al., 2016;
Brandon D. Swan et al., 2018; Vesper et al., 2002). In cortical prostheses, stimulating and
recording from the same single neurons becomes vital for successful implementation of the single
neuron-level, multi-input, multi-output model-based microstimulation (Deadwyler, 2018; Song et
al., 2018).
All of these require high spatial resolution, high signal-to-noise ratio, feedback signals
recorded from the stimulated tissue, and power efficiency, and electrode stability. For recording,
high spatial resolution and signal-to-noise ratio are necessary for differentiating single neuron
activities from background noise. For stimulation, high spatial resolution is essential for focal
delivery of electrical charge to the target neural tissue. Feedback control based on recording from
the stimulated tissue enables proper adjustment of stimulation parameters over time. Lastly, free-
roaming animal experiments and implantable neuromodulation devices both require low power
consumption and electrode stability for long-term use of the device.
55

These needs may be addressed with low-impedance microelectrodes that allow both
stimulation and recording. The main challenge of such electrodes is that the geometric area of a
recording electrode should be comparable to the size of a single neuron to record unitary activities,
but at the same time, stimulation electrodes require relatively large surface area to obtain low
electrochemical impedance that allows safe charge injection to evoke desired neural responses.
Paradoxically, reducing electrode size for high spatial resolution stimulation and recording would
result in an increase of the electrochemical impedance of the electrode-tissue interface (Stuart F
Cogan, 2008) and higher thermal noise (Suner et al., 2005). For stimulation electrodes, where the
same amount of charge must be delivered across a smaller interface, the increased impedance
results in increased electrode polarization, power consumption, and limits maximum
electrochemically reversible stimulation pulse (Merrill, Bikson and Jefferys, 2005). It is therefore
highly desirable to minimize electrochemical impedance of the electrode while keeping the
electrode area small enough for single neuron recording.
The impedance of electrode/electrolyte interfaces is generally modelled as a combination
of resistors and capacitance in parallel (Merrill, Bikson and Jefferys, 2005). The simplest being
the simplified Randles model, which consists of a resistor (representing the solution resistance) in
series with a resistor (representing the charge transfer resistance) and a capacitor (representing the
double layer capacitance) in parallel. Because impedance is proportional to resistance and
inversely proportional to capacitance, efforts to decrease interface impedance are based on
decreasing resistance and increasing capacitance.
Surface roughening is the most common method to increase the capacitance. By
roughening the surface, electrochemical surface area is increased while the geometric (or
macroscopic) surface area remains the same (Cogan, 2008). Platinum-black, a very friable coating,
56

was originally used for surface roughening, which provoked severe foreign body response
displacing neurons and disabling recording single neuron activities (Loeb et al., 1977). Other
coatings that increase real surface area include Titanium Nitride (TiN) (Weiland, Anderson and
Humayun, 2002), graphene oxide (Apollo et al., 2015), PEDOT (Boehler, Aqrawe and Asplund,
2019) and Carbon Nanotubes (CNT) (Wang et al., 2006).
To reduce resistance, either the electrolyte must be more conductive, or the charge transfer
resistance must decrease. In vivo, the electrolyte is the tissue resistance. Strategies to decrease
tissue resistance generally focus on reducing immune responses that can lead to fibrous
encapsulation of the electrode (Polikov, Tresco and Reichert, 2005). These include reducing the
size of the electrodes (Kozai et al., 2012), decreasing the mechanical mismatch between the
electrode and the tissue (Kim et al., 2013; Luan et al., 2017; Xu et al., 2018; Wang et al., 2020),
and incorporating bioactive molecules onto the surface of the electrode to attenuate the immune
response (Zhong and Bellamkonda, 2007).
To reduce the charge transfer resistance, a valence-shifting layer that can absorb and desorb
electrons and ions in a reversable manner can be incorporated into the electrode. Iridium oxide
(IrOx), due to its multiple oxide states (Robblee, Lefko and Brummer, 1983), is the most common
material with this property used in neural electrodes. Currently there are three approaches used to
make IrOx coated microelectrodes: activated iridium oxide film where a bulk iridium electrode is
oxidized by cycling it through positive and negative voltages in an aqueous solution (Beebe and
Rose, 1988), sputtered iridium oxide films, where an iridium target is used in the presence of
oxygen (Cogan, Plante and Ehrlich, 2004), and electrodeposited iridium oxide (Lu et al., 2009).
These techniques also roughen the surface, thus increasing the capacitance as well. Among those
approaches, electrodeposition has the unique advantage of being cost efficient as it does not require
57

a cleanroom and can be applied to biomedical electrodes made from almost any electrically
conductive material. Perhaps the biggest drawback to IrOx; however, is that it is a brittle material,
which can cause it to fail when stimulating with high charge densities (Cogan et al., 2004).
Electrodeposited Pt-Ir Coating (EPIC) is an electrodeposition process in which Pt and Ir
are co-deposited onto a conductive surface. EPIC maintains the advantages of electrodeposited
iridium oxide (increases surface area and lower charge transfer resistance), with the added benefit
of containing Pt, a less brittle metal than Ir, which likely contributes to the coating more robust
and less prone to delaminate compared to IrOx or PEDOT (Petrossians et al., 2011; Dalrymple et
al., 2019; Welle, 2020 ). Pt-Ir has the added advantage of having been used (in one form or another)
in FDA approved neuroelectronic devices for decades (Cogan, 2008), unlike TiN and PEDOT,
which have only been used for pacemaker applications (Schaldach, Hubmann and Hardt, 1989;
Boehler, Aqrawe and Asplund, 2019), and graphene oxide and CNT’s, which have not been used
in FDA approved devices. EPIC has demonstrated its ability to record single units through
microelectrodes (Cassar et al., 2019), as well as deliver charge through relatively large cochlear
electrodes both in vitro (Lee et al., 2018; Dalrymple et al., 2019) and in vivo (Dalrymple et al.,
2020). In this study, EPIC was evaluated for its ability to enable stimulation through electrodes
small enough to record single units. This evaluation involved in vitro and acute in vivo
electrochemical characterization including electrochemical impedance spectroscopy, cyclic
voltammetry, and polarization waveform analysis.
In this chapter, I will discuss the design and in vitro characterization of a microelectrode
array for same electrode stimulation and recording. The electrodes are originally designed for
single neuron recording and are treated with EPIC. The results demonstrate that EPIC increased
58

the electrodes’ charge storage capacitance by adding roughness and allowing reversable faradaic
reactions while maintaining their geometric area for single neuron recording.
6.2 Electrode Design
Commercially available, 8-channel micro electrode array (MEA) (Microprobes for Life
Science, Gaitherbsurg MD; platinum-iridium, 6mm length, 75µm diameter, 150µm interelectrode
spacing, ~500kΩ impedance) were used for this study. The device contained a 2 × 4 arrangement
of Pt-Ir microelectrodes, expanding into an area that covers 300 µm × 750 µm. The entire length
of each electrode was insulated with a layer of chemical vapor deposition of Parylene-C followed
by another layer of polyimide tubing around the base of the electrodes for additional stiffness. The
tip of each microelectrode was exposed by electropolishing Parylene-C to of approximately 10 µm
in length for this study (Figure 6.1b). At the base, the microelectrodes were mated to a 10 channel
Omnetics connector. The leads from the Omnetics connector were soldered onto a printed circuit
board with a surface mounted header which split the leads out for easy connection using hook
wires.
6.3 Electrode Preparation
Every other microelectrode was coated with EPIC (EPIC Medical Inc., Pasadena, CA)
using a process described previously (Petrossians et al., 2011). Due to cross talk within the
Omnetics connector, one unwanted electrode was coated. Therefore, the devices include 5 coated
and 3 uncoated microelectrodes (Figure 6.1a). Next the device was imaged with a scanning
electron microscope (SEM). Imaging was performed using a field emission SEM (Joel JSM-7001)
at 15 kV. SEM images of a coated and an uncoated electrode’s surface morphologies shown in
figures 6.1c, d, respectively, provided visual confirmation that the coating increased the effective
area while maintaining the geometric area.
59

Figure 6-1 a) Schematic showing relative locations of coated and uncoated contacts. This arrangement
allowed a side-by-side comparison of contacts with and without Pt-Ir coating. b) Optical micrograph of
Microprobes microelectrode array. SEM micrograph of c) Pt-Ir coated and, d) uncoated microelectrode
tips. c) The fractal nodules on the coated electrode dramatically increase the electrochemical surface area
while the geometric surface area remains similar to d) the uncoated electrode.
6.4 Safety Requirements
Each microelectrode underwent electrochemical characterization in Phosphate Buffer
Solution (PBS) including electrochemical impedance spectroscopy (EIS) (+/-10mV vs Ag|AgCl,
100 kHz – 0.5 Hz) and cyclic voltammetry (CV) (0.8 V to -0.6 V vs Ag|AgCl). Figure 6.2 displays
a representative voltammogram of a coated and uncoated electrode in room temperature PBS. The
cathodic capacity of the coated and uncoated electrodes was calculated from anodic to cathodic
sweeps (100 mV/s) of the cyclic voltammetry (Stuart F Cogan, 2008; Merrill et al., 2005). The
coated electrodes drew 50±3 nC (n = 10) and the uncoated electrodes drew 1.2±0.1 nC (n = 6).
Next, the geometric surface area of the electrodes was approximated by the SEM image in Figure
6.1 to be 4e-6 cm
2
. Dividing this area by the measured capacity provides the cathodic charge
storage capacitance (CSCc), which is defined as the total amount of charge available per geometric
surface area for an electrode (Stuart F Cogan, 2008). The calculated CSCc was 12.5+0.75 mC/cm2
60

and 0.3±0.025 mC/cm2 for the coated and uncoated electrodes, respectively. Thus, the coated
electrodes generated a significantly higher current than uncoated electrodes (two-sample t-test
p<0.001), likely due to lower impedance of the coated electrodes.
The artifact corresponding to peaks on the CV plot of the coated electrode in Figure 6.2 at
1.75 V (during anodic sweep) and -0.05 V (during cathodic sweep) is likely from silver
nanoparticle contamination during the coating process. This contamination may occur when some
silver nanoparticles from the reference electrode leaks into the plating solution and bond to the
electrodes surface. A detailed discussion regarding silver peaks in CV plots is discussed in (Van
der Horst et al., 2015).
Figure 6-2 Representative cyclic voltammograms for coated (black) and uncoated (red) electrodes in room
temperature PBS. Artifacts from silver leakage caused during the coating process are greyed out. The
calculated CSCC is an average of 0.3±0.025 mC/cm
2
for uncoated (n =6) and 12.5+0.75 mC/cm
2
for the
coated (n =10) electrodes.
Furthermore, voltage transient across the microelectrode in response to single biphasic
cathodic first current pulses were recorded across each microelectrode in 1/6 diluted PBS
mimicking impedance of brain tissue of approximately 0.25 S/m (Kandadai et al., 2012) to
determine the maximum electrochemically safe stimulation amplitudes given a fixed pulse
duration of 200 µs. A polarization voltage of >-700 mV was considered safe for the cathodic phase
and within water window (Stuart F Cogan, 2008). The uncoated electrodes reached this limit at ~5
61

µA; whereas the coated electrodes reached this limit at >50 µA. Table 6.1 organizes the pulse
parameters by duration, amplitude and total charge for coated and uncoated electrodes.
Test # Current (µA) Charge (nC)
1 1 0.2
2 2 0.4
3 3 0.6
4 4 0.8
5 5 1
6 10 2
7 20 4
8 30 6
9 40 8
10 50 10
Table 6-2 List of stimulation test pulse parameters (pulse amplitude and total charge delivered). The same
pulse duration (200 µs per phase) was used for all tests. Test numbers 1-5 were applied to both electrode
types (coated and uncoated). Test numbers 6-10 (in grey) were only tested on the coated electrodes because
the parameters exceeded safety limits when used on the uncoated electrodes.
6.5 Discussion
Our results show that the coated microelectrodes have on average 41× of the viable charge
for the same geometric surface area. EPIC coating increased CSCs by two mechanisms: 1)
increasing the effective area as shown in Figure 1, 2) involvement of faradic reactions from two Ir
states namely Ir
4+
+e
-
= Ir
3+
(Robblee, 1983).
However, since CSCc is obtained under low sweep rates and low current densities, it is not
an accurate measure of safe reversable charge injection capacities of the electrode during
stimulation with constant current biphasic pulses. To assess this parameter, we applied biphasic
pulses with a constant duration and increased the stimulation amplitude until the electrode
polarization crossed a water window of -700 mV (this is discussed in more details in the next
chapter when we measure the polarization voltage of the electrode in vivo).
62

What we found here was that the coated electrodes allowed 10× of the stimulation
amplitude compared to the uncoated electrodes in vitro. Subsequently, more charge per phase of
stimulation pulse may be applied to the microelectrode without causing irreversible reactions.
Widening the range of stimulation parameters is especially valuable in chronic applications where
adjustment, typically an increase, of stimulation parameters is needed over time due to changes in
neural circuitry as well as the electrode-tissue interface.

