University of Southern California Dissertations and Theses

An in vitro model of a retinal prosthesis

(USC Thesis Other)

An in vitro model of a retinal prosthesis

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

AN IN VITRO MODEL OF A RETINAL PROSTHESIS

by

Ashish Kishore Ahuja

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

May 2007

Copyright 2007 Ashish Kishore Ahuja

ii
Epigraph

...Nowhere did we stop long enough to get a
particularized impression, but the general sense of
vague and oppressive wonder grew upon me.
- Joseph Conrad

iii
Acknowledgments

There have been many people who have given me great support over the last
five years. My thesis advisor, Dr. Weiland, pushed me towards retinal
electrophysiological studies despite my initial reservations. Results from those
experiments comprise a large portion of this thesis. I thank him for his
encouragement to do something I was not initially comfortable doing and truly
introducing me to a new field.
Without the love and support of my mom, dad, and sister I would not have
been able to begin my work at USC. They have continued to be a foundation for me
during my time here.
The direction my research took was largely shaped by a few close peers.
Matthew Behrend was instrumental in the setup of the electrophysiology rig. My
research fed off a great deal from his energy and enthusiasm. I thank Jack
“whatever” Whalen for his input on the electrochemistry studies, and, more
importantly, for his ability to get me to come to the lab as early as he did and pursue
science with as much rigor as he did. I thank Dr. Sampath for his retinal
electrophysiology expertise, and Haru Okawa for his help with tissue handling.
Patrick Nasiatka, Michelle Hauer, Alan Horsager, and Ghassan Gholmieh provided
useful discussions throughout the course of my work. Dr. Tanguay always reminded
me that scientific research is an inherently emotional endeavor, not only an
intellectual one. I thank him for many great discussions. Dr. Humayun’s

iv
exuberance has been truly inspiring. He always reminded me to direct my focus to
applying knowledge rather than just accumulating it.
I also thank Vinit Ahuja, Siddhartha Khosla, Sabira Alloo, Sanjay Sethi, and
Charles Lepere for their ability to keep me balanced during the last few years.

v
Table of Contents

Epigraph ii

Acknowledgements iii

List of Figures viii

Abbreviations xi

Abstract xiii

Chapter 1. Introduction 1
1.1 Visual Prostheses Background 1
1.2 Brief Review of Relevant Retinal Anatomy and Physiology 7
1.2.1 Vertebrate retina 7
1.2.2 Remodeling in degenerate retina 9
1.3 Retinal Disease and Treatment 10
1.4 Neural Excitation 12
1.4.1 Circuit representation of neuron 12
1.4.2 Cellular excitation 13
1.4.3 Derivation of the strength-duration relationship 14
1.4.4 Spherical cell in a uniform electric field 16
1.5 Structure of this Thesis 17

Chapter 2. Design and Microfabrication of Multielectrode Arrays 21
2.1 Multielectrode Array Background 21
2.2 Microfabricated Electrode Arrays for Hippocampal and
Retinal Electrophysiology 22
2.2.1 Trisynaptic and Perforant Path MEAs 22
2.2.2 Retinal custom MEAs 24
2.3 Microfabrication Process Flow 26
2.3.1 MEAs with Au or Pt electrodes and Si
x
N
y
/SU-8
insulation 26
2.3.2 Alternate process flow for ITO leads 28
2.4 Improved SNR and stability of Si
x
N
y
/SU-8 insulation 29
2.5 CA3 Replacement in Hippocampus: In Vitro Proof-of-concept 32

Chapter 3. The Inhomogeneous Current Distribution 37
3.1 Prior Evidence of Inhomogeneous Distribution 37
3.2 Evidence of Inhomogeneous Distribution using EDS 39
3.3 Electrostatic Finite-Element Modeling of Electrode Current
Distribution 42
3.3.1 Primary current distribution: Disc flush with

vi
surrounding dielectric 42
3.3.2 Electrode geometries for current homogeneity 45
3.4 Fabrication of Recessed Microdiscs using SU-8 50
3.5 Conclusion 53

Chapter 4. Electrochemical Characterization of Electrodes 55
4.1 Capacitive and Faradaic Charge Transfer 55
4.2 Electrochemical Techniques 59
4.2.1 Cyclic Voltammetry (CV) 61
4.2.2. CV Electroplating using an ammonium
hexachloroplatinate bath 63
4.2.3 Electrochemical Impedance Spectroscopy (EIS) 65
4.3 Increased Charge Injection with Microelectrodes 67
4.3.1 Evidence of increased current using CV 68
4.3.2 Evidence of increased current using biphasic
pulsing 70
4.4 The Dependence of Spectral Impedance on Disc
Microelectrode Radius 73
4.4.1 EIS results on microelectrode discs of
varying diameter 74
4.4.2 Empirical and FEM correspondence with
Newman’s relationship 78
4.4.3 RC time constant for microelectrodes 79
4.4.4 Shift to dependence on area at lower frequencies 81
4.5 Conclusions 84

Chapter 5. Spatial Response Properties of Electrically Stimulated
Salamander Retina 88
5.1 Prior Work in Electrical Stimulation of Isolated Vertebrate Retina 88
5.2 Materials and Methods 90
5.2.1 Animal model 90
5.2.2 Tissue preparation 90
5.2.3 Custom retinal MEAs 92
5.2.4 Electrical stimulation and recording 93
5.2.5 Experimental protocol 96
5.2.6 Pharmacological agents and artifact subtraction 97
5.2.7 FEM of electrode-electrode interaction 99
5.2.8 Data analysis 100
5.3 Results 101
5.3.1 Response latency 101
5.3.2 Excitation radius vs. stimulus amplitude 103
5.3.3 Monopolar, dual monopolar, and bipolar stimulation 106
5.3.4 Frequency dependence of presynaptic and direct
ganglion cell excitation 107

vii
5.3.4 Charge and charge density thresholds using 200 µm
and 10 µm diameter electrodes 110
5.4 Conclusions 110

Chapter 6. Conclusions and Future Work 117
6.1 Summary, Conclusions, and Future Work for Microfabrication
Modeling, and Electrochemical Characterization of Microelectrodes 117
6.2 Summary and Conclusions for Array Electrophysiology Studies
Salamander Retina 122
6.3 Future Work for Array Electrophysiology Studies in Salamander
Retina 125
6.3.1 Retinal degenerate Model 125
6.3.2 Axonal vs. somatic Stimulation 126
6.3.3 Charge vs. charge Density 128
6.4 Limitations of this Experimental Approach 130

Bibliography 133

viii
List of Figures

1.1 Schematic of sub and epi-retinal prostheses approaches 4
1.2 Schematic of epi-retinal prosthesis with external camera 6
1.3 Photomicrograph and SEM of bovine retinal cross-section 8
1.4 Circuit diagram of cell membrane 13
1.5 Schematic of spherical cell intracellularly injected with current 14
1.6 Spherical cell in a uniform electric field 16
2.1 Photomicrographs of Perforant Path and Trisynaptic MEAs 24
2.2 Photomicrographs of Retinal MEAs 25
2.3 Microfabircation process flow for MEAs 29
2.4 Circuit diagram of MEA-tissue interface 30
2.5 SEM and Photomicrograph of MEAs without SU-8 31
2.6 Photomicrograph of MEA with Si
x
N
y
/SU-8 insulation 32
2.7 Schematic representation of ‘CA3 Replacement’ in hippocampus 33
2.8 Photomicrograph of CA3 Replacement cMEA 34
2.9 Amplitude vs. Input event (time) showing CA3 Replacement 36
3.1 Schematic representation of current density profile for disc electrode 38
3.2 Photomicrograph of a pulsed polycrystalline Au electrode 40
3.3 Atomic weight percent vs. disc electrode position 41
3.4 A disc electrode embedded in a semi-infinite slab of conductivity σ 43
3.5 Finite-element model showing primary current distribution 45

ix
List of Figures

3.6 Schematic of a segmented disc electrode 46
3.7 Finite-element model of hemispherical electrode 47
3.8 Schematic diagram of unrecessed and recessed disc electrodes 47
3.9 Finite-element model of variable depth recessed disc electrode 48
3.10 Finite-element model of variable depth recessed disc (side view) 49
3.11 Profilometry data for recessed disc electrode 52
3.12 Photomicrographs of variable disc MEAs with Pt recessed discs 52
4.1 Schematic representation of the capacitive double-layer 56
4.2 Energy diagram for reduction and oxidation 57
4.3 Schematic of custom three electrode Teflon electrochemical cell 60
4.4 Cyclic voltammograms of electroplated and evaporated Pt electrodes 65
4.5 Randles model of the electrode-electrolyte interface 67
4.6 Normalized cyclic voltammograms for various sized Au microelectrodes 69
4.7 Voltage drop vs. time for various sized biphasic pulsed microelectrodes 72
4.8 Voltage drop vs. disc diameter 73
4.9 Phase angle and |Z| vs. frequency for Pt microdiscs of varying diameter 77
4.10 Solution resistance vs. microdisc radius for Pt perimeter MEA 79

x
List of Figures

4.11 Phase shift at 100 kHz and time constant vs. microdisc diameter 81
4.12 |Z| vs. microdisc radius at 100 kHz, 10 kHz, 1 kHz, 100 Hz, and 10 Hz 83
4.13 First, fifth, and ninth harmonics of the Fourier series for a square wave 87
5.1 MEA with salamander retina 91
5.2 Photomicrograph of custom Retinal MEA with 200 µm stimulating pads 93
5.3 Picture of electrophysiology rig 95
5.4 Experimental flow chart for electrophysiological array studies 97
5.5 Direct stimulation of RGCs ascertained using CdCl
2
99
5.6 Post-stimulus time histogram (PSTH) showing reverberating response 102
5.7 Threshold current vs. distance from 200 µm diameter electrode edge 103
5.8 Strength-duration data for cells at different distances 105
5.9 PSTHs of monopolar, dual monopolar, and bipolar stimulation 106
5.10 Number of spikes vs. stimulus pulse frequency 109
6.1 Illustration of longitudinal vs. transverse stimulation of retina 127

6.2 Photomicrograph of variable diameter array for electrophysiology 129
studies

6.3 Cross-section of retina with schematic representation of epi-retinal 132
stimulation with 10 µm disc electrodes

xi
Abbreviations

APV - R-2-amino-5-phosphonopentanoate (NMDA receptor antagonist)

APB - DL-2-amino-phosphonobutyric acid (glutamate agonist, suppresses ON
pathway)

AMD - age related macular degeneration

AMPA receptor - α-amino-3-hydroxy-5-methylisoxazole receptor (a non-NMDA
ionotropic receptor for glutamate)

ATP - adenosine triphosphate

cMEA - conformal multielectrode array

CNV - choroidal neovascularization

CNQX - 6-cyano-7-nitroquinoxaline-2,3-dione (an AMPA/kainite receptor
antagonist)

CV - cyclic voltammetry

DG - dentate gyrus

EDS - energy dispersive spectroscopy

EECP - electrically evoked cortical potential

EIS - electrochemical impedance spectroscopy

ERG - electroretinogram

FEM - finite-element model

fEPSP - field evoked post-synaptic potential

FPGA - field programmable gate array

GABA - gamma-aminobutyric acid (an inhibitory neurotransmitter)

xii
Abbreviations

HMDS – hexamethyldisilazane (an adhesion promoter for photoresist)

ITO - indium tin oxide

KA – kainate (an agonist for the kainite receptor, an ionotropic glutamate receptor)

MEA - multielectrode array

NMDA - N-methyl-D-aspartic acid (An NMDA receptor is an ionotropic receptor for
the excitatory neurotransmitter glutamate)

PDA - Piperidinedicarboxylic acid

PECVD - plasma enhanced chemical vapor deposition

PBS - phosphate buffered saline

PET – positron emission tomography

PSTH – post-stimulus time histogram

PP - perforant path

RGC - retinal ganglion cell

RHE - reversible hydrogen electrode

RIE - reactive ion etch

RP - retinitis pigmentosa

RPE - retinal pigment epithelium

TTX - tetrodotoxin (sodium channel blocker)

xiii
Abstract

The electrode-retina interface of an epi-retinal prosthesis implant has been
studied using finite-element modeling, electrochemical methods, and retinal array
electrophysiology. The studies presented in this thesis have relied on the flexibility
of custom designed and microfabricated multielectrode arrays to investigate the
current distribution at the electrode-electrolyte interface and it’s interaction with
retinal tissue.
A flexible photolithographic microfabrication process flow has been
optimized for the fabrication of conformal MEAs for hippocampal and retinal slice
electrophysiological experiments. (Seven different electrode pad layouts were used
in the fabrication of various arrays.) Electrostatic finite-element modeling was used
to model the inhomogeneous current distribution at the electrode surface.
Using cyclic voltammetry (CV), biphasic pulsing, and electrochemical
impedance spectroscopy (EIS) it has been shown that (1) microelectrodes with
diameters ≤100 µm exhibit increased charge density, and (2) that the impedance of
microelectrodes transitions from a perimeter to an area dependent regime (e.g.,
!
1
r
"
1
r
2
) as the frequency is decreased from 100 KHz to 10 Hz. The later, proves
that the current distribution at the electrode surface is dynamic, and evolves during
the time course of a stimulus pulse.

xiv
MEAs with 200 µm diameter stimulating electrodes and 10 µm diameter
recording electrodes were employed in isolated tiger salamander retina
electrophysiology studies. Pharmacological agents were used to isolate direct
excitation of ganglion cells from excitation of other inner retinal cells. Strength-
duration data suggests that if amplitude will be used for the coding of brightness in
retinal prostheses, shorter pulses (200 µsec) will allow for a smaller region in the
area of the electrode to be excited over a larger dynamic range compared with longer
pulses (1 ms). Both electrophysiological results and electrostatic finite-element
modeling show that electrode-electrode interactions can lead to increased thresholds
for sites half way between simultaneously stimulated electrodes (29.4 ± 6.6 nC)
compared with monopolar stimulation (13.3 ± 1.7 nC). Presynaptic stimulation of
RGCs with both 200 µm and 10 µm diameter electrodes yielded threshold charge
densities of 12 ± 6 nC/cm
2
and 7.66 ± 1.30 mC/cm
2
, respectively, while the required
charge was 12.5 ± 6.2 nC and 19 ± 3.3 nC.

1

Chapter 1

Introduction

1.1 Visual Prostheses History and Background

The sensation of visual perception can be elicited by the electrical stimulation
at any one of multiple parts of the visual pathway. In 1929 Foerster exposed the
occipital pole of a subject with normal vision and was able to electrically stimulate a
region to create localized percepts termed phospenes that were spatially correlated
with the visual field (Foerster, 1929). A similar experiment was repeated in 1932 by
Krause and Schum in a subject that had suffered a gunshot wound in the area of the
left optic nerve leaving him blind for eight years (Krause & Schum, 1931). The fact
that localized phospenes could be elicited in a subject deprived of visual stimulus for
such a long period of time was significant. It was not obvious that the cortex could
create the sensation of perception after a sustained period of visual deprivation.
Brindley and Lewin were then the first to use multiple surface electrodes to stimulate
the striate cortex (also known as primary visual cortex or V1) (Brindley & Lewin,
1968). They used an array of 80 electrodes implanted in contact with the occipital

2
pole of the right cerebral hemisphere to elicit percepts in a 52 year-old subject
blinded due to retinal detachment. Each electrode was addressed via individually
connected radio receivers; specific receivers were activated by pressing a
transmitting coil of an oscillator tuned to the desired frequency on the scalp. This
was the design of the first chronically implanted prosthesis.
While surface stimulation provided significant initial results there were a
number of drawbacks. It was found that excitation in one area could create two
spatially separated phosphenes and that supratheshold excitation could create
percepts that could last for up to two minutes and also cause pain due to meningeal
stimulation (Brindley & Lewin, 1968). Electrodes placed closer than 2.4 mm apart
created phospenes that interacted with eachother (Karny, 1975). In the time domain
there was no flicker fusion frequency (Dobelle & Mladejovsky, 1974). And, perhaps
most importantly, the high thresholds limited the minimum electrode size. (1-3 mA
was required for a 1 mm
2
surface electrode (Dobelle et al., 1976).)
More recent attempts at developing a cortical prosthesis have abandoned
surface stimulation and focused on intracortical stimulation with penetrating probes
(Dobelle et al., 1976; Schmidt, 1992; Normann, 1999). In one study 38
microelectrodes were implanted near the right visual cortex in a 42 year old woman
who had been blind for 22 years due to glaucoma (Schmidt et al., 1996). Individual
phospenes were able to be disseminated using electrodes spaced 500 µm apart, a
factor of five improvement compared with surface stimulation. An excellent
experiment compared the threshold for stimulation using surface electrodes and 37.5

3
mm long intracortical electrodes in the same patient, and found that intracortical
stimulation had a threshold 10-100 times lower than surface stimulation (Dobelle et
al., 1976).
While significant advances have been made in cortical stimulation an FDA
approved device is yet to be developed. This is due to the increased complexity of
neural wiring in visual cortex and the dangers of surgical implantation. (In some
cases, epileptic seizures have occurred during stimulation (Kotler, 2002).) Because
of this electrical stimulation in retina has attracted more interest recently (Humayun
et al., 1996; Chow & Peachey, 1998; Zrenner et al., 1999; Rizzo et al., 2001).
Postmortem anatomical studies of patients with RP (Stone et al., 1992;
Santos et al., 1997) and AMD (Kim et al., 2002b) found that both inner retinal cells
(78.4%) and ganglion cells (29.7%) may be preserved even though there is
considerable photoreceptor cell death. It was first shown using corneal stimulation
that a percept could be elicited in an RP patient (Potts & Inoue, 1969). This lead to
further studies in which phosphenes were elicited by direct retinal stimulation of still
viable cells of the ganglion cell layer and inner retina (Humayun et al., 1996;
Humayun et al., 1999). Two different approaches have been taken to retinal
stimulation using an electrode array: subretinal and epiretinal.
A subretinal approach places an active or passive device in between the outer
and inner retina, and has been the approach taken by a number of research groups.
One group has implanted a silicon array consisting of 5000 microelectrode-tipped
microphotodiodes in the right eye of six patients with advanced RP (Chow &

4
Peachey, 1998; Chow et al., 2004). The aim of these implants is to convert incident
photons to a stimulating pulse via the photoelectric effect. While improvements
were reported in the detection of brightness, contrast, and shape, much of it was in a
region that did not correspond with array placement. Possible neurotrophic factors
induced by stimulation could be a factor. It is yet uncertain whether or not the
reported 8-12 nA pixel current induced over a 9×9 iridium electrode can depolarize a
cell.

Figure 1.1 Schematic illustrating sub and epi-retinal protheses approaches.

Another subretinal effort consists of an active device with 16, 200 µm
2
pads
that uses a transchoroidal polyimide coil to power the device (Zrenner et al., 1999;
Kuttenkeuler et al., 2006). Two such active devices have been implanted near the
foveal rim of 2 RP subjects that are able to discriminate vertical bars from horizontal

5
bars. It remains to be seen how well the device will be tolerated after >24 months of
implantation.
While the aim of a subretinal implant is to excite cells earlier in the retinal
pathway so as to preserve some processing (e.g., the bipolar-ganglion cell synapse),
recent work has shown that the degenerate retina undoes gross anatomical
remodeling in the inner retina which affects the both the connections these cells
make as well as the viability of the cells themselves (Marc, 2003).
Epiretinal prostheses have a longer history of successful chronic implantation
and partial restoration of visual function compared with subretinal implants.. An
4×4 array of 250-500µm diameter Pt electrodes was implanted in six patients
(Humayun et al., 2003; Humayun et al., 2005). The implant allowed for the subjects
to discern ambient light and simple shapes using a head-mounted camera and an
inductive link that transmitted both power and data. Visual psychophysics suggests
that a more densely packed electrode array with smaller sites that can address fewer
groups of cells will increase resolution. One study created pixelized images that
would correspond to 4×4, 6×10, and 16×16 electrode arrays and used sighted
subjects to show that a 16×16 array might allow patients to read 36 font type (Hayes
et al., 2003). Using a similar vision simulator, another study showed subjects could
to navigate through a maze with obstacles using a 25 × 25 array of pixels
representing the visual field (Cha et al., 1992).

6

Figure 1.2. A schematic representation of an epiretinal prothesis with external camera. Data and
power are wirelessly transmitted via an inductive link; the receiver is surgically implanted behind the
patient’s ear. A flexible electrode array is proximity coupled to the ganglion cell layer.

Since a threshold charge is required to elicit a percept and because there are
charge density limits imposed by electrochemical safety (as will be discussed in
chapter 4) and tissue heating considerations as well as challenges associated with the
routing of many leads per unit area, a minimum electrode size is defined. Current
implants employ 250-500µm diameter electrodes. Much of the work in this thesis
concerns electrochemical characterization and physiological response as a function
of decreasing electrode size so that we may better understand the limitations placed
on and advantages afforded by high-spatial density stimulating arrays.

7
1.2 Brief Review of Relevant Retinal Anatomy and Physiology

1.2.1 Vertebrate Retina
The vertebrate retina is divided into three layers containing neural cell bodies
(the outer nuclear, inner nuclear, and ganglion cell layers) and two layers containing
the synaptic interconnections between these cells (the outer and inner plexiform
layers) (Rodieck, 1998). Although a cross-sectional view of the retina with a light
microscope may give one the impression that the plexiform layers consist of tangled
braids of synapses, in reality they consist of “an ordered stack of synaptic planes,
more like a club sandwich than a plate of spaghetti.” (Masland, 2001) This highly
laminar and organized structure is reflective of the sophisticated parallel processing
that is done in the retina. A thorough treatment of the complex structure and
function of the various retinal cell types is out of the scope of this thesis, and,
therefore, a relatively brief description will be given here.
Photoreceptors (rods and cones) line the retinal pigment epithelium (RPE) at
the back of the eye and convert incoming photons to a neurochemical signal. The
RPE regulates the flow of oxygen and nutrients to the outer retina, which is the most
metabolically active tissue in the body. The fact that the retina is organized such that
light must pass through multiple cellular layers before being absorbed by the
photoreceptors is most likely due to this high metabolic rate (e.g., rods and cones are
positioned near a bed of blood vessels which would occlude light were retinal
organization reversed). Cones are responsible for color vision in daylight conditions

8
(photopic), and rods are responsible for vision in dim light (scotopic). In the most
basic sense incoming photons create a graded voltage difference across the
membrane of the photoreceptor via a phototransduction cascade. There is only one
type of rod, however, in primates, there are three types of cones (L, M, and S) that
have maximal sensitivity at different wavelengths within the visible spectrum (~400-
700nm). The retina compares the responses of these different cone types in order to
create a color image.

Figure 1.3. (Left) A photomicrograph of a retinal cross-section shows the highly laminar structure.
From top to bottom, rods and cones, outer limiting membrane (OLM), outer nuclear layer (ONL),
outer plexiform layer (OPL), inner nuclear layer (INL), inner plexiform layer (IPL), ganglion cell
layer (GCL), inner limiting membrane (ILM). (Courtesy http://webvision.med.utah.edu/.) (Right)
SEM image of retinal cross section of a bovine eye. (Courtesy Adrian Rowley.)