63

CHAPTER 7. PROPOSAL FOR IN VIVO VALIDATION
7.1 Design of Electrophysiology Experiments for in vivo Evaluation
7.1.1 Goal
In this chapter, we evaluate the in vivo performance of EPIC coatings on microelectrodes
prepared and characterized in Chapter 5. The arrays were implanted in the CA1 cell body layer of
rat hippocampi. Performance of coated electrodes was quantitatively compared with uncoated
electrodes within the same device and across different animals. The electrodes were tested for
signal-to-noise ratio, electrode polarization, power efficiency, charge injection capacity, and
capability to stimulate and record short latency and prolonged neural responses to various
electrochemically reversable stimulation parameters.
Here, we demonstrate stimulation and recording from the same electrode using
microelectrodes originally designed for single neuron recording. Our results demonstrated that
EPIC coating enabled safe stimulation above 5 µA, which enabled recordings of spontaneous
spikes, field excitatory post synaptic potentials (EPSPs), and population spikes (PS) from both the
stimulation electrode and neighboring recording electrodes. Results further showed that,
compared to uncoated electrodes, EPIC electrodes recorded neural signals with higher signal-to-
noise ratios, and generated lower voltages for given stimuli. Thus, EPIC provides a powerful tool
for monitoring and manipulating neural circuits at the single neuron level.
7.1.2 Surgical Procedure
Electrochemistry and electrophysiology experiments were conducted in dorsal hippocampi
of male Sprague-Dawley rats (n = 3, 350-450g, 3-4 months).
64

All procedures were performed in accordance with protocols approved by the Institutional
Animal Care and use Committee of the University of Southern California. The rat was pre-
anesthetized by an intraperitoneal injection of Ketamine and Xylazine cocktail. During the surgery,
anesthesia was maintained with an inhalation of isoflurane (1∼2% in pure oxygen) administered
through a nose cone from isoflurane machine. The status of anesthesia was checked frequently by
pinching the toe or footpad, and a heating pad was used to maintain and monitor the animal
temperature.
The animals were mounted onto the stereotaxic frame through ear bars. Craniotomy of 2
mm × 4 mm was performed over the right dorsal hippocampus. Dura and pia were removed before
the implantation. The electrodes were inserted at ~2.60 mm posterior to the bregma and ~2.45 mm
lateral to the midline, and it was angled ~30 degrees from the midline to match the septal-temporal
axis of the hippocampus. A micro-manipulator was employed to support and advance the electrode
2.5-3.8 mm from the surface of the cortex. A reference electrode was inserted far away from the
electrode array in the hindbrain in each experiment.
Neural signals were monitored as the electrodes were advanced into the brain for the
presence of complex spikes (a burst of 2-6 single spikes of decreasing amplitudes with <5 ms
interspike intervals (Ranck, 1973). Complex spikes serve as an electronic signature for pyramidal
neurons of the hippocampus, which help confirm placement of the electrodes in the CA1 region of
hippocampus. After the microelectrodes had reached the target location, data acquisition began.
Five sets of experiments were performed in vivo: (1) spontaneous activity recording, (2) in
vivo EIS measurement, (3) recording of voltage transient response to stimulation, (4) stimulation
and recording from the same and neighboring channels, (5) recordings from euthanized rat to
separate neural responses from artifact and noise.
65

7.1.3 Data Acquisition
All neural activities were digitized and recorded by a recording system (Digidata 1322A,
Molecular Devices) and saved by pClamp9 (Molecular devices) software using 100 kHz sampling
frequency. The recording amplifier was set to a gain of 80 dB and a filter of 300 Hz -10 kHz to
capture single unit activities. The output of the recording system was connected to a speaker to
allow for auditory discrimination between single and complex spikes activity. The recording
amplifier filter was then changed to a wideband filter of 1 Hz-10 kHz to capture single-unit as well
as multi-unit activity.
7.1.4 Signal to Noise Ratio Calculation
Activity from two microelectrodes (one coated and one uncoated electrode) were
simultaneously saved in one-minute long recordings for signal to noise ratio (SNR) analysis. The
SNR was defined as the power spectral density (PSD) (averaged from 1 Hz to 5 kHz to include
frequencies with associated power from multi and single units (Harrison, 2007) of the signal
recorded from anesthetized rat (PSDanesthetized) divided by the PSD of recordings made from the
same region after the rat was euthanized (PSDeuthanized) (Suarez-Perez et al., 2018). Using PSD to
calculate SNR eliminated the need for any assumptions about the amplitude of action potentials,
as is necessary when SNR is calculated using a chosen threshold to separate noise from neural
signal. PSD also allowed for a comparison of signal power from alive versus euthanized rats. The
signal recorded from the euthanized rat is purely noise from the electrode and the recording system.
A mixed linear model was used to determine the statistical significance of coating on SNR.
Independent t-test was applied to find out whether there is a significant difference between
uncoated and coated microelectrodes within and across animals. All results are presented as mean
+/- standard error (SE).
66

7.1.5 Electrochemical Characterization
Once spontaneous activity was recorded from all channels, the recording amplifiers were
disconnected from the microelectrodes. Each contact was then connected one at a time to Gamry
Reference 600 potentiostat (Gamry instruments, Warminster, PA, United States) to measure EIS.
The faraday cage surrounding the surgery table was used as ground, and the Pt-Ir wire implanted
in the hind brain as return electrode. The impedance as a function of frequency were plotted and
compared between coated and uncoated electrodes.
7.1.6 Voltage Transient Response
The electrodes were connected to the stimulator described in Chapter 4 one at a time using
a coaxial cable. Charge-balanced, cathodic first, biphasic single pulses were delivered to the
electrodes with each subsequent pulse having a larger total charge. A fixed pulse duration of 200
µs with no interface interval was used. Table 5.1 organizes the pulse parameters by duration,
amplitude and total charge for coated and uncoated electrodes. The voltage across the electrode in
response to each stimulation pulse was digitized at 1 MHz sampling frequency and recorded. A
duration of ~1-second ground phase was used between each pulse to allow for complete discharge
of the electrode before pulsing it with higher amplitude.
The maximum polarization in the cathodic phase across the electrode-tissue interface was
calculated. There are two factors in the transient voltage response: the ohmic voltage drop (Va)
arising from the ionic conductivity of the tissue (Rs) and the polarization across the electrode-
electrolyte interface (ΔEp), see chapter 3.4. Va and ΔEp may have some overlap, which introduces
uncertainty into ΔEp calculation. Another factor that contributes to this uncertainty arises from
limitation of current sources when loaded with high impedance such is the case with
microelectrodes. This limitation arises from an increased time constant at the output of the constant
67

current stimulator. The resultant voltage response to the applied squared current pulses is a
biphasic pulse with round corners, which makes clear ΔEp measurements with microelectrodes
more difficult.
To mitigate these challenges, we calculated ΔEp by 1) estimating Rs from EIS at >50 kHz,
2) recording voltage transient across an Rs equivalent 3) subtracting the waveform obtained from
an Rs equivalent from the electrode transient voltage waveform (Figure 7.1). All data are reported
for the cathodic phase of the pulse as stimulation pulses are cathodic first and negative ΔEp would
be larger than positive ΔEp.

Figure 7-1 Quantification of electrode polarization from current pulsing. (a) Simplified electrical equivalent
circuit model of the electrode-electrolyte interface. Electrode/electrolyte interface is modeled by a capacitor
(C dl) and a resistor in parallel (R p). The resistance of the electrolyte is modeled using a simple resistor (R s).
b) Voltage response of constant biphasic pulse (with current I) sent through a single small (5 kΩ) resistor.
There is a linear response that maintains the square wave of the biphasic pulse. c) Voltage response to
constant biphasic pulse sent through a single 50 kΩ resistor meant to approximate the Rs of tissue. Because
of the size of the resistor, the voltage response is no longer linear likely due to the large time constant at the
output of the stimulator. d) The voltage response to the 50 kΩ resistor superimposed onto the voltage
response of the coated (blue) and uncoated (red) electrodes. The polarization voltage for the uncoated
68

electrode (Ep 1) and coated electrode (Ep 2) is the subtraction between the access polarization estimated in
(c) and the voltage responses of the uncoated (red) and coated electrode (blue), respectively.
Next, power consumption associated with driving current pulses through stimulation
electrodes was computed using the equation below:
𝑃𝑜𝑤𝑒𝑟 = ∫ 𝐼𝑉
𝑇 0
(𝑡 )(𝑑𝑡 )
Where V(t) is the voltage measured from the polarization voltage curve; dt is the step size in time;
T is the pulse duration; and I is the applied current to the electrode tissue interface. Power
consumption associated with driving the coated electrodes was compared to power consumption
used by uncoated electrodes to determine if any significant power savings were gained by
application of the coating.
An independent t-test was used to determine statistical significance of electrode
polarization and power consumption between coated and uncoated microelectrodes. All results
presented as mean +/- standard error (SE).
7.1.7 Neural Response to Stimulation
Stimulation may cause prolong saturation of the recording amplifier which would mask
short latency neural response. In chapter 5, I reported a stimulus artifact suppression technique that
reduced the artifact down to ~2ms after the termination of the stimulation pulse from the stimulated
electrode. The designed stimulus artifact suppression technique (Elyahoodayan et al., 2019) was
used here to record short latency neural response to stimulation.
The two-channel recording system was connected to the stimulating electrode and another
electrode in the electrode array. With this arrangement, we could monitor the effect of stimulation
on the targeted tissue and neighboring channels. Stimulation and recording were conducted twice
using stimulation parameters summarized in table 6.11 before switching the second channel of the
69

recording system to another neighboring electrode. Thus, the stimulation electrode was pulsed 14
times using the same stimulation parameters. This experiment was repeated using an uncoated
electrode as the stimulation electrode. A duration of ~5-Seconds recovery period was used to allow
the tissue to return to base line before pulsing it with the next amplitude.
Directly evoked action potentials were recorded, and corresponding changes were observed
in the multi-unit and single unit band, including an increase in magnitude of short latency evoked
response and changes in spike rate associated with increasing stimulus amplitudes. Responses were
recovered within 2.5ms from the stimulating and neighboring electrode.
7.1.8 Spike sorting analysis
Data from microelectrode recordings from the stimulation electrode before and after each
stimulus were analyzed and activity of different neurons per microelectrode were identified by
using Plexon off-line sorter. Since neuronal firing rates show temporal variability following
stimulation, the time course of action potentials was shown as peri-event raster and peri-event
histogram for each stimulation. Peri-event raster and histograms were initially visually inspected
to identify firing patterns associated with stimulation. Successive trials were synchronized with
the stimulation artifact for EPSP responses.
7.2 Statistical Analysis to Demonstrate Functionality
7.2.1 Electrochemical Measurements
EIS data for the uncoated and coated microelectrodes in room temperature PBS to assess
baseline performance, as well as, in the CA1 region of hippocampus are shown in Figure 7.2 for
side-by side comparison. The data is demonstrated in bode-plot format (phase angle not shown) in
which the logarithm of the impedance is plotted as a function of the logarithm of frequency. At 1
70

kHz (center frequency of spike activity), the impedance for the electroplated microelectrode was
reduced by approximately 9.3× in vitro and 7.4× in vivo compared to that of the uncoated
electrodes.
At high frequencies (greater than approximately 500 kHz), impedance magnitudes showed
resistive behavior representing Rs. Rs is approximately 5±3 kΩ in vitro and 50±9 kΩ in vivo. This
is on average a 10× difference in Rs, causing an upward shift of the traces by a decade in the in
vivo plots compared to the in vitro plots. Furthermore, Rs is inversely proportional to the exposed
surface area of the electrode and the solution conductivity constant (Newman, 1965). Thus, the
variability observed between electrode impedances both in vitro and in vivo is due to the variability
in the electropolishing process to expose the electrode tip by the manufacturer. Another source of
variability in vivo only is inhomogeneity in tissue resistivity causing larger spread across traces in
comparison to the in vitro traces.

Figure 7-2 Bode plots of impedance magnitude |Z| vs. frequency recorded for Pt-Ir coated (black traces,
n=5) and uncoated (red traces, n=3) in vitro (left plot), and in rat hippocampus (right plot). 60 Hz noise
picked up by high impedance electrodes observed in vitro are greyed out. Impedance magnitude at 1 kHz
demonstrated an average 9.3× in vitro and 7.4× in vivo reduction in impedance magnitude. R s (highlighted
in yellow) approaches 5 kΩ±3 (n=8) in vitro and 50±9 kΩ (n = 8) in vivo.
7.2.2 Electrode Response to Stimulation Pulses
Figure 7.3a shows representative voltage transient response curves at the stimulation
electrode surface in response to cathodic first current pulses of 1 µA to 5 µA. After 200 µs the
71

applied current was reversed by an equal but opposite anodic pulse, resulting in voltage transient
in the positive direction. The voltage transient across the uncoated electrode resulted in a
polarization curves with masked Va (Rs and Ep segments were merged). Generally, the voltage
transient of the coated electrode showed smaller increase in voltage over the pulse interval, and
the shape of Ep was more linear and less parabolic.
Figure 5b,c show the calculated Ep and power consumption, respectively, plotted as a
function of pulse current amplitude. Data from 1 µA to 5 µA are shown as comparison between
the coated and uncoated electrodes. Data from 10 µA to 50 µA are shown for the coated electrodes
only. For Ep, a water window of -700 mV was chosen to avoid electrode potential exertion (Stuart
F Cogan, 2008). The uncoated electrode reached this window at 5 µA, whereas the coated electrode
allowed a stimulation current of 50 µA before reaching this limit.
Consistently, for all five test pulses used, the coated electrodes showed a significantly
lower Ep, and power consumption, as compared to the uncoated electrodes (p < 0.001 for all test
cases). In all scenarios tested, the coated electrode resulted an average 83% improvement in Ep
and 64% improvement in power consumption versus the uncoated electrodes. In chronic
stimulation applications, this could lead to having stable electrodes and significant power savings.
72

Figure 7-3 In vivo voltage transient of Pt-Ir coated vs. uncoated microelectrodes, in response to biphasic
current pulsing in rat hippocampus. a) Representative voltage response traces recorded from Pt-Ir coated
(black) and uncoated (red) microelectrodes, in response to biphasic current pulses (pulse duration of 200
µs, amplitudes of 1 µA to 5 µA). b) Average electrode polarization as a function of stimulus current pulses
across the coated (black) and uncoated (red) electrodes. The uncoated microelectrode surface potential
crosses the cathodic potential safety limit (U = -700 mV) at ~5 µA, whereas the coated electrode reaches
the same polarization in response to a 10× larger current pulse (i = ~50 µA). On average there is an 83%
reduction in electrode polarization for all stimuli. c) Calculated average power consumption associated with
driving the electrodes with biphasic stimulation pulses, plotted as a function of stimulus current magnitudes.
Data comparing the Pt-Ir coated (grey) and uncoated (red) power consumption are shown on the left. Higher
stimulation pulses applied to the coated electrode are shown on the right. Error bars indicate standard error
calculated for coated (n =15) and uncoated (n = 9) electrodes. There is a statistically significant
73

improvement in power consumption using the Pt-Ir coated electrodes (p < 0.001). On average there is 64%
reduction in power across all stimuli.
7.2.3 Signal to Noise Ratio of Spontaneous Neural Activity
1-minute sample plots of the signal recorded from an anesthetized rat overlaid with the
signal recorded from a euthanized rat (considered to be the baseline noise of the system) is shown
in Figure 7.4. The plot shows visual comparison of the signal and noise level of the coated and
uncoated electrodes. It is apparent that the uncoated electrodes manifested higher noise level than
the coated electrodes. A sample of complex spikes recorded from a coated and an uncoated
microelectrode is shown in Figure 7.4 inset to demonstrate proper placement of the electrode array
in the CA1 region of hippocampus. From the recording, single units were isolated at each time
point and the number of discernable units per electrode was quantified for each electrode within
the array (Figure 7.5).

Figure 7-4 Comparison of spontaneous recordings made from uncoated (top 3) and coated (bottom 5)
electrodes in the CA1 region of the hippocampus (all recordings are plotted on the same scale).
74

Superimposed grey traces are recordings made from the same electrodes after the animal has been
confirmed. Zoomed insets (right) are examples of complex spikes indicating proper placement of electrodes
in the rat CA1 region of hippocampus (euthanized recordings are omitted).