In contrast with most other neurons in response to a stimulus, rod and cone
photoreceptors hyperpolarize when stimulated. In the dark these cells continuously
release glutamate (the major excitatory neurotransmitter in the retina) at the rod-

9
bipolar and cone-bipolar cell synapse. The absorption of a photon causes a decrease
in the amount of glutamate that is released, and causes either bipolar cell
hypolarization (OFF pathway) or depolarization (ON pathway) (Rodieck, 1998).
Bipolar cells also respond to this input in a graded fashion and provide excitatory
input onto ganglion cells. In this same inner retinal layer amacrine cells provide
largely inhibitory input onto bipolar and ganglion cells. Ganglion cells receive direct
input from both of these cell classes, which serve to modify the firing rate of the cell.
It is this rate at which action potentials are initialized in the ganglion cell that
provides the information coding for all aspects of the visual field. In this, within the
context of an epi-retinal prosthesis, it becomes theoretically possible to recreate this
code via electrical stimulation if (1) we understand how the retina codes for an
arbitrary visual stimulus and (2) we can individually address each ganglion cell.

1.2.2 Remodeling in Degenerate Retina
It is important to understand the stages of retinal degeneration and the
associated anatomical and physiological changes that occur. A comprehensive study
is has been done by Marc et al., in which three phases of degeneration and
remodeling are classified (Marc, 2003). In the first two phases photoreceptor stress
and death and associated loss of tropic transport are observed. Both bipolar and
horizontal cells can actually retract dendrites while the later can sprout axonal and
dendritic processes that can reach the inner plexiform layer. Muller cells can form a
dense fibrotic layer and seal off the subretinal space, electrically isolating implants

10
placed there via the choroid. In phase 3 the number of viable cells of all classes are
depleted. Bipolar and amacrine cells can migrate up to the ganglion cell layer and
undergo neural rewiring.
Such anatomical changes manifest physiologically. Using a patch clamp
technique in a degenerate mouse model, it has been shown that rod bipolar cells lose
their sensitivity to the excitatory neurotransmitter glutamate while they increase their
response to the inhibitory horizontal cell neurotransmitter GABA (Varela et al.,
2003). These findings strongly undermine retinal prosthesis efforts that aim to excite
inner retinal cells in order to preserve some retinal processing.

1.3 Retinal Disease and Treatment

Based on a 2000 U.S. census it is estimated that age-related macular
degeneration (AMD) affects 1.75 million people over 40, and that by the year 2020
this population will grow to 3 million people (Klein et al., 1997; Friedman et al.,
2004). There are two types of AMD. Dry AMD is due to hypertropic and atropic
changes that occur in the RPE behind the macula as well as deposits that accumulate
there. Wet AMD is caused by choroidal neovascularization (CNV) and can cause
rapid vision loss due to the leakage and rupture of these fragile vessels near the
macula (Rakoczy et al., 2006). Of the two types dry AMD is more prevalent
constituting 80% of all cases.

11
Retinitis pigmentosa is a family of genetic disorders with an incidence of
1/4000 in the United States (Berson, 1993). The loss of photoreceptor viability may
my affect either central or peripheral vision first depending on whether the dystrophy
is of the cone-rod or rod-cone variety. Taken together, AMD and RP are the leading
causes of retinal blindness.
While there has been recent progress in a drug treatment for wet AMD
(Genentech, 2006), both RP and the more common dry AMD still have no available
treatment. There are a number of rescue strategies that are being investigated to
restore some level of vision to afflicted patients. Gene therapy has shown success in
treating animal models of Leber’s disease, a disease that causes retinal ganglion cell
(RGC) and axonal degeneration (Preising & Heegard, 2004). However, each genetic
mutation of the disease would require an individualized therapy. Retinal fetal sheet
transplantation, which is aimed at replacing a degenerate PR layer, has been limited
by the number of useful synaptic connections that is established between the donor
and host tissue (Lund, 2001), although recent work has shown that the number of
viable connections made may be maximized if donor cells are taken from the
developing retina during the peak of rod genesis (MacLaren et al., 2006). There has
also been interest in expressing channelrhodopsin-2 in ganglion cells so that they,
themselves, are light sensitive. While success has been shown in vivo in degenerate
model rats, the channel is expressed by using a viral vector which may pose threats
of it’s own (Bi et al., 2006).

12
1.4 Neural Excitation

1.4.1 Circuit representation of a neuron
A neuron has a resting membrane potential of approximately -70 mV with
respect to the extracellular space. The energy required to keep this potential
difference across the cell membrane comes from ATP driven ion pumps which keep
the a higher K
+
concentration inside the cell and a higher Na
+
outside of the cell
(Kandel et al., 2000). At rest the cell membrane is 100 times more permeable to
potassium than sodium due to open potassium leak channels. Therefore, more
potassium ions are free to move down the concentration gradient compared with
sodium ions. (Hence, the negative resting potential.) When a cell is electrically
stimulated by an adjacent neuron, voltage gated ion channels temporally open and
allow for an influx of Na
+
and a delayed efflux of K
+
. The cell is depolarized and an
action potential is elicited that regeneratively propagates down the cell axon.
Stimulation using external electrodes differs from this situation since both the current
source and sink are outside of the cell membrane. In order to depolarize the cell, the
cell membrane, which can be modeled as an RC circuit with associated time
constants, must be charged sufficiently. If the potential gradient induced across the
membrane by the external source is above threshold, an action potential is elicited.

13

Figure 1.4. Circuit diagram of a cell membrane includes conductances and electrochemical driving
forces associated with sodium, chloride, and potassium ions channels in parallel with a sodium-
potassium pump and the bilipid membrane capacitor (modified from Kandell).

Neuromuscular excitation has been studied extensively over the past century,
and, therefore, the theory on excitation of myelinated and unmyelinated fibers has
been well developed (Ranck, 1975). However, since retinal stimulation targets the
cell soma in order to produce punctate phospenes that are spatially correlated with
the visual field, theory on the stimulation of cell soma will be reviewed. (The
possibility of axonal stimulation will be discussed in chapter 6.)

1.4.2 Cellular excitation
The extracellular excitation of a cell is complex because the relationship
between the transmembrane voltage and induced current is nonlinear (Struijk, 2004).
During the initialization of an action potential, the opening of a threshold number of
sodium channels begins a cascade of many more channels opening regeneratively.
When using a large stimulating disc electrode this complexity is compounded by the

14
fact that the local spatial voltage gradient is not homogeneous; field fringing at the
edges of the disc will induce greater gradients compared with the disc center. For
these reasons, the analytical treatment of extracellular stimulation has been done
only by making simplifications in order to yield a tractable problem. These
solutions, nevertheless, give us important insight into the physical basis of neural
excitation.

1.4.3 Derivation of the Strength-Duration relationship
For a passive spherical cell with a penetrating current carrying wire, the total
current, I
S
, is the sum of the capacitive component (displacement current) and the
resistive component (ionic current). (Note: Although in retinal stimulation the
current source (i.e., the stimulating electrode) is extracellular, important neural
stimulation parameters (i.e., the rheobase and chronaxie) will be derived here by
pacing the source within the cell. This derivation closely follows that presented by
Struijk (Struijk, 2004).)

!
I
S
=I
C
+I
I
=C
m
"V
m
"t
+
V
m
R
m

For a current pulse of length T, the time
dependent transmembrane voltage will be

!
V
m
(t) =I
S
R
m
(1"e
"t/#
m
)

defined for
!
0"t"T, where the membrane
time constant
!
"
m
=R
m
C
m
.
Figure 1.5. Schematic of a spherical cell
intracellularly injected with current I
S

15

In order to reach the threshold voltage for depolarization,
!
V
th
, a threshold amount of
charge,
!
Q
th
, is required. However, since excitation requires the charging of the
capacitive cell membrane with time constant
!
"
m
, the cell also discharges over time.
Therefore, even for an infinitely long pulse, there is a minimum current amplitude
required, termed the rheobase. (i.e., since the membrane does not hold charge
indefinitely, the rate at which charge is delivered has a minimum.)

!
lim
t"0
V
m
(t) =V
th
=lim
t"0
(I
rh
R
m
(1#e
#t/$
m
)) =I
rh
R
m

or
!
I
rh
=
V
th
R
m

when
!
t =T and at the threshold current we obtain the strength-duration relationship
(
!
I
th
vs. T).
!
I
th
=
I
rh
1"e
"T /#
m

A parameter that is often referenced is the chronaxie that is defined to be the pulse
duration that corresponds to twice the current amplitude of the rheobase. Also,

!
Q
th
=
TI
rh
1"e
"T /#
m

16
We note that as
!
T"0, the threshold charge reaches a minimum indicating that
shorter pulses are more charge efficient for stimulation. In addition, with regard to
retinal stimulation, it has been shown with modeling using NEURON software
(Greenberg et al., 1999) and with patch clamp recording (Fried et al., 2006; Margalit
& Thoreson, 2006) that shorter pulse lengths preferentially excite ganglion cells and
longer pulses may target bipolar cells due to their longer time constants.

1.4.4 Spherical cell in a uniform electric field
Studies on red blood cells in 1923 concluded that a cell consisted of a well-
conducting interior surrounded by a comparatively poorly conducting membrane
(Fricke, 1923, 1924; Cole, 1933). If we exclude charge sources, Laplace’s equation
can be applied

Figure 1.6. A spherical cell in a uniform
field

!
"
2
V =0

For the case of a nonconducting sphere
in a medium of conductivity σ and
uniform electric field
!
E
"
, we can use
Laplace’s equation in polar coordinates

!
"
2
V =
1
r
#
#r
(r
2
#V
#r
) +
1
r
2
sin$
#
#$
(sin$
#V
#$
) =0

17
subject to the boundary conditions
!
"V
"x
#$E as
!
x" ±# and
!
"V
"r
=0 for
!
r =a.
The only solution to this differential equation is
!
V =Arcos" +Bcos"/r
2

with
!
A ="E and
!
B ="
a
3
E
2

At the surface of the sphere
!
V ="
3Ea
2
cos#

Note that for when the field is perpendicular to the membrane the transmembrane
voltage is maximized. Also, for a given field strength a larger diameter cell will
have a higher induced transmembrane voltage, meaning that these cells have lower
charge thresholds.

1.5 Structure of this thesis

This thesis is primarily focused on the electrode-tissue interface for a retinal
prosthesis. The order of the chapters also tracks the process of extracellular
stimulation from the electrode to the electrochemical transfer of charge at the
electrode-electrolyte interface to the excitation of ganglion cells and other retinal
cells within the inner retina.

18
In chapter two the background of multielectrode technology for the purposes
of stimulating and recording neural responses is presented. Such MEAs have been
designed and fabricated for use in hippocampal and retinal electrophysiology
experiments. (The former comprises work done in conjunction with Dr. Theodore
Berger’s and Dr. Armand Tanguay’s research groups). The benefits of electrode
arrays that are conformal to the hippocampal cytoarchitecture are presented. The
application of this flexibility of design to custom retinal arrays is discussed within
the context of near-term epiretinal implants. These MEAs were employed for all of
the retinal electrophysiological experiments presented in chapter 5. The
microfabrication process flow is described in detail, including the electrodeposition
of high-surface area Pt using an ammonium hexachloroplatinate bath to increase
electrode charge delivery. (The ammonium hexachloroplatinate chemistry has been
developed by Dr. John Whalen. (Whalen et al., 2005))
The MEA-tissue interface is presented as an RC network; modifications that
have been made to the array fabrication to minimize the shunt capacitance and
increase the recorded neural response amplitude are discussed. Finally, results of
hippocampal ‘CA3 Replacement’ proof-of-concept experiments employing cMEAs
as the interface between FPGA/VLSI hardware and rat hippocampal slice are
presented.
In chapter three, evidence is given for an inhomogeneous (primary) current
distribution at the disc electrode surface, and the associated adverse effects with
regard to electrode corrosion are discussed. The analytical treatment of the primary

19
current distribution is reviewed. Finite-element modeling is used to model the
current density distribution at the electrode surface of recessed microdiscs, and such
recessed microdiscs are fabricated using SU-8 epoxy-based photoresist.
In chapter 4 the electrode-retina interface is represented electrochemically.
The results presented here extend the electrostatic modeling discussed in chapter 3 in
the time domain since all data is experimentally obtained and subject to
electrochemical diffusion, convection, and migration. Cyclic voltammetry (CV),
current pulsing, and electrochemical impedance spectroscopy (EIS) are introduced as
means of characterizing the electrode. Studies on electrodes of varying diameter
allow for investigation into evolution of the current density profile at the electrode
surface as a function of time. The investigation of the dependence of delivered
charge and electrode impedance on size is relevant to future retinal prostheses since
increased spatial resolution will require smaller electrodes to individually address
fewer cells (Humayun, 2001; Palanker et al., 2005)
In chapter 5 all methods and results of isolated retina electrophysiology
experiments using custom MEAs are presented. These experiments were setup to
mimic, as closely as possible, the electrode-retina interface in a retinal prosthesis
subject, and gain insight into which elements are excited during different stimulus
protocols. The pros and cons of the chosen animal model (tiger salamander) are
discussed. The experimental setup and process flow are described. This includes the
MATLAB code used for spike detection and plotting of post-stimulus time
histograms (written with Michelle Hauer). All pharmachological agents used for the

20
purposes for isolating direct ganglion cell excitation are listed. The experiments
allowed for investigation of (1) the radial extent of excitation for varying stimulus
strength and pulse duration, (2) the frequency dependence of excitation of ganglion
cells vs. inner retinal cells, and (3) the dependence of threshold charge and charge
density on electrode size.
In chapter 6 the pros and cons of isolated retinal electrophysiology will be
discussed. Multiple wild conjectures will then be recklessly posed with regard to
elicited percepts in patients based on results presented in chapter 5. Suggestions for
future work will also be presented.

21

Chapter 2

Design and Microfabrication of Multielectrode Arrays

2.1 Multielectrode array background
Single cell recording techniques such as patch clamp recording and
extracellular wire recording have given us valuable insights into single cell
dynamics. However, the central nervous system uses the concerted firing of many
cells to relay and process information (Meister et al., 1995; Brivanlou et al., 1998;
Fries et al., 2001; Shlens et al., 2006). In order to study physiological activity of
neuronal ensembles a number of research groups developed various planar and
penetrating multielectrode array technologies for use in in vitro (Gross et al., 1985;
Novak & Wheeler, 1986; Wheeler & Novak, 1986; Normann, 1999; Wagenaar &
Potter, 2004) and in vivo (Normann, 1999; McCreery et al., 2002) applications.
Planar arrays generally consisted of rows and columns of disc electrodes that formed
a rectangular pattern. These arrays consisted of 60 electrodes spaced by 50-200µm
and interfaced with commercially available preamplification boards and data
acquisition systems (MultichannelSystems, Reutlingen, Germany; Plexon Inc.,
Dallas, Texas; Alpha Med Sciences, Osaka, Japan). The number of electrodes was

22
defined by lead routing and data acquisition limitations. Also, since these are
commercially available, only 1-2 versions of the array are available and the full
capabilities of photolithography are not exploited. Here we describe conformal
multielectrode arrays (cMEAs), which are designed to mimic the cytoarchitecture of
the cortical region being studied. This technology was developed to assist in studies
of hippocampus, towards a cortical prosthesis for that region (Berger et al., 2005)
(This is the only chapter in which results pertaining to hippocampal
electrophysiology will be discussed.) The flexibility afforded by the technique was
then directly used for the design of custom retinal MEAs allowing for all of the
experiments discussed in chapter 5.

2.2 Microfabricated electrode arrays for hippocampal and retinal
electrophysiology

2.2.1 Trisynaptic and Perforant Path MEAs
The hippocampus is a region of cortex that has a role in the conversion of
short-term memory to long-term memory and in spatial navigation (Milner, 1970).
Input from entorhenial cortex enters the trisynaptic pathway of the hippocampus
(Dentate Gyrus (DG) – CA3 – CA1) via the perforant pathway (PP) (Kandel et al.,
2000). In order to electrophysiologically characterize these three subregions and the

23
nonlinear interaction between them, two different cMEAs were fabricated, both
using the Si
x
N
y
-SU-8 dual insulation process flow as described in section 2.2.
The perforant path cMEA consists of a 3 × 20 array of either 28 µm diameter
Pt disc electrodes (616 µm
2
) or oval electrodes with twice the geometric area
(1232 µm
2
), and was designed in order to target neural pathways in close proximity
(i.e., lateral and medial perforant pathway in DG). The 50 µm spacing between these
two afferent pathways makes it difficult to consistently electrically stimulate
individual pathways with conventional glass probes. This array allowed for the
study of the nonlinear dynamics in PP and the effects of 17beta-Estradiol (estrogen)
on pair-pulse facilitation (Kim et al., 2006). Specifically it was found that 17beta-
estradiol potentiates synaptic transmission in each subfield of the hippocampal slice,
with the greatest magnitude of potentiation at the associational/commissural fibers in
CA3. This may have clinical implications on the association between estrogen
levels and learning and memory.
The ‘Trisynaptic’ cMEA, as shown in Figure 2.1, includes nine sets of seven
linearly-spaced 28 µm diameter pads with a 50 µm center-to-center spacing, each set
spanning one of the key input/output regions of the DG, CA3, and CA1 regions of rat
hippocampus, thereby allowing for a complete diagnostic assessment of the
nonlinear dynamics of the trisynaptic hippocampal circuitry (Gholmieh et al., 2006).

24

Figure 2.1 Microphotographs of conformally-mapped multielectrode arrays (cMEAs). (Left)
Perforant Path cMEA consisting of Pt disc (616 µm
2
) and oval (1232 µm
2
) electrodes. (Right)
Trisynaptic’ cMEA aligned with an acute hippocampal slice, showing a trisynaptic placement of nine
subarrays distributed over all three subfields, i.e., DG (left), CA3 (right), and CA1 (top).

2.2.2 Retinal custom MEAs
A near-term retinal prosthesis faces the challenge of replacing 100 million
photoreceptors and all of the early visual processing that is done in the outer and
inner retina with 16-1000 large electrodes. Although ganglion cells code for
brightness and contrast by varying firing frequency, a prosthesis may also vary the
amplitude of the stimulus pulse to code for such features. An increase in intensity,
however, will also increase the excitation area and variables such as brightness and
phosphene size may be inherently coupled. Understanding how electrical stimulus
threshold, varies for neurons at different radial distances from stimulating pads will
allow us to better understand this relationship as well as provide insight into visual
psychophysical results with retina implant patients.

25
Custom MEAs were designed and fabricated to allow for a study of the
relationship between excitation area and distance. The MEAs have large stimulating
pads (75-200µm diameter, 500 and 150µm center-to-center spacing, respectively)
and smaller recording electrodes (8-10µm diameter) that are arranged with radial
symmetry with respect to the recording sites. These MEAs were fabricated using the
multilayer Si
x
N
y
insulation process flow as described in section 2.3. This procedure
allowed for the relatively small 8µm diameter pads to be photolithographically
defined. This could not be achieved with the more viscous SU-8 photoresist that was
employed for all other fabricated MEAs

Figure 2.2. Photomicrographs of two different custom Pt retinal multielectrode arrays. (Left) 200
µm diameter Pt electroplated stimulating electrodes with 10 µm diameter stimulating/recording
electrodes. (Center) 75 µm diameter Pt stimulating electrodes with 8 µm diameter
stimulating/recording electrodes. (Right) 200 µm array with tiger salamander retinal slice.

26
2.3 Microfabrication Process Flow

2.3.1 MEAs with Au or Pt leads and electrodes and Si
x
N
y
/SU-8

insulation
All photolithographic masks were laid out using Tanner L-Edit CAD
software (Tanner Research, Pasadena, Ca). Float glass substrates were first cleaned
using acetone and methanol and dried under streaming nitrogen gas. Clarion 5214
positive/negative photoresist (Clariant Inc. Long Beach, Ca) was spin coated at 3000
rpm using HMDS adhesion promoter. Photoresist was softbaked at 120°C to remove
excess solvent, followed by photomask alignment and subsequent UV exposure (42
mJ, λ = 365 nm) to define the underlying trace pattern. A 45 second hardbake was
then performed at 120°C, followed by development in 4:1 diluted AZ5214 chemical
developer for 20-30 seconds.
Polycrystalline Pt (99.999%, Int. Adv. Mat., Spring Valley, NY) (200nm, 5
A/sec) was deposited on a Cr adhesion layer (30nm, 2 A/sec) using electron beam
evaporation (pressure – 1 × 10
-7
Torr). Following deposition, substrates were
immersed in acetone overnight to perform photoresist liftoff. After the Pt traces
were defined, the insulation layer was deposited in one of the following two ways:

1. A 300nm thick Si
x
N
y
insulating layer was deposited by plasma-enhanced
chemical vapor deposition (PECVD) at 275° C (30W, 20 sccm (standard cubic
centimeters per minute) Silane, 40 sccm NH
3
, 60 sccm N
2
, 350 mT). Substrates
were cleaned using acetone and methanol, washed with DI water and dried with

27
streaming nitrogen gas. SU-8 2001 epoxy-based photoresist was spun on at 3000
rpm to give a 1.5 micron thick film . As softbake was done at 65° C for 4 minutes
and then brought up to 95° C for 1 minute. (The temperature is ramped up to
minimize stresses induced in the thick SU-8 layer.) A second photolithographic step
defined electrode vias in the negative SU-8 resist. (68 mJ, λ = 365 nm). A hardbake
was done at 65° C for 1 minute and then brought up to 95° C for 4 minutes.
Chemical development was done using SU-8 Developer for 1-3 minutes.
Since SU-8 is an extremely durable resist it provided protection of electrode
leads during the subsequent reactive ion etch (RIE) of the underlying Si
x
N
y
using
CF
4
gas (100W, 200mT).

2. A 1µm thick Si
x
N
y
insulating layer was deposited by plasma-enhanced
chemical vapor deposition (PECVD) at 275° C (30W, 20 sccm Silane, 40 sccm NH
3
,
60 sccm N
2
, 350 mT). This was separated into three different depositions in order to
relieve thermal stresses induced in the film. (i.e. the temperature was brought back to
25° C between depositions.) A second photolithographic step was used to define
vias to the underlying Pt electrodes with the same parameters as the first. Si
x
N
y
,
masking the electrodes and contact pads, was reactive ion etched (RIE) using CF
4
.

To ensure that no organic residue was left on the thin-film microelectrodes, a
final oxygen ashing procedure was performed on all microelectrode arrays. 2-point

28
electrical resistance measurements were sometimes used to confirm Pt surface
exposure.