Figure 7-5 Example of acute single unit recordings from each electrode. Red, blue and green traces represent
different single units. White and light grey background indicate coated and uncoated electrodes
respectively.
A linear mixed model was used to determine the statistical significance of the coating and
the filter on PSD’s calculated from PSDanesthetized and, PSDeuthanized. Because the same electrodes
were used with different rats, the animal was included as a random effect. The linear model had
3 significant effects (Figure 7.6a). 1. As expected, PSDeuthanized was significantly lower than
PSDanesthetized (χ²(1)=67.95, p=2.2e-16), which validated our decision to use these recordings as our
baseline noise for SNR calculations. 2. For all frequencies, PSDanesthetized was significantly higher
in the coated electrodes compared to uncoated electrodes (χ²(1)=3.87, p=0.049). 3. PSDeuthanized
was significantly lower in the uncoated electrodes compared to coated electrodes (χ²(1)=14.35,
p=0.00015) (in contrast to the opposite relationship for spontaneous activity described in effect
number 2). This can be seen in Figure 7.6a in the PSD values especially near -40 dB for PSDeuthanized
made from coated electrodes. When comparing the PSDeuthanized versus the PSDanesthetized traces for
coated electrodes, there is a difference in magnitude greater than the standard error across all
75

frequencies (Figure 7.6a). In contrast, the difference between PSDeuthanized and PSDanesthetized for the
uncoated electrodes is smaller and approaches zero with increasing frequency.
SNR was defined as PSDanesthetized - PSDeuthanized. A linear model with the coating and filter
as fixed variables and animal as a random variable was fit to the SNR data (Figure 7.6b). The
results of the linear model showed a significant effect of the coating (χ²(1)=14.2, p<0.00016), with
coated microelectrodes having higher SNR (coated=9.09±1.53 dB, uncoated=1.90±0.50 dB,
Figure 7.6c). The filter used (1 Hz or 300 Hz) did not have a significant effect on PSD (χ²(1)=1.57,
p=0.21).

Figure 7-6 SNR of spontaneous activity. a) PSD from 1-5 kHz of spontaneous neural recordings in the CA1
region of rat hippocampus (PSD anesthetized) and activity after the animal was euthanized (PSD euthanized) using
either a 1 Hz HPF (left) or 300 Hz HPF (right). b) The SNR from 1-5 kHz for uncoated (red) and coated
(black) electrodes for recordings made with a 1 Hz HPF (left) or a 300 Hz HPF (right). SNR approaches 1
dB for frequencies above ~3 kHz for uncoated electrodes when using a 1 Hz HPF and above 4 kHz when
using a 300 Hz HPF. All results presented as mean ± SE for uncoated (n = 6) and coated (n = 10) electrodes.
c) Mean SNR calculated from neural recording data (300Hz-5kHz filtered) for uncoated (n=12) and coated
76

(n=20) electrodes. The effect of the coating was statistically significant in the mixed linear model
(p<0.0002), but the effect of the filter was not (p=0.81).
7.2.4 Short Latency Neural Response to Stimulation
Short latency extracellular evoked responses were obtained from the CA1 cell body layers
following stimulation. A total of 80 response curves were generated to monitor changes 2.5 ms
following the initiation of 10 separate stimulation pulses across a coated stimulation electrode and
recordings across all 8 electrodes. The results in Figure 7.7 can be separated into two categories
depending on whether only PSs were potentiated or PSs plus fEPSPs were potentiated. At low
amplitudes (1-5 µA), PSs are potentiated in the absence of fEPSP. At amplitudes above 10 µA,
potentiation of PSs was accompanied by potentiation of fEPSP.

Figure 7-7 1 Electrically evoked fEPSPs and PS recorded from all eight electrodes in response to biphasic
current pulsing through electrode # 2 in anesthetized rat CA1 region of hippocampus. “U” and “C” written
on the left side of the plot represent the uncoated and coated electrodes respectively followed by the
electrode number (electrode configuration as sketched in Figure 1a). The vertical lines to the left of each
response for C2 represent a time 2.5 ms after the initiation of the stimulation pulse (pulse amplitude labeled
above each line). Stimulus pulse amplitudes of 1 µA to 5 µA induced potentiation of PS without EPSP.
77

Pulse amplitudes of 10 µA to 50 µA cause potentiation of PS accompanied with EPSPs. An example of
stimulus artifact (red) followed by neural response (EPSP and PS) is presented on the right.
The amplitude of PS was measured as the difference in voltage between the nadir of the
PS trough and the mean in voltage between the fEPSP peaks on either side of the negative
deflection (Figure 7.8 inset) (Gholmieh et al., 2004). The input-output response curves were
generated using 1-50 µA stimulation amplitudes. Statistics were performed on raw values of PS
amplitude determined from average waveforms (n=4). When only PS was potentiated there was
almost no change in the recorded amplitude with increasing stimulation amplitude. In contrast,
when fEPSP plus PS were potentiated the calculated amplitude increased with the stimulus
amplitude and saturated at 40 µA (Figure 7.9).

Figure 7-8 1 Input-output curves of PS amplitude in the CA1 region of rat hippocampus (n = 4). PS
amplitude corresponding to the stimulation electrode (blue), and the neighboring coated (black) and
uncoated (red) electrodes are shown. Inset is a schematic diagram of a PS (negative trough) accompanied
by an fEPSP (positive hump). The PS amplitude is computed as the difference between the nadir of the PS
and the interpolated value of the fEPSP at that same time (as shown in red dash line).
7.2.5 Prolong Effect of stimulation
Figure 7.9 shows characteristics of neurons recorded from the stimulation electrode (coated
and uncoated) presenting each firing pattern 14 times before and after selected stimulation pulses
78

as a peri-event raster. From the raster plot corresponding to the coated electrode, we found two
type of responses to stimulation: excitation only and inhibition-excitation. Excitation only activity
demonstrates an increase in firing rate proceeding stimulation at amplitudes below 5 µA.
Inhibition-excitation happens at and above 10 µA, which triggers activation of interneurons
followed by excitation. A surprising finding is the long wave of inhibition in some trials (not all)
of up to 1.5 seconds, followed by excitation, which may have clinical and pathophysiological
implications not yet understood. From the uncoated electrode, excitation-only activity was
observed as the electrode was limited to low amplitude stimulation pulses for safety. In both cases,
there are some similarities and variabilities across different trials apparent within a stimulus in the
raster plot which may be resulting from stimulating a dynamic neurophysiological mechanism.
The neuronal responses to electrical stimulation were also classified based on peri-stimulus
time histogram (PSTH). The response patterns were clustered using 50 ms bin intervals and
averaged across 14 trials. Initial inhibition observed in some trials in the raster plot is masked in
the PSTH as there seems to be an increase in spikes rate in other trials. However, it is clear from
the PSTH that there is an increase in spike rate at around 3 second post stimulation observed from
the coated electrode.
79

Figure 7-9 1 Neuronal firing patterns before and after stimulation from the coated (left) and uncoated (right)
electrode recorded from the stimulation electrode in the CA1 region of rat hippocampus neuron. The red
vertical lines represent the time of stimulation with its corresponding magnitude written above it. The plots
consist of peri-event raster and its corresponding peri-event histogram below it. Each dot in the raster plot
represents the occurrence of a single action potential in the recorded neuron for 14 trials. The peri-histogram
represents spike counts accumulated per 5 ms bins.

80

7.3 Discussion
In this chapter, we quantified the in vivo performance of microelectrodes in an array
electroplated with Pt-Ir and further compared them with uncoated microelectrodes on the same
array. Results showed that coated electrodes exhibited superior performance compared with
uncoated electrodes in terms of SNR, power consumption, electrode polarization, and charge
storage capacitance. Quantitative analysis indicated substantial improvements of coated electrodes
over uncoated electrodes due to reduction of impedance.
Lower electrochemical impedance magnitude of a microelectrode improves recording
performance by reducing thermal noise and thereby increasing SNR. Here, we evaluated SNR by
recording from each electrode before and after the animal was euthanized. The signals recorded
after euthanasia were considered noise arising from the electrode-tissue interface and the recording
system. Since the same recording system was used to record from each electrode, the only variable
contributing to noise was the electrode. We then quantified SNR using power spectral density
analysis, which demonstrated statistically significant improvement. Results showed that lower
impedance of coated electrodes extended to above 1kHz, which is the frequency range of single-
unit and multi-unit activity and LFP’s that are biomarkers in closed-loop neuromodulation.
Minimizing electrode impedance is highly desirable in chronic neuro-stimulation
applications as our results suggested that coated electrodes exhibited higher power efficiency and
lower electrode polarization. Improved power efficiency and polarization voltage are due to the
fact that power and voltage are directly proportional to electrode impedance. Improved power
efficiency is essential in free-roaming battery-powered animal experiments and battery-operated
implantable neural modulation/prosthetic systems(Berger et al., 2012, 2005; Lo et al., 2017;
Miranda et al., 2015; Song and Berger, 2013; Zhou et al., 2019). Improved polarization voltage
81

results in long-term stability of the electrode because continuous high polarization at the
electrode/electrolyte interface can lead to dissolution (McHardy et al., 1980), corrosion (M.
Schuettler, 2002), and/or deformation of the electrode (Ordonez et al., 2015).
To assess the charge injection capacity of the electrodes, we applied biphasic pulses with
a constant duration and increased the stimulation amplitude until the electrode polarization crossed
a predefined water window of -700mV. What we observed with the uncoated electrodes was that
the voltage transient response exhibited a masked Rs response with visible asymmetry of the
biphasic pulses. Rs response is masked because of limitations of the current source being loaded
with high impedance, increasing the rise time at the output of the current source and causing round
edges in response to a square pulse. Furthermore, asymmetry is a result of a small double layer
capacitor in the electrical equivalent circuit of the electrode-electrolyte interface, which dominated
the transient response over the response due to Rs. On the other hand, the coated microelectrodes
exhibited a typical transient response recorded from macroelectrodes such as the ones shown in
Figure 3.4. Overall, the transient voltage from coated microelectrodes were lower with faster
discharge period due to symmetry in biphasic pulses.
To obtain a more accurate estimate of the polarization voltage, we applied the same
biphasic pulses across a resistor that mimicked Rs and subtracted the waveform from the transient
voltage response across electrodes. What we found was that the coated electrodes allowed 10× of
the stimulation amplitude compared to the uncoated electrodes. Subsequently, more charge per
phase of stimulation pulse may be applied to the microelectrode without causing irreversible
reactions. Widening the range of stimulation parameters is especially valuable in chronic
applications where adjustment, typically an increase, of stimulation parameters is needed over time
due to changes in neural circuitry as well as the electrode-tissue interface.
82

It is important to note that in general the electrochemical impedance magnitude of the
coated microelectrodes is reduced as result of increasing the effective area of the electrode and not
the geometric surface area. Therefore, the coated microelectrodes could inject a larger range of
reversable stimulation pulses to the tissue while maintaining the ability to record single unit
activity.
In conclusion, EPIC coating allowed us, for the first time, to use microelectrodes designed
for single unit recording as stimulation electrodes. We demonstrated this capability in immediate
and prolonged neural responses to stimulations by recording fEPSPs, PSs and spontaneous spikes
from the same and neighboring microelectrodes in response to varying stimulation parameters.
Thus, EPIC coated microelectrodes offer the capability of closed-loop neural stimulation to and
recording from the same microelectrodes and provides a powerful tool for monitoring and
manipulating neural circuits at the single neuron level.

83

CHAPTER 8: INTEGRATED MICROFLUIDIC CHANNELS
8.1 Introduction
Extraneural cuffs are among the least invasive peripheral nerve interfaces as they remain
outside the nerve. However, compared with more invasive interfaces, these electrodes may suffer
from lower selectivity and sensitivity since the targeted nerve fibers are more distanced from the
electrodes. Dr. Ellis Meng’s group developed a lyse-and-attract cuff electrode (LACE) by
microfabrication aiming to improve selectivity and sensitivity while maintaining a cuff format. Its
engineering design is described in (Cobo et al., 2019) . LACE is a hybrid cuff that integrates both
microelectrodes and microfluidic channels. The ultimate goal is to increase fascicular selectivity
and sensitivity by focal delivery via the microchannels of (1) lysing agent to remove connective
tissue separating electrodes from nerve fibers, and (2) neurotrophic factors to promote axonal
sprouting of the exposed nerve fibers into microfluidic channels where electrodes are embedded.
In this chapter, I focus on demonstrating in vivo function of microfluidics and
microelectrodes in acute preparations in which we evaluate the ability to focally remove
connective tissue and record and stimulate with microchannel-embedded microelectrodes neural
activity in rat sciatic nerves. Surgical implantation demonstrates preservation of LACE function
following careful and minimal handling. In vivo electrical evaluation demonstrates the ability of
microelectrodes placed within microfluidic channels to successfully stimulate and record
compound action potentials from rat sciatic nerve. Furthermore, collagen-rich epineurium was
focally removed following infusion of collagenase via microchannels and confirmed via
microscopy. Thus, the feasibility of using a cuff having integrated microelectrodes and
microfluidics to stimulate, record, and deliver drug to focally lyse away the epineurium layer was
demonstrated in acute experiments on rat sciatic nerve.
84

Figure 8-1 Peripheral nerve interfaces can be used for a) closed-loop control of prostheses or limbs after
spinal cord injury, and b) modulation of various organs.
8.2 Methods
8.2.1 Electrode Design and Fabrication
LACE (Fig. 8.2) is a flexible thin-film polymer device featuring metal electrodes for
recording and stimulation, microfluidic channels for drug delivery, and an adjustable locking
mechanism for precise fitting around the nerve. It is fabricated from Parylene C using surface
micromachining techniques on a temporary silicon wafer carrier. Four microfluidic channels are
spaced 350 µm apart on the circumference of the nerve, each with a 20 µm H  250 µm W cross-
section and 200 µm-diameter outlets. The electrode layout is shown in Figure 8.3. Eight metal
electrodes (300 µm  1500 µm) made from 200 nm-thick platinum are arranged as a bipolar pair
inside each microfluidic channel. One electrode of each pair is directly beneath the microfluidic
outlet while the second is approximately 4 mm away from the outlet. A ninth electrode of the same
dimensions is located outside the channels (the surface electrode). The locking mechanism consists
85

of a tab which is led by suture through a slit which catch on a set of ratchet-like teeth. The series
of locking teeth allow for adjustable sizing to fit nerves of 1.1 to 1.5 mm diameter in 0.1 mm
increments.

Figure 8-2 Fabricated Parylene C cuff electrode with integrated microfluidic channels. A red photoresist
layer is shown in the channels for visibility.

Figure 8-3 Magnified view of the microfluidic outlets and the electrode arrangement (left) and illustrated
cross section of one channel and its electrodes (right). Electrodes E1, E3, E5, and E7 are partially occluded
by the microfluidic channels with the outlets directly above the electrodes. Electrodes E2, E4, E6, and E8
86

are fully occluded by the channel (not used in experiments described in this paper). The surface electrode
(ES) is fully exposed.
8.2.2 Surgical Technique
Surgery was performed on male Wistar rats (>9 weeks old, 300-400 g). All animal
experiments were approved by the Institutional Animal Care and Use Committee (IACUC) and
the Department of Animal Resources of the University of Sothern California (DAR, USC).
Animals were anesthetized with 0.2 mL/100 g of ketamine and xylazine. Anesthesia was
maintained throughout the surgical procedure by inhalation of a mixture of oxygen and isoflurane.
A midsagittal incision was made at the right thigh. Junction between muscles was identified
and pulled apart by blunt dissection. The sciatic nerve was exposed; blood was rinsed off using 1 
phosphate-buffered saline (PBS). LACE was then implanted around the right sciatic nerve with a
4.0 suture needle. The suture was knotted at one end and the needle was threaded through a hole
in the electrode tab (Fig. 8.4a). After LACE was guided under and then around the nerve, the suture
needle was passed through the first slit and locked in place by pulling on the suture at one end and
the FFC cable connected to the electrode at the opposite end (Fig. 8.4b). This technique allows for
easy implantation with minimal to no damage to the electrode or tissue.

Figure 8-4 a) LACE threaded underneath sciatic nerve. b) LACE locked around the sciatic nerve.
87

8.2.3 Recording Set-Up
The connection scheme for the electrophysiological recording setup is illustrated in Figure
8.5a. Evoked compound action potentials (CAP), a summation of action potentials generated by
multiple fibers, were recorded differentially between the surface electrode and E7. A large
platinum wire was placed on the exposed muscle of the animal near the incision as ground. The
recording amplifier (A-M systems, model 1700) was set to 80 dB gain with a 10 Hz – 10 kHz band
pass filter. All signals were digitized and recorded with a recording system (Digidata 1322A,
Molecular Devices) and data were saved with pClamps9 software (Molecular Devices) using a
100 kHz sampling frequency.
Bipolar needle electrodes were inserted into the nerve 20 mm from the recording site and
were connected to an external stimulator (Multichannel Systems STG4000). Biphasic, cathodic-
first pulses with a fixed duration of 200 µs (Stuart F. Cogan, 2008c), and amplitudes of 100 – 700
µA in steps of 50 µA were delivered.
After normal CAPs were recorded, a negative control experiment was performed to verify
the neural nature of the recorded signals. Lidocaine was applied to the sciatic nerve distal from the
stimulation site. Lidocaine is a local anesthetic that prevents the generation and conduction of
action potentials in a nerve by binding to and blocking fast inactivation voltage-gated Na+
channels (Vedantham and Cannon, 1999). Therefore, application of lidocaine would abolish neural
activity to demonstrate that the recorded data was indeed neural activity, not noise or artifact from
stimulation.
The nerve was then stimulated while the leg muscle was observed for twitching. Twenty
minutes after application of lidocaine, no muscle twitch in response to nerve stimulation was
observed. Nerve recordings were then performed with the same stimulation paradigm using LACE
88

(Figure 8.5b). Subsequently, while nerve conduction was still blocked by lidocaine, the leg muscle
was stimulated directly using needle electrodes, producing a twitch. While directly stimulating the
muscle, recording from the nerve was performed to verify whether EMG activity contaminated the
previous CAP recordings (Figure 8.5c).
8.2.4 Stimulation Set-Up
Figure 8.7a illustrates the stimulation set-up. The rat sciatic nerve is composed of tibial and
peroneal divisions which innervate the thigh, leg and foot muscle. For stimulation testing, the
stimulus was applied to E5 as the working electrode and surface electrode as the return. Motor
responses were monitored from compound muscle potentials of the extensor digitorum longus
muscle recorded by bipolar needle electrodes. The bipolar recording electrodes were positioned to
record differentially along the longitudinal direction of the muscle fibers. Again, a negative control
was performed by applying lidocaine to the nerve and recording responses from the same muscle
(Figure 8.7b).
8.2.5 Drug Delivery
Before implantation, the microfluidic channels were primed with collagenase as described
previously (Cobo et al., 2019). Once LACE was implanted, collagenase was infused at 0.1 µL/min
for 10 minutes. This flow rate was previously determined by benchtop experiments with dye on a
nerve phantom, which resulted in localized delivery with no diffusion (Cobo et al., 2019). The
reaction time of collagenase to lyse the epineurium is approximately one hour (Rydevik et al.,
1985). Thus, one hour after delivery, the nerve was excised, and the rat was euthanized with
compressed carbon dioxide gas. The tissue was fixed by immersion in 10% formalin solution for
an hour in 4
o
C to preserve morphology. It was then transferred to 30% sucrose solution to
dehydrate it overnight. Two-photon (2P) second harmonic generation (SHG) microscopy was used
89

to examine the nerve structure. SHG does not require labeling with molecular probes and is thus a
non-destructive technique to image tissue anatomy (Vijayaraghavan et al., 2014). Collagen is a
noncentrosymmetric molecule emitting exactly half of the wavelength of the excitation wavelength
(Campagnola et al., 2002; Zoumi et al., 2002) and provides sufficient signal for SHG imaging.
Since the peripheral nerve is rich in collagen, SHG is a suitable technique to image collagen fibrils
at high resolution (Sinclair et al., 2012).
The tissue was whole mounted in 1% clear agarose gel. SHG imaging was carried out on a
commercial scanning Zeiss LSM510NLO two-photon microscope. The excitation wavelength was
set to 900 nm and SHG signal was collected using a 420-460 nm bandpass filter. All images were
acquired with a 25x/1.05 numerical aperture water immersion lens. The nerve was tile-scanned
with a field of view of 600x600 µm. The axial and lateral resolution was set to 10 µm.
After 2P-SHG imaging, agarose was removed, and the tissue was mounted in optimal
cutting temperature (OCT) compound and frozen at -80
o
C. It was then sectioned at 30 µm on a
cryostat and treated by H&E staining to observe the quality and location of lysing activity. Stained
samples were imaged using a DM2500 microscope and a DFC450 digital camera (Leica
Microsystem, Cambridge, UK). The lysed regions were measured by Leica Application Suite
Version 4.
8.3 Results
8.3.1 Fabrication Results
The electrodes were fabricated, packaged, and characterized as described in (Cobo et al.,
2019). Benchtop tests verified function of the microfluidic channels and locking mechanism.
Cyclic voltammetry (CV) was performed in 0.05 M H2SO4 to clean the electrodes of any residues
90

that may remain after fabrication. CV and electrochemical impedance spectroscopy (EIS) were
then performed in phosphate buffered saline (PBS) to assess stimulation characteristics. From
integration of the CV with a scan rate of 250 mV/s, cathodal charge storage capacity of the
platinum electrodes was 1.1 ± 0.1 mC/cm2 (mean ± SE, n = 4) and impedance at 1 kHz was 1.8 ±
0.11kΩ (n = 6). Devices were disinfected by soaking for 15 minutes in isopropanol.
8.3.2 Acute Recording Test
Evoked CAP responses were recorded from rat sciatic nerves using LACE (n = 3). Figure
8.5d shows a representative example of overlaid CAP responses elicited by stimuli with varying
amplitudes. The CAP responses increase in magnitude with the stimulus magnitude, indicating
that stronger stimuli recruit more nerve fibers which constructively add to produce a larger CAP
response at the recording site. Since the recording is bipolar, there are two phases to the CAP
activity: a positive followed by a negative phase. This arises from the connection scheme, having
the recording electrode proximal to the stimulation electrode connected to the non-inverting input
of the differential amplifier and the distal electrode connected to the inverting input. The temporal
dispersion of the CAP waveform is due to variability in conduction velocity across nerve fibers
(D. H. Liang et al., 1991; Maynard et al., 1997). Larger myelinated axons have faster conduction
velocity due to smaller axial resistance. Also, large diameter axons have lower threshold to
extracellular current due to smaller axial resistance, allowing more efficient passage of
longitudinal extracellular current (Hodgkin Alan Lloyd et al., 1946; Huxley and Stämpeli, 1949).
Figure 8.5e verifies the neural nature of the recorded signals. After application of lidocaine,
elimination of responses following stimulation confirmed that the recorded signals shown in
Figure 8.5a were indeed neural. Furthermore, the leg muscle was stimulated with a needle
91

electrode to evoke EMG activity. Figure 8.5f demonstrates that, after application of lidocaine to
the nerve proximal from the neuromuscular junction, only stimulation artifacts were recorded by

Figure 8-5 Acute in vivo recording setup is shown on the right of each panel, and the corresponding
recordings from LACE are shown on the left. a) LACE was used to record neural activities from the nerve.
E7 was used as the working electrode and SE was used as the reference electrode. Stimulation pulses were
applied to the nerve 20 mm from the recording site using needle electrodes. The stimulation pulses were
charge-balanced anodic-first biphasic pulses with a fixed duration of 200 µs, and pulse amplitudes ranging
from 100 µA to 700 µA in steps of 50 µA. The recording amplifier was set to a gain of 80 dB, with 10 Hz-
10 kHz band pass filter. b) Lidocaine was applied to the nerve distal from the stimulation electrodes and (a)
was repeated. c) Lidocaine was again applied to the nerve, the leg muscle was stimulated at 700 µA, and
activity was recorded from the nerve using LACE. d) CAPs elicited with stimulations of varying amplitude
were recorded using LACE. e) and f) are verification of recorded nerve response. e) Application of lidocaine
92

to the nerve abolished neural responses while stimulation artifacts were unchanged. f) EMG artifacts are
absent during direct muscle stimulation.
LACE, while no neural response was observed. This result further verifies that the signals shown
in Figure 8.5a were not EMG or contaminated by EMG.
Each neural response to its respective stimulus was rectified and integrated. The mean over
three trials for each response was computed and the obtained values were normalized by the
maxima and plotted versus the stimulus magnitude (Figure 8.6a). Here, the threshold stimulation
current for eliciting CAP was 100 µA. The plot demonstrates a steep slope for stimuli below 200
µA and a very shallow slope for stimuli greater than 200 µA suggesting gradual saturation of
number of recruited fibers above 200 µA.
Figure 8.6b illustrates the latency-to-onset of CAP versus the stimulation amplitude. Since
large diameter fibers are recruited first and have faster conduction velocity than small diameter
fibers, this plot is relatively flat. At a distance of 20 mm between stimulation and recording
electrodes, the latencies provide an approximation of conduction velocities of 16-21 m/s.
Conduction velocities are consistent with values previously reported in rat sciatic nerve
experiments (Xue et al., 2015; Yu et al., 2014).

93

Figure 8-6 Recruitment curve of sciatic nerve fibers. Error bars are standard error mean values from three
rats. a) Rectified and integrated values of each CAP activity following stimulation. The integrated CAPs
are normalized by the maxima and plotted versus the stimulus intensity. b) Latency to onset versus the
stimulus intensity.
8.3.3 Acute Stimulation Test
The stimulation tests after acute implantation verified that LACE can deliver enough
charge to the nerve to cause muscle contractions. Figure 8.7c shows an example of recordings from
extensor digitorum longus muscle with increasing stimulation intensity. The recording is from
multiple motor unit action potentials, adding up temporally and spatially to generate complex
waveforms. The maximum stimulus applied across the electrode is 0.04 µC. This is well below the
94

charge storage capacitance of LACE at ~4.5 µC, ensuring delivery of electrochemically reversable
stimulation pulses.
As before, Figure 8.7d is a verification of recorded neural activities. After application of
lidocaine, the absence of any activity following stimulation confirms that the recorded signals are
biological.

Figure 8-7 Acute in vivo stimulation setup is shown on the right panel, and corresponding recorded evoked
EMG shown on the left panel. a) E5 was used as the working stimulation electrode and SE was used as the
reference. Anodic-first stimulation pulses set to a fixed pulse duration of 200 µs and amplitudes ranging
from 40 µA to 200 µA were applied. EMG was recorded bipolarly using needle electrodes in the extensor
digitorum longus muscle in the rat’s right hind leg. The recording amplifier was set to a gain of 60 dB with
10 Hz-10 kHz band pass filter. b) Lidocaine was applied proximal from the neuromuscular junction and (a)
was repeated. c) EMG activity recorded following nerve stimulation. d) Verification of the recorded muscle
95

response. Application of lidocaine to the nerve abolished EMG response while the stimulus artifact was
unchanged.
8.3.4 Drug Delivery
A rendering of LACE installed on a nerve is illustrated in Figure 8.8a. Histological results
demonstrated effects of collagenase treatment on the morphology of sciatic nerves. Tissue closest
to the microfluidic channel openings demonstrated obvious disruption of epineurium (Figure 8.8b)
whereas tissue further away from the microfluidic channel openings was unaffected (Figure 8.8C).
The lysed region was ~300 µm of the nerve’s perimeter, which is comparable to the diameter of
the microfluidic channel outlet (200 µm). This result suggests that the chosen flow rate and
duration produced a focal distribution of collagenase effects on the surface of the nerve with ~50
µm diffusion beyond the edge of the outlet.
Tile-scanned SHG images were acquired using a 900 nm excitation wavelength. The
acquired scans were then z-stacked in Imaris image analysis software to construct a 3D rendering
(Figure 8.9a). The green regions indicate emission from collagen fibers, while diminished signal
is representative of collagen absence. A dark circular region with a diameter of 290 nm, enlarged
in Figure 8.9b, showed a clear, focal removal of collagen. Again, this shape and dimension is
comparable to the outlet of the LACE microfluidic channel (Figure 8.9c) and agrees with the lysed
region found in transverse histological sections.
96

Figure 8-8 Demonstration of the effect of lysing agent on exterior connective tissue of the nerve. a)
Rendering of the LACE installed on the nerve. b) Histology slides after delivery of collagenase show that
it effectively disrupted the epineurium near the microchannel outlet. c) Away from the outlets, the
epineurium was intact.

97

Figure 8-9 a) 2P-SHG images of whole mount nerve tissue with excitation wavelength of 900 nm. b) A
circular dark region, magnified, indicates absence of collagen fibers. Comparison of the dimensions of the
dark region (290 µm diameter) with c) the microchannel outlet (200 µm diameter) demonstrates focal
delivery.