2.3.2 Alternate process flow for Indium-Tin Oxide conductive leads
Indium-tin oxide (ITO) is a transparent conductor and has been used for the
fabrication of MEAs when it is important to visualize cell location and growth on the
array (Gross et al., 1985). Such ITO based MEAs have been fabricated for both
hippocampal culture and slice studies (Figure 2.1). The alternate process flow for
these devices is as follows (Figure 2.3).
ITO coated float glass substrates (Delta Technologies, R
sheet
= 15-20
Ω/square) were first cleaned using acetone and methanol and dried under streaming
nitrogen gas. A single Si
x
N
y
insulating layer was deposited by PECVD at 275° C
(30W, 20 sccm Silane, 40 sccm NH
3
, 60 sccm N
2
, 350 mT) that served as a
protective layer against the subsequent ITO acid etch. The first photolithographic
step and chemical development was done as described above to define electrode
leads. The Si
x
N
y
layer was reactive ion etched using CF
4
(100W, 200mT). In order
to etch out the underlying ITO and hence define the conductive leads, an
HNO
3
:HSO
4
:H
2
O bath was used. The entire substrate was placed in the bath for
approximately 20 minutes. Full etching was confirmed with optical microscope and
an ohmmeter.
The final Si
x
N
y
layer was deposited and the second photolithographic step
and chemical development was done to open electrode vias as described above.

29

Figure 2.3 MEA microfabrication process flow using ITO transparent conductive leads and Si
x
N
y
/SU-
8

insulation.

2.4 Improved SNR and stability of dual Si
x
N
y
/SU-8

insulation

The tissue-MEA interface can be represented as an RC network (Figure 2.4).
For the purposes of this discussion, the electrode-electrolyte impedance can be
modeled using discrete circuit elements. While the details of the electrode-electrolyte
interface are more complex and will be discussed in detail in chapter 4, here we will
focus on the importance of decreasing the shunt capacitance (C
S
) in order to increase

30
the signal-to-noise ratio of recorded responses. This proved to be especially
important in hippocampal slice experiments in which field evoked post-synaptic
potentials (fEPSPs) were recorded from a population of cells. In contrast with all-or-
none action potentials (‘spikes’) generated by single retinal ganglion cells, fEPSPs
contain information in both amplitude and interpulse spacing, and, therefore, it is not
sufficient to merely detect signals above threshold. For this reason the employment
of a dual insulation layer proved to be absolutely critical in characterizing the
nonlinear dynamics between different subregions of hippocampus.

Figure 2.4 An RC circuit representation of the MEA-tissue interface. Here V
S
is the neural source
signal, R
S
is the solution resistance, R
P
is the polarization resistance, and C
dl
is the capacitive double-
layer. The later three comprise the electrode-electrolyte interface. (Modified from Gross et al., 1995)

While silicon nitride is an excellent ceramic insulator with good adhesion to
glass, it is difficult to deposit stable thick films (>2 µm) with few pinholes. Figure
2.5 shows an electrode insulted with a 0.2 µm thick layer of silicon nitride that has

31
been electroplated with Pt using an ammonium hexachloroplatinate bath. The fact
that the insulated lead has been plated is due to pinholes in the thin-film. The fact
that this film does not prevent the reduction of Pt ions on the electrode lead would
suggest that neurons above the leads could similarly be excited or recorded from.
SU-8 is an epoxy-based negative photoresist that is available in a range of
viscosities so that it can be spun on at different thicknesses and high aspect ratios.
While it is a durable, non-toxic material it has poor adhesion to float glass as can be
seen in figure 2.5. Here an MEA is shown that has only been coated with SU-8 2000
(1.5 µm) and has been soaked in phosphate buffered saline and pulsed with cathodic-
first biphasic pulses (0.1 mC/cm
2
) for 72 hours. It is clear that the film readily
delaminates in these conditions.

Figure 2.5 (Left) An SEM image of a Pt electroplated electrode and silicon nitride insulated lead.
(Right) A photomicrograph of an MEA that has been pulsed at 0.1 mC/cm
2
in PBS for 72 hours.

A Si
x
N
y
/SU-8 dual insulation was developed to combine the advantages of
these individual layers. This technique yields a thermally and mechanically stable

32
insulation layer with good adhesion to the underlying substrate. The SU-8 seals up
pinholes in the underlying silicon nitride and increases the thickness of the insulation
to decrease the shunt capacitance. Figure 2.6 shows an MEA in which some of the
electrodes have been plated with Pt and been soaked in PBS and pulsed as above.
This array does not show the delamination or electroplating of electrode leads as was
seen in arrays with only one of the two layers.

Figure 2.6 (Left) A Si
x
N
y
/SU-8 dual insulation layered MEA with five electroplated Pt electrodes.
(Right) Recorded EPSPs and population spikes from DG of a rat hippocampal slice with (left) and
without (right) dual insulation. In this case the dual insulation allows for approximately a factor of
two improvement in SNR (i.e., 600µV to 1300µV).

2.5 CA3 Replacement in Hippocampus: In Vitro proof-of-concept

This microfabrication process flow employing an Si
x
N
y
/SU-8 dual insulation
layer allowed for the fabrication of ‘CA3 Replacement’ cMEAs that served as the
interface between FPGA/VLSI hardware and hippocampal slice in seminal CA3
replacement experiments (Figure 2.7) (Ahuja et al., 2004). These experiments are an

33
initial step in developing a neural prostheses aimed at replacing damaged neurons in
the hippocampus with a biomimetic system comprised of silicon neurons (Berger et
al., 2005). Such an implant could serve as a biomedical remedy for the cognitive and
memory loss that accompanies Alzheimer's Disease and the speech and language
deficits that result from stroke.

Figure 2.7 Schematic representation of ‘CA3 Replacement’ of damaged hippocampus with VLSI
biomimetic hardware (Courtesy Dr. Theodore Berger)

The ‘CA3 Replacement’ cMEA included two different circular pad sizes: 28
µm diameter pads were grouped in series to form sets of stimulating pads in (DG)
(three at a time) and CA1 (two at a time), while 36 µm diameter pads are used for
recording in DG, CA3, and CA1. By grouping sets of stimulating pads in series the
overall surface area was maximized to increase charge injection over previous
generations by a factor of three (i.e., ‘Trisynaptic’ cMEAs). The stimulating pads

34
were been placed only in DG and CA1 in order to interface with the FPGA/VLSI
hardware that replaced the CA3 region in these experiments.

Figure 2.8 (Left) The electrode pad placement was determined by averaging the trisynaptic
cytoarchitecture over ten hippocampal rat slices. (Right) Optical photomicrograph ‘CA3 Replacement’
cMEA. Pt stimulating (36µm diameter) and recording electrodes (28µm diameter) are arranged in both
the DG (lower left) and CA1 (upper middle) regions, while two sets of recording electrodes are provided
in the CA3 (lower right) region for pre-replacement characterization.

The major intrinsic circuitry of the hippocampus is preserved in 500 µm thick
slices cut transverse to the longitudinal axis. Input signals from perforant path (PP)
excite dentate gyrus granule cells, which excite CA3 pyramidal cells via mossy
fibers. The processed signal is then outputted to CA1 pyramidal cells by the
Schaffer collaterals. In order to capture the nonlinear dynamics of the CA3 region
sprague-dawley rat hippocampal slices were characterized using a
mutlichannelsystems stimulation amplifier, preamplification board, and data
acquisition software as well as ‘CA3 Replacement’ cMEAs. A random impulse train

35
(1200 pulses) was input into PP and the trisynaptic responses of DG, CA3, and CA1
were analyzed so that the nonlinear relationship between DG and CA3 could yield a
computational model that was used to program an FPGA. Then the connection
between CA3 and CA1 was transected by cutting the Schaffer collaterals; the same
random impulse train was then used to stimulate PP, and the output of DG was sent
to the FPGA and this output was then sent to CA1. In this way the ability of the
FPGA to replace function could be assessed by comparing EPSPs generated by CA1
output in these two cases.
The results of this experiment show excellent agreement between the
recorded EPSP amplitude in an intact slice compared with a transected slice with
CA3 Replacement hardware (Figure 2.8). The average normalized mean square
error (NMSE) is less than 10%.

36

Figure 2.9 (Left) System integration of an FPGA-based nonlinear model of the CA3 hippocampal region
with a hippocampal slice in which the output from the dentate gyrus (mossy fibers in the hilus) have been
transected, thus eliminating the normal propagation of activity from dentateCA3CA1. The FPGA-
based model bi-directionally communicates with the living slice through a ‘CA3 Replacment’ cMEA.
(Right) A comparison of hippocampal CA1 output in response to the FPGA model of CA3 with the
output of CA1 in response to the biological CA3. Examples of results from two random impulse train
experiments are shown in the two panels, respectively: amplitudes of population EPSPs recorded from
the CA1 region are shown as a function of 50 impulses chosen from among 2,400 impulses of the
random trains (1,200 administered before transecting inputs to CA3; 1,200 administered after
transection). Time intervals between impulses are not represented in the figures; only “Input Event”
number (sequence of sample impulses) is shown to “collapse” the x-axis. (Figure courtesy of Dr.
Theodore Berger)

37

Chapter 3

The Inhomogeneous Current Distribution

3.1 Prior evidence of inhomogeneous distribution

While it would be optimal for retinal prostheses to be able to address
individual ganglion cells independently there are number of reasons why current
implants employ arrays consisting of 16-60 electrodes with diameters ranging from
50-500µm. Photolithography and electron beam lithography allow for the
fabrication of even submicron electrodes, however, the routing of leads to a highly
dense array remains difficult, especially if limited to the plane of the electrode itself
(Weiland & Humayun, 2005). There are also electrochemical safety limits that are
imposed by the stimulating electrode material. Stimulation currents that exceed
these predefined limits can cause electolysis and change the local pH, inject
neurotoxic heavy metal ions nearby cells via metal dissolution (Brummer et al.,
1977), and cause electrode corrosion. Stimulation outside of this safe-limit,
therefore, adversely affects both the tissue as well the electrode.

38
Stimulation with disc electrodes has given evidence of an inhomogeneous
distribution at the electrode surface (also termed the primary distribution within
electrochemical literature). It was noticed that during electrosurgery (e.g., the
destruction of cancer cells with the use of high-frequency voltages) patients suffered
burns around the perimeter of the electrode (Overmyer et al., 1979; Shah & Webster,
1979). Cochlear implant electrodes and cardiac defibrillator electrodes have been
reported to show preferential corrosion at the electrode edge (Wiley & Webster,
1982a; Rubinstein et al., 1987). This implies that the perimeter of a disc electrode
exhibits a higher localized current density compared with the electrode center. This
is detrimental from the standpoint of maximizing charge injection per unit area while
still remaining below electrochemical safety limits: the current density at the
electrode edge may exceed these limits while the current density at the electrode
center is well below this value.

Figure 3.1 Schematic representation of current density profile for a disc electrode showing charge
accumulation at the electrode periphery.

39
Here we will show empirical evidence for preferential corrosion at the
electrode edge due to this distribution using energy dispersive spectroscopy of
current pulsed Au electrodes. The theory of the inhomogeneous current distribution
will then be reviewed. Then we will discuss recessed electrodes as a means of
engineering a more homogeneous current distribution by controlling the boundary
conditions for the electromagnetic field at the electrode surface. This will be
confirmed and quantified using the quasi-static electromagnetic module within
FEMLAB finite-element modeling software. Finally microfabricated recessed
microelectrodes will be presented using a variable thickness SU-8 insulation layer.

3.2 Evidence of inhomogeneous distribution using EDS

Cyclic voltammetric results of Au in H
2
SO
4
(as will be presented in chapter
4) show that anodic formation of gold oxide forms ca. 1.3, 1.55, and 1.7 V vs.
reversible hydrogen electrode (RHE). Although the aim of a biphasic pulse is to
reduce all oxidized species at the electrode surface so that there is no net charge
introduced (Lilly et al., 1955), a metal oxide layer will still be present after sustained
pulsing (Figure 3.2). This is due to unrecoverable charge in the form of irreversible
Faradaic reactions such as gas evolution and diffusion (Merrill et al., 2005). A
place-exchange process where an oxygen atom on the surface changes place with an
underlying metal atom in favor of a total lower energy may also contribute
(Angerstein-Kozlowska et al., 1989). In this case a thick oxide can form (i.e., many

40
monolayers thick) (Tremiliosi-Filho et al., 1992) which may limit the ability of a
pulse of the opposite polarity from reducing all of the oxide formed on the electrode
surface during the first half of the pulse.

Figure 3.2 Photomicrograph of an unpulsed (top) and a pulsed (bottom) polycrystalline Au electrode
(72 hours in PBS at 0.1 mC/cm
2
)

In EDS an electron gun targets the specimen to be studied. The generated x-
ray energy is dependent on the energy level spacing of the material itself. The local
elemental composition can be determined to within ~0.1 wt. % (Goldstein et al.,
2003). It is ideal for studying the oxide composition at the electrode surface because
inelastic scattering in the metal lattice allows only the atoms on the material surface
to contribute significantly to the detected signal. Because it is the electrode surface
that participates in Faradaic charge transfer, EDS provides relevant characterization.

41

Figure 3.3 Au and Au oxide surface composition data obtained from EDS longitudinal and lateral line
scans across a 200 µm diameter Au microdisc that has been pulsed for 72 hours in PBS at 0.1
mC/cm
2
.

Using metal oxidation as a marker for the local overpotential that was
induced during electrode pulsing, energy dispersive spectroscopy was employed to
quantify the Au
2
O
3
concentration as a function of the radial distance on the electrode
surface (Figure 3.3). Electron-beam evaporated thin-film polycrystalline Au
electrodes (200 and 244 µm diameter) were pulsed with biphasic current pulses using
a MultichannelSystems STG 8000 stimulus generator in phosphate buffered saline
(PBS) for 72 hours at 0.1 mC/cm
2
and 50 Hz. A large Pt wire mesh (Alpha Aesar,
99.99%) was used as a counter electrode. EDS data was obtained by scanning
longitudinally and laterally across the electrode. The average Au-oxide composition
by atomic weight at the electrode edge was 42.5±2.1% compared with 30.8±0.6% at
the center. In contrast, the Au composition by atomic weight at the electrode edge

42
was 41.9±3.0% compared with 58.1±1.2% at the center. As the percentage of
available Au sites decreases at the edges of the disc compared with the center, the
number of active sites that can participate in Faradaic charge transfer also decreases.
The use of EDS in determining the oxide content profile across the electrode surface
should be utilized in characterizing the effectiveness of yielding a more
homogeneous current distribution by means of varying pulse parameter (discussed in
chapter 4) or electrode recess (discussed in section 3.3.2).

3.3 Electromagnetic finite-element modeling of electrode current
distributions

3.3.1 Primary current distribution: disc flush with surrounding dielectric
The analytical solution for the primary current distribution on a disc electrode
can be derived by solving Laplace’s equation (Wiley & Webster, 1982a; Rubinstein
et al., 1987)
!
"
2
V =0

using the following boundary conditions for a disk of radius a sitting on a semi-
infinite slab with conductivity σ, as shown in figure 3.4.

!
V =V
0
when z = 0, r
!
" a

43
!
dV
dz
=0 when z = 0, r> a

and for
!
r"#,z"#.
The solution can be represented as a superposition of terms with the form
!
e
"k|z|
J
0
(kr)
!
V(r,z) = A(k)e
"k|z|
J
0
(kr)dk
0
#
$

where z is distance from electrode, k is the wave vector, and J
0
is the zeroth order
Bessel function. The solution for potential can then yield an expression for current
distribution using
!
E ="#V and
!
J ="E (Heitz & Kreysa, 1996).

!
I(r) =
I
m
2(1"
r
2
a
2
)

where I
m
is the mean current.

Figure 3.4 A disk electrode of radius a embedded in a semi-infinite slab of conductivity σ.

44
Closer to the edge of the electrode (
!
r"a) the current approaches infinity
(
!
I"#), which, clearly, cannot happen in a physical system. This is due to the fact
that the equation assumes the electrode to have an edge completely perpendicular to
the planar surface and also because overvoltages are ignored in the derivation (Heitz
& Kreysa, 1996). In reality, a fabricated microelectrode will always have some
curvature to the edge, since either sputtering, electrodeposition, or metal evaporation
will yield an edge with some slope. Although seemingly minor, the details of this
boundary condition can have a large effect on the electric field and current
distribution. This will be discussed further when solutions are proposed for
controlling the primary current distribution on the electrode by properly defining the
architecture (section 3.3.2).
A quasi-static finite-element model was used to illustrate the primary current
distribution at the disc surface (FEMLAB, Comsol Inc., Los Angeles, CA). The
simulations utilized the conductive media electrostatics module within FEMLAB.
Tetrahedral elements were used to define the model meshes. The disc was flush with
the surrounding insulating dielectric and centered beneath the base of a larger
hemispherical volume of saline medium. The radius of the medium surrounding the
electrode was large enough to approximate an infinite volume. The hemispherical
surface defining the upper boundary served as the ground electrode (zero-potential
boundary). Figure 3.5 shows results of a 30 µm diameter gold thin-film electrode (σ
= 4.1 x 10
7
S/m) immersed in a bulk medium with a conductivity of saline solution, σ
= 1.54 S/m (Johnson et al., 2005). The ITO lead (σ = 2 x 10
6
S/m) and SU-8

45
insultion layer (ε
r
= 3)

thicknesses were 0.8 and 1.6 µm, respectively. A 100 µA of
current was injected from the underlying indium tin oxide lead. The primary
distribution current is shown taking a slice 1 µm above and parallel to the disc
surface to avoid current singularities at the edge (Figure 3.5).

Figure 3.5. A finite-element electromagnetic model in the quasi-static regime showing the primary
current distribution on a 30 µm diameter gold disk electrode.

3.3.2 Electrode geometries for current homogeneity
A number of approaches have been suggested in order to give a more uniform
charge distribution at the electrode surface. A segmented disc electrode consists of n
concentric annuli in which neighboring annuli are connected via n resistors (Wiley &
Webster, 1982a, b). The current distribution over the entire electrode can be
effectively tuned by setting the correct value of each resistor, R
n
, and by injecting
current in to the center disc of the electrode (Figure 3.6). Unfortunately, for practical
purposes the fabrication of such electrodes is difficult since the resistors would have
to be fabricated in plane with specific values fixed during fabrication and current
would have to be injected into the center disc.

46

Figure 3.6. Geometry of a segmented disc electrode for engineering current distribution. For the
purposes of illustration it is has been divided into three areas. The current in the n
th
annuli must pass
through a series resistors, R
n
. The ohmic drop across the resistor reduces the potential in the n
th
annuli
to help level the current density profile (from Wiley & Webster, 1982).

A hemispherical electrode has also been proposed (Deconinick, 1992). In
this case, the spherical symmetry of the structure would allow for charge to reside on
the surface evenly if immersed in saline solution. This is illisutrated in a finite-
element model of a hemispherical electrode with a radius of 15 µm (Figure 3.7).
This geometry also increases electrode surface area for a given planar geometric area
and would decrease ohmic drop compared with a disc electrode of the same
diameter. However, fabrication of hemispheric microelectrodes of specific radii is
not as simple as fabrication of microdiscs. It is also likely that in contact with tissue
the current distribution would deviate from being completely uniform because the
pressure of the tissue on the electrode would not be uniform and this contact area is
known to increase the local electrochemical impedance in vivo (Shah et al., 2004).

47

Figure 3.7. Cross-sectional plot of normalized current density across the surface of a hemispherical
electrode at a height of 60 mm and 80 mm above the electrode plane. The structure exhibits a
uniform surface charge.

From a microfabircation standpoint the most practical solution proposed has
been recessing the electrode below the dielectric insulation (Rubinstein et al., 1987).
This changes the boundary conditions at the electrode edge giving a more uniform
current distribution. The effect can intuitively be understood by noting that as the
surrounding dielectric layer is increased, the electrical pathways with the least
resistance, those by the edge where we observe electric field fringing, are taken
away. The remaining pathways have a more uniform solution resistance, and,
therefore, the current does not have a single, highly preferential path.

Figure 3.8 A schematic representation of resistive pathways for a metal electrode immersed in saline
solution. In an unrecessed disc (left) the path at the electrode edge is that of low resistance compared
to the recessed case (right). The aperture in the recessed case still exhibits current density peaks at the
edge.

48
Rubinstein used Green’s method of moments to show that the effect of
recessing the electrode approaches a maximum when the ratio of the recess depth to
the pad radius is approximately 1/3. FEMLAB modeling of a 32 µm diameter disc
electrode (with the same material parameters as defined in section 3.3.1) over a range
of recess depths (1.2, 3.6, 6 µm) indicates that an insulation layer thickness of 6 µm
can decrease the peak current density by 45% compared with an unrecessed
electrode (Figure 3.9).

Figure 3.9 Plot of current density at a height of 1.0 mm above the electrode plane for a 32 mm
diameter disk electrode with varying SU-8 insulation layer thicknesses: (top left) SU-8 layer 0.4 mm
below electrode plane, (middle left) 2.0 mm above electrode plane, and (bottom left) 4.4 mm above
electrode plane; (right) cross-sectional current density plots for the three cases.

49
Current peaks still persist at the aperture of the recessed electrode (Figure
3.10). This is true for the same reason the unresecessed disc exhibits a nonuniform
distribution. Recessing the electrode will, therefore, not change the field profile at
the tissue itself but should prevent preferential corrosion. It has been shown,
however, that current hot spots due to the primary distribution are significantly
decreased within tens of microns of a 200 µm disc electrode yielding a Gaussian far-
field distribution (Behrend et al.).

Figure 3.10 Side view of electrostatic FEMLAB simulations of unrecessed, 3 µm recessed, and 6 µm
recessed cases illustrating the movement of current ‘hot spots’ off of the electrode surface to the
recess aperture (indicated using white arrows).

50
3.4 Fabrication of recessed microdiscs using SU-8 photoresist

In order to provide current uniformity the recess depth must be on the order
of a third of the radius of the disc electrode. Susserman et al. used discs varying in
diameter ranging from 0.5-35 mm recessed in stacked removable plexiglass plates
immersed in 0.9% NaCl solution to confirm analytical and modeling results
(Suesserman et al., 1991). The experimental setup also consisted of a Pt/Ir
(90%/10%) wire (50.8 µm diameter) that was able to be moved along the surface of
the electrode with a x-y-z stepper motor and a buffer amplifier to detect local
potential perturbations due to a sinusoidal varying potential placed on the electrode.
Experimental data showed negligible potential difference at the electrode surface
edge compared with the center, while the potential difference at the recessed
electrode aperture edge was 2-2.5 compared with the center. This is consistent with
modeling results presented here (Figures 3.9 and 3.10). Although the Pt/Ir wire had
a large reported impedance (1.66 MΩ at 1KHz), it would have still perturbed the
local field that it measured. Therefore, for the purposes of measuring the
effectiveness of recessed microdiscs in reducing preferential corrosion, using EDS to
quantify the local metal-oxide content across the surface is preferable.
FEMLAB modeling suggests that an insulating layer thickness of 2-33 µm is
required to even out the current distribution for 10-200 µm diameter discs. Such
recessed microdiscs were fabricated employing the microfabrication process
described in section 2.3. SU-8 epoxy-based photoresist is available in a range of

51
viscosities. High aspect-ratio structures can then be fabricated by controlling the
spin-speed during application of the photoresist (Figure 3.11). Figure shows.
Using SU-8 2002, 2005, 2007 and 2010 recessed 50, 100, 150, 200, and 250 µm
diameter polycrystalline Pt discs were fabricated (Figure 3.12). The recess depth
was measured using a DEKTAK profilometer (Figure 3.11). 1, 5, 9, 12.5, and 20 µm
recess depths were successfully achieved.
Electrodes smaller than 50 µm diameter (e.g., 25 µm diameter) were difficult to
pattern. This may have been because the UV light scattered in the thicker SU-8 film,
cross-linking the polymer resist in areas that were intended to remain soluble during
chemical development. Another possibility is that since chemical development is
longer for smaller electrodes with high aspect ratio vias due to microfluidic effects,
these electrodes were underdeveloped. Unfortunately, since all electrode sizes were
fabricated on the same substrate one development time (usually 60-100 seconds) had
to be chosen for all discs.