98

8.4 Discussion
In this chapter, we developed the surgical technique to implant LACE in rat sciatic nerves
with minimal handling of the device and the tissue. The adjustable locking mechanism allowed for
simple adjustment of the LACE around nerves of various diameters. After implantation, device
integrity and function were maintained. This enabled demonstration of LACE’s capability to
record neural activity from rat sciatic nerves, and to deliver charge to stimulate the nerve and evoke
muscle response. The recorded signals were confirmed to have neural origin by comparing evoked
potentials before and after application of lidocaine to the nerve. Leg muscle innervated by the
implanted nerve was also stimulated to confirm absence of EMG artifact in the recorded nerve
signals. The bipolar recording configuration, using an electrode partially inside the microfluidic
channel and a surface electrode, obtained neural activities in response to increasing stimulation
pulses similar to recordings previously reported using the same set-up (Xue et al., 2015; Yu et al.,
2014). Furthermore, stimulation of the nerve through LACE was successfully demonstrated in
muscle activity recorded with needle electrodes in the rat hind leg.
Although cuff electrodes have been used successfully in animals and humans (Graczyk et
al., 2016; Guiraud et al., 2016), there has been limited success in recording nerve fibers with high
sensitivity and stimulating them with high selectivity. Some peripheral interfaces achieve higher
acute performance by sacrificing the long-term health of the nerve. Microscale drug delivery, as
demonstrated using the LACE, may offer an alternative strategy. We demonstrated the capability
of LACE to deliver controlled boluses of drug focally to biochemically manipulate the electrode-
tissue interface. Histological analysis and 2P-SHG imaging suggest that collagenase lysed the
epineurium layer focally with minimal diffusion from the microfluidic outlet. This is in
99

comparison to previous work, where an uncontrolled large bolus of 1ml of collagenase was applied
around the tibial nerve to study the effect of collagenase on nerve tissue.
Following the focal lysing of epineurium which has now been demonstrated, LACE
provides an ideal setup to potentially take advantage of the phenomenon of collateral sprouting of
nerve fibers. Application of neurotrophic factors, such as NGF and methylcobalamin, has been
shown to enhance collateral sprouting from an intact nerve (McCallister et al., 2001). Collateral
sprouting of nerve fibers following microsurgical removal of the epineurium layer from the nerve
has been demonstrated in both sensory and motor fibers (Liu et al., 2014; Šámal et al., 2006). In
contrast to these microsurgical techniques, LACE has the advantageous capability of
enzymatically removing the connective tissue, thereby reducing the risk of surgical nerve injury
(Rydevik et al., 1985). The long-term response of the nerve to collagenase and the ability to induce
axonal sprouting remain to be investigated.
The main objective of this study was to test LACE and its feasibility to perform acute in
vivo stimulation, recording and drug delivery on rat sciatic nerve. This study is the first
demonstration of the LACE’s potential to offer a new, drug delivery-based extraneural approach
to peripheral nerve interfacing for developing hybrid bionic systems and bioelectronic medicines.
Future work beyond this thesis includes design of packaging for chronic LACE
implantation including electrical and drug delivery components. It also includes performing
functional testing of LACE in chronic implantation conditions including nerve health and stability
of the interface after delivery of collagenase to elucidate possible level of axonal inflammation
and fibrosis. It further includes evaluation of electrodes for cross-talk and impedance spectra to
quantify leakage current, electrode performance, and wiring integrity for chronic
electrophysiological recording and electrical stimulation experiments. We expect that these results
100

will be one step closer to offering a solution for selective nerve stimulation and recording for
biomedicine application.

101

CHAPTER 9: SYSTEM MINIATURIZATION
9.1 System Description
Thus far, I have described the system design and an in vivo evaluation of a bidirectional
neurostimulator and recording system including hardware design considerations, electrode design
and practical application for acute neuroscience research. The in vivo experiments have required
large rack mounted amplifiers and data acquisition systems with long wires connecting the
electrodes to remote amplifiers. For chronic experiments, long wires would restrict free-roaming
behavioral experiments and add extra noise to the recordings.
To facilitate more sophisticated investigations into neural activity controlled by behavior,
I have discussed the design and implementation of a small, lightweight stimulation and recording
system to permit free behaving electrophysiology experiments. The design includes an ASIC
multi-channel neuro-stimulation and recording chip from Intan Technologies coupled with a
wireless microcontroller for data telemetry. A mechanical design to house the PCB and mount it
on rat’s skull is also described. The system will be powered by two batteries carried on the rat’s
back as a backpack and short wires will be routed from the back of the animal to its skull where
the PCB connected to electrodes resides (Figure 9.1). Design fabrication and in vivo evaluation
will be performed in future studies and is not included in this thesis work.
102

Figure 9-1 Illustration of a miniaturized multi-channel neurostimulation and recording module with
wireless data telemetry housed and mounted on the animal’s skull. The power source is two-coin batteries
mounted on the animal’s back and routed to the PCB.
9.2 PCB Prototype
An ASIC chip (RHS2116) developed by Intan technologies capable of stimulation and
recording through 16 independent channels is utilized. The chip is a 7 mm x 7 mm 44-pin QFN
package. This chip is programmed by a wireless microcontroller (ESP32) with a Wi-Fi module.
The microcontroller was programmed using Arduino integrated development environment (IDE)
to configure stimulation parameters and settings on the recording amplifier. Other components
used include a flash memory (8Mb) to store the programmed data, and a USB to serial UART
interface (FT231x) (Figure 9.2). The PCB also includes an on-board charger to recharge the
batteries. During this time and when the PCB is not being used in experiments, a switch will
disconnect the batteries from the main components.
103

Figure 9-2 PCB layout (designed in Altium Designer) integrating a 16 channel neurostimulation and
recording module chip (Intan) with a wireless microcontroller (ESP32) for data telemetry. The PCB also
includes an antenna chip for wireless data telemetry, a flash to store the programmed data, and FTDI to
bridge USB to UART commands, two regulators for battery powering, and an on-board charger to recharge
batteries.
9.3 Miniaturized Design
A flex-rigid PCB was designed to miniaturize the early prototype. One of the most
important factors to consider when implementing a flex-rigid PCB is bend radius. Factors that
influence bend radius include copper thickness, copper geometry and whether insulating coverlay
layers are used. A thicker flex region improves mechanical robustness but trades off bend radius.
Another factor effecting the bend radius is the length of the flex region. A longer flex region
increases the bend radius by reducing the stress on coper traces. However, a longer region trades
off with the desire for short flex regions for miniaturization.
104

Here we designed five double-side four-layer rigid PCBs, each 12x12 mm, connecting to
each other via a four-layer flex material (Polyimide) as shown in Figure 9.3. The flex region is 200
µm thick and 3 mm long which allowed bending of the flex region to 90
o
to create a cube like
shape (Figure 9.3). The folding feature allows for miniaturization of the PCB to fit within a
compact volume of 18 x 18 x 15 mm. Polygon Pour was used on the top layer, bottom layer, and
two middle layers for connection to +Vcc, -Vcc, and ground respectively.

Figure 9-3 Rendering of rigid-flex PCB consisting of double-sided 4-layer PCB designed in Altium
Designer.
9.4 Chassis Design
We designed a housing to mount the PCB on the rat’s skull using SolidWorks demonstrated
in Figure 9.4. The housing is a snap-fit design because it is a mating component that does not
require additional hardware fasteners such as screws and bolts. It consists of two parts as shown
in Figure 9.4 a, b. Figure 9.4a consists of cantilevers, or protruding parts, which are designed to
catch on undercuts, or depressions, in Figure 9.4b. During the joining operation, the cantilevers
catch on the undercuts in the mating component. After the joining operation, the snap-fit feature
returns to a stress-free condition. The joint may be separated by applying a force to both opposite
105

sides of the cube to separate the components. For this design, plastic is a suitable material because
it is flexible and also allows for wireless communication.
Figure 9.4c demonstrates the complete housing design with jointed mating components.
The purpose of the through holes are to secure the PCB on the pieces where connectors reside. The
connectors include an Omnetics for connecting to the electrodes, power connector and µUSB for
reprogramming the chips. An opening on top of the housing is also placed for a switch for power
control.

Figure 9-4 SolidWorks rendering of PCB housing. a) Top piece with protruding cantilevers. b) Bottom
piece with undercuts. The PCB with the Omnetics connectors first gets secured in (b) using four 1 mm
screws. c) Jointed mating components (width: 20 mm; length: 20 mm; height: 17 mm). Wall thickness is 1
mm. The openings are for connectors and a switch for power control as labeled.

106

Figure 9-5 Rendering of assembled folded PCB inside the housing.
To power the system, two-coin batteries were connected in series to obtain ±3.7V. Each
battery is connected to a regulator with a fixed output of ±3.3V. The batteries are rechargeable
with approximately 500 mAhr capacity. The batteries each have a mass of 14 g and a 35 mm
diameter. Using low-mass batteries is desirable for the animal, however it trades off low capacities
and high internal resistance. Here, depending on the stimulation setting, the batteries would allow
for device operation of 1-5 hrs.
9.5 Chassis Prototype Test in Rat
To test whether the housing design is a good fit for mounting on rat’s skull, we 3D printed
the design, and a mock-up of Omnetics connectors using Acrylonitrile Butadiene Styrene
(ABS) printing material. The two pieces of the housing were printed separately and then snap fitted
on to each other. The mockup connector was super glued to the housing and allowed to dry over
night before implantation.
107

The animal was anesthetized as described in previous chapters and placed on the surgery
table. A nose cone administered isoflurane and a stereotaxic set-up frame was mounted through
ear bars. Five small holes, for anchor screws, were drilled in one hemisphere of the skull and
craniotomy was marked on the contralateral hemisphere (Figure 9.6 a). The mock-up Omnetics
connector was placed above the marked craniotomy region and a thin layer of dental cement was
applied around the connector to hold it in place. The housing with a mating Omnetics connector
was then glued to the connector on the skull. Additional dental cement was then applied to the
connector to further secure it.
The batteries were attached to a rodent jacket (Lomir Biomedical Inc.) by Velcro and worn
by the animal. In the future design, short wires from the batteries would be routed along the neck
of the animal to the main PCB residing on the skull.

Figure 9-6 Mock-up implantation process of the housing on rat’s skull. a) Five anchoring screws on the left
hemisphere of the skull and a marked craniotomy region on the contralateral hemisphere. b) 3D printed
Omnetics connector over the craniotomy region secured in place with dental cement. c) Omnetics connector
108

glued to the housing is glued to the mating connector on the skull. d) Batteries attached to a rodent jacket
via Velcro and worn on the back of the animal.
The animal was then recovered on warm bedding and then returned to its animal holding
once able to walk around normally. Animal was monitored for pain, inflammation, and infection
for up to two weeks. Up to 4 weeks, the animal was also monitored daily for free-roaming behavior
around its cage with the housing and batteries.
One day post-surgery, the animal seemed to be in pain and was not moving around
normally. We administered 0.3 mg buprenorphine and the animal and was returned to its cage. The
following day, the animal was free roaming the cage and did not seem to have modified behavior
caused by the head stage and batteries for up to 4 weeks when the animal was euthanized.

109

CHAPTER 10: CONCLUSION AND PROSPECT
The system presented in this thesis is a novel, versatile, and cost-efficient neural interface
technology for small animal neuroscience research. The design process and in vivo evaluation
demonstrate feasibility of use in rat experiments. The system described in this work consisted of
several pieces including PCB schematic and layout design, electrode design, and integrated
microfluidic channels that were prototyped, systematically tested and validated for use in acute
studies. The results of this work presented capability of the system to record immediate and
prolonged neural responses to stimulations by recording fEPSPs, PSs and spontaneous spikes from
the same and neighboring microelectrodes in response to varying stimulation parameters.
Integrated microfluidic channels were also separately tested and demonstrated its capability to
deliver drug focally for chemical manipulation of the interface while maintaining the ability to
record from and stimulate the neural tissue. Thus, the system offers the capability of closed-loop
neural stimulation to and recording from the same interface and provides a powerful tool for
monitoring and manipulating neural circuits at the single neuron level to small population of
neurons.
Future work concerns engineering effort for chronic implantation including electrical and
drug delivery components. The requirements include integration of a wireless embedded system
for wireless data transmission, and fabrication of rigid-flex PCB for mounting on rat’s skull.
Wireless data transmission will consume large power and the battery-operated system will be
operable for short experiments. To improve power efficiency for chronic studies, we propose
transmitting data based on user needs, thereby reducing data rate. To do this, three options may be
provided to the user as follows:
110

1) Users interested in spike activity only: This technique is taking advantage of slow spike rate.
The user would choose a voltage threshold above which spike activity would be recognized. By
comparing the signal with this threshold, the time at which the signal goes beyond would be
stamped and transmitted. In this way, the user will have time stamps of spikes and the rest of the
signal, which is mostly consistent of noise would not be transmitted. For users interested in spike
waveform as well as time stamps, data may be saved 3 ms before and after the signal crosses the
threshold. Thus, a 6 ms waveform tagged with a time at which it occurred would be transmitted.
2) Users interested in LFP only: This technique is taking advantage of low frequency spectrum of
LFP. In this case all sampled data will be saved and transmitted. However, since sampling rate
maybe as low as 1 kHz, data rate would be substantially reduced.
3) Option to record raw data for users interested in the entire signal.
We expect that next generation design including these features will be one step closer to offering
a complete solution for small animal free-roaming neuroscience research.

111

REFERENCES
Ackermans, L., Temel, Y., Cath, D., van der Linden, C., Bruggeman, R., Kleijer, M., Nederveen,
P., Schruers, K., Colle, H., Tijssen, M.A.J., Visser-Vandewalle, V., 2006. Deep brain
stimulation in Tourette’s syndrome: Two targets? Movement Disorders 21, 709–713.
https://doi.org/10.1002/mds.20816
Badia, J., Boretius, T., Andreu, D., Azevedo-Coste, C., Stieglitz, T., Navarro, X., 2011.
Comparative analysis of transverse intrafascicular multichannel, longitudinal
intrafascicular and multipolar cuff electrodes for the selective stimulation of nerve
fascicles. Journal of Neural Engineering 8, 036023. https://doi.org/10.1088/1741-
2560/8/3/036023
Banni, S., Carta, G., Murru, E., Cordeddu, L., Giordano, E., Marrosu, F., Puligheddu, M., Floris,
G., Asuni, G.P., Cappai, A.L., Deriu, S., Follesa, P., 2012. Vagus Nerve Stimulation
Reduces Body Weight and Fat Mass in Rats. PLoS ONE 7, e44813.
https://doi.org/10.1371/journal.pone.0044813
Bard AJ, F.L., 2001. Electrochemical Methods: Fundamentals and Applications, 2nd ed. Wiley,
New York.
Beebe, X., Rose, T.L., 1988. Charge injection limits of activated iridium oxide electrodes with
0.2 ms pulses in bicarbonate buffered saline (neurological stimulation application). IEEE
Transactions on Biomedical Engineering 35, 494–495. https://doi.org/10.1109/10.2122
Berger, T.W., Ahuja, A., Courellis, S.H., Deadwyler, S.A., Erinjippurath, G., Gerhardt, G.A.,
Gholmieh, G., Granacki, J.J., Hampson, R., Hsaio, M.C., Lacoss, J., Marmarelis, V.Z.,
Nasiatka, P., Srinivasan, V., Song, D., Tanguay, A.R., Wills, J., 2005. Restoring lost
cognitive function. IEEE Engineering in Medicine and Biology Magazine 24, 30–44.
https://doi.org/10.1109/MEMB.2005.1511498
Berger, Theodore W., Hampson, R.E., Song, D., Goonawardena, A., Marmarelis, V.Z.,
Deadwyler, S.A., 2011. A cortical neural prosthesis for restoring and enhancing memory.
Journal of Neural Engineering 8. https://doi.org/10.1088/1741-2560/8/4/046017
Berger, Theodore W, Hampson, R.E., Song, D., Goonawardena, A., Marmarelis, V.Z.,
Deadwyler, S.A., 2011. A cortical neural prosthesis for restoring and enhancing memory.
Journal of neural engineering 8, 46017. https://doi.org/10.1088/1741-2560/8/4/046017
Berger, T.W., Song, D., Chan, R.H.M., Marmarelis, V.Z., 2010. The Neurobiological Basis of
Cognition: Identification by Multi-Input, Multioutput Nonlinear Dynamic Modeling.
Proceedings of the IEEE 98, 356–374. https://doi.org/10.1109/JPROC.2009.2038804

112

Berger, T.W., Song, G.A., Chan, R.H.M., Shin, D., Marmarelis, V.Z., Hampson, R.E., Sweatt,
A.J., Heck, C.N., Liu, C.Y., Wills, J., LaCoss, J., Granacki, J.J., Gerhardt, G.A.,
Deadwyler, S.A., 2012. Role of the Hippocampus in Memory Formation : Restorative
Encoding Memory Integration Neural Device As a Cognitive Neural Prosthesis. IEEE
Pulse 3, 17–22. https://doi.org/10.1109/MPUL.2012.2205775
Bragin, A., Jando, G., Nadasdy, Z., Hetke, J., Wise, K., Buzsaki, G., 1995. Gamma (40-100 Hz)
oscillation in the hippocampus of the behaving rat. The Journal of Neuroscience 15, 47
LP – 60. https://doi.org/10.1523/JNEUROSCI.15-01-00047.1995
Campagnola, P.J., Millard, A.C., Terasaki, M., Hoppe, P.E., Malone, C.J., Mohler, W.A., 2002.
Three-Dimensional High-Resolution Second-Harmonic Generation Imaging of
Endogenous Structural Proteins in Biological Tissues. Biophysical Journal 82, 493–508.
https://doi.org/10.1016/S0006-3495(02)75414-3
Caparso, A.V., Durand, D.M., Mansour, J.M., 2009. A Nerve Cuff Electrode for Controlled
Reshaping of Nerve Geometry. Journal of Biomaterials Applications 24, 247–273.
https://doi.org/10.1177/0885328208097426
Caravaca, A.S., Tsaava, T., Goldman, L., Silverman, H., Riggott, G., Chavan, S.S., Bouton, C.,
Tracey, K.J., Desimone, R., Boyden, E.S., Sohal, H.S., Olofsson, P.S., 2017. A novel
flexible cuff-like microelectrode for dual purpose, acute and chronic electrical interfacing
with the mouse cervical vagus nerve. Journal of Neural Engineering 14, 066005.
https://doi.org/10.1088/1741-2552/aa7a42
Chaturvedi, A., Foutz, T.J., McIntyre, C.C., 2012. Current steering to activate targeted neural
pathways during deep brain stimulation of the subthalamic region. Brain Stimulation 5,
369–377. https://doi.org/10.1016/j.brs.2011.05.002
Cheng, C., Tsai, P., Yang, T., Cheng, W., Yen, T., Luo, Z., Qian, X., Chen, Z., Lin, T., Chen,
W., Chen, W., Liang, S., Shaw, F., Chang, C., Shih, F., Hsin, Y., Lee, C., Ker, M., Wu,
C., 2017. A fully integrated closed-loop neuromodulation SoC with wireless power and
bi-directional data telemetry for real-time human epileptic seizure control, in: 2017
Symposium on VLSI Circuits. pp. C44–C45.
https://doi.org/10.23919/VLSIC.2017.8008541
Cobo, A.M., Larson, C.E., Scholten, K., Miranda, J.A., Elyahoodayan, S., Song, D., Pikov, V.,
Meng, E., 2019. Parylene-Based Cuff Electrode With Integrated Microfluidics for
Peripheral Nerve Recording, Stimulation, and Drug Delivery. Journal of
Microelectromechanical Systems 28, 36–49.
https://doi.org/10.1109/JMEMS.2018.2881908
Cogan, Stuart F, 2008. Neural Stimulation and Recording Electrodes. Annual Review of
Biomedical Engineering 10, 275–309.
https://doi.org/10.1146/annurev.bioeng.10.061807.160518
113

Cogan, Stuart F., 2008a. Neural Stimulation and Recording Electrodes. Annual Review of
Biomedical Engineering 10, 275–309.
https://doi.org/10.1146/annurev.bioeng.10.061807.160518
Cogan, Stuart F., 2008b. Neural Stimulation and Recording Electrodes. Annual Review of
Biomedical Engineering 10, 275–309.
https://doi.org/10.1146/annurev.bioeng.10.061807.160518
Cogan, Stuart F., 2008c. Neural Stimulation and Recording Electrodes. Annual Review of
Biomedical Engineering 10, 275–309.
https://doi.org/10.1146/annurev.bioeng.10.061807.160518
Cogan, Stuart F, Guzelian, A.A., Agnew, W.F., Yuen, T.G.H., McCreery, D.B., 2004. Over-
pulsing degrades activated iridium oxide films used for intracortical neural stimulation.
Journal of Neuroscience Methods 137, 141–150.
https://doi.org/10.1016/j.jneumeth.2004.02.019
Cogan, S F, Plante, T.D., Ehrlich, J., 2004. Sputtered iridium oxide films (SIROFs) for low-
impedance neural stimulation and recording electrodes, in: The 26th Annual International
Conference of the IEEE Engineering in Medicine and Biology Society. pp. 4153–4156.
https://doi.org/10.1109/IEMBS.2004.1404158
Conway, B.E., 1991. TECHNICAL PAPERS’ ELECTROCHEMICAL SCIENCE AND
TECHNOLOGY Transition from “Supercapacitor” to “Battery” Behavior in
Electrochemical Energy Storage 138.
D. H. Liang, G. T. A. Kovacs, C. W. Storment, R. L. White, 1991. A method for evaluating the
selectivity of electrodes implanted for nerve simulation. IEEE Transactions on
Biomedical Engineering 38, 443–449. https://doi.org/10.1109/10.81563
Dalrymple, A.N., Huynh, M., Robles, U.A., Marroquin, J.B., Lee, C.D., Petrossians, A., Whalen,
J.J., Li, D., Parkington, H.C., Forsythe, J.S., Green, R.A., Poole-Warren, L.A., Shepherd,
R.K., Fallon, J.B., 2019. Electrochemical and mechanical performance of reduced
graphene oxide, conductive hydrogel, and electrodeposited Pt–Ir coated electrodes: an
active in vitro study. Journal of Neural Engineering 17, 16015.
https://doi.org/10.1088/1741-2552/ab5163
Deadwyler, R.E.H. and D.S. and B.S.R. and D.F. and A.S.D. and B.M.R. and X.S. and R.T.W.
and M.R.W. and D.E.C. and A.W.L. and H.M.-C. and G.P. and M., 2018. Developing a
hippocampal neural prosthetic to facilitate human memory encoding and recall. Journal
of Neural Engineering 15, 36014.
del Valle, J., Navarro, X., 2013. Chapter Two - Interfaces with the Peripheral Nerve for the
Control of Neuroprostheses, in: Geuna, S., Perroteau, I., Tos, P., Battiston, B. (Eds.),
International Review of Neurobiology. Academic Press, pp. 63–83.
https://doi.org/10.1016/B978-0-12-420045-6.00002-X
114

DeMichele, G.A., Troyk, P.R., 2003. Stimulus-resistant neural recording amplifier, in:
Proceedings of the 25th Annual International Conference of the IEEE Engineering in
Medicine and Biology Society (IEEE Cat. No.03CH37439). pp. 3329-3332 Vol.4.
https://doi.org/10.1109/IEMBS.2003.1280857
Elyahoodayan, S., Jiang, W., Xu, H., Song, D., 2019. A Multi-Channel Asynchronous
Neurostimulator With Artifact Suppression for Neural Code-Based Stimulations.
Frontiers in Neuroscience 13, 1011. https://doi.org/10.3389/fnins.2019.01011
Elyahoodayan, S., Larson, C., Meng, E., Song, D., 2018. Acute electrophysiological and drug
delivery tests of Parylene cuff electrode with embedded microfluid channels, in: 43rd
Neural Interfaces Conference. Presented at the Neural Interface Conference, Wiley,
Minneapolis, MN.
Ewing, S.G., Porr, B., Riddell, J., Winter, C., Grace, A.A., 2013. SaBer DBS: A fully
programmable, rechargeable, bilateral, charge-balanced preclinical microstimulator for
long-term neural stimulation. Journal of Neuroscience Methods 213, 228–235.
https://doi.org/10.1016/j.jneumeth.2012.12.008
Gholmieh, G., Courellis, S., Dimoka, A., Wills, J.D., LaCoss, J., Granacki, J.J., Marmarelis, V.,
Berger, T., 2004. An algorithm for real-time extraction of population EPSP and
population spike amplitudes from hippocampal field potential recordings. Journal of
Neuroscience Methods 136, 111–121. https://doi.org/10.1016/j.jneumeth.2004.01.006
Giagka, V., Serdijn, W.A., 2018. Realizing flexible bioelectronic medicines for accessing the
peripheral nerves – technology considerations. Bioelectronic Medicine 4.
https://doi.org/10.1186/s42234-018-0010-y
Graczyk, E.L., Schiefer, M.A., Saal, H.P., Delhaye, B.P., Bensmaia, S.J., Tyler, D.J., 2016. The
neural basis of perceived intensity in natural and artificial touch. Science Translational
Medicine 8, 362ra142. https://doi.org/10.1126/scitranslmed.aaf5187
Grill, W.M., Mortimer, J.T., 1996. Non-invasive measurement of the input-output properties of
peripheral nerve stimulating electrodes. Journal of Neuroscience Methods 65, 43–50.
https://doi.org/10.1016/0165-0270(95)00143-3
Guiraud, D., Andreu, D., Bonnet, S., Carrault, G., Couderc, P., Hagège, A., Henry, C.,
Hernandez, A., Karam, N., Le Rolle, V., Mabo, P., Maciejasz, P., Malbert, C.-H.,
Marijon, E., Maubert, S., Picq, C., Rossel, O., Bonnet, J.-L., 2016. Vagus nerve
stimulation: state of the art of stimulation and recording strategies to address autonomic
function neuromodulation. Journal of Neural Engineering 13, 041002.
https://doi.org/10.1088/1741-2560/13/4/041002
Hamani, C., McAndrews, M.P., Cohn, M., Oh, M., Zumsteg, D., Shapiro, C.M., Wennberg,
R.A., Lozano, A.M., 2008. Memory enhancement induced by hypothalamic/fornix deep
brain stimulation. Annals of Neurology 63, 119–123. https://doi.org/10.1002/ana.21295
115

Hambrecht, F.T., 1995. Visual prostheses based on direct interfaces with the visual system.
Bailliere’s clinical neurology 4, 147–165.
Hampson, R.E., Song, D., Chan, R.H.M., Sweatt, A.J., Riley, M.R., Gerhardt, G.A., Shin, D.C.,
Marmarelis, V.Z., Berger, T.W., Deadwyler, S.A., 2012. A nonlinear model for
hippocampal cognitive prosthesis: Memory facilitation by hippocampal ensemble
stimulation. IEEE Transactions on Neural Systems and Rehabilitation Engineering 20,
184–197. https://doi.org/10.1109/TNSRE.2012.2189163
Hampson, R.E., Song, D., Opris, I., Santos, L.M., Shin, D.C., Gerhardt, G.A., Marmarelis, V.Z.,
Berger, T.W., Deadwyler, S.A., 2013. Facilitation of memory encoding in primate
hippocampus by a neuroprosthesis that promotes task-specific neural firing. Journal of
neural engineering 10, 66013. https://doi.org/10.1088/1741-2560/10/6/066013
Harrison, R.R., 2007. A Versatile Integrated Circuit for the Acquisition of Biopotentials, in:
2007 IEEE Custom Integrated Circuits Conference. pp. 115–122.
https://doi.org/10.1109/CICC.2007.4405694
Hodgkin Alan Lloyd, Rushton William Albert Hugh, Adrian Edgar Douglas, 1946. The electrical
constants of a crustacean nerve fibre. Proceedings of the Royal Society of London. Series
B - Biological Sciences 133, 444–479. https://doi.org/10.1098/rspb.1946.0024
Hoffer, J.A., Kallesøe, K., 2001. How to Use Nerve Cuffs to Stimulate, Record or Modulate
Neural Activity, in: Moxon, K.A., Chapin, J.K. (Eds.), Neural Prostheses for Restoration
of Sensory and Motor Function. CRC Press, pp. 139–175.
Holden, L.K., Finley, C.C., Firszt, J.B., Holden, T.A., Brenner, C., Potts, L.G., Gotter, B.D.,
Vanderhoof, S.S., Mispagel, K., Heydebrand, G., Skinner, M.W., 2013. Factors affecting
open-set word recognition in adults with cochlear implants. Ear and hearing 34, 342–360.
https://doi.org/10.1097/AUD.0b013e3182741aa7
Hottowy, P., Skoczeń, A., Gunning, D.E., Kachiguine, S., Mathieson, K., Sher, A., Wiącek, P.,
Litke, A.M., Dąbrowski, W., 2012. Properties and application of a multichannel
integrated circuit for low-artifact, patterned electrical stimulation of neural tissue. Journal
of Neural Engineering 9, 66005. https://doi.org/10.1088/1741-2560/9/6/066005
Houweling, A.R., Brecht, M., 2008. Behavioural report of single neuron stimulation in
somatosensory cortex. Nature 451, 65–68. https://doi.org/10.1038/nature06447
Huxley, A.F., Stämpeli, R., 1949. Evidence for saltatory conduction in peripheral myelinated
nerve fibres. The Journal of Physiology 108, 315–339.
https://doi.org/10.1113/jphysiol.1949.sp004335
IRNICH, W., 1980. The Chronaxie Time and Its Practical Importance. Pacing and Clinical
Electrophysiology 3, 292–301. https://doi.org/10.1111/j.1540-8159.1980.tb05236.x
116