52
SU-8
Product
Viscosity
(cSt)
Spin
Speed
(rpm)
Thickness
(µm)
3000 2
2000 2.5
1000 3

2002

7.5
500 5
3000 6
2000 7

2005

45
1000 7.5
3000 8.5
2000 10

2007

140
1000 12.5
3000 13
2000 15

2010

380
1000 20

Figure 3.11 (Left) Table showing viscosity for different SU-8 chemistries and the corresponding
spin speed required to obtain films of desired thicknesses. (Right) Profilometery scan across a
variable disc sized MEA with SU-8 2002 photoresist spun on at 500 rpm giving a 5 µm recess (units
are in angstroms).

Figure 3.12 Photomicrographs of variable disc MEAs with Pt recessed discs. Electrode diameters are
50, 100, 150, 200, and 250 µm and insulation thicknesses are (from left to right) 1, 5, 9, 12.5, and 20
µm.

53
3.5 Conclusion

In this chapter FEMLAB modeling was used to show that a more uniform
current distribution could be achieved at the disc microelectrode surface employing a
thick SU-8 insulation layer. Such recessed discs have been fabricated over a range
of electrode sizes. This insulation layer has proven to be robust after months of
hippocampal slice testing as described in chapter 2.
It has been reported that after continuous pulsing of other IrOx stimulating
electrodes for stimulation of the central auditory system, failure was observed after a
few million pulses (10 mA, electrode area 1000-4000 µm
2
) (Najafi, 1994). “This
was mainly due to the very high current density around the edges of the electrode.
This phenomenon has been reported by several investigators for similar disc
electrodes. The very high current density around the perimeter of the site causes
delamination of the iridium oxide stimulating surface as it is subject to high current
levels.” It is important to place weight on such reports of more fully developed
neural prostheses. Especially with regard to long-term pulsing of disc electrodes it is
likely the same phenomenon will be observed in retinal stimulating electrodes after
time. Recessed disc microelectrodes may provide a practical solution to preferential
corrosion and delamination.
Future work should consist of long term current pulsing of these electrodes in
PBS solution and intermittent characterization using EDS, cyclic voltammetry, and
electrochemical impedance spectroscopy. The later two techniques will be discussed

54
in detail in the next chapter with regard to characterizing neural stimulating
electrodes.

55

Chapter 4

Electrochemical Characterization of Electrodes

4.1.1 Capacitive and Faradaic charge transfer
The electrode-tissue is interface is electrochemical. Once a metal electrode is
placed in an ionic solution, a monolayer of specifically adsorbed solvent molecules
and ions is formed on the surface. This monolayer separates the metal surface from
solvated cations and anions; the formed diffuse layer extends out to a region of the
solution where the concentration gradient reaches a constant value. The region near
the electrode but in solution constitutes what is known as the capacitive double layer
due to this separation of charge (Figure 4.1) (Bard & Faulkner, 1980). The charging
and discharging of this layer via anodic and cathodic current pulses during neural
stimulation causes a redistribution of charge in the solution and induces current flow
in the extracellular space in the form of ionic charge movement. This form of
stimulation is optimal because only electrostatic forces are involved and there is no
introduction of neurotoxic materials, such as heavy metal ions or hydroxyl groups,
which may cause local pH shifts. The specific capacitance of the double layer is
C
DL
=10-20 µF/cm
2
due to the fact that the thickness is one water molecule. This

56
value is fixed because only a certain number of water molecules can form a
monolayer on the metal surface per unit area. While this is a relatively large specific
capacitance, the breakdown voltage is low due to the fact the layer is only a few
angstroms thick. Therefore, the maximum charge density that can be transferred by
strictly capacitive means is limited to approximately 20 µC/cm
2
(Roblee & Rose,
1990). Charge density in excess of this can lead to the breakdown of the thin
dielectric layer formed by water and Faradaic charge transfer. Although the limits
for capacitive charge transfer can be increased by using capacitive electrodes with
high surface morphology such as TiN (Hammerle et al., 2002; Weiland et al., 2002)
or metal-oxides with an increased dielectric constant such as Ta/Ta
2
O
5
(Guyton &
Hambrecht, 1973; Donaldson, 1974; McCreery et al., 1988), and, therefore,
increased charge storage capacity, the limits are often still below those required for
neural stimulation.

Figure 4.1. Schematic representation of the capacitive double-layer at the electrode-electrolyte
interface. (IHP) inner Helmholtz plane. (OHP) outer Helmholtz plane. (modified from (Bard &
Faulkner, 1980))

57
When anodic or cathodic current stimulation is large enough, oxidation or
reduction reactions will be induced at the metal surface. This method of stimulation
is almost always required to reach excitation thresholds. When the potential is
driven to increase the electron energy level in the metal, electrons will be transferred
to vacant molecular orbitals for a given species in the solution (Bard & Faulkner,
1980). When a pulse of the opposite polarity is injected, the reverse process will
occur and electrons in molecular orbitals in the solution will be transferred to the
lower energy level of the metal is shown in figure 4.2.

Figure 4.2. Energy diagram for reduction (left) and oxidation (right) of species A. (Modified from
Bard and Faulkner.)

Because platinum is the most common noble metal used for electrodes, we
will investigate the Faradaic charge transfer mechanisms, and the associated oxide
formation, that play a role in stimulation with platinum. Both oxide formation and
H-atom plating are reversible reactions, meaning any oxidized species can be
reduced with a pulse of the opposite polarity. Together with double-layer charging
they allow for charge injection limits of 300-350 µC/cm
2
for platinum (Brummer &
Turner, 1977).

58

oxide formation: Pt + H
2
O ↔ PtO + 2H
+
+ 2e
-
H-atom plating: Pt + H
+
+ e
-
↔ Pt-H

Irreversible Faradaic reactions involve products that will diffuse from the electrode
surface and cannot be reduced. Platinum chloride, for example, is more soluble than
platinum because it has a greater electrochemical driving force to stay in solution.

Pt + 4Cl
-
→ [PtCl
4
]
-2
+ 2e
-

Hydrogen and oxygen gas evolution is also generally irreversible since gases can
rapidly diffuse from the electrode surface.

2H
2
O + 2e
-
→ H
2
+ 2OH
-
2H
2
O → O
2
+ 4H
+
+ 4e
-

Safe stimulation requires the reversible transfer of charge so that neither
electrode nor tissue is damaged. Within the context of retinal prostheses, smaller
electrodes (<250 µm diameter) will be required in future implants in order to
individually address smaller number of cells and increase spatial resolution. While
there have been multiple electrophysiological studies that investigated retinal
stimulation in animals and in subjects with various sized electrodes (as will be

59
discussed in chapter 5), comparatively little attention has been paid to the
dependence electrochemical charge injection and impedance as a function of
decreasing electrode size. In this chapter the results of experiments using MEAs
consisting of thin-film Pt electrodes with varying diameter (25, 50, 100, 150, 250,
300, 450, 500, and 550 µm) are presented. By using cyclic voltammetry (CV),
biphasic pulsing, and electrochemical impedance spectroscopy (EIS) it is shown that
(1) microelectrodes with diameters <100 µm exhibit increased charge density due
most likely to a hemispherical compared with a planar diffusion profile, and (2) that
the impedance of microelectrodes transitions from a perimeter to an area dependence
(e.g.,
!
1
r
"
1
r
2
) as the frequency is decreased from 100 KHz to 10 Hz. The latter,
proves that the current distribution at the electrode surface is dynamic, and evolves
during the time course of a stimulus pulse.

4.2 Electrochemical techniques

All electrochemical techniques described here, and used for the collection of
data presented in section 4.3 and 4.4, used a Gamry FAS1 potentiostat (Gamry
Instruments, Warminster, PA) or a current pulse generator (MultichannelSystems
STG 2008 stimulus amplifier) and a custom made three electrode Teflon
electrochemical cell (Figure 4.3). A potentiostat is a device that controls the
potential between the working and reference electrode by injecting the required

60
current through counter electrode. The working electrode is held at virtual ground,
providing a return current path. Typically, and in these experiments, the
microelectrode is the working electrode, a Ag/AgCl (3 M NaCl) electrode is the
reference electrode that holds a constant potential value (U
eq
= 220 mV vs. standard
hydrogen electrode), and a large Pt wire mesh (Alpha Aesar, 99.99%) is the large
surface area counter electrode that assures that only the potential drop across the
microelectrode-electrolyte interface contributes significantly to the total potential
drop (i.e., an electrode with increased area has an increased number of active sites
for charge transfer, decreasing ohmic drop). This setup allowed for both
electroplating of Pt films using ammonium hexachloroplatinate and the surface
characterization of electrodes with CV, biphasic pulsing, and EIS.

Figure 4.3. (Left) Schematic of custom three electrode Teflon electrochemical cell (RE – reference
electrode, CE – counter electrode, WE – working electrode). (Right) Photomicrograph of variable
disc diameter MEA with polycrystalline thin-film evaporated Pt and SiNx/SU-8 dual insulation layer.

61
4.2.1.1 Cyclic voltammetry
In cyclic voltammetry the electrode potential is linearly ramped between two
points repeatedly and the current due to redox reactions induced at the electrode
surface is recorded. The potential for reducing and oxidizing certain species in the
analyte is specific to the binding energy of the valence electrons of that molecule. In
this way a cyclic voltammogram is an electrochemical fingerprint of the material
being studied. (The electroactive species in the analyte also contribute to the
voltammogram.) Information can be obtained regarding the composition, crystalline
structure, and surface morphology of a given electrode. Within the context of neural
stimulation, it is critical that all reactants produced at the electrode surface can be
retrieved as the electrode potential is cycled in the opposite direction. The main use
of CV in characterizing stimulating electrodes then is to obtain the maximum charge
injection an electrode can deliver while still remaining in a safe, reversible potential
window. This range in which O
2
and H
2
gas evolution is avoided and in which
successive voltammograms overlap to show repeatability is the ‘water window’
(Stamford, 1990). Given the voltage scan rate, the total charge delivered by an
electrode within one anodic sweep within the predefined potential limits can be
found by simply integrating the current over time. Cyclic voltammetry, however,
does not capture diffusion and mass-transport limits to current that may occur during
a stimulus pulse. Therefore, biphasic pulsing must be also used to accurately assess
safe-stimulation limits. While the maximum charge density is a figure of merit for

62
stimulating electrodes, it is the potential at the electrode surface that drives Faradaic
reactions and determines this value.
For all experiments with Au and Pt films cyclic voltammetry was done in 250
mM H
2
SO
4
bubbled with N
2
(in order to displace anodically evolved O
2
(Germain et
al., 2004)) and cycled at 200 mV/sec. In contrast with phosphate buffered saline,
which contains many electroactive species (e.g., sodium chloride, sodium phosphate,
potassium phosphate), sulfuric acid allows for cyclic voltammgrams with sharp,
well-defined redox peaks due to the limited number of electroactive species. The
hydrogen content also allows for very clean hydrogen adsorption and desporption
peaks, which can be used to calculate electrode surface area. This scan rate was used
to increase SNR (e.g. a higher scan rate increases the current (
!
i =C(dV dt)), and
quickly achieve a repeatable voltammogram. Experiments were conducted in a
Faraday cage to minimize 60-cycle noise. The potential range for Au was U = [-0.3
V to 1.5 V] and the range for Pt was U = [-0.2 V to 1.25 V]. After fabrication it was
necessary to cycle each electrode 200-300 times before steady-state was obtained
and data was collected. Initial changes in the voltammogram may have been due to
the initial annealing of the thin-film using current pulses, burning off of residual
photoresist, electro-smoothing (Hu & Liu, 1999), electron migration or a place
exchange process (Angerstein-Kozlowska et al., 1989).

63
4.2.1.2 Cyclic voltammetry electroplating using an ammonium hexachloroplatinate
bath
In order to minimize impedance and increase charge capacity cyclic
voltammetry was employed in the electroplating of stimulating and recording
electrodes in an ammonium hexachloroplatinate bath. Minimizing the impedance
increased the achievable SNR for recording electrodes and the maximum injected
charge for stimulating electrodes for the electrophysiological experiments discussed
in chapter 5. Compared with more conventional Pt acidic plating baths, ammonium
hexachloroplatinate allows for the reduction of Pt ions at a near-neural (pH = 7.8)
which avoids the codeposition of hydrogen that leads to hydrogen embrittlement
(Whalen et al., 2005). In addition, this chemistry does not contain lead acetate, as is
common in other Pt plating solutions, which is cytotoxic (Tang et al., 1996). The
solution composition was 17 mM (NH
4
)
2
PtCl
6
+ 250 mM Na
2
HPO
4
. 50 mL of
solution was prepared by dissolving sodium hydrogen phosphate (Alpha aesar,
1.75g) in Millipore deionized water and then adding ammonium hexachloroplatinate
(Alpha aesar, 0.3875 g).
The four-electron reduction of Pt during plating is dictated by the following
half-cell reactions

!
PtCl
6
2"
+2e
"
#PtCl
4
2"
+2Cl
"
U
eq
= 0.526 V (Ag/AgCl)

!

!
PtCl
4
2"
+2e
"
#Pt +4Cl
"
U
eq
= 0.558 V (Ag/AgCl)

64

The potential range for deposition was U = [-0.8 V to 0.6 V] vs. Ag/AgCl. The scan
rate was 200 mV/sec, and the typically 180 cycles yielded a stable film. A stable
film was confirmed with the cyclic voltammogram obtained during the deposition
and with visual inspection.
A comparison of cyclic voltammograms for a 200 µm diameter e-beam
evaporated polycrystalline thin-film Pt electrode and an electroplated Pt electrode
with increased surface roughness is shown in figure 4.4. The increased number of
active sites is evidenced by the greater amount of area (delivered charge) underneath
the CV for the electroplated electrode. A thick conductive Pt oxide begins to form at
0.6 V vs. Ag/AgCl which is fully reduced in during the cathodic sweep at 0.5 V. The
two peaks at -0.18 V and -0.02 V correspond the to adsorption and desorption of
hydrogen. Because hydrogen atoms form a monolayer on Pt, the real surface area of
the electrode can be found by integrating the current due to H-atom plating and
subtracting the current due to double layer charging (Roblee & Rose, 1990):
!
Q
H
= (i"i
DL
)dt =
(i"i
DL
)dU
dU/dt
"0.2
0.1
# #

65

Figure 4.4. Cyclic voltammetry of electroplated and thin-film evaporated electrodes in a 250 mM
H
2
SO
4
analyte bubbled with N
2
and cycled at 200 mV/sec.

4.2.2 Electrochemical Impedance Spectroscopy
Electrochemical impedance spectroscopy (EIS) uses a perturbative sinusoidal
voltage (usually ±10 mV rms) to determine the electrode impedance over a specified
frequency range. For a given driving potential, the resulting current amplitude and
phase lag is recorded (Barsoukov & Macdonald, 2005).

!
Z =
U(t)
I(t)
=
U
o
cos("t)
I
o
cos("t#$)
=Z
o
cos("t)
cos("t#$)

This data is represented using either a Nyquist plot (-Im Z vs. Re Z) or a Bode plot
(log |Z| vs. log ω, φ vs. log ω).
As was discussed in chapter 2, the electrode-electrolyte interface can be
described as an RC circuit. The simplest model of the interface is the Randels model

66
which consists of a capacitor, representing the double-layer capacitance (C
DL
), in
parallel with a resistor, representing the charge-transfer resistance (R
CT
), both in
series with another resistor, representing the solution resistance (R
S
) (Bard &
Faulkner, 1980) (Figure 4.5). Values for each of these elements can be attained from
EIS data. The high frequency limit of the spectral impedance, where there is no
phase shift, represents the solution resistance. As the frequency decreases, the
impedance represents the contribution of the double-layer capacitance; the
corresponding phase angle shifts close to 90°. The low frequency impedance
magnitude again begins to flatten out, and represents the resistance to faradic charge
transfer. The phase angle again approaches zero. It is important to note that since
only a small amplitude voltage signal is used, we are operating in the double-layer
charging region of the corresponding CV curve for that material. (This is taken to be
the linear portion of the electrochemical I-V curve (Barsoukov & Macdonald,
2005).) R
S
will be independent of the potential, but C
DL
and R
CT
will not. Since a
stimulus pulse will drive the electrode potential to non-linear portion of the I-V curve
these limitations must be kept in mind when making conclusions based on EIS data
for the purposes of neural stimulation. Still, the combination of CV, EIS, and
biphasic current pulsing gives a complete picture of electrochemical reactions
occurring at the electrode surface during stimulation.
For all EIS experiments, potentiostatic impedance spectra from 100 kHz to
10 mHz (U = 0.0 V vs. open circuit, ΔU = +/-10 mV rms) were collected using a
Gamry FAS1 potentiostat (Gamry Instruments, Warminster, PA). Experiments were

67
performed in phosphate buffered saline (PBS) solution to mimic biological chemistry
(9 g/L sodium chloride, 0.8 g/L sodium phosphate diabasic, 0.14 g/L potassium
phosphate monobasic) at room temperature and in a Faraday cage.

Figure 4.5. Randles model of the electrode-electrolyte interface. R
CT
is charge transfer resistance, C
DL
is the double-layer capacitance, and R
S
is the solution resistance.

4.3 Increased charge injection with microelectrodes

Microelectrodes on the order of <100 µm in diameter have been of great
interest in electroanalytical chemisty because of a decreased ohmic drop in the
vicinity of the electrode, a fast steady-state response due a relatively small RC time
constant, and an increased current density due to increased mass transport (Stulik et
al., 2000). Because of the electroactivity of certain neurotransmitters it is also
possible to electrochemically detect micromolar concentrations of biological
signaling agents such as dopamine, adrenaline, and serotonin via fast cyclic
voltammetry using small enough microelectrodes (Stamford, 1989, 1990). While
increased charge injection has been shown analytically and with simulations (Forster,
2005), no rigorous experimental study has been performed over a range of electrode

68
sizes. For the purposes of a higher resolution retinal prostheses, increased mass
transport for microelectrodes would allow for smaller stimulating electrodes to safely
elicit neuronal responses.
Using a variable sized disc MEA with thin-film e-beam evaporated
polycrystalline Au microelectrodes, charge injection limits were investigated as a
function of microelectrode size using cyclic voltammetry and galvanostatic biphasic
pulsing.

4.3.1 Evidence of increased current density using cyclic voltammetry
Cyclic voltammograms were obtained for Au microelectrodes with diameters
of radii of 25, 50, 100, 150, 250, 300, 450, 500, and 550 µm in 250 mM H
2
SO
4

bubbled with N
2
and cycled at 200 mV/sec. The potential range was U = [-0.3 V to
1.5 V]. Cyclic voltammograms were normalized for electrode area, so that injected
charge density could be compared (Figure 4.6). Macroelectrodes exhibit a largely
planar diffusion profile, whereas microelectrodes have a hemispherical diffusion
profile. The greater contribution of the edge effect in smaller electrodes allows for
increased charge injection (Forster, 2005). Because of the planar diffusion profile,
microelectrodes >100 µm in diameter had a constant anodically delivered charge
density of approximately 860 µC/cm
2
. 100, 50, and 25 µm diameter electrodes had
charge densities of 2053, 2676, and 6110 µC/cm
2
, respectively.

69

Figure 4.6. (Top) Cyclic voltammograms normalized with respect to area of various sized Au
electrodes in a 250 mM H
2
SO
4
analyte bubbled with N
2
and cycled at 200 mV/sec. (Middle)
Corresponding anodic charge density delivered vs. electrode diameter. (Dashed line is linear
extrapolation of constant charge density for electrodes with diameter >100 µm.) (Bottom) Schematic
diagram illustrating (left) planar and (right) hemispherical diffusion profiles for larger and smaller
microelectrodes, respectively. (modified from (Stulik et al., 2000)).

70
4.3.2 Evidence of increased current density using biphasic pulsing
While CV is conventionally used to obtain charge injection limits for
stimulating electrodes, because the contribution of mass transport to the limiting
current is time dependent it is important to verify increased charge density of
microelectrodes with biphasic pulsing. 25, 50, 100, 150, 200, and 250 µm diameter
Au microelectrodes were pulsed with 0.5, 2, and 3 ms per phase cathodic-first
biphasic pulses in PBS (Figure 4.7). The current amplitude was adjusted such that
the charge density was constant for each electrode for each charge duration (0.5
mC/cm
2
); the charge density limit was set by the minimum current amplitude that
could be output from the stimulus amplifier (1 µA). This was done assuming a
homogeneous current distribution and dividing charge per phase by geometric area.
This simplification had to be made since including the functional form of the primary
distribution in normalization is extremely difficult especially given the dynamic
nature of the current profile as is described in the next section. The potential drop
across the electrode-electrolyte interface was measured using an Tektronix TDS
5034B digital phosphor oscilloscope in math mode and probing across the electrode
and the Pt wire mesh counter. In this two electrode setup it was assumed that
potential was dropped entirely across the comparatively high impedance electrode.
Pt wire mesh position was held constant during for all measurements. The input
pulse was monitored using a 6.8 kΩ sense resistor in series with the electrode. (This
value was high enough value such that the current pulse amplitude could reliably be

71
distinguished from baseline noise for all electrode sizes, but was not so high so as to
perturb the rectangular shape of the pulse.)
The potential drop across the electrode-electrolyte interface decreases with
electrode size for biphasic stimulus pulses of constant current density for 0.5, 2, and
3 ms per phase pulsewidths. This is consistent with cyclic voltammetry results. 0.5
ms pulses yielded a voltage drop of 1.7 V for 25 µm diameter electrodes compared
with 3.3 V for 200 µm diameter electrodes. For 2 and 3 ms pulses this decrease was
comparatively less, yielding a voltage drop of 2.35 V for 25 µm diameter electrodes
and 3.25 V for 250 µm diameter electrodes. While these voltages are outside of the
water window for Au thin-film electrodes, they were so because of minimum current
amplitude limitations of the stimulus amplifier, and these recordings were stable over
10,000 pulses. Therefore, in conjunction with the cyclic voltammetry results, are
indicative of a general trend that microelectrodes <100 µm diameter exhibit higher
charge injection limits.