Joëls, M., Fernhout, B., 1993. Decreased Population Spike in CA1 Hippocampal Area of
Adrenalectomized Rats after Repeated Synaptic Stimulation. Journal of
Neuroendocrinology 5, 537–543. https://doi.org/10.1111/j.1365-2826.1993.tb00519.x
Kandadai, M.A., Raymond, J.L., Shaw, G.J., 2012. Comparison of electrical conductivities of
various brain phantom gels: Developing a “brain gel model.” Materials Science and
Engineering C 32, 2664–2667. https://doi.org/10.1016/j.msec.2012.07.024
Kang, X., Liu, J.-Q., Tian, H., Yang, B., Nuli, Y., Yang, C., 2015. Self-Closed Parylene Cuff
Electrode for Peripheral Nerve Recording. Journal of Microelectromechanical Systems
24, 319–332. https://doi.org/10.1109/JMEMS.2014.2381634
Kelly, S.K., Ellersick, W.F., Krishnan, A., Doyle, P., Shire, D.B., Wyatt, J.L., Rizzo, J.F., 2014.
Redundant safety features in a high-channel-count retinal neurostimulator, in: 2014 IEEE
Biomedical Circuits and Systems Conference (BioCAS) Proceedings. pp. 216–219.
https://doi.org/10.1109/BioCAS.2014.6981701
Kerr, A.L., Cheng, S.-Y., Jones, T.A., 2011. Experience-dependent neural plasticity in the adult
damaged brain. Journal of Communication Disorders 44, 538–548.
https://doi.org/10.1016/j.jcomdis.2011.04.011
Koopman, F.A., Chavan, S.S., Miljko, S., Grazio, S., Sokolovic, S., Schuurman, P.R., Mehta,
A.D., Levine, Y.A., Faltys, M., Zitnik, R., Tracey, K.J., Tak, P.P., 2016. Vagus nerve
stimulation inhibits cytokine production and attenuates disease severity in rheumatoid
arthritis. Proceedings of the National Academy of Sciences 113, 8284–8289.
https://doi.org/10.1073/pnas.1605635113
Krause, M.R., Vieira, P.G., Csorba, B.A., Pilly, P.K., Pack, C.C., 2019. Transcranial alternating
current stimulation entrains single-neuron activity in the primate brain. Proceedings of
the National Academy of Sciences 116, 5747–5755.
https://doi.org/10.1073/pnas.1815958116
Larson, C.E., Meng, E., 2019. A Review for the Peripheral Nerve Interface Designer. Journal of
Neuroscience Methods 108523. https://doi.org/10.1016/j.jneumeth.2019.108523
Lawrence, S.M., Larsen, J.O., Horch, K.W., Riso, R., Sinkjaer, T., 2002. Long-term
biocompatibility of implanted polymer-based intrafascicular electrodes: Implanted
Polymer-Based Intrafascicular Electrodes. Journal of Biomedical Materials Research 63,
501–506. https://doi.org/10.1002/jbm.10303
Lee, B., Zubair, M.N., Marquez, Y.D., Lee, D.M., Kalayjian, L.A., Heck, C.N., Liu, C.Y., 2015.
A Single-Center Experience with the NeuroPace RNS System: A Review of Techniques
and Potential Problems. World Neurosurgery 84, 719–726.
https://doi.org/10.1016/j.wneu.2015.04.050
Lega, B., Burke, J., Jacobs, J., Kahana, M.J., 2016. Slow-Theta-to-Gamma Phase-Amplitude
Coupling in Human Hippocampus Supports the Formation of New Episodic Memories.
Cerebral Cortex 26, 268–278. https://doi.org/10.1093/cercor/bhu232
117

Leventhal, D.K., Cohen, M., Durand, D.M., 2006. Chronic histological effects of the flat
interface nerve electrode. Journal of Neural Engineering 3, 102–113.
https://doi.org/10.1088/1741-2560/3/2/004
Leventhal, D.K., Durand, D.M., 2003. Subfascicle Stimulation Selectivity with the Flat Interface
Nerve Electrode. Annals of Biomedical Engineering 31, 643–652.
https://doi.org/10.1114/1.1569266

Limnuson, K., Lu, H., Chiel, H.J., Mohseni, P., 2015. A bidirectional neural interface SoC with
an integrated spike recorder, microstimulator, and low-power processor for real-time
stimulus artifact rejection. Analog Integrated Circuits and Signal Processing 82, 457–470.
https://doi.org/10.1007/s10470-015-0489-z
Little, S., Pogosyan, A., Neal, S., Zavala, B., Zrinzo, L., Hariz, M., Foltynie, T., Limousin, P.,
Ashkan, K., FitzGerald, J., Green, A.L., Aziz, T.Z., Brown, P., 2013. Adaptive deep brain
stimulation in advanced Parkinson disease. Annals of neurology 74, 449–457.
https://doi.org/10.1002/ana.23951
Liu, H.-F., Chen, Z.-G., Fang, T.-L., Arnold, P., Lineaweaver, W.C., Zhang, J., 2014. Changes of
the donor nerve in end-to-side neurorrhaphies with epineurial window and partial
neurectomy: A long-term evaluation in the rat model. Microsurgery 34, 136–144.
https://doi.org/10.1002/micr.22167
Lo, Y., Kuan, Y., Culaclii, S., Kim, B., Wang, P., Chang, C., Massachi, J.A., Zhu, M., Chen, K.,
Gad, P., Edgerton, V.R., Liu, W., 2017. A Fully Integrated Wireless SoC for Motor
Function Recovery After Spinal Cord Injury. IEEE Transactions on Biomedical Circuits
and Systems 11, 497–509. https://doi.org/10.1109/TBCAS.2017.2679441
Loeb, G.E., Peck, R.A., 1996. Cuff electrodes for chronic stimulation and recording of peripheral
nerve activity. Journal of Neuroscience Methods 64, 95–103.
https://doi.org/10.1016/0165-0270(95)00123-9
Lu, Y., Wang, T., Cai, Z., Cao, Y., Yang, H., Duan, Y.Y., 2009. Anodically electrodeposited
iridium oxide films microelectrodes for neural microstimulation and recording. Sensors
and Actuators B: Chemical 137, 334–339. https://doi.org/10.1016/j.snb.2008.11.036
M. Schuettler, T.S., 2002. Corrosion of Thin-Film Electrodes In Vitro : Comparison of Disk and
Lily Pad Corrosion of Thin-Film Electrodes In Vitro : Comparison of Disk and Lily Pad
Design. IFESS 8–11.
Månsson, K.N.T., Salami, A., Frick, A., Carlbring, P., Andersson, G., Furmark, T., Boraxbekk,
C.-J., 2016a. Neuroplasticity in response to cognitive behavior therapy for social anxiety
disorder. Translational Psychiatry 6, e727.
118

Månsson, K.N.T., Salami, A., Frick, A., Carlbring, P., Andersson, G., Furmark, T., Boraxbekk,
C.-J., 2016b. Neuroplasticity in response to cognitive behavior therapy for social anxiety
disorder. Translational Psychiatry 6, e727.
Marks, W.B., Loeb, G.E., 1976. Action currents, internodal potentials, and extracellular records
of myelinated mammalian nerve fibers derived from node potentials. Biophysical Journal
16, 655–668. https://doi.org/10.1016/S0006-3495(76)85719-0
Maynard, E.M., Nordhausen, C.T., Normann, R.A., 1997. The Utah Intracortical Electrode
Array: A recording structure for potential brain-computer interfaces.
Electroencephalography and Clinical Neurophysiology 102, 228–239.
https://doi.org/10.1016/S0013-4694(96)95176-0
McCallister, W.V., Tang, P., Smith, J., Trumble, T.E., 2001. Axonal regeneration stimulated by
the combination of nerve growth factor and ciliary neurotrophic factor in an end-to-side
model. The Journal of Hand Surgery 26, 478–488.
https://doi.org/10.1053/jhsu.2001.24148
McCreery, D.B., Shannon, R. V, Moore, J.K., Chatterjee, M., 1998. Accessing the tonotopic
organization of the ventral cochlear nucleus by intranuclear microstimulation. IEEE
Transactions on Rehabilitation Engineering 6, 391–399.
https://doi.org/10.1109/86.736153
McHardy, J., Robblee, L.S., Marston, J.M., Brummer, S.B., 1980. Electrical stimulation with Pt
electrodes. IV. Factors influencing Pt dissolution in inorganic saline. Biomaterials 1,
129–134. https://doi.org/10.1016/0142-9612(80)90034-4
McIntyre, C.C., Chaturvedi, A., Shamir, R.R., Lempka, S.F., 2015. Engineering the Next
Generation of Clinical Deep Brain Stimulation Technology. Brain Stimulation 8, 21–26.
https://doi.org/10.1016/j.brs.2014.07.039
Merrill, D.R., Bikson, M., Jefferys, J.G.R., 2005. Electrical stimulation of excitable tissue:
Design of efficacious and safe protocols. Journal of Neuroscience Methods 141, 171–
198. https://doi.org/10.1016/j.jneumeth.2004.10.020
Miranda, R.A., Casebeer, W.D., Hein, A.M., Judy, J.W., Krotkov, E.P., Laabs, T.L., Manzo,
J.E., Pankratz, K.G., Pratt, G.A., Sanchez, J.C., Weber, D.J., Wheeler, T.L., Ling, G.S.F.,
2015. DARPA-funded efforts in the development of novel brain–computer interface
technologies. Journal of Neuroscience Methods 244, 52–67.
https://doi.org/10.1016/j.jneumeth.2014.07.019
Nag, S., Sikdar, S.K., Thakor, N.V., Rao, V.R., Sharma, D., 2015. Sensing of Stimulus Artifact
Suppressed Signals From Electrode Interfaces. IEEE Sensors Journal 15, 3734–3742.
https://doi.org/10.1109/JSEN.2015.2399248

119

Navarro, X., Krueger, T.B., Lago, N., Micera, S., Stieglitz, T., Dario, P., 2005. A critical review
of interfaces with the peripheral nervous system for the control of neuroprostheses and
hybrid bionic systems. Journal of the Peripheral Nervous System 10, 229–258.
https://doi.org/10.1111/j.1085-9489.2005.10303.x
Newman, J., 1965. Resistance for Flow of Current to a Disk John.
O’Leary, G., Groppe, D.M., Valiante, T.A., Verma, N., Genov, R., 2018. NURIP: Neural
Interface Processor for Brain-State Classification and Programmable-Waveform
Neurostimulation. IEEE Journal of Solid-State Circuits 53, 3150–3162.
https://doi.org/10.1109/JSSC.2018.2869579
Ordonez, J.S., Rudmann, L., Cvancara, P., Bentler, C., Stieglitz, T., 2015. Mechanical
deformation of thin film platinum under electrical stimulation, in: 2015 37th Annual
International Conference of the IEEE Engineering in Medicine and Biology Society
(EMBC). pp. 1045–1048. https://doi.org/10.1109/EMBC.2015.7318544
Petrossians, A., Whalen III, J.J., Weiland, J.D., Mansfeld, F., 2011. Electrodeposition and
Characterization of Thin-Film Platinum-Iridium Alloys for Biological Interfaces. Journal
of The Electrochemical Society 158, D269. https://doi.org/10.1149/1.3559477
Picaza, J.A., Hunter, S.E., Cannon, B.W., 1977. Pain Suppression by Peripheral Nerve
Stimulation. Stereotactic and Functional Neurosurgery 40, 223–234.
https://doi.org/10.1159/000102446
Plachta, D.T.T., Gierthmuehlen, M., Cota, O., Espinosa, N., Boeser, F., Herrera, T.C., Stieglitz,
T., Zentner, J., 2014. Blood pressure control with selective vagal nerve stimulation and
minimal side effects. Journal of Neural Engineering 11, 036011.
https://doi.org/10.1088/1741-2560/11/3/036011
Polikov, Vadim S, Tresco, P.A., Reichert, W.M., 2005. Response of brain tissue to chronically
implanted neural electrodes. Journal of Neuroscience Methods 148, 1–18.
https://doi.org/10.1016/j.jneumeth.2005.08.015
Polikov, Vadim S., Tresco, P.A., Reichert, W.M., 2005. Response of brain tissue to chronically
implanted neural electrodes. Journal of Neuroscience Methods 148, 1–18.
https://doi.org/10.1016/j.jneumeth.2005.08.015
Priori, A., Foffani, G., Rossi, L., Marceglia, S., 2013. Adaptive deep brain stimulation (aDBS)
controlled by local field potential oscillations. Experimental Neurology 245, 77–86.
https://doi.org/10.1016/j.expneurol.2012.09.013
Ranck, J.B., 1973. Studies on single neurons in dorsal hippocampal formation and septum in
unrestrained rats: Part I. Behavioral correlates and firing repertoires. Experimental
Neurology 41, 462–531. https://doi.org/10.1016/0014-4886(73)90290-2
120

Robblee, L.S., 1983. Activated Ir: An Electrode Suitable for Reversible Charge Injection in
Saline Solution. Journal of The Electrochemical Society 130, 731.
https://doi.org/10.1149/1.2119793
Rodríguez, F.J., Ceballos, D., Schu¨ttler, M., Valero, A., Valderrama, E., Stieglitz, T., Navarro,
X., 2000. Polyimide cuff electrodes for peripheral nerve stimulation. Journal of
Neuroscience Methods 98, 105–118. https://doi.org/10.1016/S0165-0270(00)00192-8
Rolston, J.D., Gross, R.E., Potter, S.M., 2009a. A low-cost multielectrode system for data
acquisition enabling real-time closed-loop processing with rapid recovery from
stimulation artifacts. Frontiers in neuroengineering 2, 12.
https://doi.org/10.3389/neuro.16.012.2009
Rolston, J.D., Gross, R.E., Potter, S.M., 2009b. A low-cost multielectrode system for data
acquisition enabling real-time closed-loop processing with rapid recovery from
stimulation artifacts. Frontiers in neuroengineering 2, 12.
https://doi.org/10.3389/neuro.16.012.2009
Russell, C., Roche, A.D., Chakrabarty, S., 2019. Peripheral nerve bionic interface: a review of
electrodes. International Journal of Intelligent Robotics and Applications 3, 11–18.
https://doi.org/10.1007/s41315-019-00086-3
Rydevik, B., Brown, M., Ehira, T., Nordborg, C., 1985. Effects of collagenase on nerve tissue.
An experimental study on acute and long-term effects in rabbits. Spine 10, 562—566.
Salam, M.T., Velazquez, J.L.P., Genov, R., 2016. Seizure Suppression Efficacy of Closed-Loop
Versus Open-Loop Deep Brain Stimulation in a Rodent Model of Epilepsy. IEEE
Transactions on Neural Systems and Rehabilitation Engineering 24, 710–719.
https://doi.org/10.1109/TNSRE.2015.2498973
Šámal, F., Haninec, P., Raška, O., Dubový, P., 2006. Quantitative assessment of the ability of
collateral sprouting of the motor and primary sensory neurons after the end-to-side
neurorrhaphy of the rat musculocutaneous nerve with the ulnar nerve. Annals of
Anatomy - Anatomischer Anzeiger 188, 337–344.
https://doi.org/10.1016/j.aanat.2006.01.017
Schiefer, M.A., Polasek, K.H., Triolo, R.J., Pinault, G.C.J., Tyler, D.J., 2010. Selective
stimulation of the human femoral nerve with a flat interface nerve electrode. Journal of
Neural Engineering 7, 026006. https://doi.org/10.1088/1741-2560/7/2/026006
Schläpfer, T.E., Kayser, S., 2014. Deep brain stimulation for treatment-resistant depression.
Klinische Neurophysiologie 45, 113–117. https://doi.org/10.1055/s-0034-1375605
Schultz, A.E., Kuiken, T.A., 2011. Neural Interfaces for Control of Upper Limb Prostheses: The
State of the Art and Future Possibilities. PM&R 3, 55–67.
https://doi.org/10.1016/j.pmrj.2010.06.016
121