72

Figure 4.7. Biphasic current pulses (bottom of each graph) and corresponding voltage drop (top of
each graph) for discs of varying diameter injected with constant current density (corresponding to a
charge density of 0.5 mC/ cm
2
) immersed in PBS with a Pt wire mesh counter electrode. (Top) 0.5
ms, (Middle) 2 ms, (Bottom) 3 ms per phase stimulus pulses.

73

Figure 4.8. Maximum voltage drop across electro-electrolyte interface for Au microelectrode discs of
varying diameter for using 0.5, 2, and 3 ms per phase biphasic pulses.

4.4. The dependence of spectral impedance on disc microelectrode
radius

The impedance of microelectrodes increases as the area decreases. For a
given stimulus current pulse, this increased impedance will increase the voltage drop
across the electrode-electrolyte interface compared with a larger electrode. This
voltage drop is undesirable because it can induce irreversible electrochemical
reactions that can inject neurotoxic heavy metal ions or byproducts into the
extracellular space (Brummer et al., 1977; Roblee & Rose, 1990). These processes
can also change the surface morphology and composition of the electrode itself. In
addition, if the device is to be used for the purposes of recording a neural response or
for impedance based sensing (Wegener et al., 2000; Yang et al., 2003), increased

74
impedance will increase the noise on the electrode, making it difficult to record
microvolt signals from neurons.
The analytical solution for the primary current distribution from Laplace’s
equation was discussed in chapter 3. Electrostatic finite-element modeling was used
to confirm Newman’s relationship between the solution resistance, R
S
, and electrode
radius, r, to be R
S
=1/4κr, where κ is conductivity (Newman, 1966). Here evidence
that the time-dependent current distribution is dynamic and departs from this
electrostatic solution is presented. It is shown how impedance does scale inversely
with radius (i.e., it follows Newman’s relationship) in the high-frequency regime
(~100 kHz). However, the impedance transitions from radius dependence to area
dependence as the frequency is decreased to 10 Hz. Moreover, smaller
microelectrodes (<50 µm), have an RC time constant small enough to exhibit area
dependence even at relatively high-frequencies of 100 kHz (i.e., frequencies at which
larger electrodes are in a solution resistance regime.) By analyzing data at various
frequencies this transition is characterized.

4.4.1 EIS results on microelectrode discs of varying diameter
Impedance data was consistent with previously reported results for platinum
microelectrodes in PBS solution (Figure 4.9) (Duan et al., 2001; Franks et al., 2005).
Except for the three smallest microelectrodes, the impedance magnitude was low and
frequency independent over the frequency range from 30 kHz to 100 kHz. At
frequencies below 30 kHz, the impedance magnitude increased with decreasing

75
frequency and the phase angle value shifts towards –90°. These changes are
indicative of a transition towards capacitive charging processes at the electrode
surface. It is common to observe a second transition into another frequency-
independent resistive region near 1 Hz associated with Faradaic charge transfer at the
electrode surface (Mailley et al., 2004). However, it has been shown for platinum
electrodes in PBS (Duan et al., 2001; Franks et al., 2005)

and other solutions
(Mailley et al., 2004), that the low frequency response is highly dependent upon
electrode initial surface potential and surface condition, and can exhibit either
resistive or capacitive behavior.
Diameter affected the impedance response of microelectrodes over the entire
frequency range tested. The frequency independent range steadily decreased from
10 kHz - 100 kHz for the 750 µm diameter microelectrode, to 30 kHz -100 kHz for
the 200 µm diameter microelectrode. The three smallest microelectrodes showed no
frequency independent response over the high frequency range tested. The
capacitive charging region of the impedance magnitude showed a shift to higher
frequencies with decreasing microelectrode size. This observation has been
qualitatively observed in comparing electrodes of different surface area (Mailley et
al., 2004; Franks et al., 2005).

Impedance measured at 100kHz was taken as the empirical solution
resistance, R
S
, except as noted below. Although the phase angle at this frequency is
typically near or equal to zero, in the case of the three smallest microelectrodes
(≤50µm) the phase angle remained greater than 20
o
, consistent with other reports

76
(Duan et al., 2001). This result is expected due to the smaller RC time constants of
these electrodes, as will be discussed. However, for the purposes of estimating the
solution resistance a phase shift closer to zero was required (i.e.,
!
["|Z|"f]
100KHz
#0). To minimize error in approximating R
S
for microelectrodes at
with these diameter, impedance data was re-plotted in the complex plane (data not
shown). Z
img
vs. Z
real
(Nyquist) plots of impedance for platinum microelectrodes in
solution have a semicircular shape, with the arc of the semicircle becoming more or
less depressed depending on frequency dispersion effects caused by the ionic
composition of the solution. The high-frequency intercept of the semi-circle is equal
to R
S
. Here, the high frequency data was linearly extrapolated to approximate the
solution resistance. The values obtained using this extrapolation method differ from
the actual impedance value at 100 kHz by less than 3%. R
S
values for all
microelectrodes were plotted vs. diameter in figure 4.10.
Data collected from simulations performed on finite element models with
different microelectrode diameters were also compared to assess effect of diameter
on current distribution. R
S
values obtained from FEMLAB modeling were plotted
vs. microelectrode diameter, Figure 4.10, to compare with empirical data and values
generated from the Newman model.

77

Figure 4.9. (Top) Phase angle vs. frequency, and (bottom) |Z| vs. frequency for Pt microdiscs of
varying diameter in phosphate buffered saline. (Inset) Magnification of high frequency range of
eight largest electrodes.

78
4.4.2 Empirical and finite-element modeling correspondence with Newman’s
radius vs. solution resistance relationship
There was excellent correspondence between the measured dependence of R
S

on microelectrode diameter, the finite-element model prediction of the resistance,
and Newman’s analytically derived function (Figure 4.10). The fact that both
Newman’s analysis and the modeling only considered electromagnetic effects and
neglected mass transport, confirmed the fact that the primary current distribution is
solely responsible for the observed 1/r dependence of the solution resistance. This
dependence follows the reasoning that the variation in current density across the
electrode surface should cause a corresponding variation in impedance across the
surface. Indeed, others have shown the RC time constant itself varies radially across
the electrode (Jorcin et al., 2004; Oldham, 2004).

79

Figure 4.10. Solution resistance (i.e., impedance at 100 KHz) vs. microdisc radius for fabricated Pt
MEAs, finite-element modeling results, and electrostatic analytical solution.

4.4.3 RC time constant for microelectrodes
Microelectrodes have received much interest for electrochemical sensing in
large part because of their ability of reach steady-state quickly compared with larger
electrodes (Stulik et al., 2000; Sandison et al., 2002; Oldham, 2004). The RC time
constant is given by
!
R
s
C ="rC
o
/4# , where R
S
is the solution resistance, C is the
double-layer capacitance, r is the microelectrode disc radius, C
o
is the specific
double-layer capacitance, and κ is the conductivity (Sandison et al., 2002). The fact
that the 22 µm, 50 µm, and 100 µm electrodes show fairly capacitive behavior at
100 kHz (e.g., phase angles of 59.3° ± 2.96°, 46.2° ± 2.30°, 26.28° ± 1.31°,

80
respectively) is consistent with the dependence of charging time constant on
radius.
!

As the radius is decreased the double-layer is charged faster, and, therefore,
even at the highest measured frequency, τ is less than the period of the sinusoidal
voltage perturbation. For example, assuming κ = 1.54 S/m and C
o
= 20 µF/cm
2
then
τ

= 1.12 µs for the 22 µm diameter electrode. Considering the period at 100 kHz is
10 µsec, we can expect that a double-layer should be fully charged and discharged
even at this frequency. Because there is a linear relationship between τ and
microelectrode disc radius, we expect that an electrode with a diameter of 250 µm (τ

= 12.80 µs) would not fully charge the double-layer at this frequency, and that is
indeed what we observe as the phase of this electrode and those of greater diameter
are <20° at 100 kHz. (Figure 4.11)

81

Figure 4.11. (Left axis) Phase shift at 100 kHz vs. microdisc diameter for three different electrode
arrays showing how smaller electrodes behave largely capacitively compared with larger electrodes.
(Right axis) Time constant vs. disc diameter. The dashed line indicates half the period of the
sinusoidally varying voltage at 100 kHz.

4.4.4 Shift to dependence on area at lower frequencies
Data was analyzed from 100 KHz to 10 Hz in decade increments and
corresponding |Z| vs. diameter plots were obtained at each of these frequencies.
(Figure 4.12) Three different sets of Pt microelectrode discs of varying diameter
were characterized for the purposes of repeatability. The arithmetic means of the
three data sets were then plotted and a power series curve fit was used to determine
the functional form of this average. We see good correspondence for all of these
electrode sets, and there is a gradual transition from approximately a 1/R

82
dependence (i.e., 1/R
1.05
) to approximately a 1/R
2
dependence (i.e., 1/R
1.97
). At 10
kHz, for an electrode diameter range of 22 µm-650 µm, we observe a 1/R
1.56
dependence indicating that most of the electrodes are in largely area driven regime
within one decade of frequency from the solution resistive regime.

83

Figure 4.12. |Z| vs. microdisc radius at 100 kHz, 10 kHz, 1 kHz, 100 Hz, and 10 Hz in PBS and
corresponding power series curve fits for three different sets of variable sized microdiscs. (Bottom
right) Power series fit exponent vs. frequency indicating shift to area dependence of impedance at
lower frequencies.

The physical basis for this transition is as follows. As the frequency
decreases there is a longer time available for a given electrode to reach steady-state.

84
As the time of anodic/cathodic bias is increased the assumption made in Newman’s
derivation of the primary distribution, no longer holds for two reasons: (1) the
solution side of the double layer is no longer isopotential as it charges, and (2)
concentration gradients arise due to carrier depletion. When the period of the
perturbation is long compared to the electrode time constant a more uniform
charging over the surface takes place. This is consistent with the fact that the exact
solution of the primary distribution predicts singularities at the electrode that are,
clearly, not physically achievable. Indeed, concentration gradients must be taken
into account to realize a physically meaningful solution (Wagner, 1951; Newman,
1966; Wiley & Webster, 1982b, a; Rubinstein et al., 1987). Because reactants are
preferentially consumed at the electrode periphery, a non-uniform concentration
gradient forms across the electrode surface. Once these effects are taken into
account the known secondary current distribution results (Heitz & Kreysa, 1996).
Because of this effect we observe a transition to an area dependent impedance as the
period of the sinusoidally varying voltage approaches the RC time constant for a
microelectrode of a specific radius, and as the capacitive double-layer is charged.

4.4 Conclusions

In this chapter CV, biphasic pulsing, and EIS were used to electrochemically
characterize polycrystalline thin-film evaporated Au and Pt microelectrodes. Each
of these methods reveals specific insights into the nature of charge transfer at the

85
electrode-electrolyte interface. CV of 25, 50, 100, 150, 250, 300, 450, 500, and 550
µm diameter Au microelectrodes in a 250 mM H
2
SO
4
analyte shows that electrodes
≤100 µm diameter exhibit an increased current density. This is most likely due to
increased mass trasport to the electrode edge, which contributes more significantly as
electrode size is decreased. This effect is confirmed using 0.5, 2, and 3 ms per phase
biphasic pulsing of these microelectrodes and noting a decreased voltage drop across
the electrode-electrolyte interface for smaller microelectrodes for a fixed charge
density of 0.5 mC/cm
2
. These results suggest that as microelectrode size decreases in
order to target smaller numbers of retinal cells, an increased charge density will be
able to used without inducing irreversible reactions compared with larger
microelectrodes. It is not known whether or not this effect will be diminished when
the microelectrode is placed in close contact with retinal tissue as this may change
the diffusion layer profile. This contact may the subject of further work.
EIS was used to characterize electrodes of varying diameter in order to
understand how the RC time-constant and impedance magnitude vary as a function
of frequency and disc size. At the high frequency limit (100 Khz) the impedance
varies approximately as 1/R, as predicted by the analytical solution. As the
frequency is decreased this relationship shifts to a 1/R
2
or area dependence.
Because the Fourier transform of a square wave is composed of an infinite
number of odd integer harmonics, it is possible to apply the dependence of
impedance on sinusoidal frequency to a square wave stimulus pulse. The Fourier
series for an ideal square wave is (Katznelson, 1976)

86

!
x(t) =
4
"
sin(2k#1)$t
2k#1
k=1
%
&

While an ideal square pulse is not practically achievable because it would require
infinite bandwidth, the more sinusoidal frequency components that are included yield
a better approximation of a square wave. The lower order harmonics ( or frequency
components) define the pulse frequency while the higher order harmonics determine
how well the pulse edges can mimic an ideal square wave (Figure 4.13). Taken
together with the results presented in section 4.4.4, this suggests that at the onset of a
stimulus square pulse the current is delivered primarily at the electrode edge, and this
distribution evolves during the time course of the pulse to include the electrode
center. In chapter 3 recessed electrodes were presented as means of minimizing
preferential corrosion at the electrode edge due to the primary distribution. Based on
EIS data presented in this chapter, it may also be possible to utilize the center of the
disc electrode for a greater portion of the stimulus if sinusoidal pulses are employed.
It is interesting to note that early cortical stimulation studies used sine wave stimuli
(Brindley & Lewin, 1968; Brindley & Rushton, 1974), and it has been shown in
isolated retina stimulation that sine wave pulses have a lower threshold for
stimulation compared with square wave pulses of the same amplitude (Suzuki et al.,
2004).

87

Figure 4.13 The first, fifth, and ninth harmonics of the Fourier series for a square wave. The
idealized square wave, which is comprised of an infinite number of odd harmonics, is represented
with the dotted line. It is the higher order harmonics that capture the transients at the pulse rise and
fall.

88

Chapter 5

Spatial response properties of electrically stimulated
salamander retina

5.1 Prior work in electrical stimulation of isolated vertebrate
retina

The response of the isolated vertebrate retina to electrical stimulation has been
the subject of extensive study. Many investigations employed single metal wire or
conical shaped electrodes for extracellular stimulation and recording (Doty &
Grimm, 1962; Dowson & Radke, 1977; Humayun et al., 1994; Weiland et al., 1999;
Jensen et al., 2003; O'Hearn T et al., 2006). Others used these electrodes for
stimulation but recorded input excitatory and inhibitory current into ganglion cells
using a whole cell patch clamp technique (Fried et al., 2006; Margalit & Thoreson,
2006). Multielectrode array stimulating and recording which most accurately
mimics the electrode-tissue interface in a retinal prosthesis implant has also been
investigated using stimulating electrodes which varied in diameter over a large range
(6-1500 µm) (Zrenner et al., 1999; Grumet et al., 2000; Stett et al., 2000; Sekirnjak

89
et al., 2006). These studies primarily investigated charge density thresholds for
various sized electrodes, and the relative threshold for somatic excitation compared
with axonal stimulation.
Here custom microfabricated multielectrode arrays were employed that were
designed to mimic the size and pitch of near-term prostheses (200 µm diameter, 500
µm center-to-center spacing) while still being able record single-unit responses from
ganglion cells. In this way it was possible to study the spatial spread of the
stimulation field and it’s dependence on pulse duration and amplitude. The
interaction of the electric field between adjacent electrodes was also investigated and
results were confirmed using FEMLAB finite-element software. Although electrode-
electrode interactions have received little attention thus far, because complex
stimulation patterns will be used to encode visual information, they will prove to set
physical constraints on how an image can be ‘recreated’ as a stimulation field at the
array-retina interface. Pharmacological agents have also been used to study the
frequency dependence of direct stimulation of ganglion cells and stimulation of
presynaptic cells (e.g., bipolar and amacrine cells). In order to assess how threshold
charge and charge density depends on electrode size, results are presented comparing
stimulation with both 200 µm and 10 µm electrodes. This dependence has a large
impact on the feasibility of future high-resolution prostheses.

90
5.2 Materials and Methods

5.2.1 Animal model
RGCs in the tiger salamander form a monolayer that is uniformly spatially
distributed, and the lack of vitreous facilitates recording from all classes of cells
(Segev et al., 2006). The functional retinal organization in all vertebrates is also
similar, although the average cell diameter is larger (20 µm) compared with
mammals (10-12 µm). In theory, cells with larger somas have lower stimulus
thresholds for extracellular stimulation (Horch & Dhillon, 2004), however, it has
been reported that the site of preferential stimulation for an epi-retinal prosthesis is
the axon hillock of the cell which is located where the axon bends out of the plane of
the nerve fiber layer and into the ganglion cell layer (Schiefer & Grill, 2006). While
the former may preclude making conclusions on absolute stimulus threshold, relative
thresholds for ganglion cells should still be valid. In addition, our primary focus in
here is to study the stimulating electric field profile using varying current sources
and sinks which should be as valid in an amphibian model as a mammalian one.
Other studies investigating the effect of electrical stimulation of retina for prostheses
applications have also used a salamander model (Margalit & Thoreson, 2006).

5.2.2 Tissue preparation
All animal protocols were approved by the Institutional Animal Care and Use
Committee at the University of Southern California. Data was collected from 22

91
larval tiger salamander retinas (Ambystoma tigrinum) using established methods
(Werblin, 1978). Tiger salamanders were kept at 5° C in a 12 hour light/dark cycle.
Salamanders were rapidly decapitated and pithed in dim room light. Eyes were
enucleated and eyecups were placed in bicarbonate solution (in mM): 110 NaCl, 2
KCl, 30 NaHCO
3
, 1.5 CaCl
2
, 1.6 MgCl
2
, 10 Glucose, 0.01 EDTA that is bubbled
with 5% CO
2
/95% O
2
. 1mm
2
square pieces of retina were cut and placed on custom
made plunger outfitted with a dialysis membrane. The plunger was then inserted
into cylindrical mount which is silicone glued onto the MEA itself; the retinal
ganglion cell (RGC) side of the tissue was gently pressed onto the array while being
observed through an inverted microscope (Figure 5.1). The retina was perfused at 3-
5 ml/min with oxygenated bicarbonate solution.

Figure 5.1 (Left) Custom Retinal MEA and MEA polycarbonate mount and plunger. The plunger is
outfitted with a microporous dialysis membrane using an o-ring. Plunger slots allow for oxygen
containing perfusate to deliver proteins and ions to the retina when it gently pressed against the MEA
(Right)

92
5.2.3 Custom Retinal MEAs
Custom multielectrode arrays were designed with four, 200 µm diameter
stimulating electrodes (500 µm center-to-center spacing), and 56, 10 µm diameter
recording electrodes arranged with radial symmetry with respect to the stimulating
electrode edge (i.e., at distances of 50, 100, and 150 µm from the stimulating
electrode edge) (Figure 5.2). The average center-to-center spacing of recording
electrodes was approximately equal to arrays used to sample every ganglion cell in a
given retinal patch in the tiger salamander (Segev et al., 2006).
Details of the two photomask photolithographic process flow for the fabrication
of MEAs with Pt leads and silicon nitride/ SU-8 insulation were described in chapter
2 and in prior work (Ahuja et al., 2004; Gholmieh et al., 2006). Electroplating of
electrodes using an ammonium hexachloroplatinate bath (17 mM (NH
4
)
2
PtCl
6
+250
mM Na
2
HPO
4
)
,
a Princeton 2263 potentiostat, and a custom three electrode
electrochemical cell (Whalen et al., 2005) was described in detail in chapter 4.
Briefly, deposition was done using cyclic voltammetry (U = [-0.8 V to 0.6 V] vs.
Ag/AgCl, scan rate 200 mV/sec for 180 cycles). Cyclic voltammetry allowed for a
resetting of the diffusion layer in each scan so that a ‘mushroom’ shaped electrode
profile was avoided, which is observed in static potential electroplating. Here, the
resulting film was geometrically well confined to the area defined by the
photolithographic process flow (Hung et al., 2002).

93

Figure 5.2. (Top) Photomicrograph of custom retinal multielectrode array with 200 µm diameter
stimulating Pt-electroplated electrodes and 10 µm diameter electron-beam evaporated Pt thin-film
recording electrodes. (Bottom) Schematic of retina, gently pressed ganglion cell side down on
custom MEA with microporous dialysis membrane.

5.2.4 Electrical stimulation and recording
Biphasic cathodic-first current pulses were generated using a stimulus
generator (STG 2001, Multichannel Systems, Reutlingen, Germany). Up to two
electrodes could be selected as stimulating electrodes and the stimuli programmed
independently. RGC responses were amplified by 1100 using an MEA 1060
preamplification board (Multichannel Systems, Reutlingen, Germany). Definition of

94
stimulus pulse parameters and data collection was done using software from
MultiChannel Systems (MC Stimulus 2.0.6.0, MC Rack 3.4.0). Data was sampled at
20 kHz; action potentials were recorded with a 300-3000 Hz band pass filter.
Symmetric biphasic cathodic-first stimulus pulses varied in amplitude from 1-
500µA, and time per phase varied from 20µsec-1ms.

95

Figure 5.3 (Top) Schematic illustrating cellular stimulation and extracellular recording. (Bottom)
Electrophysiological setup.

96
5.2.5 Experimental Protocol
The full experimental process flow is shown in figure 5.4. After the retina was
placed RGC side down on the array a 80 ms full-field light flash (4×10
3
cd/m
2
) was
given to find which electrodes were well coupled to cells. This information assured
that adjacent electrodes were recording from different cells. Only electrodes
showing a signal-to-noise ratio (SNR) >20:1 were studied. All electrical stimulation
protocols used trains of 50 biphasic pulses. A stimulus value was deemed above
threshold if it elicited a response over 75% of the time (i.e., ≥ 38 spikes elicited in a
50 pulse train.) With the exception of stimulus frequency studies, the pulse
repetition rate was 2.5Hz. With the exception of strength-duration studies, all pulses
were 400 µsec per phase, cathodic-first biphasic pulses with no interpulse spacing.

97

Figure 5.4 Experimental flow chart for electrophysiological array studies.

5.2.6 Pharmacological agents and artifact subtraction
The following pharmacological agents were used for the purposes of isolating
the source of the neural response (i.e., presynatically driven responses vs. direct
excitation of RGCs) by mixing with the perfusate. CNQX (antagonist to
AMPA/KAR, 75µM), APV (competitive antagonist to NMDAR, 400µM), and PDA
(agonist to NMDAR, 1mM) were used in nine experiments to block NMDA and
non-NMDA mediated glutamatergic synaptic input to RGCs (Stett et al., 2000;
Sekirnjak et al., 2006). CdCl
2
(Ca
+
channel blocker, 1mM) was also used to abolish

98
synaptic transmission. These manipulations allowed for the direct stimulation of
ganglion cells to be separated from longer latency stimulation of presynaptic cells.
Because direct stimulation has a latency < 7 ms, it usually superimposed on the
stimulus artifact. A method of artifact subtraction has been described by others and
is used here (Figure 2) (Fried et al., 2006; Sekirnjak et al., 2006). An artifact for a
pulse which does not evoke a response can be subtracted from an artifact with a
superimposed response to reveal the elicited spike. TTX (Na
+
channel blocker) was
also employed for the purposes of artifact subtraction as described in the referenced
works.