Sessolo, M., Khodagholy, D., Rivnay, J., Maddalena, F., Gleyzes, M., Steidl, E., Buisson, B.,
Malliaras, G.G., 2013. Easy-to-fabricate conducting polymer microelectrode arrays.
Advanced Materials 25, 2135–2139. https://doi.org/10.1002/adma.201204322
Shepherd, R.K., Hardie, N.A., Baxi, J.H., 2001. Electrical Stimulation of the Auditory Nerve:
Single Neuron Strength-Duration Functions in Deafened Animals. Annals of Biomedical
Engineering 29, 195–201. https://doi.org/10.1114/1.1355276
Sinclair, E.B., Andarawis-Puri, N., Ros, S.J., Laudier, D.M., Jepsen, K.J., Hausman, M.R., 2012.
Relating applied strain to the type and severity of structural damage in the rat median
nerve using second harmonic generation microscopy. Muscle & Nerve 46, 895–898.
https://doi.org/10.1002/mus.23443
Song, D., Berger, T.W., 2013. Hippocampal Memory Prosthesis BT - Encyclopedia of
Computational Neuroscience, in: Jaeger, D., Jung, R. (Eds.), . Springer New York, New
York, NY, pp. 1–5. https://doi.org/10.1007/978-1-4614-7320-6_558-1
Song, D., Chan, R.H.M., Marmarelis, V.Z., Hampson, R.E., Deadwyler, S.A., Berger, T.W.,
2009. Nonlinear modeling of neural population dynamics for hippocampal prostheses.
Neural Networks 22, 1340–1351. https://doi.org/10.1016/j.neunet.2009.05.004
Song, D., Chan, R.H.M., Marmarelis, V.Z., Hampson, R.E., Deadwyler, S.A., Berger, T.W.,
2007. Nonlinear Dynamic Modeling of Spike Train Transformations for Hippocampal-
Cortical Prostheses. IEEE Transactions on Biomedical Engineering 54, 1053–1066.
https://doi.org/10.1109/TBME.2007.891948
Song, D., Robinson, B.S., Hampson, R.E., Marmarelis, V.Z., Deadwyler, S.A., Berger, T.W.,
2018. Sparse Large-Scale Nonlinear Dynamical Modeling of Human Hippocampus for
Memory Prostheses. IEEE Transactions on Neural Systems and Rehabilitation
Engineering 26, 272–280. https://doi.org/10.1109/TNSRE.2016.2604423
Steigerwald, F., Müller, L., Johannes, S., Matthies, C., Volkmann, J., 2016. Directional deep
brain stimulation of the subthalamic nucleus: A pilot study using a novel
neurostimulation device. Movement Disorders 31, 1240–1243.
https://doi.org/10.1002/mds.26669
Suarez-Perez, A., Gabriel, G., Rebollo, B., Illa, X., Guimerà-Brunet, A., Hernández-Ferrer, J.,
Martínez, M.T., Villa, R., Sanchez-Vives, M. V, 2018. Quantification of Signal-to-Noise
Ratio in Cerebral Cortex Recordings Using Flexible MEAs With Co-localized Platinum
Black, Carbon Nanotubes, and Gold Electrodes. Frontiers in Neuroscience 12, 862.
https://doi.org/10.3389/fnins.2018.00862
Sun, F.T., Morrell, M.J., 2014. Closed-loop Neurostimulation: The Clinical Experience.
Neurotherapeutics 11, 553–563. https://doi.org/10.1007/s13311-014-0280-3

122

Suner, S., Fellows, M.R., Vargas-Irwin, C., Nakata, G.K., Donoghue, J.P., 2005. Reliability of
signals from a chronically implanted, silicon-based electrode array in non-human primate
primary motor cortex. IEEE Transactions on Neural Systems and Rehabilitation
Engineering 13, 524–541. https://doi.org/10.1109/TNSRE.2005.857687
Swan, Brandon D., Gasperson, L.B., Krucoff, M.O., Grill, W.M., Turner, D.A., 2018. Sensory
percepts induced by microwire array and DBS microstimulation in human sensory
thalamus. Brain Stimulation 11, 416–422. https://doi.org/10.1016/j.brs.2017.10.017
Swan, Brandon D, Gasperson, L.B., Krucoff, M.O., Grill, W.M., Turner, D.A., 2018. Sensory
percepts induced by microwire array and DBS microstimulation in human sensory
thalamus. Brain Stimulation 11, 416–422. https://doi.org/10.1016/j.brs.2017.10.017
Tan, D.W., Schiefer, M.A., Keith, M.W., Anderson, J.R., Tyler, D.J., 2015. Stability and
selectivity of a chronic, multi-contact cuff electrode for sensory stimulation in human
amputees. Journal of Neural Engineering 12, 026002. https://doi.org/10.1088/1741-
2560/12/2/026002

Tran, N., Bai, S., Yang, J., Chun, H., Kavehei, O., Yang, Y., Muktamath, V., Ng, D., Meffin, H.,
Halpern, M., Skafidas, E., 2014. A Complete 256-Electrode Retinal Prosthesis Chip.
IEEE Journal of Solid-State Circuits 49, 751–765.
https://doi.org/10.1109/JSSC.2014.2298037
Van der Horst, C., Silwana, B., Iwuoha, E., Somerset, V., 2015. Synthesis and Characterization
of Bismuth-Silver Nanoparticles for Electrochemical Sensor Applications. Analytical
Letters 48, 1311–1332. https://doi.org/10.1080/00032719.2014.979357
Vedantham, V., Cannon, S.C., 1999. The Position of the Fast-Inactivation Gate during Lidocaine
Block of Voltage-gated Na
+
Channels. The Journal of General Physiology 113, 7–16.
https://doi.org/10.1085/jgp.113.1.7
Velliste, M., Perel, S., Spalding, M.C., Whitford, A.S., Schwartz, A.B., 2008. Cortical control of
a prosthetic arm for self-feeding. Nature 453, 1098–1101.
https://doi.org/10.1038/nature06996
Venkatraman, S., Elkabany, K., Long, J.D., Yao, Y., Carmena, J.M., 2009. A system for neural
recording and closed-loop intracortical microstimulation in awake rodents. IEEE
Transactions on Biomedical Engineering 56, 15–22.
https://doi.org/10.1109/TBME.2008.2005944
Vesper, J., Chabardes, S., Fraix, V., Sunde, N., Østergaard, K., 2002. Dual channel deep brain
stimulation system (Kinetra) for Parkinson's disease and essential tremor: a
prospective multicentre open label clinical study. Journal of Neurology, Neurosurgery
& Psychiatry 73, 275 LP – 280. https://doi.org/10.1136/jnnp.73.3.275
123

Vijayaraghavan, S., Huq, R., Hausman, M.R., 2014. Methods of peripheral nerve tissue
preparation for second harmonic generation imaging of collagen fibers. Methods 66,
246–255. https://doi.org/10.1016/j.ymeth.2013.08.012
Voges, J., Koulousakis, A., Sturm, V., 2007. Deep brain stimulation for Parkinson’s disease.
Acta Neurochirurgica, Supplementum 171–184. https://doi.org/10.1007/978-3-211-
33081-4_19
Wagenaar, D.A., Potter, S.M., 2002. Real-time multi-channel stimulus artifact suppression by
local curve fitting. Journal of Neuroscience Methods 120, 113–120.
https://doi.org/10.1016/S0165-0270(02)00149-8
Weiland, J.D., Cho, A.K., Humayun, M.S., 2011. Retinal prostheses: Current clinical results and
future needs. Ophthalmology 118, 2227–2237.
https://doi.org/10.1016/j.ophtha.2011.08.042
Wichmann, T., 2000. A digital averaging method for removal of stimulus artifacts in
neurophysiologic experiments. Journal of Neuroscience Methods 98, 57–62.
https://doi.org/10.1016/S0165-0270(00)00190-4
Xue, N., Sun, T., Tsang, W.M., Delgado-Martinez, I., Lee, S.-H., Sheshadri, S., Xiang, Z.,
Merugu, S., Gu, Y., Yen, S.-C., Thakor, N.V., 2015. Polymeric C-shaped cuff electrode
for recording of peripheral nerve signal. Sensors and Actuators B: Chemical 210, 640–
648. https://doi.org/10.1016/j.snb.2015.01.006
Yu, H., Xiong, W., Zhang, H., Wang, W., Li, Z., 2014. A cable-tie-type parylene cuff electrode
for peripheral nerve interfaces. IEEE, pp. 9–12.
https://doi.org/10.1109/MEMSYS.2014.6765560
Zabara, J., 1985. Peripheral control of hypersynchronous discharge in epilepsy.
Electroencephalography and Clinical Neurophysiology 61, S162.
https://doi.org/10.1016/0013-4694(85)90626-1
Zhou, A., Johnson, B.C., Muller, R., 2018. Toward true closed-loop neuromodulation: artifact-
free recording during stimulation. Current Opinion in Neurobiology 50, 119–127.
https://doi.org/10.1016/j.conb.2018.01.012
Zhou, A., Santacruz, S.R., Johnson, B.C., Alexandrov, G., Moin, A., Burghardt, F.L., Rabaey,
J.M., Carmena, J.M., Muller, R., 2019. A wireless and artefact-free 128-channel
neuromodulation device for closed-loop stimulation and recording in non-human
primates. Nature Biomedical Engineering 3, 15–26. https://doi.org/10.1038/s41551-018-
0323-x
Zoumi, A., Yeh, A., Tromberg, B.J., 2002. Imaging cells and extracellular matrix in vivo by
using second-harmonic generation and two-photon excited fluorescence. Proceedings of
the National Academy of Sciences 99, 11014–11019.
https://doi.org/10.1073/pnas.172368799
124

Abstract (if available)

Abstract New technologies have emerged in recent years to treat various neurological disorders via electrical neuromodulation or replace a lost function via neural prosthesis. For example, deep brain stimulation has provided therapy to patients with Parkinson’s disease, epilepsy, severe depression, and dystonia. Cochlear implants and retinal prothesis have successfully provided hearing and some visual acuity to patients with sensory impairment by electrical stimulation of the cochlea or retina using appropriate patterns of pulses. Hippocampal memory prothesis has also shown success in restoring memory by reinstating the input-out properties of neurons, which requires delivery of neural-code-like stimulation patterns to the brain tissue with multiple electrodes. ❧ Current technology to support such systems lack feedback control and the ability to precisely modulate and monitor neural activities. These needs may be met with high density recording and modulation (electrical and chemical) systems with high resolution, stimulation and recording from the same electrode for feedback control, ability to deliver complex pattern of stimulation, and power efficient interfaces. Furthermore, miniaturized wireless systems is essential for chronic free-roaming applications. ❧ To better support neuromodulation and neural prosthesis, we present the following neural interface systems: 1) a highly configurable multichannel asynchronous neural-code-based stimulator with ability to stimulate and record from the same electrode near simultaneously, 2) implantable electrode design with high signal-to-noise-ratio, power efficiency, and spatial resolution for both recording and stimulation, 3) design and performance of flexible electrodes with microfluidic channels to inject drug focally with controlled doses for chemical modulation of the electrode-tissue interface, 4) system miniaturization for mounting on rat’s skull for future free-roaming small animal neuroscience research. The body of work presented int this thesis starts with background review of the need for neural interface technology, to technology design, prototyping, and testing on benchtop and then in vivo rats, concluding with the plan for next generation design and validation. We expect the final device to be a valuable tool for studying neurobiological basis of cognitive functions and a step closer to meet the unmet needs of next generation neural interface technologies.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Bidirectional neural interfaces for neuroprosthetics

PDF

Fabrication and packaging of three-dimensional Parylene C neural interfaces

PDF

Multi-region recordings from the hippocampus with conformal multi-electrode arrays

PDF

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

PDF

Manipulation of RGCs response using different stimulation strategies for retinal prosthesis

PDF

The electrochemical evaluation of Parylene-based electrodes for neural applications

PDF

Towards a high resolution retinal implant

PDF

Ultrasound neuromodulation and its applications for noninvasive vision restoration

PDF

Simulating electrical stimulation and recording in a multi-scale model of the hippocampus

PDF

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

PDF

On the electrophysiology of multielectrode recordings of the basal ganglia and thalamus to improve DBS therapy for children with secondary dystonia

PDF

Parylene C bioMEMS for implantable devices with electrochemical interfaces

PDF

Detection and decoding of cognitive states from neural activity to enable a performance-improving brain-computer interface

PDF

Low-power, dual sampling-rate, shared-architecture ADC for implantable biomedical systems

PDF

Theoretical foundations and design methodologies for cyber-neural systems

Asset Metadata

Creator Elyahoodayan, Sahar (author)

Core Title Next generation neural interfaces for neural modulation and prosthesis

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 08/08/2020

Defense Date 08/07/2020

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

Tag artifact suppression,brain,cuff electrode,deep brain stimulation,drug delivery,electrochemistry,electrophysiology,hippocampal memory prosthesis,intracortical recording,intracortical stimulation,multiplexing,OAI-PMH Harvest,peripheral nerve,Pt-Ir electrodeposition

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Berger, Theodore (committee chair), Song, Dong (committee chair), Hashemi, Hossein (committee member), Meng, Ellis (committee member)

Creator Email Elyahood@usc.edu,sa.elyahoo@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-364407

Unique identifier UC11666322

Identifier etd-Elyahooday-8902.pdf (filename),usctheses-c89-364407 (legacy record id)

Legacy Identifier etd-Elyahooday-8902.pdf

Dmrecord 364407

Document Type Dissertation

Rights Elyahoodayan, Sahar

Type texts

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

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

Repository Name University of Southern California Digital Library

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