99

Figure 5.5 (Top) ON and OFF light response of a cell to an 80 ms full-field shutter flash before (control) and
after CdCl
2
application. (Middle) Proof-of-concept of artifact subtraction method to reveal spikes corrupted by
stimulus artifact. (Bottom) Direct ganglion cell excitation is confirmed by noting persistence of recorded spikes
before (control) and after CdCl
2
application.

5.2.7 Finite-element modeling of electrode-electrode interaction
Finite element model simulations were designed with FEMLAB software
(Comsol Inc., Los Angeles, CA). The simulations utilized the conductive media

100
electrostatics module within FEMLAB; Tetrahedral elements were used to define the
model meshes (as described in chapter 3). Microelectrodes were modeled as 200 µm
diameter circular discs (500 µm center-to-center spacing), flush and centered beneath
the base of a larger hemispherical volume of saline medium. The radius of the
medium surrounding the electrode was large enough to approximate an infinite
volume. The hemispherical surface defining the upper boundary served as the
ground electrode (zero-potential boundary). Microelectrode conductivity was 5x10
7

S/m (an isopotential surface) and hemispherical medium conductivity was 1.54 S/m
(Johnson et al., 2005). This model represents two adjacent flat microelectrode discs
immersed in saline solution.
A fixed current of 30 µA was injected into the lower boundary surface of the
microelectrodes.

5.2.7 Data Analysis
Data from electrode recordings was converted to ASCII format using
MC_DataTool V 2.1 and post-stimulus time histograms (PSTHs) were generated
using custom written MATLAB code. The threshold requirement for detecting
spikes was 40 µV. The latency of a spike was defined as the mean of the time points
of the downward and upward crossings of the voltage waveform with the 40 µV
threshold. Standard deviations, best fit curves, and student t-tests were calculated
using EXCEL and Igor Pro V 5.03 software. All plots were also generated using
Igor Pro.

101
5.3 Results

5.3.1 Response latency
Spikes were elicited by both the direct excitation of ganglion cells and the
excitation of bipolar cells and their subsequent glutamatergic excitation of ganglion
cells. The former had latencies of 0.5 – 7 ms with respect to the onset of the
stimulus pulse while the later had latencies of 3 - 400 ms. The source of excitation
was ascertained using both a CNQX/APV/PDA cocktail (in the concentrations
described in 5.2.5) and cadmium chloride. This method has been employed by
others to determine the origin of elicited spikes (Jensen et al., 2005; Margalit &
Thoreson, 2006; Sekirnjak et al., 2006). Engaging inner retinal cells sometimes
resulted in reverberating activity that could last up to 400 µsec (Figure 5.6).

102

Figure 5.6. Post-stimulus time histogram for two cells representative of reverberating response to
stimulation. (50 pulse train of 400 µsec cathodic first biphasic pulses.)

103

5.3.2 Excitation radius vs. stimulus amplitude
Presynaptic excitation of ganglion cells further from the stimulating site had a
higher threshold for stimulation compared with excitation of cells closer to the
stimulating electrode. Cells 50 µm from the electrode edge had a threshold current
of 27.75 ± 8.66 µA compared with cells 433 µm away that had a threshold of 68 ±
23.09 µA (p < 0.01; n = 42 cells; latency of first spike ranged from 3-16 ms) (Figure
5.7).

Figure 5.7 Threshold current vs. distance from 200 µm diameter stimulating electrode edge for
presynaptic excitation (n = 42 cells).

104
Strength-duration data for cells 100, 150, and 235 µm away from the
stimulating electrode edge (n = 9 cells total) was collected for 60, 100, 200, 400,
600, 800, and 1000 µs biphasic pulses (one cell was able to be excited with a 20 µs
pulse; in other instances the minimum pulse duration was set by the maximum
amplitude that could be achieved without saturating the preamplifiers.)
Two figures of merit for strength-duration curves are the rheobase and the
chronaxie. The rheobase is the minimum current required to stimulate a cell,
regardless of duration. (Such a minimum arises from the fact that the cell membrane
will discharge with time, and, therefore, a certain rate of charge delivery is required.)
The chronaxie is the time corresponding to twice the rheobase. Power curve fitting
(
!
y =y
o
+Ax
p
) was in done IGOR Pro software to determine chronaxie and rheobase
values, where the rheobase was defined by the y-axis offset (
!
y
o
) of the curve fit.
Average rheobase values were 6.5 ± 0.7, 27.8 ± 16.0, and 46.2 ± 53.5 µA and
average chronaxie values (
!
y
o
/A
p
) were 179 ± 7.1, 481 ± 95.0, and 600 ± 77.8 µs
for cells 100, 150, and 235 µm away from the stimulating electrode edge,
respectively.

105

Figure 5.7 (Top) Strength-Duration data for cells at different distances (100, 150, and 235 µm)
from 200 µm diameter electrode edge. (Inset) Chronaxie and rheobase values vs. distance from
stimulating electrode edge. (Bottom) log-log plot of strength-duration data.

106
5.3.3 Monopolar, dual monopolar, and bipolar stimulation
For experiments in which a cell was well coupled to a 10 µm recording
electrode in between and equidistant from adjacent 200 µm stimulating electrodes,
the relative charge thresholds for monopolar, dual monopolar (i.e., adjacent
electrodes driven equally and simultaneously), and bipolar stimulation were
compared.

Figure 5.8 (PSTHs) of subthreshold (15 nC/phase) (left column) and suprathreshold (20 nC/phase)
(middle column) stimulation for monopolar (top row), dual monopolar with adjacent stimulating pads
in parallel (second row), and bipolar stimulation (third row). (Right column) Electrostatic finite-
element modeling results showing the paths of electric field lines for (top) monopolar, (middle) dual
monopolar, and (bottom) bipolar stimulation using neighboring stimulating electrodes.

107
Monopolar stimulation required an average charge threshold of 13.3 ± 1.7
nC, compared with 29.4 ± 6.6 nC for dual monopolar stimulation, and 10.0 ± 3.4 nC
for bipolar stimulation (P < 0.02 in comparing the statistical difference of all three
data sets with each other; n = 5 cells). The fact that stimulating two adjacent
electrodes simultaneously with the same polarity increased the charge threshold by
approximately 212% was consistent with finite-element modeling results of
electrostatic interactions between electrodes. Figure 5.8 shows simulation results
from two 200 µm diameter Pt electrodes stimulated in monopolar, dual monopolar,
and bipolar configurations.
The relationship between the membrane voltage (V
m
) and external uniform
electric field for a passive nonconducting spherical cell is
!
V
m
="
3
2
Ercos#
, where r is
the cell radius (Cole, 1933). A threshold voltage must be induced across the
membrane to trigger voltage-gated ion channels and hence the local field vector is a
critical parameter in excitation. The finite-element model predicts a ‘dead region’ at
the midpoint between stimulation sites for the dual monopolar configuration due to
the field of one electrode perturbing the field of the other, which increases the
threshold relative to monopolar stimulation in this location.

5.3.4 Frequency dependence of presynaptic and direct ganglion cell excitation
As the frequency of stimulation is increased to 10 Hz ± 4.67 Hz the
response rate of all presynapitcally driven responses approaches zero (n=6) while

108
ganglion cells can consistently be excited up to 500 Hz (n=3). 400 µs per phase pulses
were delivered at pulse repetition rates ranging from 2.5-500Hz. Both a
CNQX/APV/PDA cocktail and the calcium channel blocker cadmium chloride (1 mM)
were used in separate experiments to distinguish the frequency limits of excitation of
presynaptic cells and ganglion cells.

109

Figure 5.9. (Top) Number of spikes elicited in a 50 pulse stimulus train vs. pulse frequency using 400 µsec
cathodic first biphasic pulses (n = 6). (Bottom) Representative histograms (left) and raw data (right) of
suppression of presynaptically driven responses above 10 Hz for a single cell. The raw data (right) depicts
reponses to successive pulses at pulse frequencies of (from bottom to top) 6.25, 8.33, and 12.5 Hz. The first
pulse of the train is at the bottom of the graph.

110
5.3.5 Charge and charge density thresholds using 200 µm and 10 µm electrodes
Presynaptic excitation of the same ganglion cell (n=5) with both 200 µm and
10 µm diameter electrodes with 400 µsec pulses yielded threshold charge densities of
12 ± 6 nC/cm
2
(average cell-electrode edge distance of 87.5 µm) and 7.66 ± 1.30
mC/cm
2
(average cell-electrode edge distance of 68.8 µm),

respectively. The average
charge required, however, was 12.5 ± 6.2 nC for the 200 µm diameter electrode and
19 ± 3.3 nC for the 10 µm diameter electrode, suggesting that a threshold amount of
charge, not charge density, must be injected into the extracellular space to elicit a
presynaptically driven response.

5.4 Conclusions

Short and long latency spikes that correspond to the direct excitation of ganglion
cells and the excitation of inner retinal cells presynaptic to the RGC layer,
respectively, have been reported by others, and the latencies observed here are in good
agreement with those studies (Jensen et al., 2003; Fried et al., 2006; Sekirnjak et al.,
2006). Reverberating activity lasting 100 ms has been reported in the stimulation of
rabbit retina with conical shaped electrodes (Fried et al., 2006), and with transcorneal
stimulation in cat retina using concentric bipolar electrodes (Shimazu et al., 1999).
Using 200 µm stimulating electrodes we observed similar activity lasting as long as
400 ms. It has been suggested that this activity is due to the excitation of both bipolar
cells, which release the excitatory transmitter glutamate, and amacrine cells, which

111
provide feedforward inhibition to ganglion cells and feedback inhibition to bipolar
cells. The longer lasting reverberating response observed in this study may be due to
the fact that much larger stimulating electrodes were used. These electrodes have an
excitation field that permeates deeper into the retina than smaller, conical electrodes,
engaging a larger portion of the inner retina. Studies in a retinal degenerate animal
will be required to see if this reverberation is curtailed due to anatomical remodeling
and changes in the physiological response of cells (Marc, 2003; Varela et al., 2003).
Our results agree with earlier investigations of threshold versus distance. Ziv
investigated the target cell location using threshold data from the stimulation of
RGCs in rabbit to build a mathematical model of the excitation field around a
conical-tipped electrode (Ziv et al., 2005). It was reported that the threshold had a
1/r
0.84
– 1/r
3.19
dependence on distance from the stimulation tip, in slight contrast with
the strict 1/r
2
dependence predicted by Coulomb’s law. We report a 1/r
1.69
dependence based on threshold amplitude vs. distance from the stimulating edge data
(Figure 5.6) using 200 µm diameter Pt electrodes. This size electrode does not allow
it to be treated as a point source, however, the relationship is in good agreement with
the investigation using conical-tipped electrodes.
The strength-duration data suggests a strategy for coding brightness. If
amplitude will be used for the coding of brightness or gray-level in near-term
prostheses, shorter pulses will allow for a smaller region in the area of the electrode
to be excited over a larger dynamic range. For example, for a 1 ms pulse, the
threshold current required to stimulate a cell 150 µm away (27 µA) is 4.75 times

112
greater than that required to stimulate a cell 100 µm away (6 µA). Comparatively,
for a 200 µs pulse, the threshold current required stimulate a cell 150 µm away (136
µA) is 10.5 times greater than that required to stimulate a cell 100 µm away (13 µA).
This trend is also reflected in the chronaxies of the strength-duration curves at
various distances from the stimulating electrode: sites proximal to the stimulating
electrode have lower chronaxies than those sites further away. These shorter pulses
will more effectively decouple amplitude dependent parameters from excitation area.
A similar conclusion has been drawn using simulations of the excitation of nerve
fibers (Grill & Mortimer, 1996) and in deep brain stimulation (Rizzone et al., 2001),
where shorter pulses more effectively target sites in close proximity to the electrode.
There was greater variability in stimulus threshold for sites farther
away from the stimulating electrode. The standard deviation for the stimulation of
cells 433 µm away is 2.6 times greater than that for cells 50 µm away. This is also
shown clearly in the log-log plots of the strength-duration data (Figure 5.7) where
curves for all cells 50 and 100 µm away are more closely correlated with each other
than all cells greater than 100 µm away. Because this data presented in Figure 5.7
involves the excitation of inner retinal cells, this variability may be due to the fact
that a greater portion of the nonlinear retinal network is engaged at greater distances
from the stimulating site, making the ganglion cell response more complex at these
distances.
Electric field interaction leads to an area of decreased excitability in between
two simultaneously activated electrodes. The difference in the PSTHs between

113
monopolar, dual monopolar, and bipolar stimulation also indicates that a ganglion
cell’s response is dependent on the stimulation source and sink. It is likely that
because multiple inner retinal cells converge on a single ganglion cell, the response
of that ganglion cell is dependent on the different portions of this inner retinal
network that are excited.
Electrode-electrode interactions have been observed in cochlear implant
patients. In psychophysical tests conducted on these patients it was found that
stimulus schemes that allowed for only interleaved pulsing between adjacent
electrodes resulted in improved scores in the closed-set identification of consonants
and the open-set recognition of words and sentences when compared with schemes
incorporating simultaneous pulsing between adjacent electrodes (Lawson et al.,
1993; Kwon & van den Honert, 2006). These interactions then should not be
ignored when making assumptions about the overall field profile at the
multielectrode array-retina interface. At the same time the multielectrode array-
retina spacing must be minimized in order observe this effect. As can be seen in the
finite-element results, at some distance the z-components of these fields align such
that no zone of inactivity should observed. Vector summation in CI subjects in
regions between electrodes has also been reported presumably due to an increased
electrode-cochlea spacing compared with electrode-retina spacing in this study
(White et al., 1984). Another retinal stimulation study employing multiple 10 µm
diameter electrodes with 60 µm center-to-center spacing has showed that each
electrode can stimulate cells in the same manner (i.e., same targeted cell and

114
response latency) whether all electrodes are driven individually or at the same time
(Sekirnjak et al., 2006). It is clear, therefore, that electrode size and pitch will also
determine how much of a factor electrode-electrode interactions will be.
The retina relays visual information to higher cortical centers by varying spike
timing. The maximum frequency at which certain classes of ganglion cells can
generate spikes is approximately 260 Hz, and so it is important that cells can be
excited at this rate if we hope to truly mimic retinal function (O'Brien et al., 2002).
We have found that presynaptic excitation is suppressed at 10 ± 4.67 Hz, while
ganglion cells can directly be excited up to 500 Hz. Our results for presynaptic
excitation are in excellent agreement with the two other studies that investigated the
frequency dependence of stimulation using conical electrodes (40 µm
2
) and Pt
electroplated microdiscs (6-25 µm diameter) (Fried et al., 2006; Sekirnjak et al.,
2006) which also report the suppression of presynaptic excitation of RGCs at ~10Hz
in both the rat and rabbit. We have, however, found an even higher maximum firing
rate of RGCs compared with previous reports. The suppression of presynaptic
excitation has been attributed to comparatively longer inhibitory input current from
amacrine cells (
!
"100 ms) compared to the excitatory bipolar input. This value of
100 ms is in good agreement with a cutoff frequency of approximately 10Hz. This
result suggests that although an epiretinal stimulating electrode can stimulate inner
retinal cells giving rise to a longer latency response, there exists a wide frequency
range in which RGCs are addressed without input from other cells. This would
eliminate complex input due to the reported reverberating response.

115
Understanding the dependence of threshold charge and charge density on
decreasing electrode size will be critical in designing higher resolution prostheses. A
recent review has compiled the results of studies using different sized stimulating
electrodes in retinal stimulation experiments in animals and in subjects (Sekirnjak et
al., 2006). While results in animals generally show that threshold charge decreases
with electrode diameter, the threshold charge density increases. This suggests a
complex relationship between thresholds and electrode size, possibly due to an
inhomogeneous current distribution at the surface of the electrode that evolves
during the time course of the stimulus pulse (Wagner, 1951; Rubinstein et al., 1987).
The trend for thresholds required for eliciting phosphenes in human subjects is less
clear. A study on the perceptual thresholds in three subjects implanted with an
epiretinal device showed there was no significant difference in charge thresholds
between 250 µm and 500 µm diameter electrodes (Mahadevappa et al., 2005). We
observe a similar dependence on absolute charge when stimulating the same RGC
with 10 µm and 200 µm diameter electrodes with similar edge-to-edge spacing
between stimulating and recording electrodes. This relative ratio of electrode sizes
used in this study is the largest reported thus far, and represents sizes that will be
used in near-term implants to those that might be used to address individual cells.
Because comparative thresholds were obtained from stimulating the same target cell
using the same preparation and the same array, the tissue-array spacing was
effectively held constant. While stimulation with large electrodes was approximately
30 times below established electrochemical safe-charge injection limits of 0.35

116
mC/cm
2
(Brummer et al., 1977), stimulation with small electrodes was
approximately 20 times greater than this limit. Further array studies with a large
range of electrode sizes fabricated on the same substrate will provide deeper insight
into the dependence of stimulation on charge and charge density.

117

Chapter 6

Conclusions and Future Work

The electrode-retina interface of an epi-retinal prosthesis implant has been
studied using finite-element modeling, electrochemical methods, and retinal array
electrophysiology. All of the studies presented in this thesis have relied on the
flexibility of custom designed and microfabricated multielectrode arrays to
investigate questions regarding the current distribution at the electrode-electrolyte
interface and it’s interaction with retinal tissue.

6.1 Summary, conclusions and future work of microfabrication,

modeling, and electrochemical characterization of microelectrodes

In chapter 2, the microfabrication process flow of planar MEAs was
described in full detail, and the use of one of these designs (CA3 Replacement
cMEA) in proof-of-concept CA3 replacement experiments in rat hippocampus was
discussed. A dual SiNx/SU-8 insulation layer was employed to decrease the array
shunt capacitance in order to increase recorded fEPSP amplitude. The continued
successful and repeatable electrophysiological testing conducted with these MEAs

118
over a time period of months was critical to making sure that the arrays were robust
and fairly biocompatible (at least for in vitro studies.)
In chapter 3, it was shown that an inhomogeneous current distribution at the
disc electrode surface causes preferential corrosion at the electrode periphery. This
was verified using energy dispersive spectroscopy to show that even after pulsing
with charge-balanced biphasic pulses, a residual metal-oxide remains on the
perimeter of the disc electrode. The recessed disc electrode was presented as means
of changing the electromagnetic boundary conditions at the electrode surface and
providing a more homogeneous current distribution. The effect of the recess depth
on the decrease in current density peaks at the disc edge was quantified using the
conductive media electrostatics module within FEMLAB, and this result
corresponded well with analytical results reported by others (Rubinstein et al., 1987;
Suesserman et al., 1991). However, these investigations focused on larger
electrodes; the required recess depth, which is approximately 1/3 the disc radius,
would have been difficult to implement in a practical device. Because electrodes for
cortical and retinal stimulation are on the order of 50-250 µm, effective recessed
electrodes can be attained using a thick insulating film. Such recessed Pt disc
microelectrodes were fabricated over a range of diameters (50, 100, 150, 200, and
250 µm) and recess depths (1, 5, 9, 12.5, and 20 µm) using SU-8 epoxy-based
photoresist of different viscosities (SU-8 2002, 2005, 2007 and 2010) spun-on at
different spin speeds. The depth was confirmed using profilometry.

119
Recessing the electrode essentially pushed the primary current distribution to
the aperture (Figure 3.10). Therefore, the effective electrode-retina distance will not
increase by the distance of the recess depth. (This proximity is critical to targeting
smaller groups of cells.) At the same, if this primary distribution excites cells
located near the electrode periphery before cells located near the electrode center, the
recess will not rectify this. The homogeneity should only minimize preferential
corrosion and not affect the excitation field profile at the retina itself. These results
should be tempered by the fact that the tissue itself was not included in the finite-
element modeling. We would expect this to change the boundary conditions.
(Evidence of this is given by the fact that electrochemical impedance of Pt and IrOx
stimulating electrodes is increased when placed in contact with the retina compared
with when the electrodes are left in the vitreous humor (Shah et al., 2004).) Future
work in modeling the current distribution should focus on carrying the analysis into
the transient domain in which the chemical engineering and electromagnetic modules
in FEMLAB are coupled and in which the anisotropic conductivity of retinal tissue is
taken into account. This will allow for a square wave stimulus to be used; the
evolution of the current distribution as a function of time can then be visualized.
Future work regarding microfabricated recessed electrodes should include the
long term pulsing and CV, EIS, and EDS characterization. Najafi has reported that in
long-term biphasic pulsing of unrecessed electrodes, corrosion at the electrode
periphery does cause device failure. It is possible that as long-term reliability
becomes more critical, the use of high-aspect ratio SU-8 could yield more stable

120
electrodes. A recess may have additional benefits as well. For electroplated
electrodes, a recess will provide a well for confinement of the grown film, which can
protrude out of the photolithographically predefined area. Especially as future arrays
become more dense this may be a practical method of making sure that adjacent
plated electrodes do not electrically short each other out. Highly dendritic
electroplated Pt which maximizes charge carrying capacity as determined by CV is
also inherently mechanically fragile (Whalen, 2004). A recess can also provide
mechanical protection for the film and create a virtual electrode at the aperture.
Chapter 4 discussed results obtained from electrochemical characterization of
variable diameter MEAs. Cyclic voltammograms normalized for geometric
electrode area and biphasic pulsing suggested that microelectrodes <100 µm in
diameter exhibit an increased current density. This is most likely due to the fact that
these electrodes have a hemispherical diffusion profile compared with the planar
profile of larger electrodes providing increased mass transport (Figure 4.8) (Stulik et
al., 2000; Forster, 2005). This is promising for future prostheses where
microelectrodes <100 µm in diameter may be used. The charge injection of such
electrodes will be greater than those predicted by merely linearly extrapolating the
charge density of larger electrodes over the same potential range (Figure 4.8).
EIS was used to show that at 100 KHz the impedance has a 1/r dependence
on electrode radius as is predicted by the primary current distribution (Newman,
1966). As the frequency decreases from 100 KHz to 10 Hz this relationship shifts to
a 1/r
2
dependence. According to Fourier analysis a square pulse consists of an

121
infinite number of odd-harmonics in frequency space; the higher frequencies define
the leading and falling transients of the pulse. The EIS data then suggests that at the
onset of the stimulus pulse the current is driven at the electrode edge, and this current
is then driven to the electrode center during the time course of the pulse. It is also
shown from the phase angle vs. frequency data that smaller electrodes exhibit
smaller RC time constants, and, therefore, move into an area dependent regime faster
than larger electrodes. (Figure 4.13) (Electrodes with diameters ≤100 µm are in the
defined capacitive regime (i.e., phase angle > 45°) at the highest frequency tested
(100 KHz).) This data was used to conclude that a sin wave stimulus would drive
current to center of the electrode more effectively than a square wave pulse of equal
amplitude. (This is because the former would consist of only the base frequency of
the later, which would consist of a number of higher order harmonics.) As the
electrode-retina spacing is minimized via a number of possible techniques, the
inhomogeneous distribution may result in donut shaped excitation fields. In this case
a sin wave stimulus may provide a more uniform excitation area compared with a
square wave pulse. (EIS characterizes the impedance at the electrode-electrolyte
interface in the linear portion of the I-V curve, and so this must be kept in mind
when making conclusions for stimulus pulses that will operate in the non-linear
portion of this curve. However, EIS has been established by others as a means of
electrode characterization for the purposes of neural stimulation (Franks et al.,
2005). It has also been postulated that the low frequency (10 mHz) charge-transfer
resistance is the first to deviate from the linear small-signal value (e.g., more that a

122
10% change compared with the small-signal value) (McAdams & Jossinet, 1992),
which is well below the typical bandwidth of a stimulus pulse (1-10 KHz).)

6.2 Summary and conclusions for array electrophysiology studies
in retina

In chapter 5 microelectrode array electrophysiology results were discussed.
Custom retinal arrays designed according to the microfabrication process flow
outlined in chapter 2 were used in spatial electrical stimulation studies in isolated
tiger salamander retina. The size and center-to-center spacing of the stimulating
electrodes in these arrays (200 µm and 500 µm, respectively) closely mimicked the
size and spacing of electrodes in near-term retinal implants. Pharmacological agents
were used to isolate direct excitation of ganglion cells from the excitation of other
inner retinal cells. Although direct RGC stimulation was observed, practically all of
the recorded spikes were generated by presynaptic stimulation, and were abolished
with the application of a CNQX/APV/PDA cocktail or cadmium chloride.
Numerous studies have reported responses to electrical stimulation that are
segregated into short and long latency responses, which correspond to direct and
presynaptic excitation of RGCs, respectively. In rabbit 125 µm diameter electrodes
were used to show that epi-retinal stimulation with 1 ms per phase biphasic pulses
yields a short (< 5 ms) and long (8-60 ms) latency response, with the long latency
response being abolished with the application of glutamate receptor antagonists

123
(CNQX/NBQX) (Jensen et al., 2003). Studies using smaller electrodes also resulted
in short and long latency responses (although the exact definition of what time frame
constitutes a short and long latency response varies from one study to the next.)
Using transretinal stimulation with a 10 µm diameter wire electrode placed in the
vitreous humor of anaesthetized rabbits, reverberating ‘burst’ activity was observed
at the 5-15 ms, 30-40 ms, and 60-70 ms post-stimulus pulse time windows (Crapper
& Noell, 1963). Interestingly enough a single spike event attributed to direct
stimulation was observed at 0.5 ms at amplitudes above those required for
presynaptic excitation. It is counterintuitive that a cell closer to the stimulating site
should have a higher threshold compared with a cells further away, although it has
been proposed that cells exhibiting a graded response (i.e., bipolar cells) might have
a lower threshold than RGCs due to cell geometry (Greenberg, 1998). Still another
study also stimulating with approximately the same sized electrodes (6-25 µm
diameter) found that a late or ‘doublet’ response (5-7 ms latency) was always
accompanied by an early spike due to direct excitation (Sekirnjak et al., 2006). The
early response always had the lowest threshold, which is in contrast with the results
of Noell and Crapper.
Several studies have shown that short pulses target ganglion cells and longer
pulses target inner retinal cells due to the longer chronaxies of graded potential cells
(Greenberg, 1998; Fried et al., 2006; Margalit & Thoreson, 2006). The definition of
what constitutes short and long pulses, however, varies. Fried used whole-cell patch
clamp recording in rabbit retina to show that 0.15 ms pulses excite ganglion cells and

124
pulses > 1 ms excite bipolar and amacrine cells. Margalit and Thoreson used the
same technique in tiger salamander retina to show that 0.5 ms pulses evoke transient
inward currents into ganglion cells and pulses > 5 ms evoke sustained inward
currents that were blocked by picrotoxin and strychnine (suggesting GABA
a/c
and
glycine receptor mediated synaptic transmission) and were also blocked in separate
experiments by the NMDA antagonist MK801 and the KA/AMPA antagonist NBQX
(Margalit & Thoreson, 2006). Taken together this means that both excitatory and
inhibitory mechanisms are engaged with epiretinal stimulation. (The investigators
attributed the differences in pulse widths required for selective stimulation to the fact
that salamander cell size is larger than mammalian cell size (i.e., 20 µm and 10-12
µm, respectively).) A more recent investigation in which anesthetized rabbits were
epiretinally stimulated with 0.25 and 2 ms pulses (500 µm diameter bipolar
electrodes), and electrically evoked cortical potentials were recorded from the
occipital surface (EECPs), found that while pharmacological manipulations (NBQX,
D-AP7, APB) decreased ERG and light induced cortical potentials, EECPs were
unaffected (Shah et al., 2006). It was concluded that inner retinal elements do not
contribute significantly to the EECP, but may do so at even longer duration pulses
(10 ms) based on prior in vitro studies by the same investigators (Jensen et al.,
2005). It is difficult to believe that no inner retinal cells are excited with stimulation
of large electrodes. However, the fact that cortical potentials are not diminished by
effectively blocking several parallel pathways to the RGC layer is interesting. It is
known that higher visual processing regions such as LGN can mediate the input in to

125
V1. Therefore, a spike that is part of a larger retinal code may not contribute to a
cortical potential or elicited percept. This is one limitation and valid point of
criticism of drawing conclusions from isolated retina studies and making hypotheses
regarding elicited percepts.
Strength-duration data suggests that if amplitude will be used for the coding
of brightness or gray-level in near-term prostheses, shorter pulses (200 µsec) will
allow for a smaller region in the area of the electrode to be excited over a larger
dynamic range compared with longer pulses (1 ms) (Figure 5.8). This is contingent
upon the fact that an increase in brightness is due to an individual cell responding
differently to suprathreshold versus threshold stimulation (e.g., spike rate increases
with amplitude). The alternative is that more cells in the vicinity of the electrode are
recruited, in which case the above hypothesis will not hold. From preliminary
patient testing the underlying neural mechanism is yet uncertain.

6.3 Future work for array electrophysiology studies in retina

6.3.1 Retinal degenerate model
Retinal degeneration results in remodeling of the remnant inner retina
including cell migration, the formation of anomalous dendritic connections, and
changes in cell physiology (Marc, 2003; Varela et al., 2003). Such changes have
been shown to increase thresholds for stimulation in electrophysiology studies. In
comparing epiretinal stimulation using Pt wire electrodes in normal and retinal

126
degenerate mice (rd1/rd1, strain C3H/HeJ), it was found that the response latency
was longer and threshold significantly higher for the later (Suzuki et al., 2004;
O'Hearn T et al., 2006). For these reasons, future array studies should use
degenerate model mice (rd1) or rats (P23H or S334ter) (Steinberg et al., 1996; Lee,
2003). An important question that needs to be addressed is the dependence of
stimulus threshold on the extent of degeneration. This may play a role in
determining when implantation in a patient would be optimal.

6.3.2 Axonal vs. Somatic stimulation
A retinal prosthesis attempts to take advantage of the fact that there is spatial
correspondence between a region of the visual field and ganglion cell location.
However, ganglion cell axons, which form the optic nerve bundle, are also
electrically excitable, and axonal excitation may not create punctate phosphenes well
correlated with the visual field. Modeling has shown that ganglion cells have a
slightly lower threshold (by 20%-73% depending on neuronal model used) compared
with their axons (Greenberg et al., 1999). Experimental results in rabbit have shown
that cathodic pulses yield a mean threshold for somatic stimulation which is half that
of axonal stimulation (Jensen et al., 2003). One reason why the cell soma may be a
preferential site for spike initiation is because there is an increased number of
voltage-sensitive sodium channels (Na
v
1.6) at the initial segment compared with the
distribution along the rest of the fiber (Boiko et al., 2003). Excitation of axons is
nevertheless a legitimate concern for an epi-retinal approach.

127
Axonal stimulation of myelinated and unmyelinated fibers has been well
investigated (Ranck, 1975; Fang & Mortimer, 1991; McIntyre & Grill, 1999). The
lowest thresholds are achieved when the source and sink electrodes create a voltage
gradient parallel to the fiber (Rushton, 1927). This longitudinal orientation yields
thresholds 4-5 times below those achieved with the transverse orientation in
stimulation of dorsal columns in cat (Rudin & Eisenman, 1954). As electrode size
decreases it may be possible to minimize axonal stimulation if cell somas are
stimulated with a bipolar pair transverse to the radial fan out of the nerve fibers
(Figure 6.1). Both configurations can be adequately compared using an in vitro array
setup. Retrograde florescent dye loading of retinal cells would help visualize cell
somas and axons with respect to the stimulating electrodes.

Figure 6.1 Electrode geometry for longitudinal and transverse stimulation setups with respect to RGC
axons. Transverse stimulation should yield higher thresholds for axonal stimulation compared with
longitudinal stimulation. Because RGC axons form a bundle, axonal stimulation may lead to
anomalous percepts such as line-shaped phosphenes. (Confocal microscopy image of mouse retina
stained with 10% solution of dextran-tetramethylrhodamine, Courtesy of Matthew Behrend)

128
6.3.3 Charge vs. Charge Density
Presynaptic stimulation of the same ganglion cell with both 200 µm and 10
µm diameter electrodes yielded threshold charge densities of 12 ± 6 nC/cm
2
and 7.66
± 1.30 mC/cm
2
, respectively, while the required charge was 12.5 ± 6.2 nC and 19 ±
3.3 nC. It is slightly counterintuitive to think that charge is the mitigating factor in
stimulation since is the local field gradient that triggers voltage-gated channels.
Studies in three retinal implant subjects found no significant difference in the
thresholds for 250 and 500 µm diameter electrodes. However a recent review of 32
publications on the stimulation of retina (in vitro, in vivo, and in subjects) shows that
threshold charge does decrease with electrode size (i.e., a line of constant charge
density (0.2 mC/cm
2
) fits the surveyed data with a correlation coefficient with
R
2
=0.75) (Sekirnjak et al., 2006). This data comprises many studies with many
variations in experimental methods (e.g., animal model, electrode type (wire vs.
planar array), and electrode-retina spacing). In order to truly answer the question of
charge vs. charge density an array should be used with varying stimulating electrode
size. Figure 6.2 shows a microfabricated array with two sets of 200, 150, 100, 50,
25, and 10 µm Pt stimulating electrodes with 10 µm diameter recording electrodes
placed in proximity of the stimulating electrode edge (8 µm edge-to-edge spacing)
(These arrays have been fabricated according to the microfabrication process flow
outlined in Chapter 2). We know from the ‘Custom Retinal MEAs’ discussed in
chapter 5 that these sized recording electrodes can record single unit activity with
good SNR (> 20 SNR). These stimulating electrodes include sizes that will used be

129
in near-term implants to those that might be used to target single cells. It is likely
that smaller electrodes will also show a larger difference in the excitation of RGCs
and presynaptic cells than larger electrodes due to the respective electric field
profiles. As electrophysiology and visual psychophysics move forward and more
patients are implanted with retinal prostheses, we will begin to better understand how
the excitation of different cell classes effects the elicited percept.

Figure 6.2 Photomicrograph of Variable Diameter MEA with 200, 150, 100, 50, 25, and 10 µm Pt
stimulating electrodes with 10 µm diameter recording electrodes placed in proximity of the
stimulating electrode edge (8 µm edge-to-edge spacing).

130
Limits of this experiment and Future Prospects

In vitro stimulation of the isolated retina has a one obvious drawback in that
cortical mechanisms cannot be studied. This experimental setup has enabled a
number of elegant studies on fundamental neural mechanisms. However, as we
move forward it is important to pull from both retinal electrophysiology and the
emerging field of neural prosthetics. Cochlear prostheses and early retinal
prostheses remind us that there is immense plasticity in the adult auditory and visual
cortex and stimulus deprived cortices. In a famous experiment in which a subject
wore glasses with image inverting prisms for a period of time, it was found that the
visual cortex rectifies the image perceived by the subject within two weeks (Stratton,
1897). This is powerful example of the way in which the cortex can respond to
changes in sensory input. Cross-modal plasticity has also been observed in visual
and auditory cortices of deaf and blind subjects (Giraud et al., 2001; Fine et al.,
2003). It was found using fMRI that a portion of the dormant area (i.e. auditory
cortex in blind subjects and visual cortex in deaf subjects) is recruited by the other
still functioning sensory modality. Cochlear implant patients also show improved
performance in auditory tasks with time post-implantation. It remains to be seen to
what extent plasticity and learning will help retinal implant subjects. For all of these
reasons fMRI, PET, and visual psychophysics will be vital in understanding the
origin of electrically elicited percepts in conjunction with retinal electrophysiology.

131
It is important to apply years of basic retinal neurobiological and
electrophysiological research to future implants, and the same time be willing to let
go findings that are only pertinent to a healthy retina and visual cortex. The
importance and feasibility of trying recreate the retinal code for any given visual
scene by addressing individual RGCs with small electrodes needs to be assessed.
This scheme is essentially the ‘holy grail’ for a retinal prothesis, however, there are
two realities that may preclude an implant from restoring fine grain vision. Firstly, it
has been reported that 30% of RGCs survive in post-mortem morphological studies
of RP paients (Stone et al., 1992; Santos et al., 1997; Kim et al., 2002a), and it is
most likely the case that this number is strongly dependent on the advancement of
the disease. In either case, a good percentage of RGCs are lost, preventing full
restoration of sight. Secondly, while it is said that the RGC layer consists of a
monolayer of cells, it is not a monolayer in the strict sense. There is some stacking
of cells making it practically impossible to address RGCs closer to the inner retina
without exciting cells closer to the electrode (figure 6.3). If the limits of electrical
stimulation preclude this then perhaps this goal will be left to other burgeoning
technologies that lend themselves more naturally to selectivity (e.g., gene therapy
(Preising & Heegard, 2004) or the expression of channelrhodopsin-2 (a naturally
occurring light sensitive ion channel) in specific cell classes (Bi et al., 2006).)
Despite these concerns, early retinal implants have been extremely successful
in restoring limited vision to patients. The immediate obstacles are those concerning
implant packaging, biocompatibility, array-retina proximity, and mechanical and

132
electrochemical safety. However, as these engineering issues are dealt with over
time, devices that will afford greater spatial resolution will be implanted in patients.
This field will be in the unique position of giving hope to those suffering from retinal
blindness worldwide, while also allowing us to study the ability of the human brain
to interface with an artificial device.

Figure 6.3 Cross-section of retina with schematic depiction of small epi-retinal electrodes. Can cell
B be selectively excited without exciting cell A? (image modified from
http://webvision.med.utah.edu/)

133

Bibliography

Ahuja A, Nasiatka P, Song D, TW B & Tanguay A. (2004). A biomimetic electronic
prosthetic for hippocampus: Planar conformal multielectrode arrays for
VLSI/hippocampal slice interface. In Soc of Neurosci. San Diego.

Angerstein-Kozlowska H, Conway BE, Tellefsen K & Barnett B. (1989).
Stochastically-gated surface processes involving anions in oxidation of Au: Time
resolution of processes down to 0.25% coverages and 50 microsecond time-scales.
Electrochemica Acta 34, 1045-1056.

Bard AJ & Faulkner LR. (1980). Electrochemical Methods. John Wiley and Sons,
Inc.

Barsoukov E & Macdonald JR. (2005). Impedance Spectroscopy: Theory,
Experiment, and Applications. Wiley Interscience Publications.

Berger TW, Ahuja A, Courellis SH, Deadwyler SA, Erinjippurath G, Gerhardt GA,
Gholmieh G, Granacki JJ, Hampson R, Hsaio MC, LaCoss J, Marmarelis VZ,
Nasiatka P, Srinivasan V, Song D, Tanguay AR & Wills J. (2005). Restoring lost
cognitive function. IEEE Eng Med Biol Mag 24, 30-44.

Berson EL. (1993). Retinitis pigmentosa. The Friedenwald Lecture. Invest
Ophthalmol Vis Sci 35, 1659.

Bi A, Cui J, Ma YP, Olshevskaya E, Pu M, Dizhoor AM & Pan ZH. (2006). Ectopic
Expression of A Microbial-Type Rhodopsin Restores Visual Responses in Mice with
Photoreceptor Degeneration. Neuron 50, 23-33.

Boiko T, Van Wart A, Caldwell JH, Levinson SR, Trimmer JS & Matthews G.
(2003). Functional specialization of the axon initial segment by isoform-specific
sodium channel targeting. J Neurosci 23, 2306-2313.

Brindley GS & Lewin WS. (1968). The sensations produced by electrical stimulation
of the visual cortex. J Physiol 196, 479-493.

Brindley GS & Rushton D. (1974). Implanted stimulators of the visual cortex as
visual prosthetic devices. TransAm Acad Ophthalmol Otolaryngol 78, 741-745.

134
Brivanlou IH, Warland DK & Meister M. (1998). Mechanisms of concerted firing
among retinal ganglion cells. Neuron 20, 527-539.

Brummer SB, McHardy J & Turner MJ. (1977). Electrical stimulation with Pt
electrodes: Trace analysis for dissolved platinum and other dissolved electrochemical
products. Brain Behav Evol 14, 10-22.

Brummer SB & Turner MJ. (1977). Electrical stimulation with Pt electrodes: II-
estimation of maximum surface redox (theoretical non-gassing) limits. IEEE Trans
Biomed Eng 24, 440-443.

Cha K, Horch KW & Normann RA. (1992). Mobility performance with a pixelized
vision system. Vision Res 32, 1367-1372.

Chow AY, Chow VY, Packo K, H,, Pollack JS, Peyman GA & Schuchard R. (2004).
The artifical silicon retina microchip for the treatment of vision loss from retinits
pigmentosa. Arch Ophthalmol 122, 460-469.

Chow AY & Peachey NS. (1998). The subretinal microphotodiode array retinal
prosthesis. Ophthalmic Res 30, 195-198.

Cole KS. (1933). Electric conductance of biological systems. In Cold Spring Harbor
Symp Quant Biol pp. 107-116.

Crapper DR & Noell WK. (1963). Retinal Excitation and Inhibition from Direct
Electrical Stimulation. J Neurophysiol 26, 924-947.

Deconinick J. (1992). Lecture notes in Engineering: Current distribtutions and
electrode shape changes in electrochemical systems Springer-Verlag, Berlin.

Dobelle WH & Mladejovsky MG. (1974). Phosphenes produced by electrical
stimulation of human occipital cortex, and their application to the development of a
prosthesis for the blind. J Physiol 243, 553-576.

Dobelle WH, Mladejovsky MG, Evans JR, Roberts TS & Girvin JP. (1976).
"Braille" reading by a blind volunteer by visual cortex stimulation. Nature 259, 111-
112.

Donaldson PE. (1974). The stability of tantalum-pentoxide films in vivo. Med Biol
Eng 12, 131-135.

Doty RW & Grimm FR. (1962). Cortical responses to local electrical stimulation of
retina. Exp Neurol 5, 319-334.

135
Dowson WW & Radke ND. (1977). The electrical stimulation of the retina by
indwelling electrodes. Invest Ophthalmol Vis Sci 16, 249-252.

Duan YY, Clark GM & Cowan RSC. (2001). Factors determining and limiting the
impedance behavior of implanted bioelectrodes. Proceedings of SPIE: Smart
Structures and Devices 4235, 498-508.

Fang ZP & Mortimer JT. (1991). Alternate excitation of large and small axons with
different stimulation waveforms: an application to muscle activation. Med Biol Eng
Comput 29, 543-547.

Fine I, Wade AR, Brewer AA, May MG, Goodman DF, Boynton GM, Wandell BA
& MacLeod DI. (2003). Long-term deprivation affects visual perception and cortex.
Nat Neurosci 6, 915-916.

Foerster O. (1929). Beitrage zur Pathophysiologie der Sehbahn und der Spehsphare.
J Psychol Neurol 39, 435-463.

Forster RJ. (2005). Microelectrodes - Retrospect and Prospect. In Electroanalytical
Methods, ed. Bard AJ & Stratman M. Wiley Publishing.

Franks W, Schenker P, Schmutz A & Hierlemann A. (2005). Impedance
characterization and modeling of electrodes for biomedical applications. IEEE Trans
Biomed Eng 7, 1295-1302.

Fricke H. (1923). The electric capacity of cell suspensions. Physic Rev 21, 708-709.

Fricke H. (1924). A mathematical treatment of the electrical conductivity of colloids
and cell suspensions. J Gen Physiol 6, 375-384.

Fried SI, Hsueh HA & Werblin FS. (2006). A method for generating precise
temporal patterns of retinal spiking using prosthetic stimulation. J Neurophysiol 95,
970-978.

Friedman DS, O'Colmain BJ, Munoz B, Tomany SC, McCarty C, de Jong PT,
Nemesure B, Mitchell P & Kempen J. (2004). Prevalence of age-related macular
degeneration in the United States. Arch Ophthalmol 122, 564-572.

Fries P, Neuenschwander S, Engel AK, Goebel R & Singer W. (2001). Rapid feature
selective neuronal synchronization through correlated latency shifting. Nat Neurosci
4, 194-200.

Genentech I. (2006). New Lucentis ranibizumab injection. In
http://www.lucentis.com/lucentis/about_lucentisjsp.

136

Germain PS, Pell WG & Conway BE. (2004). Evaluation and origins of the
difference between double-layer capacitance behaviour at Au-metal and oxidized Au
surfaces. Electrochem Acta 49, 1775-1788.

Gholmieh G, Soussou W, Han M, Ahuja A, Hsiao MC, Song D, Tanguay AR, Jr. &
Berger TW. (2006). Custom-designed high-density conformal planar multielectrode
arrays for brain slice electrophysiology. J Neurosci Methods 152, 116-129.

Giraud AL, Price CJ, Graham JM, Truy E & Frackowiak RS. (2001). Cross-modal
plasticity underpins language recovery after cochlear implantation. Neuron 30, 657-
663.

Goldstein J, Newbury D, Joy D, Lyman C, Echlin P, Lifshin E, Sawyer L & Michael
J. (2003). Scanning electron microscopy and x-ray microanalysis. Kluwer
Academic/Plenum Press, New York.

Greenberg R. (1998). Analysis of electrical stimulation of the vertbrate retina - work
towards a retinal prosthesis. Dissertation, Department of Biomedical Engineering,
Johns Hopkins University.

Greenberg RJ, Velte TJ, Humayun MS, Scarlatis GN & de Juan E, Jr. (1999). A
computational model of electrical stimulation of the retinal ganglion cell. IEEE
Trans Biomed Eng 46, 505-514.

Grill WM, Jr. & Mortimer JT. (1996). The effect of stimulus pulse duration on
selectivity of neural stimulation. IEEE Trans Biomed Eng 43, 161-166.

Gross GW, Wen WY & Lin JW. (1985). Transparent indium-tin oxide electrode
patterns for extracellular, multisite recording in neuronal cultures. J Neurosci
Methods 15, 243-252.

Grumet AE, Wyatt JL, Jr. & Rizzo JF, 3rd. (2000). Multi-electrode stimulation and
recording in the isolated retina. J Neurosci Methods 101, 31-42.

Guyton DL & Hambrecht FT. (1973). Capacitor electrode stimulates nerve or muscle
without oxidation-reduction reactions. Science 181, 74-76.

Hammerle H, Kobuch K, Kohler K, Nisch W, Sachs H & Stelzle M. (2002).
Biostability of micro-photodiode arrays for subretinal implantation. Biomaterials 23,
797-804.

137
Hayes JS, Yin VT, Piyathaisere D, Weiland JD, Humayun MS & Dagnelie G.
(2003). Visually guided performance of simple tasks using simulated prosthetic
vision. Artif Organs 27, 1016-1028.

Heitz E & Kreysa G. (1996). Principles of Electrochemical Engineering. VCH
Publishers, New York.

Horch KW & Dhillon GS. (2004). Neuroprosthetics: Theory and Practice. World
Scientific Publishing Co. , River Edge, NJ.

Hu C-C & Liu K-Y. (1999). Voltammetric investigation of platinum oxides. I.
Effects of aging on their formation/reduction behavior as well as catalytic activities
for methanol oxidation. Electrochemica Acta 44, 2727-2738.

Humayun M, Propst R, de Juan E, Jr., McCormick K & Hickingbotham D. (1994).
Bipolar surface electrical stimulation of the vertebrate retina. Arch Ophthalmol 112,
110-116.

Humayun MS. (2001). Intraocular retinal prosthesis. Trans Am Ophthalmol Soc 99,
271-300.

Humayun MS, de Juan E, Jr., Dagnelie G, Greenberg RJ, Propst RH & Phillips DH.
(1996). Visual perception elicited by electrical stimulation of retina in blind humans.
Arch Ophthalmol 114, 40-46.

Humayun MS, de Juan E, Jr., Weiland JD, Dagnelie G, Katona S, Greenberg R &
Suzuki S. (1999). Pattern electrical stimulation of the human retina. Vision Res 39,
2569-2576.

Humayun MS, Freda R, Fine I, Roy A, Fujii G, Greenberg RJ, Little J, Mech B,
Weiland J & de Juan J, E. (2005). Implanted intraocular retinal prosthesis in six blind
subjects. In ARVO. Fort Lauderdale, Fl.

Humayun MS, Weiland JD, Fujii GY, Greenberg R, Williamson R, Little J, Mech B,
Cimmarusti V, Van Boemel G, Dagnelie G & de Juan E. (2003). Visual perception
in a blind subject with a chronic microelectronic retinal prosthesis. Vision Res 43,
2573-2581.

Hung A, Zhou D, Greenberg R & Judy JW. (2002). Micromachined electrodes for
retinal prostheses. In EMBS. Houston.

Jensen RJ, Rizzo JF, 3rd, Ziv OR, Grumet A & Wyatt J. (2003). Thresholds for
activation of rabbit retinal ganglion cells with an ultrafine, extracellular
microelectrode. Invest Ophthalmol Vis Sci 44, 3533-3543.

138

Jensen RJ, Ziv OR & Rizzo JF, 3rd. (2005). Thresholds for activation of rabbit
retinal ganglion cells with relatively large, extracellular microelectrodes. Invest
Ophthalmol Vis Sci 46, 1486-1496.

Johnson AM, Sadoway DR, Cima MJ & Langer R. (2005). Design and testing of an
impedance-based sensor for monitoring drug-delivery. Jour of Electrochem Soc 152,
H6-H11.

Jorcin JB, Orazem ME, Pebere N & Tribolle B. (2004). CPE analysis by local
impedance analysis. In EIS Symposium. Coco Beach, Fl.

Kandel ER, Schwartz JH & Jessel TM. (2000). Principles of Neural Science: 4th
Edition. McGraw-Hill Professional Publishing, New York.

Karny H. (1975). Clinical and physiological aspects of the corticla visual prosthesis.
Surv Ophthalmol 20, 47-58.

Katznelson Y. (1976). An introduction to harmonic analysis Dover Publications,
Inc., New York.

Kim MT, Soussou W, Gholmieh G, Ahuja A, Tanguay A, Berger TW & Brinton RD.
(2006). 17beta-Estradiol potentiates field excitatory postsynaptic potentials within
each subfield of the hippocampus with greatest potentiation of the
associational/commissural afferents of CA3. Neuroscience 141, 391-406.

Kim SY, Sadda S, Humayun MS, de Juan E, Jr., Melia BM & Green WR. (2002a).
Morphometric analysis of the macula in eyes with geographic atrophy due to age-
related macular degeneration. Retina 22, 464-470.

Kim SY, Sadda S, Pearlman J, Humayun MS, de Juan E, Jr., Melia BM & Green
WR. (2002b). Morphometric analysis of the macula in eyes with disciform age-
related macular degeneration. Retina 22, 471-477.

Klein R, Klein BE, Jensen SC & Meuer SM. (1997). The five-year incidence and
progression of age-related maculopathy: the Beaver Dam Eye Study. Ophthalmology
104, 7-21.

Kotler S. (2002). Vision Quest. In Wired Magazine.

Krause F & Schum H. (1931). Die epiliptischen Erkankungen. In Neue Deutsche
Shirurgie, ed. Kunter H. Stuttgart.

139
Kuttenkeuler C, Wilke R, Volker M, Sachs H, Gabel VP, Gekeler F, Besch D, Bartz-
Schmidt K & Zrenner E. (2006). Subretinal implant study group. Subretinal chronic
active multi-electrode arrays in blind patients: fundus appearance, optical coherence
tomography and angiography. In ARVO. Fort Lauderdale, FL.

Kwon BJ & van den Honert C. (2006). Dual-electrode pitch discrimination with
sequential interleaved stimulation by cochlear implant users. J Acoust Soc Am 120,
EL1-6.

Lawson DT, Wilson BS & Finley CC. (1993). New processing strategies for
multichannel cochlear prostheses. Prog Brain Res 97, 313-321.

Lee D. (2003). Photoreceptor degeneration in Pro23His and S334TER transgenic
rats. Adv Exp Med Biol 55, 297-302.

Lilly JC, Hughes JR, Alvord EC, Jr. & Galkin TW. (1955). Brief, noninjurious
electric waveform for stimulation of the brain. Science 121, 468-469.

Lund RD. (2001). Cell Transplantation as a treatment for retinal disease. Prog Retin
Eye Res 20, 415-449.

MacLaren RE, Pearson RA, MacNeil A, Douglas RH, Salt TE, Akimoto M, Swaroop
A, Sowden JC & Ali RR. (2006). Retinal repair by transplantation of photoreceptor
precursors. Nature 444, 203-207.

Mahadevappa M, Weiland JD, Yanai D, Fine I, Greenberg RJ & Humayun MS.
(2005). Perceptual thresholds and electrode impedance in three retinal prosthesis
subjects. IEEE Trans Neural Syst Rehabil Eng 13, 201-206.

Mailley S, Hyland M, Mailiey P, McLaughlin JA & McAdams ET. (2004). Thin film
platinum cuff electrodes for neurostimulation: in vitro approach of safe
neurostimulation parameters. Bioelectrochemistry 63, 359-364.

Marc RE. (2003). Neural remodeling in retinal degeneration. Prog Retin Eye Res 22,
607.

Margalit E & Thoreson WB. (2006). Inner retinal mechanisms engaged by retinal
electrical stimulation. Invest Ophthalmol Vis Sci 47, 2606-2612.

Masland RH. (2001). The fundamental plan of the retina. Nat Neurosci 4, 877-886.

McAdams ET & Jossinet J. (1992). A physical interpretation of Schwan's limit
current of linearity. Ann Biomed Eng 20, 307-319.

140
McCreery DB, Agnew WF & Bullara LA. (2002). The effects of prolonged
intracortical microstimulation on the excitability of pyramidal tract neurons in the
cat. Ann Biomed Eng 30, 107-119.

McCreery DB, Agnew WF, Yuen TG & Bullara LA. (1988). Comparison of neural
damage induced by electrical stimulation with faradaic and capacitor electrodes. Ann
Biomed Eng 16, 463-481.

McIntyre CC & Grill WM. (1999). Excitation of central nervous system neurons by
nonuniform electric fields. Biophys J 76, 878-888.

Meister M, Lagnado L & Baylor DA. (1995). Concerted signaling by retinal
ganglion cells. Science 270, 1207-1210.

Merrill DR, Bikson M & Jefferys JG. (2005). Electrical stimulation of excitable
tissue: design of efficacious and safe protocols. J Neurosci Methods 141, 171-198.

Milner B. (1970). Memory and the medial temporal regions of the brain. Academic,
New York.

Najafi K. (1994). Inductively Powered Microstimulator, pp. 1-80. University of
Michigan, Ann Arbor.

Newman JS. (1966). Current distribution on a rotating disc below the limiting
current. J Electrochem Soc 113, 1235-1236.

Normann RA, Maynard E.M., Rousche, P.J., et al. (1999). A neural interface for a
cortical vision prosthesis. Vision Res 39, 2577-2587.

Novak JL & Wheeler BC. (1986). Recording from the Aplysia abdominal ganglion
with a planar microelectrode array. IEEE Trans Biomed Eng 33, 196-202.

O'Brien BJ, Isayama T, Richardson R & Berson DM. (2002). Intrinsic physiological
properties of cat retinal ganglion cells. J Physiol 538, 787-802.

O'Hearn T M, Sadda SR, Weiland JD, Maia M, Margalit E & Humayun MS. (2006).
Electrical stimulation in normal and retinal degeneration (rd1) isolated mouse retina.
Vision Res.

Oldham KB. (2004). The RC time 'constant' at a disk electrode. ElectrochemComm
6, 210-214.

141
Overmyer KM, Pearce JA & DeWitt DP. (1979). Measurements of temperature
distributions at the electro-surgical dispersive electrode sites. Trans ASME, J
Biomech Eng 101, 66-72.

Palanker D, Vankov A, Huie P & Baccu S. (2005). Design of a high-resolution
optoelectronic retinal prosthesis. 2, S105-120.

Potts AM & Inoue J. (1969). The electrically evoked response (EER) of the visual
system. II. Effect of adaptation and retinitis pigmentosa. Invest Ophthalmol 8, 605-
612.

Preising MN & Heegard S. (2004). Recent advances in early-onset severe retinal
degeneration: more than just basic research. Trends Mol Med 10, 51-54.

Rakoczy EP, Yu MJ, Nusinowitz S, Chang B & Heckenlively JR. (2006). Mouse
models of age-related macular degeneration. Exp Eye Res 82, 741-752.

Ranck JB, Jr. (1975). Which elements are excited in electrical stimulation of
mammalian central nervous system: a review. Brain Res 98, 417-440.

Rizzo JF, 3rd, Wyatt J, Humayun M, de Juan E, Liu W, Chow A, Eckmiller R,
Zrenner E, Yagi T & Abrams G. (2001). Retinal prosthesis: an encouraging first
decade with major challenges ahead. Ophthalmology 108, 13-14.

Rizzone M, Lanotte M, Bergamasco B, Tavella A, Torre E, Faccani G, Melcarne A
& Lopiano L. (2001). Deep brain stimulation of the subthalamic nucleus in
Parkinson's disease: effects of variation in stimulation parameters. J Neurol
Neurosurg Psychiatry 71, 215-219.

Roblee LS & Rose TL. (1990). Electrochemical guidelines for selection of protocols
and electrode materials for neural stimulation. In Neural Prosthesis: Fundamental
Studies, ed. Agnew WF & McCreery DB. Englewood Cliffs.

Rodieck RW. (1998). The first Steps in Seeing. Sinauer Associates, Inc., Sunderland.

Rubinstein JT, Spelman FA, Soma M & Susserman MF. (1987). Current density
profiles of surface mounted and recessed electrodes for neural prostheses. IEEE
Trans Biomed Eng 34, 864-875.

Rudin DO & Eisenman G. (1954). The action potential of spinal axons in vitro. J
Gen Physiol 37, 505-538.

142
Rushton WA. (1927). The effect upon the threshold for nervous excitation of the
length of nerve exposed, and the angle between current and nerve. J Physiol 63, 357-
377.

Sandison ME, Anicet N, Glidle A & Cooper JM. (2002). Optimization of the
geometry and porosity of microelectrode arrays for sensor design Anal Chem 74,
5717-5725.

Santos A, Humayun MS, de Juan E, Jr., Greenburg RJ, Marsh MJ, Klock IB &
Milam AH. (1997). Preservation of the inner retina in retinitis pigmentosa. A
morphometric analysis. Arch Ophthalmol 115, 511-515.

Schiefer MA & Grill WM. (2006). Sites of neuronal excitation by epiretinal
electrical stimulation. IEEE Trans Neural Syst Rehabil Eng 14, 5-13.

Schmidt EM, Bak MJ, Hambrecht FT, Kufta CV, O'Rourke DK & Vallabhanath P.
(1996). Feasibility of a visual prosthesis for the blind based on intracortical
microstimulation of the visual cortex. Brain 119 ( Pt 2), 507-522.

Schmidt EM, Bak, M.J., Hambrecht, F.T. (1992). Feasibility of a visual prosthesis
for the blind based on intracortical microstimulation of the visual cortex. Brain 119,
507-522.

Segev R, Puchalla J & Berry MJ. (2006). Functional organization of ganglion cells in
the salamander retina. J Neurophysiol 95, 2277-2292.

Sekirnjak C, Hottowy P, Sher A, Dabrowski W, Litke AM & Chichilnisky EJ.
(2006). Electrical stimulation of mammalian retinal ganglion cells with
multielectrode arrays. J Neurophysiol 95, 3311-3327.

Shah HA, Montezuma SR & Rizzo JF. (2006). In vivo electrical stimulation of rabbit
retina: Effect of stimulus duration and electrical field orientation. Exp Eye Res.

Shah RN & Webster JG. (1979). Burns under electrosurgical dispersive electrodes.
In Proc AAMI Annu Meeting. Las Vegas, NV.

Shah S, Chu A, Zhou D, Greenberg R, Guven D, Humayun MS & Weiland JD.
(2004). Intraocular impedance as a function of the position in the eye, electrode
material, and electrode size. In EMBS San Francisco.

Shimazu K, Miyake Y & Watanabe S. (1999). Retinal ganglion cell response
properties in the transcorneal electrically evoked response of the visual system.
Vision Res 39, 2251-2260.

143
Shlens J, Field GD, Gauthier JL, Grivich MI, Petrusca D, Sher A, Litke AM &
Chichilnisky EJ. (2006). The structure of multi-neuron firing patterns in primate
retina. J Neurosci 26, 8254-8266.

Stamford JA. (1989). In vivo voltammetry - prospects for the next decade. Trends in
Neurosciences 12, 407-412.

Stamford JA. (1990). Fast cyclic voltammetry: Measuring transmitter release in 'real
time'. Jour of Neuroscience Methods 34, 67-72.

Steinberg RH, Flannery JG, Naash M, Oh P, Matthes MT, Yasumura D, Lau-
Villacorta C, Chen J & LaVail NM. (1996). Transgenic rat models of inherited
retinal degeneration caused by mutant opsin genes. Invest Ophthalmol Vis Sci, S698.

Stett A, Barth W, Weiss S, Haemmerle H & Zrenner E. (2000). Electrical multisite
stimulation of the isolated chicken retina. Vision Res 40, 1785-1795.

Stone JL, Barlow WE, Humayun MS, de Juan E, Jr. & Milam AH. (1992).
Morphometric analysis of macular photoreceptors and ganglion cells in retinas with
retinitis pigmentosa. Arch Ophthalmol 110, 1634-1639.

Stratton GM. (1897). Vision without inversion of retinal image. Psychological
Review 4, 341-360.

Struijk JJ. (2004). Passive models of excitable cells. In Neuroprosthetics: Theory
and Practice, ed. Horch KW & Dhillon GS. World Scientific Publishing, Singapore.

Stulik K, Amatore C, Holub K, Marecek M & Kutner W. (2000). Microelectrodes:
Definitions, Characterization, and Applications. Pure App Chem 72, 1483-1492.

Suesserman MF, Spelman FA & Rubinstein JT. (1991). In vitro measurement and
characterization of current density profiles produced by non-recessed, simple
recessed, and radially varying recessed stimulating electrodes. IEEE Trans Biomed
Eng 38, 401-408.

Suzuki S, Humayun MS, Weiland JD, Chen SJ, Margalit E, Piyathaisere DV & de
Juan E, Jr. (2004). Comparison of electrical stimulation thresholds in normal and
retinal degenerated mouse retina. Jpn J Ophthalmol 48, 345-349.

Tang HW, Yan HL, Hu XH, Liang YX & Shen XY. (1996). Lead cytotoxicity in
primary cultured rat astrocytes and Schwann cells. J Appl Toxicol 16, 187-196.

144
Tremiliosi-Filho G, Jerkiewicz G & Conway BE. (1992). Characterization and
significance of the sequence of stages of oxide film formation at platinum generated
by strong anodic polarization. Langmuir 8, 658-667.

Varela C, Igartua I, De La Rosa EJ & De La Villa P. (2003). Functional
modifications in rod bipolar cells in a mouse model of retinitis pigmentosa. Vision
Res 43, 879-885.

Wagenaar DA & Potter SM. (2004). A versatile all-channel stimulator for electrode
arrays, with real-time control. J Neural Eng 1, 39-45.

Wagner C. (1951). Theoretical analysis of the current distribution under circular
dispersive electrodes. J Electrochem Soc 98, 116-119.

Wegener J, Keese CR & Giaever I. (2000). Electric cell-substrate impedance sensing
(ECIS) as a noninvasive means to monitor the kinetics of cell spreading to artificial
surfaces. Exp Cell Res 259, 158-166.

Weiland JD, Anderson DJ & Humayun MS. (2002). In vitro electrical properties for
iridium oxide versus titanium nitride stimulating electrodes. IEEE Trans Biomed Eng
49, 1574-1579.

Weiland JD & Humayun MS. (2005). A biomimetic retinal stimulating array. IEEE
Eng Med Biol Mag 24, 14-21.

Weiland JD, Humayun MS, Dagnelie G, de Juan E, Jr., Greenberg RJ & Iliff NT.
(1999). Understanding the origin of visual percepts elicited by electrical stimulation
of the human retina. Graefes Arch Clin Exp Ophthalmol 237, 1007-1013.

Werblin FS. (1978). Transmission along and between rods in the tiger salamander
retina. J Physiol 280, 449-470.

Whalen JJ. (2004). Platinum nanowire microelectrode arrays for neurostimulation
applications: fabrication, characterization, and in-vitro retinal cell stimulation. In
Materials Science and Engineering, pp. 293. The Johns Hopkins University,
Baltimore.

Whalen JJ, Weiland JD & Searson PC. (2005). Electrochemical deposition of
platinum from aqueous ammonium hexachloroplatinate solution. J Electrochem 152,
C738-C743.

Wheeler BC & Novak JL. (1986). Current source density estimation using
microelectrode array data from the hippocampal slice preparation. IEEE Trans
Biomed Eng 33, 1204-1212.

145

White MW, Merzenich MM & Gardi JN. (1984). Multichannel cochlear implants:
channel interactions and processor design. Arch Otolaryngol 110, 493-501.

Wiley JD & Webster JG. (1982a). Analysis and control of the current distribution
under circular dispersive electrodes. IEEE Trans on Biomed Eng 29, 381-385.

Wiley JD & Webster JG. (1982b). Distributed equivalent circuit models for circular
dispersive electrodes. IEEE Trans on Biomed Eng 29, 385-389.

Yang L, Ruan C & Li Y. (2003). Detection of viable Salmonella typhimurium by
impedance measurment of electrode capacitance and medium resistance. Biosensors
and Bioelectron 19.

Ziv OR, Rizzo JF & Jensen RJ. (2005). In vitro activation of retinal cells: estimating
location of stimulated cell by using a mathematical model. J Neural Eng 2, S5-S15.

Zrenner E, Stett A, Weiss S, Aramant RB, Guenther E, Kohler K, Miliczek KD,
Seiler MJ & Haemmerle H. (1999). Can subretinal microphotodiodes successfully
replace degenerated photoreceptors? Vision Res 39, 2555-2567.

Abstract (if available)

Abstract The electrode-retina interface of an epi-retinal prosthesis implant has been studied using finite-element modeling, electrochemical methods, and retinal array electrophysiology. The studies presented in this thesis have relied on the flexibility of custom designed and microfabricated multielectrode arrays to investigate the current distribution at the electrode-electrolyte interface and it's interaction with retinal tissue.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Manipulation of RGCs response using different stimulation strategies for retinal prosthesis

PDF

Saliency based image processing to aid retinal prosthesis recipients

PDF

Modeling retinal prosthesis mechanics

PDF

Towards a high resolution retinal implant

PDF

PET study of retinal prosthesis functionality

PDF

Developing a neural prosthesis for hippocampus: proof-of-concept using the in vitro slice

PDF

Effect of continuous electrical stimulation on retinal structure and function

PDF

Prosthetic visual perception: retinal electrical stimulation in blind human patients

$Intraocular camera for retinal prostheses: refractive and diffractive lens systems$

PDF

Intraocular camera for retinal prostheses: refractive and diffractive lens systems

PDF

Intraocular and extraocular cameras for retinal prostheses: effects of foveation by means of visual prosthesis simulation

PDF

Prosthetic vision in blind human patients: Predicting the percepts of epiretinal stimulation

PDF

Next generation neural interfaces for neural modulation and prosthesis

PDF

Viewing the picture we paint

PDF

Investigation of the electrode-tissue interface of retinal prostheses

PDF

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

PDF

Cortical and subcortical responses to electrical stimulation of rat retina

PDF

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

PDF

Novel imaging systems for intraocular retinal prostheses and wearable visual aids

PDF

Dependence of rabbit retinal synchrony on visual stimulation parameters

PDF

A biomimetic approach to non-linear signal processing in ultra low power analog circuits

Asset Metadata

Creator Ahuja, Ashish Kishore (author)

Core Title An in vitro model of a retinal prosthesis

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Publication Date 04/23/2007

Defense Date 01/24/2007

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

Tag electrophysiology,neural engineering,OAI-PMH Harvest,retinal prosthesis

Language English

Advisor Choma, John, Jr. (committee chair), Berger, Theodore W. (committee member), Steier, William H. (committee member), Tanguay, Armand R., Jr. (committee member), Weiland, James D. (committee member)

Creator Email ashishah@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m450

Unique identifier UC1469530

Identifier etd-Ahuja-20070423 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-484970 (legacy record id),usctheses-m450 (legacy record id)

Legacy Identifier etd-Ahuja-20070423.pdf

Dmrecord 484970

Document Type Dissertation

Rights Ahuja, Ashish Kishore

Type texts

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

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu