Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Prosthetic vision in blind human patients: Predicting the percepts of epiretinal stimulation
(USC Thesis Other)
Prosthetic vision in blind human patients: Predicting the percepts of epiretinal stimulation
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
PROSTHETIC VISION IN BLIND HUMAN PATIENTS: PREDICTING THE
PERCEPTS OF EPIRETINAL STIMULATION
by
Devyani Nanduri
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)
December 2011
Copyright 2011 Devyani Nanduri
Acknowledgements
Completing such a difficult endeavor has only been possible with the guidance and support of
wonderful mentors, colleagues, friends and family. First and foremost, I would like to thank my
Ph.D. adviser, Dr. James Weiland. He has given me the freedom to think and work independently
while pushing me to complete my goals when I have needed it the most. I would like to thank
Dr. Ione Fine for remaining such an integral part of my research project despite her move to the
University of Washington in the nascent stages of my PhD. Ione has taught me how to address
very difficult scientific problems in a unique and rigorous manner and has been a constant guide
both professionally and personally
Terry Byland and Linda Morefoot, who participate in the clinical trial are truly extraordinary
individuals and have been an inspiration to me. Their commitment to the retinal prosthesis
project has significantly contributed to the development of a technology that will in turn help
many other blind individuals. I would like to thank Terry for being dedicated to my experiments
each and every week and having many words of encouragement along the way. By working with
Terry on a weekly basis over four years, I have been able perform experiments unconstrained by
testing time, which in turn has allowed me to take more risks and to develop my research questions
creatively.
Working on a PhD involving an industry run clinical trial was a unique opportunity that
presented some challenges. Having an understanding of my situation due to a similar experience,
Dr. Alan Horsager was extremely helpful in my shaping my research questions and designing
experimental methodology that would be within the scope of the clinical trial. On the 2-Sight
front, Dr. JessyDornhelpedtointegratemewiththeresearchteamatthecompanyanddeveloped
ii
the means for me to collect and analyze my data from clinical sites around the world. I am grateful
to both these individuals for all their assistance.
I would also like to thank my committee of Dr. Loeb, Dr. Humayun and Dr. Medioni for
helping me refine my research to balance clinical relevance with scientific merit. As well, I would
like to thank Dr. Geoff Boynton for all of his assistance in developing my computational models.
The people at Second Sight Medical Products, Inc. have been a large part of my research
experience. I would like to thank Dr. Robert Greenberg for providing me the opportunity to
be involved with the project and giving me feedback on my research. I would like to thank
Maura Arsiero, Francesco Merlini, Dr. Brian Coley, Fatima Anaflous and Dr. Gislin Dagnelie for
collecting my data at clinical sites around the world. I would also like to thank Dr. Arup Roy,
Dr. Matt McMahon (at NEI), Kelly McClure, Dr. Thomas Lauritzen and Dr. Ashish Ahuja (at
USC) for all their intellectual mentorship.
I have been able to collaborate and develop my research ideas with my peers at the Doheny
Vision Research Center. Specifically, I would like to thank Dr. Aditi Ray, Dr. Vivek Pradeep,
Dr. Neha Parikh, Alice Cho, Andrew Weitz and Tim Nayar for long scientific and non-scientific
discussions.
Lastly, Iwouldliketothankmyfamily, Amma, AppaandAkshayforbeingaconstantstrength
throughout this endeavor. In addition, my friends here and back home, Rahul, Brenda, Jessy,
Alan, Shruthi, Devi, Puja, Jamin, John, Freda, Michel and Jen have all kept me well-balanced
through good times and kind words.
iii
iv
Table of Contents
Acknowledgements ii
List of Tables viii
List of Figures ix
Abstract xv
Chapter 1: Introduction 1
1.1 Motivation 1
1.2 The history of visual prostheses 4
1.2.1 Primary Visual Cortex 5
1.2.2 Lateral Geniculate Nucleus Stimulation 7
1.2.3 Optic Nerve Stimulation 7
1.2.4 Retinal Implants 8
1.2.5 Extraocular Implants 9
1.2.6 Optogenetic Therapy 11
1.3 Structure and computation of the retina 12
1.3.1 Structure of the retina 12
1.3.2 Computation in the retina 14
1.4 Effects of retinal degeneration 15
1.4.1 Structural changes in diseased retina 16
1.4.2 Functional changes in diseased retina 18
1.4.3 Cortical Plasticity 19
1.5 Electrical stimulation of the retina 20
1.5.1 Stimulation Threshold 20
1.5.2 Stimulation Specificity 23
1.5.3 Spatial extent of activation 24
1.6 Clinical findings with an epiretinal prosthesis 27
1.6.1 Properties of percepts 28
1.6.2 Evidence of form vision 28
1.7 Outline of this thesis 29
Chapter 2: Experimental setup & General methods 31
2.1 System (Argus I and Argus II) 31
2.2 Subjects 33
2.2.1 Selection criteria 33
2.2.2 Subject details 34
2.2.2.1 Argus I Subject 34
2.2.2.2 Argus II Subjects 34
v
2.3 Psychophysical methods 35
2.3.1 Threshold measurement 36
2.3.1.1 Argus I Subject 36
2.3.1.2 Argus II Subjects 36
2.3.2 Drawing task 38
2.3.2.1 Argus I Subject 38
2.3.2.2 Argus II Subjects 38
2.4 Analysis 40
2.4.1 Estimating the fovea on a fundus photograph 40
Chapter 3: Phosphene reproducibility 41
3.1 Introduction 41
3.2 Materials & Methods 42
3.2.1 Subjects 42
3.2.2 Psychophysical methods 43
3.2.2.1 Control task - tactile drawing 43
3.2.2.2 Retinal stimulation 43
3.2.3 Analysis 45
3.3 Results 47
3.3.1 Control tactile drawing experiment 47
3.3.2 Retinal stimulation experiment 53
3.3.2.1 Phosphene variability 53
3.3.2.2 Phosphene size bias 62
3.4 Discussion 63
3.4.1 Phosphene variability 63
3.4.2 Phosphene bias 66
Chapter 4: Phosphenes elicited by single electrode stimulation 68
4.1 Introduction 68
4.2 Methods 69
4.2.1 Subjects 69
4.2.2 Psychophysical methods 69
4.2.2.1 Retinal stimulation 69
4.2.3 Analysis 70
4.3 Results 71
4.3.1 Phosphene descriptions 71
4.3.2 Phosphene shape analysis 74
4.4 Discussion 77
4.4.1 Shape descriptions 77
4.4.2 Shape analysis 78
4.4.2.1 Phosphene area 79
4.4.2.2 Phosphene length and elongation 80
4.4.2.3 Phosphene orientation
80
Chapter 5: Development of a predictive model
82
5.1 Introduction 82
5.2 Methods 83
5.2.1 Subjects 83
5.2.2 Psychophysical methods 83
5.2.2.1 Retinal stimulation 83
5.2.3 Computational modeling 84
vi
5.3 Results 85
5.3.1 Model validation 85
5.3.2 Single electrode stimulation 87
5.3.2.1 Percepts vs. Predictions 87
5.3.2.2 Phosphene length 89
5.3.3 Paired electrode stimulation 90
5.4 Discussion 93
Chapter 6: Frequency and amplitude modulation 97
6.1 Introduction 97
6.2 Methods 99
6.2.1 Subjects 99
6.2.2 System 99
6.2.3 Psychophysical methods 99
6.2.3.1 Control task - tactile drawing 99
6.2.3.2 Retinal stimulation 99
6.3 Results 101
6.3.1 Control tactile drawing experiment 101
6.3.2 Retinal stimulation experiment 101
6.3.2.1 Phosphene descriptions 102
6.3.2.2 Phosphene size and brightness 103
6.3.3 Computational modeling 107
6.3.4 Comparing experimental data with modeling predictions 111
6.4 Discussion 113
6.4.1 Phosphene brightness 113
6.4.2 Phosphene size 115
6.4.3 Caveats 115
6.4.4 Implications for future retinal prostheses 116
Chapter 7: Additional Experiments 118
7.1 Sinusoid stimulation 118
7.1.1 Introduction 118
7.1.2 Methods 120
7.1.2.1 Subjects 120
7.1.2.2 Retinal stimulation 120
7.1.2.3 Psychophysical methods 121
7.1.2.4 Analysis 123
7.1.3 Results 123
7.1.3.1 Qualitative assessments 123
7.1.3.2 Experiment 1: True sinusoids 123
7.1.3.3 Experiment 2: Pseudosinusoids 124
7.1.3.4 Experiment 3: Paired stimulation 124
7.1.4 Discussion 125
7.1.4.1 Single electrode stimulation 125
7.1.4.2 Paired electrode stimulation 126
7.1.4.3 Limitations 127
7.2 High frequency stimulation 127
7.2.1 Introduction 127
7.2.2 Methods 128
7.2.2.1 Subjects 128
7.2.2.2 Retinal stimulation 128
vii
7.2.2.3 Psychophysical methods 128
7.2.2.4 Analysis 129
7.2.3 Results 129
7.2.4 Discussion 130
Chapter 8: Conclusions 138
8.1 Controlling spatial properties of percepts 139
8.2 Factors affecting resolution 140
8.3 Subject variability 142
Bibliography 144
List of Tables
3.1 Drawing variability and phosphene variability. . . . . . . . . . . . . . . . . . 52
3.2 Phosphene size bias. Phosphene paired stimulation distance compared to com-
pact and elongated tactile drawing area bias . . . . . . . . . . . . . . . . . . . . . 62
5.1 Axonal stimulation average correlation. Average axonal stimulation correla-
tions for Subject 1, Subject 2 and Subject 3 tended to be significantly higher than
no axon stimulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.2 Summary of paired stimulation predictions with axon model. . . . . . . 94
6.1 Phosphene shape variability for frequency and amplitude modulation
experiments. Tactile control drawing variability with standard error (row 1 &
2) and phosphene data variability (row 3) across area, major and minor axes, and
orientation shape descriptors. Tactile variability was separated into two categories:
compact shapes (minor axis length > 50% major axis length) and elongated shapes
(minor axis length < 50% major axis length). Phosphene drawing variability (row
3) was equal to (orientation, major and minor axes) or significantly less than (area)
tactile drawing variability. Asterisks represent significantly better performance for
phosphenes than for elongated tactile shapes, * p<0.05, Students 2-tailed t-test). 102
6.2 Slope values for best fit of ’Brightness and Size vs. Amplitude and
Frequency’ plots. P values specify whether slope is significantly greater than
zero slope. The last row is mean slopes across all 9 electrodes (* p<0.05, **
p<0.01). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
viii
List of Figures
1.1 Concept of an epiretinal prosthesis. The camera captures an image of the
world and projects it onto the retina via a microelectrode array. Electrodes stimu-
late the underlying retinal cells and the subjects perceives phosphenes that can be
combined into a pixelized view of the world (adapted from Chader et al., 2009). . 3
1.2 LeRoy’s experiment in 1755. Set-up of the first experiment reported to induce
phospheneslikelythroughelectricalstimulationoftheoccipitallobes(adaptedfrom
Marg, 1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Structure of the retina. Stained retina (left) and retinal schematic (right)
with labeled rod and cone photoreceptors (R & C), outer plexiform layer (OPL),
horizontal cells (H), rod and cone bipolar cells (RB & RC), amacrine cells (A &
AII), and ganglion cells (G) (courtesy of Rachel Wong, University of Washington). 13
1.4 Retinal remodeling during RP. Human retina with RP has significant remod-
eling and cell loss including cell migration (adapted from Jones and Marc, 2005). 17
1.5 Strength duration curve. Average change in threshold (large bottom plot) and
overall charge (small top plot) as a function of pulse duration at different recording
distances for rat retina (adapted from Sekirnjak et al., 2006). . . . . . . . . . . . . 22
1.6 Axonal activation with electrical stimulation. Threshold map obtained with
calcium imaging experiment in salamander retina. Blue circle shows a 200 μm
electrode, solid red circles and streaks indicate activation of ganglion cell somas
and antidromic axons (adapted from Behrend et al., 2008) . . . . . . . . . . . . . 26
2.1 Schematic of the Argus I and Argus II arrays and systems. (A) Argus I
4x4 electrode array, (B) Argus II 6x10 electrode array, (C) Overview of the Argus
I implant, (D) Overview of the Argus II implant . . . . . . . . . . . . . . . . . . . 32
2.2 Shifted array for Subject 1. Fundus photographs with array location and
optic disc in retina clearly visible. (A) Photograph from March, 2008, with black
arrows pointing to optic disc center (right arrow) and array center (left arrow).
(B) Photograph from March, 2009 with arrows marking same retina locations in
panel A (notice that left arrow does not mark the same array location as panel A,
demonstrating a shift in the array position). . . . . . . . . . . . . . . . . . . . . . 35
ix
2.3 Argus I subject drawing task. Subject drawings on a grid screen are captured
by an external camera and recorded to a video file. Video files are analyzed offline
by tracking the location of the pen tip from frame-to-frame and translated to a
binary image. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4 Argus II subject drawing task. Subject drawings recorded by a touchscreen
monitor in x-y screen coordinates and translated to a binary image. . . . . . . . . 39
2.5 Estimating fovea location. Fundus photograph for Subject 4 shows estimate of
foveal region. Array placement is clearly in the macular region of the retina. . . . . 40
3.1 Tactile target control task. (A) Argus I tactile shapes. (B) Argus II tactile
shapes. (C) Subject feels shape before tracing on a screen . . . . . . . . . . . . . . 44
3.2 Paired phosphene distance calculation. Example from subject 1 calculates
the distance between phosphenes generated from paired electrode stimulation and
compares it to the actual distance between stimulating electrodes (based on the
geometry of the array) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3 Tactile drawing for diagonal rectangle shape. Example of Subject 4 tactile
drawing trials (5 repeats) shows an overestimation of size. . . . . . . . . . . . . . 48
3.4 Drawing bias for all subjects. Comparison of shape drawings to tactile shape
for area, major axis and minor axis and orientation descriptors. Size comparison
expressed as a ratio, orientation comparison as a difference . . . . . . . . . . . . . 49
3.5 Control task drawing variability for all subjects. . . . . . . . . . . . . . . . 50
3.6 Phosphene variability example for Subject 1. Repeated phosphene drawings
for consistent (A) and inconsistent (B) example trials . . . . . . . . . . . . . . . . 53
3.7 Phosphene variability example for Subject 2. Repeated phosphene drawings
for consistent (A) and inconsistent (B) example trials . . . . . . . . . . . . . . . . 54
3.8 Phosphene variability for Subject 1. Phosphene area (A), major axis (B),
minor axis (C) and orientation (D) variability . . . . . . . . . . . . . . . . . . . . . 56
3.9 Phosphene variability for Subject 2. Phosphene area (A), major axis (B),
minor axis (C) and orientation (D) variability . . . . . . . . . . . . . . . . . . . . . 57
3.10 Phosphene variability for Subject 3. Phosphene area (A), major axis (B),
minor axis (C) and orientation (D) variability . . . . . . . . . . . . . . . . . . . . . 58
3.11 Phosphene variability for Subject 4. Phosphene area (A), major axis (B),
minor axis (C) and orientation (D) variability . . . . . . . . . . . . . . . . . . . . . 59
3.12 Subject 1 phosphene variability over time. Minor axis variability measured
for the same 4 electrodes over the course of 430 days of testing . . . . . . . . . . . 60
x
3.13 Variability in the number of phosphenes. Experiments divided into single
electrodes, singles stimulated with pairs and pairs for Subject 1 (A), Subject 2 (B),
Subject 3 (C), Subject 4 (D). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1 Retinal stimulation pulse train. Experiment used a biphasic, 0.45 ms cathodic
firstchargebalanced20Hzstimulationpulsetrainthatvariedinamplitudebetween
1.25 - 2X threshold. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.2 Single electrode example phosphenes. The first column shows an schematic
of the subject’s array (position adjusted to match actual position in fundus pho-
tograph) with stimulating electrode marked by a red circle. The second column
represents the individual trials (each trial shown in a different color) aligned based
on their position on the reference grid. The third column shows the average draw-
ing across five trials, plotted in a gray-scaled image. The fourth column shows
the calculated mean shape descriptors across all trials with standard error values.
Area given in degrees squared visual angle; Major and Minor axis given in visual
angle degrees; Orientation given in degrees. Each row represents an example for
a different subject. Although absolute position varied (marked in different colors,
second column), phosphene shapes were very consistent from trial to trial. . . . . 72
4.3 Subject 1 single electrode phosphenes. Drawings from 13 single electrodes
overlayed on electrode array. Red circles indicate stimulating electrode, blue square
marks the approximate location of the fovea . . . . . . . . . . . . . . . . . . . . . 73
4.4 Subject 2 single electrode phosphenes. Drawing from 29 single electrodes
overlayed on electrode array. Red circles indicate stimulating electrode, blue square
marks the approximate location of the fovea . . . . . . . . . . . . . . . . . . . . . . 74
4.5 Subject 3 single electrode phosphenes. Drawing from 24 single electrodes
overlayed on electrode array. Red circles indicate stimulating electrode, blue square
marks the approximate location of the fovea . . . . . . . . . . . . . . . . . . . . . . 75
4.6 Subject 4 single electrode phosphenes. Drawing from 29 single electrodes
overlayed on electrode array. Red circles indicate stimulating electrode, blue square
marks the approximate location of the fovea . . . . . . . . . . . . . . . . . . . . . . 75
4.7 Single electrode phosphene area. Normalized phosphene area as a function of
distance to the fovea (A) and the horizontal raphe (B) for all subjects. . . . . . . 76
4.8 Single electrode phosphenes major axis length. Normalized phosphene
length as a function of distance to the fovea (A) and the horizontal raphe (B)
for all subjects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.9 Single electrode phosphene elongation. Phosphene elongation as a function
of distance to the fovea (A) and the horizontal raphe (B) for all subjects. . . . . . 78
4.10 Single electrode phosphene orientation. Phosphene orientation as a function
of electrode location relative to the horizontal raphe for all subjects. . . . . . . . . 79
4.11 Ganglion cell axon pathways. In the macular region, nerve fiber trajectories
drastically shift in orientation from inferior to superior retina. . . . . . . . . . . . 81
xi
5.1 Retinal stimulation pulse train. Experiment used a biphasic, 0.45 ms cathodic
firstchargebalanced20Hzstimulationpulsetrainthatvariedinamplitudebetween
1.25 - 4X threshold and was presented on single electrodes or pairs of electrodes. . 83
5.2 Predicting percepts with a computational model. Based on (A) a mathe-
matical model of nerve fiber trajectories in a human retina and (B) the electrode
positions relative to the optic disc (measured by looking at a subjects fundus pho-
tograph). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.3 Optimizing the model. The model was rotated about the optic disc and pre-
dicted percepts were correlated with subject drawings. (A) Subject drawing had
high correlation at an optimized model rotation and poor correlations otherwise.
(B) Optimal orientation for each subject placed the fovea (marked by colored dots)
at a location that is anatomically realistic. (C) Average model correlations dropped
off as a function of rotation away from optimal orientation. . . . . . . . . . . . . . 86
5.4 Subject 1 model prediction examples. Phosphene drawings (second column)
vs. predictions with axonal stimulation (third column) and no axonal stimulation
(fourth column) show that axonal predictions tend to match experimental data
better than no axonal stimulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.5 Subject 2 model prediction examples. While most axonal stimulation predic-
tions matched experimental data better than non-axonal stimulation predictions
(row 1 and 2), in some cases (row 3), neither axonal nor non-axonal predictions
could explain phosphene shape. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.6 Subject3modelpredictionexamples. Axonalstimulationpredictionsmatched
experimental data better than non-axonal stimulation predictions (row 1 and 2) in
most cases, while in others, non-axonal predictions had higher correlations than
the axon model prediction (row 3). . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.7 Subject 4 model prediction examples. Axonal stimulation predictions were
not always correlated to experimental data. . . . . . . . . . . . . . . . . . . . . . . 91
5.8 Phosphene Length vs. Axon Model Length. For Subject 1 and Subject 3,
phosphene length was shorter than predictions when stimulated axon tracts were
greater than 10
o
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.9 Subject1pairedelectrodemodelpredictionexamples. Phosphenedrawings
(second column) vs. predictions with axonal stimulation (third column) and no
axonal stimulation (fourth column) show that predictions with axon model are
better at predicting both the shape of phosphenes and the number of percepts. . . 94
5.10 Subject2pairedelectrodemodelpredictionexamples. Phosphenedrawings
(second column) vs. predictions with axonal stimulation (third column) and no
axonal stimulation (fourth column) show that predictions with axon model can
sometimes predict the shape and number of percepts. . . . . . . . . . . . . . . . . 95
xii
5.11 Subject3pairedelectrodemodelpredictionexamples. Phosphenedrawings
(second column) vs. predictions with axonal stimulation (third column) and no
axonal stimulation (fourth column) show that predictions with axon model predicts
the shape of phosphenes and occasionally predicts the number of percepts. . . . . 96
5.12 Subject4pairedelectrodemodelpredictionexamples. Phosphenedrawings
(second column) vs. predictions with axonal stimulation (third column) and no
axonalstimulation(fourthcolumn)showthataxonmodelpredictionsdonotalways
match the experimental data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.1 Schematic of the Biphasic pulse train. Pulse trains were varied by either
changing pulse amplitude (gray dashed arrows) or pulse frequency (black arrows). 100
6.2 Phosphene drawings from 3 different electrodes. Example phosphenes from
3 different electrodes (a single trial for each electrode) show (A) Schematic of array
highlighting the example electrodes shown in B-D (D2, C4, B3). For all three
electrodes stimulation 0.45 ms biphasic, 20 Hz pulse train for a duration of 500 ms.
(B,C) For electrodes D2 and C4 the pulse train was at 1.25X threshold. (D) For
electrode B3 the pulse train was at 3X threshold. . . . . . . . . . . . . . . . . . . 103
6.3 The effects of current amplitude and frequency for electrode D2. Row 1:
phosphene drawings at baseline stimulation, Row 2: phosphene drawings after an
increase in amplitude, Row 3: phosphene drawings after an increase in frequency . 104
6.4 The effects of amplitude and frequency on apparent brightness and size.
Panels A and C show brightness and apparent size as a function of normalized
(relative to threshold) amplitude for 9 electrodes. Panels B and D show brightness
and apparent size as a function of normalized (relative to threshold) frequency
for the same 9 electrodes. Each electrode’s data are fit with the best fit linear
regression. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.5 Normalizedsizeandbrightnessasafunctionofamplitudeandfrequency.
Averaged across 9 electrodes; data are fit using linear regression. . . . . . . . . . . 108
6.6 Model schematic. BOX 1: The time stimulus, f(t), was transformed into a
spatio-temporal representation, based on the measured electrophysiological thresh-
olds from a disk electrode. BOXES 2-5: The output of this was passed through a
modified version of the ‘Perceptual Sensitivity Model’ incorporating threshold and
suprathreshold parameters. The resulting output corresponds to a spatial bright-
ness response, B(r). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.7 Predicted percepts with model. Increasing amplitude shown in top row and
increasing frequency in bottom row. . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.8 Comparing experimental findings to model. Comparing brightness rating
and apparent size psychophysical data to model predictions for the effects of chang-
ing amplitude and frequency. Experimental data from Amplitude Modulation are
shown in dark gray and data from Frequency Modulation are in light gray. Dotted
solid lines show model predictions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
xiii
7.1 Schematic of types of waveforms for experiments. (A) True sinusoids. (B)
Pseudosinusoids. (C) Conventional biphasic pulse train. (D) Paired stimulation
with electrode 1 activated with a pseudosinusoid waveform and electrode 2 with a
biphasic pulse train (top) or with electrode 1 and 2 as asynchronously activated
with biphasic pulse trains (bottom). . . . . . . . . . . . . . . . . . . . . . . . . . . 122
7.2 Phosphene drawings with true sinusoid stimulation. Stimulating electrode
shown in column 1; sinusoid stimulation phosphene trials in columns 2 and 3;
biphasic stimulation phosphenes in column 4. . . . . . . . . . . . . . . . . . . . . 132
7.3 Phosphene drawings with pseudosinusoid stimulation. Stimulating elec-
trode shown in column 1; average phosphene drawings for 20, 40 and 100 Hz pseu-
dosinusoid stimulation shown in columns 2-4; average phosphene drawings for 20
Hz biphasic pulse train in column 5. . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.4 Pairedpseudosinusoidstimulation. Phosphenedrawingsfortwodifferentpairs
of electrodes (A & B). On electrode array schematic, pseudosinusoid stimulated
electrode 1 is indicated with a red circle, biphasic electrode 2 with a blue circle; (i)
electrode 1 pseudosinusoid stimulation phosphene drawing; (ii) electrode 2 biphasic
stimulation phosphene drawing; (iii) electrode 1 stimulated with pseudosinusoid
and electrode 2 stimulated with biphasic pulse train; (iv) electrode 1 and electrode
2 stimulated with biphasic pulse trains. . . . . . . . . . . . . . . . . . . . . . . . . 134
7.5 Inferior retina phosphenes. Phosphenes produced from electrodes located in
the inferior retina stimulated with high frequency biphasic pulse trains. . . . . . . 135
7.6 Superior retina phosphenes. Phosphenes produced from electrodes located in
the superior retina stimulated with high frequency biphasic pulse trains. . . . . . . 136
7.7 Foveal retina phosphenes. Phosphenes produced from electrodes located in
near the fovea or horizontal raphe stimulated with high frequency biphasic pulse
trains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
xiv
Abstract
Over the last 8 years it has been shown that an implantable retinal prosthesis can partially restore
visual capabilities to blind humans. With current electrode arrays it is only possible to stimulate
groups of cells rather than individual single cells with spatio-temporal precision. While artificial
vision from a retinal prosthesis is unable to completely replicate the neural response patterns of
normal vision, by stimulating groups of cells with electrodes patients see electrically elicited visual
percepts. Thus this work will focus mainly on the form of percepts created with single electrodes
in a prosthesis given certain varying stimulation parameters, and the development of a model to
predict how stimuli applied produce certain percepts with single and paired prosthesis electrodes.
Ideally we would want this model to resemble a digital display with independent pixels in
order to individually control each percept seen by the subject in relation to the stimuli. We find
however that the model that best fits the data does not resemble a digital display but instead the
nerve fiber bundle trajectories in the human retina.
The work presented here gives insight into the factors affecting form perception with a micro-
electronic retinal prosthesis. Specifically, by directly measuring the shapes of visual percepts from
single and paired electrodes at different stimulation parameters, the building blocks of prosthetic
vision are understood. The incentive is that we can use this information to develop a strategy
that can control the percepts from each individual prosthesis electrode and piece them together
in an organized way to best represent the visual world.
xv
Chapter 1
Introduction
1.1 Motivation
Visual impairment ranging from slight impairment to complete blindness affects millions of people
worldwide. In the developed world, one of the major causes of vision loss is photoreceptors dis-
eases such as retinitis pigmentosa (RP) and age-related macular degeneration (AMD). Currently,
over 15 million or approximately 1/4000 people are afflicted with a photoreceptor diseases and
these numbers are expected to rise as the population ages (Chader, 2005; Chader etal., 2009).
While both RP and AMD are characterized by the death of photoreceptors which eventually lead
to complete vision loss, the remaining neural cells in the retina are relatively spared, though
somewhat disorganized (Jones etal., 2005). The prevalence of these various vision disabilities has
given rise to a number of therapies that attempt to interface with and activate the remaining
visual system.
Proposedsolutionscurrentlyunderinvestigationincludeanoptogeneticstrategyinvolvinglight
sensitive proteins such as Channelrhodopsin-2 (ChR2), gene replacement therapy, and epiretinal
and subretinal visual prostheses. Each therapy has its strengths. Gene replacement technology
has been successful clinically in treating one type of RP (i.e., Leber’s Congenital Amaurosis;
RPE65 mutation)(Hauswirth etal., 2008). However gene therapy treatment is limited in scope
1
by the specificity to a particular gene mutation and the presence of residual photoreceptors,
requiringinterventionduringinitialstagesofphotoreceptordegeneration. Optogenetictechnology,
providing light sensitivity functionality to remaining neural cells in the retina, has shown success
at the preclinical stage. However, a clinically viable therapy will require light stimulation more
than 5 orders of magnitude greater than the threshold of cone photoreceptors (Doroudchi etal.,
2011; Lagali etal., 2008). Visual prosthesis implants that can activate the visual system at various
stages such as the cortex, optic nerve or retina are also being developed. Strategically, retinal
prostheses provide an intervention at an earlier stage in the visual stream than other types of
visual prostheses. While subretinal prostheses have been able to provide implant recipients with
low-level spatial vision, groups have yet to implant a fully chronic device in subjects (Wilke etal.,
2011). There is also evidence that the degenerate bipolar-ganglion cell synapse and subretinal
gliosis of a diseased retina could make it difficult both to place the array, to stimulate bipolar
cells, and for bipolar cells to transfer useful information to ganglion cells ((Marc and Jones,
2003; Marc etal., 2003).
An epiretinal prosthesis is the only fully chronic solution currently available to partially restore
sight to people who are blind from severe retinitis pigmentosa across multiple RP gene mutations.
Since 2002, Second Sight Medical Products Inc. has implanted 6 subjects with the first generation
of the device (Argus I System) and 32 subjects with the second generation of the device as part of
a multi-site clinical trial (Ahuja etal., 2011; Mahadevappa etal., 2005). Most recently, the device
has also been approved for sale in parts of Europe. Analogous to a cochlear implant, the system
electrically stimulates the inner retina ganglion cell layer using an implanted microelectrode array.
Whentheretinaiselectricallystimulated, thesubjectperceivedaspotoflightcalleda’phosphene’.
Inanidealsituation, avisualprosthesissystemwouldrestorevisionbytakinganimageofwhat
is seen in the visual field (captured in real time by a video camera) and representing each pixel
with a phosphene. The final image would look much like a gray-scale digital scoreboard, where
each phosphene produced by an electrode can be thought of as a pixel that varies in brightness.
2
As the technology improves, pixel/electrodes size would decrease and the resolution of the system
would increase. A schematic depicting the concept of how such an epiretinal prosthesis could
potentially produce artificial vision is shown in Figure 1.1.
Figure 1.1: Concept of an epiretinal prosthesis. The camera captures an image of the world
and projects it onto the retina via a microelectrode array. Electrodes stimulate the underlying
retinal cells and the subjects perceives phosphenes that can be combined into a pixelized view of
the world (adapted from Chader et al., 2009).
Current clinical research has shown that an epiretinal prosthesis is clinically feasible and can
provide limited low-level spatial vision to implant recipients (Ahuja etal., 2011; Caspi etal., 2009).
In addition, clinical investigations of the stimulus space have provided insight into stimulation
thresholds, brightness parameters and spatio-temporal interactions between electrical pulses and
electrodes (Horsager etal., 2009; Horsager etal., 2010; Horsager etal., 2011; Greenwald etal.,
2009). Although patients implanted with these devices have been shown to be able to perform
some simple visual tasks, little is known about the spatial qualities of the resulting percepts.
3
There is a need for a systematic analysis of how stimulating the retina creates the phosphenes in
space to form a representation of the world.
The goal of this thesis is to thoroughly understand the building blocks of artificial form per-
ception by directly measuring the visual experience of epiretinal prosthesis subjects. In particular,
this work quantifies the relationship between an electrical stimulus to its elicited percept in the
contextofpotentialmechanismsmediatingphospheneshapeandtheimplicationsforfuturehigher
resolution prosthesis devices.
1.2 The history of visual prostheses
The first reported use of electrical stimulation to create the sensation of vision was over two cen-
turies ago. In 1755, Charles LeRoy, a chemist and physician briefly induced the first documented
phosphenes by discharging a Leyden jar (acting as a high-voltage capacitor) through the head of a
blind boy (LeRoy, 1755). One set of wires was wound around the patient’s head, likely contacting
the skull near the occipital lobes, while the other was connected to his leg completing the current
pathway (see Figure 1.2). When the wires leads were place in the Leyden jar, the shocks provoked
‘terrible cries’ and the patient reported ‘flames descending rapidly before his eyes’ (Marg, 1991).
Almost two centuries later, during cases of occipital lobe surgeries, phosphenes were elicited un-
intentionally (Lowenstein and Borchardt, 1918; Krause, 1924; Bradley etal., 2005). A concerted
attempt to deliberately induce visual sensations through electrical stimulation and carefully ex-
plore how phosphenes are elicited began in the mid 20th century. These early studies evolved
into the concept for a visual prosthesis device; a full-scale system to restore sight to the blind by
intervening with the normal visual signaling pathway. The next few pages provide an overview of
different types of visual prosthesis systems that have been in development over the last 50 years.
Each strategy intervenes at a different stage along the visual pathway and has its strengths and
weaknesses.
4
Figure 1.2: LeRoy’s experiment in 1755. Set-up of the first experiment reported to induce
phosphenes likely through electrical stimulation of the occipital lobes (adapted from Marg, 1991)
.
1.2.1 Primary Visual Cortex
In the mid-20th century, Penfield, while studying the neural origins of epilepsy, discovered that
electrical stimulation of the cortex caused subjects to perceive visual percepts described as stars,
wheels, discs, spots, streaks, and lines (Penfield, 1947; Rasmussen and Penfield, 1947; Penfield
and Rasmussen, 1952; Penfield and Jasper, 1954). More than a decade later, Brindley and Lewin
implanted the first chronic electrode array in the visual cortex for the purpose of restoring vision
(Brindley and Lewin, 1968a; Brindley and Lewin, 1968b; Brindley, 1970). In the procedure, an
array of 80 platinum disc electrodes was placed on the surface of the occipital pole of the subject,
a 52-year-old woman (Brindley and Lewin, 1968a; Brindley and Lewin, 1968b; Brindley, 1970). In
1972, another subject was implanted with an 80-electrode array (Brindley etal., 1972). All these
5
subjects were able to see independent and distinct visual percepts which varied in size and some
spatial mapping and threshold experiments were performed. However, experiments attempting
to provide functional form vision through combining phosphenes into letters and shapes proved
unsuccessful.
In the later 1970’s, there was another attempt to develop a chronic cortical prosthesis by the
vision scientist William Dobelle. Two blind subjects were implanted with a hexagonal array of
64 platinum disc 1mm
2
electrodes (Dobelle, 1974; Dobelle etal., 1974; Dobelle and Mladejovsky,
1974). Spatial mappings of phosphenes from stimulating various electrodes indicated that stim-
ulation far from the occipital pole led to peripheral phosphenes, and stimulation clustered on
the surface of the visual cortex led to clustered phosphenes (Evans etal., 1979; Henderson etal.,
1979; Dobelle etal., 1979). Further studies found that simultaneous and interleaved pulse stimuli
on pairs of electrodes separated by as much as 1 mm apart resulted in fused dumbbell-shaped
or elongated line percepts (Bak etal., 1990). Another clinical cortical prosthesis study using
penetrating microelectrodes in the visual cortex showed that an increase in current causes a de-
crease in phosphene size (Schmidt etal., 1996). These results suggest that cortical stimulation
using multi-electrode stimulation could be complicated by electrode-electrode and lateral neural
interactions.
One of the advantages of a cortical prosthesis device over other visual prosthesis devices
is that it can be used to treat a wider variety of visual impairment diseases (Margalit etal.,
2002). At present, work continues to progress in this area with several groups (Schmidt etal.,
1996; Bradley etal., 2005; Fernández etal., 2005; Troyk etal., 2003; Normann etal., 2009; Nor-
mann etal., 1999). Unfortunately, difficulties with safety and surgical techniques may limit the
clinical viability of this solution (Kotler, 2002; Girvin, 1988).
6
1.2.2 Lateral Geniculate Nucleus Stimulation
The lateral geniculate nucleus (LGN) is located in the thalamus of the brain and serves as a
relay center for visual information from the retina through to V1. There are two LGNs for
each vertically segmented hemifield. Pezaris and Reid are currently investigating the feasibility of
implanting a prosthesis in the LGN (Pezaris and Reid, 2009; Pezaris and Eskandar, 2009). Recent
behavioral research in nonhuman primates has demonstrated that visual or electrical stimulation
of LGN specific receptive fields can cause highly localized and repeatable eye movements to the
location in space represented by the receptive field of that neuron (Pezaris and Reid, 2007).
Like cortex and retina, retinotopy in the LGN is maintained. Thus, an electrode array im-
plantedintheLGNwouldbeabletoproduceproperlyspatiallymappedpercepts(Schneideretal.,
2004). Although seemingly highly invasive, surgical implant procedures would be similar to that
of Deep Brain Stimulation (DBS) of the thalamus in human patients for the treatment of Parkin-
son’s disease, and would likely have the same established safety and implant protocol. A large
disadvantage is that a full field prosthesis would require bilateral LGN implants (Pezaris and
Eskandar, 2009).
1.2.3 Optic Nerve Stimulation
The optic nerve is the collection of ganglion cell axons that project information from the retina
to higher processing centers in the cortex. The concept of stimulating the optic nerve to restore
vision is based upon the same principles used to develop cuff electrodes for nerve fiber recordings
(Hoffer etal., 1981), stimulation of muscle fibers (Baratta etal., 1989; Fang and Mortimer, 1991;
Loeb and Peck, 1996) and the vagus nerve (Woodbury and Woodbury, 1991). Using a cuff
electrode, onegrouphasshownthatopticnervestimulationcangeneratevisualperceptsinasingle
blind subject (Veraart etal., 1998; Brelén etal., 2005). Furthermore, variability in phosphene
position was reported to be ~5-10° visual angle (Obeid etal., 2010). One disadvantage to this
7
approach is that the optic nerve does not maintain retinotopy (Fitzgibbon and Reese, 1996),
and could compromise the spatial mapping of the percepts. Although experiments have shown
that the location of a percept can be changed with stimulation parameters such as frequency
and pulse width, this does not necessarily improve the spatial mapping capabilities of the device
(Veraart etal., 2003).
1.2.4 Retinal Implants
Retinal prosthesis devices strive to target neural cells early on in the visual pathway prior to
neural processing in the LGN (Dan etal., 1996; Wang etal., 2007b) and cortex (Hubel and Wiesel,
1962; Hubel and Wiesel, 1968).
Subretinal electrode arrays are implanted in the subretinal space between the pigment epithe-
lial cells and the remaining retina in order to mimic the functionality of the degenerated photore-
ceptors. There are several groups working on slightly different subretinal techniques (Chow and
Chow, 1997; Chow etal., 2004; Palanker etal., 2005; Jensen and Rizzo, 2006; Jensen and Rizzo,
2008; Zrenner etal., 1997; Zrenner, 2002; Zrenner etal., 1999). In theory, subretinal stimulation
should utilize the computational processing of the retinal circuitry by targeting the retinal bipolar
cells, as opposed to an intervention further along in the visual system. However, there is evidence
that the degenerate bipolar-ganglion cell synapse and subretinal gliosis of a diseased retina could
make it difficult to place the array and stimulate bipolar cells (Marc and Jones, 2003; Jones and
Marc, 2005). Successful clinical findings have been reported by Zrenner’s group. Their system can
operate by either direct stimulation of the retina with a wire across the skin or through the use
of a photodiode microchip to substitute for the function of photoreceptors by sensing incoming
light and electrically activating the neural retina. In early studies from Chow’s group, this light-
based pathway had been limited by insufficient power supply to photodiodes. Presently, clinical
studies by Zrenner’s group have been reported on acute and semi-chronic implant subjects. With
light activation, these subjects were able to discriminate bright objects on a dark table, describe
8
patterns and discriminate letters and short words (Zrenner etal., 2011; Benav etal., 2010). With
direct stimulation, two subjects reported whitish round dot-like percepts and one subject reported
elongated short whitish or yellowish lines (Wilke etal., 2011; Zrenner etal., 2011). Subjects were
also able to distinguish vertically and horizontally oriented electrode stimulations from each other
(Wilke etal., 2011). Results are extremely promising and will need to be confirmed in chronically
implanted subjects.
Epiretinal implants are implanted on the vitreal front surface of the retina in contact with the
ganglion cell layer. While the goal has been to stimulate remaining neural circuitry from bipolar
to ganglion cells (Greenberg, 1998), electrophysiology studies seem to indicate that ganglion cells
are primarily activated (Sekirnjak etal., 2008). There are several groups that have worked on this
approach with varying degrees of success (Feucht etal., 2005; Walter etal., 2005; Gerding etal.,
2007; Humayun etal., 1999a). Thus far, the most advanced prosthesis project is a collaborative
effort led by Dr. Mark Humayun and Second Sight Medical Products Inc. (2-Sight). Over
the last ten years, as part of two separate 2-Sight clinical trials, more than 30 subjects have
been implanted chronically with epiretinal arrays (Ahuja etal., 2011; Humayun etal., 2003).
More recently, the device has been approved to be marketed in parts of Europe for advanced RP
patients. Surgically, epiretinal implants are simpler than subretinal implants, but there are several
concerns. Stimulation of the ganglion cell layer may not utilize computational processing ability
of other neural cell layers. Furthermore, stimulation of passing ganglion cell axon fibers could
lead to elongated and poorly localized percepts (Freeman etal., 2011). A detailed description of
the Argus I and Argus II systems is given in section 2.1.
1.2.5 Extraocular Implants
Extraocular electrical stimulation has a long history dating back to early experiments from Brind-
ley’s group (Brindley, 1964; Brindley, 1962). Brindley used a makeshift corneal electrode to apply
stimulating pulse trains directly to his own eye and discovered that electrical impulses interfered
9
with his natural vision (Brindley, 1964; Brindley, 1962). A scientist from Cambridge, Carpen-
ter, also performed some extraocular experiments on a human subject and was able to elicit
phosphenes by passing alternating currents of 100 Hz (Carpenter, 1972; Carpenter, 1973). In more
recent times, a supra-choroidal prosthesis is being developed Takeshi Morimoto etal., ‘Chronic im-
plantation of newly developed suprachoroidal-transretinal stimulation (STS) prosthesis in dogs.’,
Invest Ophthalmol Vis Sci, (Jul 2011), hURL: http://dx.doi.org/10.1167/iovs.10-6971i; Hi-
royuki Kanda etal., ‘Electrophysiological studies of the feasibility of suprachoroidal-transretinal
stimulation for artificial vision in normal and RCS rats.’, Invest Ophthalmol Vis Sci, 45:2 (Feb
2004); Takashi Fujikado etal., ‘Evaluation of phosphenes elicited by extraocular stimulation in
normals and by suprachoroidal-transretinal stimulation in patients with retinitis pigmentosa.’,
Graefes Arch Clin Exp Ophthalmol, 245:10 (Oct 2007), hURL: http://dx.doi.org/10.1007/
s00417-007-0563-zi; Kentaro Nishida etal., ‘Efficacy of suprachoroidal-transretinal stimulation
in a rabbit model of retinal degeneration.’, Invest Ophthalmol Vis Sci, 51:4 (Apr 2010), hURL:
http://dx.doi.org/10.1167/iovs.09-4120i. In this system, the electrode array is passed
through the sclera and sits beneath the choroid, while the electronics are implanted outside of
the eye. Surgically, this implant is relatively simple and the electrode to neural tissue distance
causes the safe charge level to increase by a factor of three compared to direct retinal stimulation
(Nakauchi etal., 2007). A number of animal studies have established safety and efficacy of ex-
traocular and suprachoroidal trans-retinal stimulation (Chowdhury etal., 2005; Morimoto etal.,
2011; Kanda etal., 2004; Nakauchi etal., 2005; Nishida etal., 2010; Shivdasani etal., 2010). Dur-
ing preliminary clinical studies, suprachoroidal stimulation was able to induce phosphenes in two
subjects with RP (Fujikado etal., 2007). While these findings show promise, with such far dis-
tance between stimulating electrode to the neural tissue, the development of a high-resolution
extraocular device remains challenging.
10
1.2.6 Optogenetic Therapy
Optogenetics is relatively new sub-field of gene therapy that combines genetics and optics to
provide cells with a completely new function by rendering them light sensitive. Cells are injected
with an adeno-associated virus (AAV) which delivers genes to the neural cell. The genes are able
to transduce channelrhodopsin, a light sensitive ion-channel into neural cells. These transduced
cellscannowactsimilartophotoreceptorsbytakinginalightstimulusandproducinganelectrical
response that continues through the rest of the visual system. Although a very different approach,
optogeneticshassomesimilaritieswithvisualprostheses, namelytheattempttocontrolthetiming
of signaling between neural cells in the retina. By rendering cells (i.e. - bipolar or ganglion cells)
further downstream from photoreceptors to be light sensitive, local retinal circuits will be driven
in an unnatural way, leading to complex retinal responses, much like electrical stimulation of the
visual system. Likely, a clinical solution would also require an external device (i.e. - a pair of
glasses with a camera), that will capture the image of the world and translate the image into an
amplified blue light signal to drive bipolar or ganglion cells. Thus far, pre-clinical development
has shown safety and proof-of-concept in both electrophysiological and behavioral experiments
(Lagali etal., 2008; Doroudchi etal., 2011; Thyagarajan etal., 2010; Lin etal., 2008). In vivo
experiments that introduce light sensitive protein channels, such as channelrhodopsin (ChR2), to
ON and OFF bipolar or ganglion cells in a degenerate mouse retina show a renewed responsiveness
tolightstimulus(Lagalietal.,2008;Doroudchietal.,2011;Bietal.,2006;Thyagarajanetal.,2010;
Lin etal., 2008; Zhang etal., 2009). Furthermore, optogentically treated animals have measurable
visually evoked potentials (VEPs) and optomotor responses (Tomita etal., 2010; Tomita etal.,
2009), and are able to perform behavioral tasks requiring residual vision (Lagali etal., 2008;
Doroudchi etal., 2011; Thyagarajan etal., 2010; Lin etal., 2008). Unfortunately, light stimulation
at levels that are 5 orders of magnitude greater than the threshold of cone photoreceptors will be
11
necessary to activate the ChR2 channels (Schnapf etal., 1987) and the operable range of usable
light levels could be severely limited (Wang etal., 2007a).
1.3 Structure and computation of the retina
In normal vision, light is focused by the cornea and lens on retinal photoreceptors located at the
back of the eye in front of the pigment epithelium. The photoreceptors transduce this information
into an electrical signal which is then processed through the various layers of the retina and finally
sent to higher centers in the brain for processing via the optic nerve, a bundle of ganglion cell
axon fibers. Unlike the conventional sensory receptors such as the cochlea, the retina is part of the
central nervous system and is able to compute and filter information about the visual image, such
as contrast, motion, luminance and color of the image. Electrical stimulation interfaces with these
neurons, albeitinanunnaturalway, withtheeventualgoalofproducingactivationpatternswithin
these cells that match the spatio-temporal activity of the normal retina. An understanding of the
highly complex structure and functioning of the retina will give insight into the cells potentially
being activated, thereby giving a partial explanation for both single electrode percept properties
and the spatiotemporal interactions during multi-electrode stimulation.
1.3.1 Structure of the retina
The retina consists of 5 main types of cells, photoreceptors consisting of rods and cones, neural
cells consisting of horizontal cells, bipolar cells, amacrine cells, and ganglion cells (Rodieck, 1988).
These cells are divided into three layers of nerve cell bodies and two layers of synapses. The outer
nuclear layer contains cell bodies of the rods and cones, the inner nuclear layer contains cell bodies
of the bipolar, horizontal and amacrine cells and the ganglion cell layer contains cell bodies of
ganglion cells and sometimes displaced amacrine cells. The outer plexiform layer (OPL) connects
rods and cones to some bipolar cells and horizontal cells. The inner plexiform layer (IPL) contains
12
synapses of bipolar cells to ganglion cells and amacrine cells which interact with bipolar cells to
modulate ganglion cell responses (Oyster, 1999; Rodieck, 1988). Based upon a combination of
morphology and physiology studies, these main types of cells have been further classified into at
least 54 different subtypes (Masland, 2001). There are 4 different photoreceptors (rods and L,
M, and S cones), 2 types of horizontal cells, 9 to 11 bipolar cells (Ghosh etal., 2004), at least
29 different amacrine cells (MacNeil and Masland, 1998), and 10 to 17 ganglion cells (Dacey,
1994; Field and Chichilnisky, 2007).
Figure 1.3: Structure of the retina. Stained retina (left) and retinal schematic (right) with
labeled rod and cone photoreceptors (R & C), outer plexiform layer (OPL), horizontal cells (H),
rod and cone bipolar cells (RB & RC), amacrine cells (A & AII), and ganglion cells (G) (courtesy
of Rachel Wong, University of Washington).
Although retinal stimulation could activate the entire network, epiretinal implants are lo-
cated closest in proximity to the ganglion cell layer and as such are more likely to interface with
IPL circuitry and ganglion cells directly. Thus, this review focuses on cells that synapse at the
IPL, bipolar, amacrine and ganglion cells. Specifically, it is important to consider the density,
distribution, receptive field size and anatomy of the ganglion cells.
13
In total there are one million ganglion cells in the retina. The distribution of these cells is non-
uniform with cell density at its highest in the parafoveal region and dropping off with eccentricity
towards the periphery (Curcio and Allen, 1990). The foveal region measures 200 μm and contains
the foveal pit, a region devoid of inner retinal cells to provide photoreceptors greater access to
light, and the foveal rim, a thicker region of the retina containing six layers of ganglion cells
(Polyak, 1941). Along with changes in cell density, ganglion cells with different receptive field
sizes and stratification of IPL are distributed across the retina (Field and Chichilnisky, 2007).
Generally cells with smaller receptive fields are located closer to the fovea. Neighboring cells tend
to have differing functionalities and are distributed as overlapping mosaics (ibid.; Wässle etal.,
1981).
The retina can be divided into four quadrants, inferior and superior (dividing the upper and
lower visual field) and temporal and nasal. The horizontal raphe is a geometric line which divides
the inferior and superior regions of the retina. Ganglion cells located at the front surface of the
retina send their output through axons that lie on the inner surface of the retina to the cortex
via the optic nerve. For each ganglion cell soma in the retina, its axon fibers traverses to the
optic disk following a pattern that curves around the fovea without crossing the horizontal raphe
(Naito, 1989; Oyster, 1999). Axon fibers are in fact the closest neural component to epiretinal
electrodes.
1.3.2 Computation in the retina
There are about 130 million photoreceptors across the retina that receive the light input and
only one million ganglion cells that output the electrical signal via axons to the cortex. Thus,
there is a significant amount of processing that happens within the layers of the retina in or-
der to compress the signal. Specifically, rather than containing information for each individual
pixel/photoreceptors, neurons respond to pixel differences. Bipolar cells have ON and OFF re-
sponse properties, but instead of a spiking response, they have a graded change in potential and
14
pass the signal onward to ganglion cells. At the IPL synapse, amacrine cells filter the response
of the bipolar cells before the signal is passed on to the ganglion cell layer. Amacrine cells create
both feedback and feed forward circuitry within bipolar cells to provide excitatory and inhibitory
pathways and can have both narrow and wide inhibition (>600 μm) (Roska etal., 2006). The
complex integration system results in unique temporal spiking patterns of ganglion cells (Roska
and Werblin, 2001). Response kinetics can adapt to fluctuations in light intensity (Zaghloul etal.,
2007; Zaghloul etal., 2005; Baccus and Meister, 2002; Rieke, 2001; Chander and Chichilnisky,
2001) for both fast adaptation (less than 100 ms) or slower (on the order of 10 s) time courses.
Thus, someganglioncellresponsesaregreaterforregionsofanimagewheretheluminancechanges
abruptly across space. Ganglion cells can have receptive fields with either ON center and OFF
center response behavior, where different light intensities in the center of a receptive field can
either yield an increase or decrease in firing rate (Field and Chichilnisky, 2007; Barlow etal.,
1964). In addition, there are many other common ganglion cell types, such as ON-OFF cells and
directionally sensitive cells (Barlow etal., 1964) that respond to other properties of the image,
such the presence of edges.
Taken together, these findings suggest that an epiretinal prosthesis with large electrodes could
potentially stimulate complex IPL inhibitory circuitry and activate many ganglion cell types,
fibers of passage from peripheral ganglion cells, and different number of ganglion cells depending
on the location of the stimulation electrode.
1.4 Effects of retinal degeneration
Retinitis pigmentosa describes a group of genetic diseases with different causes and biological
mechanisms, but with similarities in clinical symptoms and biological consequences (Daiger etal.,
2007). Retinal degenerative diseases can be inherited through autosomal-recessive, autosomal-
dominant or X-linked genes and other inheritance means. Research into the different triggers of
15
retinal degeneration (RD) from about 50% of all cases has revealed more than 200 different genetic
causes (Ayuso and Millan, 2010), with each gene mutating to cause multiple forms of the disease
(Daiger etal., 2007). Correspondingly, many different animal genotype and phenotype mutations
of RD exist (Chang etal., 2002). Despite the diversity in the genes and mutations causing pho-
toreceptor cell death, the structural changes in the retina after the onset of RP are remarkably
similar across the different genetic causes (Jones and Marc, 2005). Structural modifications in
the retina are accompanied by functional changes to neuronal processes (Margolis and Detwiler,
2011). Furthermore, over the last 20 years, the effects of long-term visual deprivation on cortical
reorganization have been explored, but are still being fiercely debated (Wandell and Smirnakis,
2009).
1.4.1 Structural changes in diseased retina
A great number of animal morphology studies have revealed distinctive stages in the reorgani-
zation of the neuronal retina during retinal degeneration, though the time course of the disease
can vary substantially (Jones and Marc, 2005). Figure 1.4 taken from (ibid.) shows the effects
of remodeling. First, when photoreceptors begin to degenerate, outer segments are shortened
(Li etal., 1998), rod and cone photoreceptors can bypass connections to horizontal cells (and
bipolar cells) and sprout ectopic neurite projections into the inner nuclear layer (sometimes as
far as the ganglion cells) (Jones etal., 2003; Peng etal., 2000; Sullivan etal., 2007). A small
number of rods are also known to migrate into the inner retina (Jones and Marc, 2005). Sub-
sequently, substantial photoreceptor cell death is accompanied by the formation of a subretinal
glial seal consisting of Müller cells and the initiation of neuronal cell death (Marc and Jones,
2003; Jones etal., 2003). Lack of photoreceptor glutamatergic input causes significant death and
rewiring of the inner nuclear layer that matches human morphology studies (Jones etal., 2005).
Cone bipolar cell dendrites eventually atrophy and horizontal cells develop sprouting processes
into the IPL (Strettoi and Pignatelli, 2000; Strettoi etal., 2003; Strettoi etal., 2002; Gargini etal.,
16
2007). Amacrine cells can sometimes migrate into the ganglion cell layer or towards the OPL
near the glial seal (Jones etal., 2005; Marc etal., 2003)and massive rewiring increases the span of
connectivity within the IPL leading to the formation of recurrent bipolar-bipolar synapses (Jones
and Marc, 2005; Barhoum etal., 2008). Despite these many changes, laminar organization of
the different layers including presynaptic proteins necessary for synaptic function is preserved
(Phillips etal., 2010). While much of the presynaptic connectivity is altered, in rd-1/rd-1 and
rd10 mouse models, ganglion cell structure including its dendritic stratification and projections
to higher visual centers is maintained (Margolis etal., 2008; Mazzoni etal., 2008).
Figure 1.4: Retinal remodeling during RP. Human retina with RP has significant remodeling
and cell loss including cell migration (adapted from Jones and Marc, 2005).
Human morphology studies of degenerate retina have focused largely on quantifying the cell
loss within the layers of the retina and eccentricities to the fovea. Within the macular region
INL cell loss is 20-60% , whereas in the ganglion cell layer (GCL) cell loss ranges between 20-
80% (Santos etal., 1997; Stone etal., 1992). Within the extramacular retina, IPL cell loss was
approximately60%andGCLcelllossrangedfrom70-80%(Humayunetal., 1999b). Thedifference
between degenerated retina extramacular and macular cell loss is hardly surprising considering
17
that RP tends to progress from the periphery towards the fovea. Although cell death is more
substantial for the GCL compared to the INL, the structure of ganglion cells is considerably
better preserved than the presynaptic cells (Margolis etal., 2008; Mazzoni etal., 2008). While
optical coherence tomography (OCT) studies with RP patients have shown a thinning of the
retinal nerve fiber layer compared to normal retina, no mention is made about changes to the
axon pathway patterns (Walia etal., 2007).
1.4.2 Functional changes in diseased retina
Unsurprisingly, such drastic structural changes to the retina are coupled to a number of func-
tional changes to both presynaptic and postsysnaptic cells. Lack of photoreceptor input alters
ON bipolar cell mGluR6 and iGluR glutamate mediated signaling pathways and causes amacrine
and ganglion cells to show spontaneous and stronger iGluR mediated responses (Marc etal.,
2007; Varela etal., 2003; Ye and Goo, 2007), although slower degenerate models may maintain
rod bipolar glutamate sensitivity (Barhoum etal., 2008). ERG response studies show a decrease
in bipolar cell response that precedes morphological changes (Gargini etal., 2007). Data suggests
that ON bipolar cells may even shift behavior to resemble OFF bipolar cells (Stasheff, 2008)
and bipolar-bipolar recurrent synapses could compromise signaling pathways in the inner retina
(Marc etal., 2003). However, in vivo studies in a degenerate mouse retina that introduce photo-
sensitivity to ON bipolar cells (through genetically encoded neuromodulators) show ganglion cell
spiking activity in response to light stimulus (Lagali etal., 2008; Doroudchi etal., 2011). Further-
more, these optogentically treated animals are able to perform behavioral tasks requiring func-
tional vision (Lagali etal., 2008; Doroudchi etal., 2011). Together these findings imply that the
bipolar-ganglion cell functional pathway remains functional at some level despite degenerative ef-
fects. Though ganglion cell morphology is maintained (Margolis etal., 2008; Mazzoni etal., 2008),
degeneration-induced spontaneous oscillatory spiking activities in both ON and OFF RGCs have
been observed, however intrinsic excitatory and inhibitory response properties are still present
18
(Margolis etal., 2008; Stasheff, 2008; Kolomiets etal., 2010). The cause of the spontaneous gan-
glion cell activity has been linked to presynaptic mechanisms and resonant bipolar-amacrine cell
pathways that arise in the absence of photoreceptor input (Margolis and Detwiler, 2011).
1.4.3 Cortical Plasticity
Whenablindspotisintroducedtotheretinaviaalesion, thevisualsystemperceptuallyfillsinthis
‘blind spot’ to complete the image of the world (Wandell and Smirnakis, 2009; Murakami etal.,
1997). Similarly as retinal degeneration progresses, the visual system has compensatory mech-
anisms in order to optimally represent the visual world. It seems plausible that compensatory
mechanisms would result in long term changes to the structure and function of the visual cortex.
Recent literature states the conflicting viewpoints on the existence and nature of adult visual
cortical plasticity (Wandell and Smirnakis, 2009). Some studies over the last 20 years have sup-
ported the concept that the removal of sensory input to V1 results in a change to receptive field
size and cortical topography (Gilbert and Wiesel, 1992). In the majority of these experiments a
retinal lesion was induced in an animal and the response within the cortical lesion projection zone
(LPZ) was measured. Cells within the LPZ are initially silenced, but eventually begin to respond
to visual stimuli placed outside of retinal lesion. These findings imply a shifting in location for
the receptive field and possibly a change in the receptive field size (Gilbert etal., 1990; Gilbert
and Wiesel, 1992; Dreher etal., 2001; Calford etal., 2000; Gilbert etal., 2009; Giannikopoulos
and Eysel, 2006). Response changes within the LPZ may be due to lateral axonal inputs from
cells outside of the LPZ and changing dendritic spines (Keck etal., 2008). However, these re-
sults mainly relied on electrophysiological recordings from single and populations of cells. On the
other hand, fMRI studies, which can record long-term cortical activity over a wide field of view,
have presented a contrasting viewpoint. Imaging studies in the cortex of primates with retinal
lesions show an initial silencing of the LPZ that is not followed by recovery in response (Smir-
nakis etal., 2005). Human fMRI studies with AMD and RP patients have both shown that cortex
19
corresponding to the region of retinal degeneration respond during task-based stimuli, but not
passive viewing (Masuda etal., 2008; Masuda etal., 2010; Liu etal., 2010). These findings suggest
top-down mechanisms (which are present prior to degeneration) contribute to cortical responses
rather than bottom-up sensory processing or newly developed lateral connections. Based on these
inconsistent views, the potential for adult cortical plasticity to facilitate the development of vision
restoration strategies is still fairly uncertain.
1.5 Electrical stimulation of the retina
Given that visual information is encoded in the retina by the precise temporal firing patterns
of ganglion cells, the ability to control a prosthesis subject’s perceptual experience will require
an understanding of how electrical stimulation interfaces with the different cells of retina and
modulates their collective response. Electrophysiological findings can give insight into which cells
are activated by different types of stimuli and whether changes to the stimuli, be it the pulse train
parameters, the location of stimulus, or the shape and configuration of electrodes, can modulate
thefinalfiringoutputofganglioncells. Furthermore, electricfieldmodelsandinvitroexperiments
can be used to provide insight into the factors which will drive the spatial extent of activation for
various stimuli.
1.5.1 Stimulation Threshold
Threshold quantifies the minimum stimulus necessary to yield the desired response. In psy-
chophysical experiments, threshold is the minimum stimulus required to produce a perceptual
change. In visual electrophysiology experiments, threshold refers to the minimum amount of
stimulus (charge, voltage, amplitude or photons) needed to elicit a spike or change in potential
from a neural cell. Ganglion cell responses are measured as spikes, while the response of bipolar
and amacrine cells are measured as a graded change in potential. An equally important factor
20
limiting the stimulus is the charge density limit for electrical stimulation. Safety requirements set
limits to charge density in order to avoid damage to electrode material or neural tissue (Brum-
mer and Turner, 1977; Brummer etal., 1983; Shannon, 1992). Together, thresholds and charge
density limits (in humans) provide the upper and lower bounds for stimulation respectively, and
thus define the dynamic range of possible stimulation intensities .
In retinal electrophysiology experiments, the electrode is either positioned on the epiretinal
or subretinal surface. Conventionally, ganglion cell spikes are measured in vitro with epiretinal
recording electrodes. Additional techniques to measure cell responses are through whole-cell patch
clamp recordings, in vivo recordings of evoked responses in the superior colliculus (Chan etal.,
2011) and calcium imaging studies (Behrend etal., 2009). In calcium imaging studies, ganglion
cells are retrogradedly loaded with a calcium dye indicator and response activation is quantified
as the degree that a cell fluorescence in response to a stimulus giving a clear picture of the spatial
extent of activation.
To characterize response properties of the retina, strength duration curves, that provide the
relationship between threshold amplitude and pulse duration, have been measured across different
species. Figure 1.5 shows a strength duration curve for a rat retina (values are averaged across 25
cells shown with a solid line). The rheobase defines the threshold current for a current pulse of
infinite duration and is the asymptote value of the curve. The chronaxie is the stimulus duration
time for which the threshold current is at double the amplitude of the rheobase and is also the
most energy efficient stimulus. Average threshold currents across different species (i.e. rat, pig
and monkeys) with 6-25μm electrodes revealed that thresholds for a 0.05 ms pulse were less than
10 μAs and decreased with increasing pulse duration, however total charge increased with pulse
duration (Sekirnjak etal., 2006). Chronaxies values ranged between 100 μs and 400 μs implying
thatshorterpulsedurationsaremorechargeefficient(Friedetal., 2006). Thresholdsalsoincreased
by a factor of 2-3 across the different electrode sizes (6-25μm) and with distance from stimulating
electrode (see Figure 1.5 dashed line ) demonstrating that current attenuates with distance from
21
stimulating electrode (Sekirnjak etal., 2006). Comparable to ~10μA, a separate study with rabbit
retina measured thresholds for a 40 μm electrode as ~15 μA (Fried etal., 2006).
Figure 1.5: Strength duration curve. Average change in threshold (large bottom plot) and
overall charge (small top plot) as a function of pulse duration at different recording distances for
rat retina (adapted from Sekirnjak et al., 2006).
Thus far, threshold studies thus far have shown conflicting results on the effect of retinal
degeneration on threshold. Thresholds for rd1 mice were approximately an order of magnitude
higher than wild-type (Jensen and Rizzo, 2008; Goo etal., 2011). Similarly, increases in threshold
(by nearly a factor of 2) were also noted for in vivo experiments recording evoked responses
22
from the superior colliculus of the RD rat (Chan etal., 2011). However, an in vitro study with
very small 10 μm electrodes showed no change in threshold between the P23H rat and normal
(Sekirnjaketal., 2009). Withapossibleincreaseinthresholdsfordegenerateretina, shortduration
pulses are more likely to have a charge density that is within the safety limits and could provide
a greater dynamic amplitude range compared to longer duration pulses.
1.5.2 Stimulation Specificity
Given that the retina consists of many different neural cell and subclasses of cells with varying
cell bodies and dendritic sizes, it is conceivable that threshold would also vary across cell types.
Thus, when designing a retinal prosthesis, the ability to target specific retinal cells based on their
response properties might allow greater control over the subject’s perception.
Thresholds for subretinal stimulation with anodic pulses are lower than cathodic pulses and
threshold variability is less with biphasic stimulation (Jensen and Rizzo, 2009). Cathode-first
pulses preferentially stimulation ON retinal ganglion cells for epiretinal stimulation since thresh-
olds for OFF retinal ganglion cells is higher than ON cells (Jensen and Rizzo, 2006; Jensen etal.,
2005a).
Recordings from ganglion cells typically generated two types of responses, a short (less than
2-5 ms) and a long latency (greater than 2 or 8 ms) response defined as the time between the
stimulus pulse onset and the start of the first observed cell spike. Long latency spikes disappeared
when synaptic blockers were used to eliminate the effects of presynaptic cells. This demon-
strates that long latency spiking can be attributed to presynaptic cells response and short latency
spikes can be attributed to direct activation of ganglion cells (Sekirnjak etal., 2006; Ahuja etal.,
2008; Shah etal., 2006; Jensen etal., 2005b; Jensen etal., 2005a). Based on integration properties
of cells, short duration pulses < 0.15 ms result in short latency spikes exclusively (Fried etal.,
2006; Sekirnjak etal., 2008) , while longer duration pulses > 2ms caused long latency spiking
(Sekirnjak etal., 2006; Greenberg, 1998). However some studies claim that 2 ms pulses create
23
both a short and long latency spike, indicating that ganglion cells are driven first before presy-
naptic cells (Shah etal., 2006). A recent study utilizes low frequency sinusoid stimulation (<25
Hz) to preferentially stimulate presynaptic cells, while avoiding stimulation of ganglion cells (Free-
man etal., 2010a; Freeman etal., 2010b). The slow rise of a low frequency sinusoid may be able to
trigger the direct release of presynaptic neurotransmitters rather than ganglion cell voltage gated
sodium channels (Freeman etal., 2010a). Thus some preference for stimulating presynaptic vs.
ganglion cells may be possible.
Ideally retinal stimulation will be able to produce ganglion cell spiking patterns that replicate
the temporal and spatial specificity of normal vision. One way to achieve temporal specificity is
to activate ganglion cells with a train of short duration biphasic pulses. With a train of short
duration pulses (cathodic phase <0.15 ms), a single spike per pulse is generated for frequencies
up to 250 Hz (Fried etal., 2006; Sekirnjak etal., 2008; Ryu etal., 2009a). On the other hand,
with suprathreshold stimulus amplitudes (20 μA-60 μA), electrodes with a diameter of 30 μm
generate multiple spikes for pulses < 0.5 ms (Ryu etal., 2009b; Ryu etal., 2010). Furthermore,
the number of elicited spikes for each pulse varies amongst different ganglion cell types (Cai etal.,
2011). With high frequency stimulation > 10 Hz, long latency responses likely attributed to
presynaptic mechanisms are suppressed (Sekirnjak etal., 2006; Ahuja etal., 2008). Overall, these
findings propose that the frequency of stimulation can control ganglion cell spiking with relatively
high temporal specificity. However, adaptive effects causing ganglion cell response to desensitize
with successive pulses could limit the operating range of frequencies (Ryu etal., 2009a; Jensen
and Rizzo, 2007; Freeman and Fried, 2011).
1.5.3 Spatial extent of activation
Suprathreshold levels of stimulation will be necessary to produce meaningful and reliable percepts.
Electric field current spread may activate cells beyond the diameter of the stimulating electrode.
Another concern for epiretinal implants is that percepts might be elicited by stimulation of both
24
the axon fiber as well as the axon initial segment (hillock). Such axonal fiber stimulation could
lead to phosphenes that are elongated in shape and poorly localized (Freeman etal., 2011). Elec-
trophysiology experiments and theoretical models have explored the factors likely to control the
spatial extent of retinal stimulation at suprathreshold amplitude levels.
With electric field modeling, the extracellular potential at a distance from the electrode can
be calculated theoretically. Early studies that assumed a homogeneous neural tissue medium, an
infinite return electrode and a static stimulus calculated the extracellular potential as (Wiley and
Webster, 1982a; Wiley and Webster, 1982b; Greenberg etal., 1999; Cottaris and Elfar, 2005).
Where V(r,z) is the potential at a location r (the x-y plane) and z, and a is the radius of the
electrode. The current at any given location can be assumed to be proportional to the gradient of
thepotential(WileyandWebster, 1982a). Thestaticequationaboveyieldsasharpdropincurrent
with distance from the stimulating electrode and infinite current at the edge of the electrodes. A
more accurate measure of current would take into account time varying properties of the stimulus
anditsinteractionwithneuraltissue(Behrendetal., 2008). Indeed, electrophysiologyexperiments
exploring the change in threshold with distance from the stimulating electrode show a slightly
different current spread than the above equation (Ahuja etal., 2008). To further support time
varying effects of the electric field, in vitro experiments suggest that longer pulses create a larger
electric field spread.
Intheretina, axonsfromaganglioncelloriginatingintheperipherywillpassbymorecentrally
located cells. Thus, an electrode situated on these central cells, is also situated on passing axons
fibers. Initial electrophysiology experiments and current modeling showed that the site of the
action potential was at the initial segment of the axon with close proximity to the soma. The
reason for its location has been attributed to the bend in the axon hillock (Schiefer and Grill,
2006), a thinning in the fiber at the initial segment (Carras etal., 1992) and higher density of
25
Figure 1.6: Axonal activation with electrical stimulation. Threshold map obtained with
calcium imaging experiment in salamander retina. Blue circle shows a 200 μm electrode, solid red
circles and streaks indicate activation of ganglion cell somas and antidromic axons (adapted from
Behrend et al., 2008) .
sodium channels (Wollner and Catterall, 1986). Recent studies have confirmed that the site of
the action potential is in a region located 0-40 um from the soma containing a high density
of sodium channels (Fried etal., 2009). With small electrodes, some studies have shown that
cathodic stimulation produces thresholds in the initial segment that are up to a tenth lower than
the rest of the axon (Jensen etal., 2003; Sekirnjak etal., 2008). However, calcium imaging studies
with larger electrodes have shown that axon fibers are activated by epiretinal stimulation and
that thresholds of the axon initial segment and the entire fiber are similar (Behrend etal., 2009).
Figure 1.6 shows the extent of spatial activation with large electrodes and a short pulse duration
pulse train; ganglion cell axons are clearly activated with threshold levels comparable to the initial
segment. An in vitro study has demonstrated that low frequency sinusoid stimulation is able to
26
avoidaxonalstimulationbyinitiallyactivatingpresynapticcells(Freeman etal., 2010a). Modeling
studies have suggested that the location of the electrode relative to the orientation of passing axon
fibers will effect activation (Rattay, 1989; Rattay, 1999). Electrodes placed perfectly parallel to
the orientation of passing axon fibers could avoid axonal stimulation (Rattay and Resatz, 2004).
However, implant precision that would allow placement of such an electrode configuration is not
feasible.
Taken together, electrophysiology experiments and system modeling do not seem to clearly in-
dicate optimal stimulation parameters. Temporal precision may be achievable with trains of short
duration biphasic, cathode-first pulses, but this is likely to compromise the spatial localization
of percepts due to stimulation of passing axon fibers. Sinusoidal stimulation may have improved
spatial benefits, but is limited in the temporal control of the retinal response and the inefficient
use of charge could limit the range of brightness levels that can be presented safely.
1.6 Clinical findings with an epiretinal prosthesis
Over 30 subjects have been chronically implanted with a 2-Sight epiretinal prosthesis over the past
7 years. As a result, there have been a great number of clinical experiments performed on these
subjects. Since this thesis explores clinical findings on the same pool of subjects, it is important to
provide a detailed literature review highlighting the major findings and implications to the field.
Initial epiretinal acute experiments were performed on 2 subjects over 10 years ago. Findings
suggested that threshold in the extramacular region were higher than the foveal region, suggesting
that arrays should be implanted in the macular region (Humayun etal., 1999a). Initial chronic
studies characterizing the system measured thresholds for 6 subjects using a biphasic pulse with
a cathode duration of 0.975 ms and electrode-retina distance using OCT to image the retina.
Results showed that thresholds were mainly below the safety charge density limit and increased
with an increase in electrode-retina distance (De Balthasar etal., 2008; Mahadevappa etal., 2005).
27
Though electrode size varied between either 260 and 520 μm, thresholds did not seem to change
with electrode size (De Balthasar etal., 2008).
1.6.1 Properties of percepts
In threshold experiments, subjects anecdotally described percepts as whitish or yellowish round or
oval shaped spots (Mahadevappa etal., 2005). Single electrode experiments show that perceptual
threshold and brightness can be manipulated predictably with a variety of parameters such as
frequency, pulse duration and amplitude (Horsager etal., 2009; Greenwald etal., 2009). Interest-
ingly, threshold is affected by the distribution of charge in time in a way that can be related to the
neural integration and adaptive properties of the visual system (Horsager etal., 2009). Brightness
of a single electrode percept increases as a power function with stimulation intensity, thereby
making is possible to code a dynamic range of brightness with different stimulus amplitude levels
(Greenwald etal., 2009). At high amplitudes, the brightness of phosphenes tends to saturate
possibly limiting the dynamic range.
When multiple electrodes are stimulated, percept appearance is highly complex (Horsager,
2009). Indeed, even a change in stimulation timing offset between electrodes as far as 1600 μm
can lead to a change in percept appearance (Horsager etal., 2010). These findings suggest that
spatiotemporal interactions exist at the neural level and contribute to percept appearance. These
results are further supported by findings that the brightness of a single electrode phosphene can
be changed by stimulating an additional electrode (Horsager etal., 2011). Taken together, these
results suggest epiretinal stimulation activates a highly complex system of neural cells and that
each electrode does not behave independently from others.
1.6.2 Evidence of form vision
Subjects implanted with chronic epiretinal prostheses have demonstrated functional low level
vision when using the entire system (with a camera) (Yanai etal., 2007) . In one experiment, a
28
singlesubjectisabletodiscriminateorientedvisualgratingsat4differentorientations(Caspietal.,
2009). Furthermore when video input information is scrambled, the subject was unable to perform
the task. These findings seem to indicate that the subject had form vision beyond a simple light
detector, however these results do not speak to what the subject saw during stimulation. In
another experiment, performed on a larger subject pool, subjects were able to point to a bright
squareonadarkbackgroundmonitor(Ahujaetal.,2011). Nearlyallsubjectsshowedimprovement
in the task with device on vs. device off . Subjects were also able to discriminate high contrast
letters on a dark background monitor and using a head scanning technique, discriminate a series
of letters to form words and sentences (Sahel etal., 2011). While results show promise of form
vision, the ability to create meaningful high-resolution images with future dense arrays is still
uncertain.
1.7 Outline of this thesis
This thesis strives to understand the perceptual experience of epiretinal electrical stimulation in
blind human subjects. I will explore the properties of percept shape as a function of stimulus
waveform parameters (i.e. amplitude, frequency, pulse duration, pulse shape) and location of
the stimulus on the retina. Previous studies have already explored the brightness and threshold
of electrically elicited perception as a function of pulse timing with single and multi-electrode
stimulation.
Chapter 2 describes the experimental setup and general methods common to all experiments
in subsequent chapters. Details about each of the subjects used in experiments are also presented.
Chapter 3 explores the reliability and reproducibility of single and paired electrode phosphene
shape across several shape descriptors such as area, major and minor axis length, and orientation.
But first, since experiments rely on subject’s ability to draw percepts accurately and consistently
29
across trials, a control experiment with tactile targets is performed to measure drawing error that
is later compared to phosphene variability.
Chapter 4 explores properties of single electrode phosphene shape across all four subjects. The
goal was to qualitatively compare percepts across subjects and quantitatively explore the change
in various shape properties such as area, major and minor axes, and orientation as a function of
stimulus location in the retina. Findings suggested that percept characteristics were predictable
and a model based on these findings was developed in Chapter 5.
Chapter 5 expands upon the ideas presented in Chapter 4 by developing a model based on
the anatomy of the retina with the pathways of nerve fiber tracts. We compare actual phosphene
drawings to predicted percepts generated by the model and explore factors that may affect system
resolution in paired electrode stimulation.
In Chapter 6, we explore how manipulating pulse train frequency and amplitude have different
effects on the size and brightness of phosphene appearance. Findings suggest that frequency
modulation may improve the encoding of a wide range of brightness levels without a loss of
spatial resolution.
Chapter 7 presents ways to drastically alter phosphene appearance with different stimuli. In
one subject, we show that variations on a low frequency sinusoid stimulation create significantly
different percepts than phosphenes produced in earlier chapters. In a different subject, we show
that high frequency stimulation can also cause a considerable change in phosphene shape that
was not previously observed.
Chapter 8 concludes with the significance of all previous findings and speculates the implica-
tions of these results when developing future artificial vision devices.
30
Chapter 2
Experimental setup & General methods
2.1 System (Argus I and Argus II)
The Second Sight Medical Products Inc. epiretinal prosthesis contains both intraocular (electrode
array) and extraocular (e.g., glasses, Visual Processing Unit) components. The first-generation
device (Argus I system) has 16 electrodes (see figure 2.1A), while the second-generation device
(Argus™ II system) has 60 electrodes (see figure 2.1B). The intraocular array, implanted epireti-
nally in the macular region of the retina, consists of 16 platinum disc electrodes (Argus I system)
in a 4 x 4 arrangement or 60 electrodes in a 6 x 10 arrangement (Argus™ II system) and is
contained within a clear silicone rubber platform . The array is held in place with a retinal tack.
In the Argus I system, the electrodes implanted are 260 or 520 μm in diameter (subtending 0.9
o
and 1.8
o
of visual angle, respectively) arranged in an alternating checkerboard pattern spaced 800
μm center-to-center. In the Argus II system, the electrodes implanted are 200 μm in diameter
(subtending 0.7
o
of visual angle) with 525μm center-to-center spacing. Custom software on a PC
laptop is used to program the external Visual Processing Unit (VPU), which in turn send stimulus
commands to the implant. Power and signal information are sent from the VPU through a wire to
an external transmitter coil that is attached and aligned magnetically, and coupled inductively, to
a secondary coil that is either implanted subdermally in the subject’s temporal skull behind the
31
ear (Argus I system - see figure 2.1C) or located along the sclera of the eye (Argus™ II system -
see figure 2.1D). The secondary coil provides power and signal information to an implanted pulse
generator (IPG), which decodes the signal and produces the commanded stimulus pulses. The
IPG transmits pulses to the array of electrodes via a multi-wire cable that traverses the sclera
. Stimulation can be presented using two different protocols: 1) camera mode – real-time video
captured by a miniature video camera mounted on the subject’s glasses is continuously sampled
by the VPU to match the stimulation current amplitude in each electrode to the brightness at
the corresponding area of the scene and 2) direct stimulation mode - the stimulation signal sent
to each electrode is independently controlled by the VPU. All experiments in this thesis were
performed in direct stimulation mode.
Figure 2.1: Schematic of the Argus I and Argus II arrays and systems. (A) Argus I
4x4 electrode array, (B) Argus II 6x10 electrode array, (C) Overview of the Argus I implant, (D)
Overview of the Argus II implant
32
2.2 Subjects
2.2.1 Selection criteria
As part of two separate clinical trials run by 2-Sight Medical Products Inc., over thirty-eight
subjects worldwide (six Argus I and thirty two Argus™ II subjects) have been implanted with
an epiretinal prosthesis since 2002. Although this provided us with a large subject pool to obtain
data from, with time constraints we felt that it was most efficient to collect data from a subset of
subjects. This decision was based on number of factors. Subjects were selected using the following
criteria: electrode array to retina distance, location of the array with respect to the fovea, and
single electrode threshold levels.
All subjects were required to have electrode to retina distances below 50 μm (as measured
with an Optical Coherence Tomography (OCT) tomogram) with the array pushing down towards
the retinal surface (not the retinal surface being lifted up towards the array). OCT array-retina
distances were measured by the research department at 2-Sight Medical Products Inc. The arrays
placement needed to be on the macular region of the retina (estimated with subject’s fundus
photograph). Estimated fovea location was calculated by methods explained in section 2.4.1.
Single electrode threshold levels needed to be less than 233 μA, with most being below 100 μA.
Thresholds were measured using methods explained in section 2.3.1.
We felt that subjects with minimal electrode-retina distance would have less current shunting
in the vitreous space between the array and retina, which could potentially cause inconsistencies
in measuring phosphene appearance. We also needed subject thresholds to be low in order to
perform suprathreshold experiments at an amplitude level that subjects could reliably identify
phosphenes. Furthermore, since successful implants are also assessed by similar criteria (i.e.-
macular placement, low threshold and minimal electrode-retina distance), conclusions drawn from
thesis results with this particular subset of subjects would be most relevant to future prosthesis
development.
33
2.2.2 Subject details
We performed experiments on a single subject chronically implanted with 16-channel retinal pros-
theses and three subjects chronically implanted with a 60-channel device (Second Sight ® Medical
Products, Inc.). These subjects are part of an FDA approved multi-center clinical trial. The clin-
ical trial is registered at http://www.clinicaltrials.gov. All tests were performed after obtaining
informed consent under a protocol approved by the Institutional Review Board (IRB) at each
subjects location and under the principles of the Declaration of Helsinki. For example, Subject
1 (see 2.2.2.1) was approved by the University of Southern California IRB at the Keck School of
Medicine.
2.2.2.1 Argus I Subject
Subject 1 is male, age 62, and was implanted in 2004 with the Argus I device at the USC Doheny
clinical site. Pre-operatively, the subject had no Light Perception (NLP) in both eyes, was blind
for 10.5 years before implantation and was 55 years of age when implanted. Tests were carried
out between 2008 and 2010 (4-6 years after implantation). In 2009, a retinal detachment was
discovered in the subject’s implanted eye, in a location far from the actual implant array. The
detachmentwasfixedsurgically; thresholdsandarray-retinadistancedidnotchange. Thelocation
of the array shifted slightly inferior between 2008 and 2009 as revealed by the subject’s yearly
routine fundus photographs (see Figure 2.2). The shift in the array location was not drastic (the
array remained over the macular region) and seemed to be unrelated to the retinal detachment.
Thresholds and array-retina distance did not change after array movement.
2.2.2.2 Argus II Subjects
Subject 2 is male, age 72, and was implanted in 2009 with the Argus II device at the Johns
Hopkins clinical site. Pre-operatively, the subject had NLP in the implant eye, was self reported
to be blind for 1 year before implantation and was 70 years of age when implanted. Tests were
34
Figure 2.2: Shifted array for Subject 1. Fundus photographs with array location and optic
disc in retina clearly visible. (A) Photograph from March, 2008, with black arrows pointing to
optic disc center (right arrow) and array center (left arrow). (B) Photograph from March, 2009
with arrows marking same retina locations in panel A (notice that left arrow does not mark the
same array location as panel A, demonstrating a shift in the array position).
carried out in 2010 (8 months- 1 year after implantation). Based on the subject’s OCT, the retina
has macular cysts (retinal holes). Subject 3 is female, age 47, and was implanted in 2009 with the
Argus II device at the London clinical site. Pre-operatively, the subject had NLP in the implant
eye and was 45 years of age when implanted. Tests were carried out in 2010 (6 months- 1 year
after implantation). Subject 4 is male, age 52, and was implanted in 2009 with the Argus II device
at the Manchester clinical site. Pre-operatively, the subject had BLP (bare light perception) in
the implant eye, was self reported to be blind for 21 year before implantation and was 50 years
of age when implanted. The subject seems to have good remaining vision in the implant eye and
likely has a large number of remaining photoreceptors. Tests were carried out in 2010 (6 months-
1 year after implantation).
2.3 Psychophysical methods
Due to geographic location, not all subjects were directly examined by the author of this thesis.
Instead, initialexperimentaldesign and pilot studies were conducted withSubject 1at the Doheny
35
Eye Institute of USC. For results collected on Argus II subject, experiment procedures were
written by the author of this thesis and sent to the each individual site. Trained field clinical
engineers (FCEs), employed by 2-Sight Medical Products, performed the experiments as written
in the procedure document. All Argus II drawing data were collected in raw form and sent to
the author for subsequent analysis. Argus II threshold data analysis was performed by 2-Sight
Medical products and final results forwarded to the author.
2.3.1 Threshold measurement
2.3.1.1 Argus I Subject
Perceptual thresholds were measured using custom-developed software on single electrodes us-
ing an adaptive yes-no procedure. Stimuli for measuring threshold was a charge-balanced, 0.45
ms/phase cathodic-first biphasic 20 Hz pulse trains, 500 ms in duration. There was no inter-
phase interval (i.e. - the total cathode-anode pulse width was 0.90 ms). On each trial, subjects
were asked to judge whether or not they saw a phosphene on that trial. Half of the trials were
stimulus-absent catch trials interleaved randomly with the stimulus-present trials. Current am-
plitude was varied using a three-up-one-down staircase procedure to find the threshold current
amplitude needed for the subjects to see the stimulus on 50% of stimulus-present trials, corrected
for the false alarm rate. During each staircase, only amplitude varied while all other parameters
(frequency, pulse width, pulse train duration, and the number of pulses) was held constant. Each
threshold was based on fitting a Weibull function to a minimum of 125 trials and error bars are
estimated using Monte-Carlo simulation (Wichmann and Hill, 2001).
2.3.1.2 Argus II Subjects
Perceptual thresholds were measured using custom-developed software on single electrodes using
a yes-no procedure by combining a method of constant stimuli (MOC) and adaptive stimuli
procedure. In conventional adaptive methods, the next stimulation amplitude is determined by
36
previous responses, while in a MOC stimuli, test stimuli current amplitudes are pre-determined
and presentation order is randomized. Stimuli for measuring threshold was a charge-balanced,
0.45 ms/phase cathodic-first biphasic 20 Hz pulse trains, 250 ms in duration. As in the Argus I
stimulation, there was no interphase interval (i.e. - the total cathode-anode pulse width was 0.90
ms). However in the Argus II procedure, rather than testing each electrode individually, as many
as six different electrodes were tested in a single experimental run. The entire experiment was
divided into five separate blocks of 12 trials per electrode in each block for a total of 72 stimulation
trials per block. In addition to stimulation trials, 32 catch trials in total were interspersed over
the five presentation blocks. Thus, the maximum number of trials over 5 blocks was 392 trials.
In each trial, either an electrode was selected randomly from the set, or stimulus-absent catch
trial was presented and subjects were asked to judge whether or not they saw a phosphene on
that trial. Current amplitude levels were pre-selected using the MOC stimuli procedure while all
other parameters (frequency, pulse width, pulse train duration, and the number of pulses) was
held constant .
After each block, based on all previous responses, a maximum likelihood algorithm determined
stimulationvaluesinthesubsequent block foreachelectrode. In addition, an estimate of threshold
was determined based on fitting responses with a Weibull function and error bars determined
using a Monte-Carlo simulation. If the confidence interval of the estimated threshold for a single
electrode in the group was within 95%, no further stimulation was needed for that particular
electrode in subsequent blocks and the estimated threshold was set as the final threshold value.
Trials on the other electrodes continued through to a maximum of five blocks. Results were
deemed unreliable if the false alarm rate, determined by the percentage that the subject saw a
stimulus during catch trials, was greater than 20%. Data from runs with higher false alarm rates
were removed from analysis and the runs were repeated.
37
2.3.2 Drawing task
Subjects were asked to perform a drawing task with a tactile target (see section 3.2) or when their
retina was electrically stimulated. In a given experimental run, a total of n stimulus conditions
(either tactile or retinal stimulation) were tested. Each condition was repeated for m trials (for a
total of m*n trials per experimental run). Repeated trials of the same condition were randomized
amongst other stimuli to confirm reproducibility of results.
2.3.2.1 Argus I Subject
Head movement was minimized with a chin rest. After each stimulus presentation, the subject
traced the shape on a grid screen (containing 6 inch horizontal and vertical gridlines) with a center
location aligned horizontally and vertically with the subject’s head. Drawing was carried out with
apenwhosecapwasadifferentcolorthanitsbody. Aheadmountedcamera(MisumiCMOSS588-
3T), located on the subject’s glasses, was used to record the trials to digital video recorder (DVR).
Video files were analyzed off-line to extract shape data using custom built tracking software. In
the first stage of processing, the entire image was rotated appropriately using the grid screen
background as a reference. In the second stage, vertical and horizontal gridlines, and the distance
from the subject to screen were used to set a new coordinate system in visual angle co-ordinates
(since the subject was 16 inches / 40.6 cm from the screen, 4 gridlines = 70.0 cm corresponded
to 73.8 degrees visual angle). In the third stage, the location of the pen cap was tracked (based
on its color) across each frame of the video file. Finally, a binary shape data file was built from
pen cap coordinate locations across all frames.
2.3.2.2 Argus II Subjects
Subjects were placed in a chair at a comfortable distance from a touchscreen monitor with its
center location aligned horizontally with the subject’s head. The distance from the subjects eyes
to the screen was recorded. After each stimulus presentation, the subject traced the shape on the
38
Figure 2.3: Argus I subject drawing task. Subject drawings on a grid screen are captured by
an external camera and recorded to a video file. Video files are analyzed offline by tracking the
location of the pen tip from frame-to-frame and translated to a binary image.
monitor and the experimenter advanced to the next trial (see figure 2.4). Touch screen data was
instantly recorded by custom software in X-Y coordinates to a text file. Text files were analyzed
offline to translate vector coordinates to a binary shape data file. The distance recorded from
the subject to screen was used to set a new coordinate system in visual angle. Since Subjects 2-4
were 33 , 30.0, and 30.5 inches from the screen, this corresponded to 60, 65 and 64 degrees visual
angle horizontal screen length respectively. After translating to the final visual angle coordinate
system, the binary image was used in subsequent shape analyses.
Figure 2.4: Argus II subject drawing task. Subject drawings recorded by a touchscreen
monitor in x-y screen coordinates and translated to a binary image.
39
2.4 Analysis
2.4.1 Estimating the fovea on a fundus photograph
Each subject’s fovea location relative to the electrode array was estimated by examining the
subject’s fundus photograph. In the photograph, the visible electrode array and optic disc are
used as fiducial markers. First, the photograph is scaled so that the distance between electrodes
in image pixels matches the true electrode spacing (800 μm for the Argus I system and 525 μm
for the Argus™ II system). Subsequently, by using a conversion that one degree of visual angle is
equal to 288μm on the retina, assumed without correction for shrinkage, the image can be scaled
to visual angle space (Drasdo and Fowler, 1974). The fovea is approximated to be temporal and
inferior to the optic disc. Thus the fovea is approximated at an angular distance of 15.5±1.1°
(temporal to the center of the optic disc), and at a vertical angular distance of -1.5±0.9° (inferior
to the center of the optic disc) (Rohrschneider, 2004). An example of these measurements and
calculations is given in figure 2.5.
Figure 2.5: Estimating fovea location. Fundus photograph for Subject 4 shows estimate of
foveal region. Array placement is clearly in the macular region of the retina.
40
Chapter 3
Phosphene reproducibility
3.1 Introduction
The ability of a visual prosthesis to reliably control the appearance of percepts will partially
determine the quality and accuracy of the resulting vision. Thus, it is important to explore the
stability and reproducibility of the perceptual experience of using a retinal prosthesis.
In a previous study, our group developed quantitative predictive models showing that the
perceptual threshold and apparent brightness of electrically elicited phosphenes is predictable
and accordingly, reproducible (Horsager etal., 2009; Horsager etal., 2011; Mahadevappa etal.,
2005; De Balthasar etal., 2008; Greenwald etal., 2009). Furthermore, the perceptual experience
of these devices is reliable enough for some subjects to perform a variety of complex spatial tasks
(using the camera and head scanning techniques) to varying degrees of success (Ahuja etal.,
2011; Caspi etal., 2009; Yanai etal., 2007; Wilke etal., 2011; Benav etal., 2010). However, to
date, there has been little work to examine the shape of electrically elicited percepts and quantify
the consistency of the visual experience from trial to trial.
Though animal studies have shown that electrical stimulation is capable of reliably generating
firing patterns within a small number of cells that match the stimulus pulse trains (Fried etal.,
2006), these findings may not translate to accuracy in a subject’s perceptual experience, since
41
cellular resolution of these studies is far beyond the abilities of current technology. Presently, a
single electrode simultaneously activates hundreds to thousands of cells with a wide variation of
structure (Masland, 2001) and function (Field and Chichilnisky, 2007).
Studies directly examining the variability of spatial properties of phosphenes are sparse. In
one study using an optic nerve stimulation device, phosphene positional variability was reported
to be ~5-10° visual angle (Obeid etal., 2010), but this study did not examine the shapes of these
phosphene. In another study with an acute epiretinal implant, results were not quantified, simply
reporting that repeated paired stimulation trials resulted in ‘similar’ phosphenes on 66% of trials
(Rizzo etal., 2003).
One of the first steps in exploring the stability of electrically elicited images is to determine
the consistency of its building blocks, namely the single and paired-electrode percepts. In this
chapter, we develop the methods for collecting and quantifying phosphene shape data using a
drawing task. Over the course of 3 years, phosphene shape data was collected for a series of
retinal stimulation experiments (see chapters 4, 5 and 6) using both single and paired electrodes
and a variety of stimulation parameters. Data from different experiments collected over the years
is pooled to give an overview of phosphene consistency. Our findings demonstrate that we are
able to measure the repeatability of single and paired electrode phosphenes in a visual prosthesis
patients. The variability of these basic building blocks gives insight into the degree to which we
can control a subject’s perception. The reliability of electrically elicited percepts is relevant to
the development of higher resolution arrays with form perception capabilities.
3.2 Materials & Methods
3.2.1 Subjects
This set of experiments were performed on Subject 1 over the course of 3 years (2008-2010),
and subject 2-4 over the course of 1 year (2010) . Details about each of the subjects can be
42
found in sections 2.2.2.1and 2.2.2.2. As an aside, Subject 1 is more trained in retinal stimulation
experiments than Subjects 2-4, having been implanted in 2004 and trained in psychophysical
experiments for over 6 years.
3.2.2 Psychophysical methods
3.2.2.1 Control task - tactile drawing
Our experiments rely on our subject’s ability to draw percepts accurately and consistently across
trials. However, our blind subjects lacked tactile-visual feedback for many years and as a result
wouldbelikelytoshowmorevariationindrawingthanablindfoldedsightedsubject. Wetherefore
began by performing a control experiment with tactile targets to establish baseline drawing error
for each subject. For the Argus I subject (Subject 1), the test stimuli consisted of a set of 11
tactile shapes made out of felt with a cardboard background. For Argus II subjects, methods
were refined to use 6 tactile shapes made again out of felt with a cardboard background. Argus
I and Argus II shapes are shown below in figure 3.1A & B. Subjects were asked to feel the felt
shapes, and then draw them on a board or a touchscreen (figure 3.1C). Shape data was recorded
for Subject 1 using the methods explained in section 2.3.2.1 and for Subjects 2-4 using methods
explained in section 2.3.2.2.
3.2.2.2 Retinal stimulation
Retinalstimulationexperimentswerecharge-balanced, 0.45ms/phasecathodic-firstbiphasicpulse
trains with a total stimulus duration of 500 ms (Argus I) or 250 ms (Argus II). Pulses were
charge balanced across cathodic and anodic pulses for safety reasons. All data were recorded
under photopic conditions. Data from section 7.1 were not used in the analysis. Stimulation
experiments were placed in 3 categories: (1) Single electrode stimulation (Singles), (2) Single
electrode stimulation collected with paired stimulation (Single w/Pairs), (3) Paired electrode
stimulation(Pairs). WithintheSinglescategory, stimulationfrequencyrangedfrom6Hz-120Hz,
43
Figure 3.1: Tactile target control task. (A) Argus I tactile shapes. (B) Argus II tactile shapes.
(C) Subject feels shape before tracing on a screen
with the majority of stimulation at 20 Hz and amplitude ranged from 1.25 - 7.5 x threshold, with
the majority of stimulation at 1.5-2 x threshold. Within the Singles w/Pairs category, stimulation
ranged from 1.25-4 x threshold at 20 Hz frequency. Within the Pairs category, stimulation ranged
from 1.25-4x threshold at 20 Hz and each of the two electrodes were activated synchronously and
asynchronously (i.e. pulse trains were interleaved by 25 ms). Threshold was defined in a separate
experimental run using a 20Hz pulse train using the procedure described previously in section
2.3.1. Phosphene shape was measured using methods analogous to those described above for the
tactile felt shapes in section 3.2.2.1. Within a run, each stimulus (at a particular frequency and
amplitude) was presented 5-10 times in random order amongst other test stimuli that varied in
either amplitude, frequency, or number of stimulated electrodes (single or pairs). For some trials
when stimulation was near threshold, a blank trial was reported (no phosphene was seen) and
44
these trials were excluded from the analysis. Stimulation that produced 2 or fewer repeated trials
(after excluding blank trials) were also excluded from the analysis, since this implied that the
percept elicited by stimulation was not reliably visible.
3.2.3 Analysis
Resulting binary images from each drawing trial were described using four shape descriptors:
area, major and minor axis length, and orientation. The area of each shape was obtained from
the 0th geometric moment (number of non-zero pixels in the image), while the orientation and
the lengths of major and minor axes were calculated from the eigenvalues and eigenvectors of
the centralized moments (fitting an ellipse to the shape and measuring the length of the longest
and shortest axis to this ellipse). From these, we were also able to obtain the minor to major
axes ratio, indicating degree of shape elongation. The orientation of the shape was simply the
angle of the longest eigenvector. Trial-by-trial variability across these four descriptor values was
calculatedasthestandarddeviationacrosstrials. Forarea, majorandminoraxeslength, standard
deviation was calculated as a percent, by normalizing by the mean value of that descriptor. A
large standard deviation implied that shapes varied widely on a trial-by-trial basis for a particular
descriptor, while a small standard deviation indicated that there was little variability across
trials. Variability from tactile control task was regarded as drawing error. Phosphene variability
from retinal stimulation experiment was then compared to drawing error to establish phosphene
consistency.
Tactile drawing bias was established for area, major and minor axis length descriptors, as
the ratio between the mean shape drawing of a tactile target and the actual tactile target. For
orientation, drawing bias was calculated as the difference between the mean shape drawing across
trials of the tactile target and the actual tactile shape.
A select set of paired stimulation experiments were analyzed to calculate phosphene size bias.
Pairs which consistently produce two distinct phosphenes as a result of individual electrodes being
45
activated were utilized in analysis. Criteria for consistency were: 1) Paired stimulation resulted
in 2 phosphenes for all trials, 2) All pairs (for a particular subject) were stimulated at the same
stimulation conditions (frequency, pulse duration, stimulation duration and amplitude level with
respect to threshold) and within a single month testing window, 3) Stimulation of either single
electrode from a pair resulted in a single phosphene (rather than two phosphenes) for least 60%
of trials.
The distance between each phosphene was manually measured in pixels and converted to
degrees visual angle . An example of such a measurement is shown in Figure 3.2. True electrode
pair distance was calculated from pitch of electrodes; Argus I electrode spacing = 800 μm, Argus
II electrode spacing = 525μm and converted to visual angle using the conversion that one degree
of visual angle is equal to 288μm on the retina (Drasdo and Fowler, 1974)) . Phosphene size bias
was then calculated as the ratio between phosphene distance and true electrode distance; mean
phosphene bias calculated across the set of pairs for each subject.
Subjects would occasionally see two phosphenes when a single electrode was stimulated or
more often a single phosphene when two electrodes were stimulated. Furthermore, it was not
always the case that the number of phosphenes was consistent from trial to trial. As a result,
an additional shape descriptor, number of phosphenes, was used to describe retinal stimulation
data. A large standard deviation in the number of phosphenes was another indicator of trial to
trial inconsistencies as it demonstrates that subjects see a different number of phosphenes from
trial to trial.
Subject 1 was the only subject with repeated phosphene shape drawing experiments performed
over the course of several years with a few months of breaks in between. Thus, for a set of 4
electrodes that were consistently tested over a one year period, changes in phosphene variability
were monitored.
46
Figure 3.2: Paired phosphene distance calculation. Example from subject 1 calculates the
distance between phosphenes generated from paired electrode stimulation and compares it to the
actual distance between stimulating electrodes (based on the geometry of the array)
3.3 Results
3.3.1 Control tactile drawing experiment
Visual inspections of the drawings from Subject 1 suggested that subjects may differ in drawing
capability between compact (minor axis length > 50% major axis length) and elongated shapes
(minor axis length < 50% major axis length). Thus, we subdivided our tactile (and phosphene)
data into these two shape groups. As described above, shapes were classified in terms of their
area, major and minor axes lengths, and orientation. For all shape descriptors, we calculated both
drawing bias - the ratio or difference between the tactile target and the mean shape drawings of
thattactiletarget, andvariability - thedifferences across repeateddrawing trialsfor a giventactile
target. An example of an elongated tactile shape and drawing trials from subject 4 is shown in
47
Figure 3.3. This subject overestimates the total area of the drawings to be 3.2±0.3 times larger
than the tactile shape and has a slight orientation bias of 6±2.3
o
.
Figure 3.3: Tactile drawing for diagonal rectangle shape. Example of Subject 4 tactile
drawing trials (5 repeats) shows an overestimation of size.
Drawing bias data for all subjects is shown in Figure 3.4; drawing variability data is shown
in Figure 3.5 and detailed in Table columns 2 and 7. Important results for each subject are
highlighted in the text. Drawing bias values list the average across compact and elongated tactile
shapes with standard error; variability values list the average for compact and elongated drawing
trials with standard deviation value. For tactile variability, the standard deviation value indicates
the range of drawing error across the set of tactile shapes and sets an upper bound for phosphene
variability.
Subject 1: In terms of drawing bias (figure 3.4A), both compact and elongated shapes were
drawn larger (compact = ~1.5±0.15 times larger than the tactile target; elongated =1.8±0.3)
than the actual size of the tactile shape. For compact shapes, this difference in area between the
compact shape drawings and actual tactile target was evenly distributed across major and minor
48
Figure 3.4: Drawing bias for all subjects. Comparison of shape drawings to tactile shape for
area, major axis and minor axis and orientation descriptors. Size comparison expressed as a ratio,
orientation comparison as a difference .
axes (~1.3±0.07 and 1.2±0.07 times larger respectively). For elongated shapes, the larger area
bias was due to differences in the minor axis (~1.6±0.10 times larger) rather than the major axis
(~0.9±0.07 times larger). Furthermore, the subject tended to draw elongated shapes biased by
9°±1.85° counter-clockwise. Measuring angular drawing bias for compact shapes was not possible,
since by being almost circular, they did not have a definitive major axis orientation. As far as
drawing variability was concerned (figure 3.5A), there was less area variability across trials for
compact (~17±2% error) than for elongated shapes (~34±2% error). Orientation was less variable
for elongated (~8±2° error) than for compact shapes (~22±4° error), also see Table 3.1.
49
Figure 3.5: Control task drawing variability for all subjects.
Subject 2: In terms of drawing bias (figure 3.4B), both compact and elongated shapes were
drawn smaller than the actual size of the tactile shape (compact = ~0.46±0.11 times smaller than
thetactiletarget; elongated=0.56±0.09) . For compactshapes, this difference inarea betweenthe
compact shape drawings and actual tactile target was evenly distributed across major and minor
axes (~0.63±0.05 and 0.71±0.14 times smaller respectively). For elongated shapes, the smaller
areabiaswasduetodifferencesinthemajoraxis(~0.67±0.11timessmaller)ratherthanthemajor
axis (~1.3±0.31 times larger). Furthermore, the subject tended to draw elongated shapes biased
by 16°±5.41° counter-clockwise. As far as drawing variability was concerned (figure 3.5B), area
and major axis variability were comparable for compact (area = ~20±5%, major axis = ~11±4%
error) and elongated shapes (area = ~17±9%, major axis = ~10±3% error). Minor axis was less
50
variable for compact (11±2% error) compared to elongated (28±13% error) shapes. Orientation
were less variable for elongated (~6±3° error) than for compact shapes (~23±20° error), also see
Table 3.1.
Subject 3: In terms of drawing bias (figure 3.4C), both compact and elongated shapes were
drawn larger than the actual size of the tactile shape (compact = ~1.5±0.31 times larger than
the tactile target; elongated =2.4±0.19) . For compact shapes, this difference in area between the
compact shape drawings and actual tactile target was largely due to the major axes (~1.4±0.13).
For elongated shapes, the larger area bias was primarily due to differences in the minor axis
(~1.9±0.31 times larger) rather than the major axis (~1.3±0.11times larger). Furthermore, the
subjecttendedtodrawelongatedshapesbiasedby4°±1.7°clockwise. Asfarasdrawingvariability
was concerned (figure 3.5C), variability across trials for compact (~20±8% error) and elongated
shapes (~23±3% error) were comparable. Drawings were less variable for elongated (~8±2° error)
than for compact shapes (~22±4° error), also see Table 3.1.
Subject 4: In terms of drawing bias (figure 3.4D), both compact and elongated shapes were
drawn larger (compact = ~2.5±0.36 times larger than the tactile target; elongated =2.3±0.68)
than the actual size of the tactile shape. For compact shapes and elongated shapes, this difference
in area between the compact shape drawings and actual tactile target was distributed across
major and minor axes. Furthermore, the subject tended to draw elongated shapes biased by
14°±4.3° counter-clockwise. As far as drawing variability was concerned (figure 3.5D), there was
less area variability across trials for compact (~17±2% error) than for elongated shapes (~34±2%
error). Orientation variability for compact (~9±9° error) and elongated shapes (~8±6° error) were
comparable to each other, also see Table 3.1.
51
Table 3.1: Drawing variability and phosphene variability.
52
3.3.2 Retinal stimulation experiment
3.3.2.1 Phosphene variability
Subjectswereabletodrawphosphenesfairlyconsistentlydependingonthestimulationconditions.
As an example, Figure 3.6 shows phosphene drawing trials from a consistent (electrode D2) and
inconsistent (electrode B3) stimulation condition (taken on the same day) for Subject 1. Major
and minor axis variability is much higher for electrode B3 (Major axis = 27.8%, Minor axis =
15.9%) compared to electrode D2 (Major axis = 6.8%, Minor Axis = 10.7 %). Furthermore,
variability in the number of phosphenes is higher for electrode B3 (No. Phosphenes = 0.32).
Figure 3.6: Phosphene variability example for Subject 1. Repeated phosphene drawings
for consistent (A) and inconsistent (B) example trials
Similarly, Figure 3.7 shows example phosphene drawing trials for consistent (electrode D07)
and inconsistent (electrode B09) stimulation conditions from Subject 2. Major and minor axis
variability is much higher for electrode B09 (Major axis = 47.7%, Minor axis = 38.6%) compared
53
to electrode D2 (Major axis = 11.7%, Minor Axis = 23.4 %). Qualitatively, it is apparent that for
electrode B09, the subject does not consistently see the same number of phosphenes on each trial.
Though both electrode D07 and B09 were stimulated individually, the experiment was performed
in a run that also contained paired stimulation.
Figure 3.7: Phosphene variability example for Subject 2. Repeated phosphene drawings
for consistent (A) and inconsistent (B) example trials
Phosphene variability for each subject is plotted in separate figures below (Figures 3.8-3.11).
Each subplot (A-D) shows the variability for a particular shape descriptor by subdividing experi-
ments into single, singles w/pairs, pair stimulation categories. Within each stimulation category,
54
compact and elongated phosphenes are plotted in separate blue and red bars respectively. Red
and blue solid lines indicate drawing error for compact and elongated tactile shapes; red and blue
dashed lines indicate a standard deviation above drawing error. The tactile standard deviation
suggests the range of drawing error across different tactile shapes and gives an upper bound for
drawing error.
Subject 1: Looking at Figure 3.8, compact and elongated phosphene variability (blue and red
bar graphs) is less than drawing error (red and blue solid lines) or less than a standard deviation
abovedrawingerror(redandbluedashedlines)acrossarea, majoraxis, minoraxisandorientation
with a few exceptions. Major axis compact phosphene variability for single electrodes collected
during paired stimulation (Singles w/Pairs: 17.06 ± 3.55 %) is greater than major axis compact
drawing variability (8.91± 2.66 %), though a student’s t-test did not suggest a difference between
these two values (p=0.09). Single w/Pair area variability (compact = 27.55± 4.14 %) was also
greater than tactile drawing error (17.46 ± 4.07 %), but a t-test did not suggest a difference
between these two values (p=0.07) . Specifics detailed in Table 3.1.
Subject 2: Referring to Figure 3.9, in terms of area, compact shape variability was less than
drawing error for single electrode stimulation, but greater for single w/pairs and pairs stimulation
conditions; elongated shape variability was greater than drawing error for single, single w/pairs
and pairs stimulation conditions. In terms of major axis, compact shape variability was less than
drawing error for single electrode stimulation, but greater for single w/pairs and pairs stimulation
conditions; elongated shape variability was greater than drawing error for single, single w/pairs
and pairs stimulation conditions. In terms of minor axis, compact shape variability was greater
than drawing error for single, single w/pairs and pairs stimulation conditions; elongated shape
variability was less than drawing error for single, single w/pairs and pairs stimulation conditions.
In terms of orientation, compact shape variability was less than drawing error for single, single
w/pairs and pairs stimulation conditions; elongated shape variability was greater than drawing
error for single, single w/pairs stimulation conditions and less for pairs condition. Student’s t-test
55
Figure 3.8: Phosphene variability for Subject 1. Phosphene area (A), major axis (B), minor
axis (C) and orientation (D) variability
between mean phosphene variability values and tactile drawing error for all conditions showed no
statistically significant difference between the two (p>0.05). Specifics detailed in Table 3.1.
Subject 3: Referring to Figure 3.10, in terms of area, compact and elongated shape variability
was less than drawing error for single, single w/pairs and pairs stimulation conditions; mean area
variability for single electrode compact phosphenes was statistically significantly less than drawing
error (p<0.05). In terms of major axis, compact shape variability was greater than drawing error
forsingleelectrodestimulationcondition, butlessforpairsstimulation; elongatedshapevariability
was greater than drawing error for single, single w/pairs, and pairs stimulation conditions. In
terms of minor axis, compact shape variability was less than drawing error for single and pairs
stimulationconditions; elongatedshapevariabilitywasgreaterthandrawingerrorforsingle, single
56
Figure 3.9: Phosphene variability for Subject 2. Phosphene area (A), major axis (B), minor
axis (C) and orientation (D) variability
w/pairs and pairs stimulation conditions. In terms of orientation, compact shape variability was
less than drawing error for single and pairs stimulation conditions; elongated shape variability was
greater than drawing error for single, single w/pairs and pairs stimulation conditions. Student’s t-
test between mean phosphene variability values and tactile drawing error for all conditions (except
single electrode compact phosphene area) showed no statistically significant difference between
the two (p>0.05). Specifics detailed in Table 3.1.
Subject 4: Referring to Figure 3.11, in terms of area, compact shape variability was less than
drawing error for single, single w/pairs and pairs stimulation conditions; elongated shape vari-
ability was greater than drawing error for single, single w/pairs and pairs stimulation conditions.
In terms of major axis, compact shape variability was greater than drawing error for single, pairs
57
Figure 3.10: Phosphene variability for Subject 3. Phosphene area (A), major axis (B), minor
axis (C) and orientation (D) variability
stimulation conditions; elongated shape variability was greater than drawing error for single, and
pairs stimulation conditions, but less for single w/pairs condition. In terms of minor axis, compact
shape variability was greater than drawing error for single, pairs stimulation conditions; elongated
shape variability was greater than drawing error for single, and single w/pairs stimulation condi-
tions, but less for pairs condition. In terms of orientation, compact shape variability was greater
than drawing error for single, pairs stimulation conditions (p<0.05 for all conditions); elongated
shape variability was less than drawing error for single, single w/pairs stimulation conditions and
pairs condition. Student’s t-test between mean phosphene variability values and tactile drawing
error for most conditions showed no statistically significant difference between the two (p>0.05).
Specifics detailed in Table 3.1.
58
Figure 3.11: Phosphene variability for Subject 4. Phosphene area (A), major axis (B), minor
axis (C) and orientation (D) variability
Subject 1 was tested with phosphene shape experiment extensively over 430 days of testing
and divided into three blocks of time, block 1: days 0 to 113, block 2: days 211 to 260, and
block 3: days 344 to 420. No psychophysical testing was performed with the subject in between
testing blocks. Minor axis variability from four electrodes (B3, C1, D1, D2) were tracked over
the entire period within each block. Degree of reproducibility was calculated as the percentage
of phosphene experiments per time block whose variability was below tactile drawing error. The
subject self-reported the number of hours using the device per week. Results are plotted in Figure
3.12. Subject reproducibility was higher for the first and third block of testing time compared to
the second block. In between block 1 and 2, the subject reported that he was only using the device
minimally at home. Phosphene reproducibility decreased after use of device at home decreased.
59
Figure 3.12: Subject 1 phosphene variability over time. Minor axis variability measured
for the same 4 electrodes over the course of 430 days of testing
Thevariabilityinthenumberofphosphenesfromtrial-to-trialisanotherindicatorofphosphene
stability from trial to trial. Looking at Figure 3.13, for each subject there is an increase in vari-
ability from single electrodes stimulation experiments to singles w/ pairs experiments. Subject 1
elongated shape variability increases from 0.05 ± 0.01 for Single electrode stimulation to 0.13 ±
0.01 for Singles w/Pairs; Subject 2 elongated shape variability increases from 0.30 ± 0.04 for Sin-
gle electrode stimulation to 0.48± 0.05 for Singles w/Pairs; Subject 3 elongated shape variability
increases from 0.09 ± 0.03 for Single electrode stimulation to 0.32 ± 0.04 for Singles w/Pairs;
Subject 4 elongated shape variability increases from 0.15 ± 0.02 for Single electrode stimulation
to 0.28 ± 0.07 for Singles w/Pairs. Variability in number of phosphenes for Paired stimulation
(stimulated along with single electrodes) was also higher than isolated Single electrode stimula-
tion for all subjects. For all subjects, variability for Singles w/Pair and Pairs were comparable to
60
each other. Furthermore, Subject 1 has the smallest variability compared to other subjects. As
an example, Subject 1 single electrode stimulation variability (elongated = 0.05 ± 0.01) is less
than Subject 2 (0.30 ± 0.04), Subject 3 (0.09 ± 0.03), and Subject 4 (0.15 ± 0.02). Findings are
further detailed in Table 3.1.
Figure 3.13: Variability in the number of phosphenes. Experiments divided into single
electrodes, singles stimulated with pairs and pairs for Subject 1 (A), Subject 2 (B), Subject 3 (C),
Subject 4 (D).
With a few exceptions, phosphene variability is comparable to drawing variability. These
results suggest that our subject was able to accurately and consistently report the size and orien-
tation of phosphenes elicited by retinal stimulation. Indeed, our results suggest that a major part
of the variability across drawing of phosphenes may in fact be due to drawing error rather than
variability in the elicited percept, since phosphene variability is less than a standard deviation
61
Table 3.2: Phosphene size bias. Phosphene paired stimulation distance compared to compact
and elongated tactile drawing area bias
above drawing error. Experiments in which single electrodes were stimulated along with pairs of
electrodes tended to show an increase in phosphene variability than single electrodes stimulated
without paired stimulation. Furthermore, Subject 1 had the lowest variability compared to all
other subjects for number of phosphenes drawn from trial to trial.
3.3.2.2 Phosphene size bias
Phosphene size bias was calculated as the ratio between mean paired phosphene distance (across
trials) in visual angle and actual electrode pair distance. These results are surprisingly similar to
drawing bias with control tactile shapes and are summarized in Table 3.2. Average ratio value for
Subject 1 calculated from 6 pairs; average value for Subject 2 calculated from 4 pairs; mean ratio
for Subject 3 from 6 pairs; mean ratio for Subject 4 calculated from 11 pairs. Subject 1 tended to
draw phosphenes from pairs 1.5±0.19 times further apart than actual electrode spacing, which is
similartotactileexperimentareabias(compact=1.5±0.15, elongated=1.8±0.3). Subject2drew
phosphenes closer together (ratio = 0.88±0.11) than actual electrode spacing, similar to tactile
experiment area bias (compact = 0.46±0.11, elongated = 0.56±0.3). Subject 3 overestimated
phosphene spacing by 2.9±0.5, which is similar to tactile drawing area bias (compact = 1.5±0.31,
elongated = 2.4±0.19). Subject 4 also overestimated phosphene spacing by 2.2±0.26, similar to
tactile experiment area bias (compact = 2.5±0.36, elongated = 2.3±0.68).
62
3.4 Discussion
Our reproducibility experiments had two main goals. The first goal was to determine whether
the apparent shape of phosphenes was consistent across different trials. Alongside, we explored
whether shape variability changed across many days of testing and if phosphene variability was
affected by other stimuli in the test set. The second goal was to determine whether phosphene
shape was biased to either over or underestimated percept size during the drawing task.
To date, little work has examined whether consistent phosphenes are elicited across repeated
stimulation. In a previous study by our group measuring apparent brightness for single electrode
stimulation, it was found that, for all electrodes, there was consistency in brightness ratings across
different trials in the same session (Greenwald etal., 2009). In a second study we found that
subjects could learn to correctly identify two different patterns of stimulation based on percept
appearance (Horsager etal., 2010). Optic nerve stimulation studies have reported phosphene
positional variability of ~5-10° visual angle, but did not examine the shapes of these phosphene,
or relative positions of phosphenes (Obeid etal., 2010). An acute epiretinal implant group semi-
quantitatively reported that repeated paired stimulation trials resulted in ‘similar’ phosphenes on
66% of trials (Rizzo etal., 2003). Here we directly examined the consistency of percept appearance
by asking subjects to draw phosphenes for single and paired electrode stimulation.
3.4.1 Phosphene variability
With our control task, subjects drew shapes based on tactile targets despite being blind for over
15 years and a lack of visual feedback. The ability to perform this control experiment with tactile
targetswithreasonableconsistencyandaccuracysuggeststhatourdrawingtaskprovidesareliable
measure of drawing error. Tactile drawing variability sets a lower limit to measure phosphene
variability across repeat stimulations. Phosphene variability that is less than or comparable to
63
tactile variability can be attributed to drawing error, while phosphene variability much greater
than tactile drawing variability likely suggests a change in the perceived phosphene.
Tactile drawing error differed for compact and elongated shapes. Compact shapes had, for
all subjects, less or the same variability as elongated shapes in Area, Major and Minor Axis
descriptors. Elongated shapes had less or the same variability as compact shapes in orientation
for the most part (3/4 subjects). Subject’s consistency in drawing the minor axis of elongated
shapes was less than compact shapes, since estimating the thickness of a thin, but long shape
without visual feedback is extremely difficult. Thus, with these key differences in tactile drawing
variability, we foundthatis became very important to also subdivide phosphene data and compare
variability to the corresponding compact or elongated drawing error separately.
Overall, Subject 1 phosphene variability was similar to that demonstrated for drawing tactile
shapes across orientation, area, major and minor axes. Each of Subject 2, Subject 3 and Subject 4
had phosphene variability that was higher than drawing error for one or more of area, major axis,
minor axis, and orientation shape descriptors. Statistics revealed that this was not statistically
significant. It is possible that Subject 1, who has received the most training in psychophysical ex-
periments, was able to interpret percepts elicited from electrical stimulation with greater accuracy
than lesser trained subjects who were not as accustomed to seeing phosphenes. Both a familiarity
with the experimental task and an increase in retinal stimulation within previous days or months
may decrease phosphene variability and increase reproducibility. Indeed, Subject 1 had high re-
producibility (73%) when self-reported home use of the device was high (on a daily basis). On
the other hand, when Subject 1 was tested after a 4 month break in which his self-reported home
use was minimal (once a month), phosphene reproducibility reduced to 50%. Reproducibility
increased back to previous values (82%) after at home use increased once again.
The ability to generate electrically elicited percepts which are stable in appearance is likely
consistent with electrophysiology data showing that short electrical pulses elicit ganglion cell
firing with remarkable reliability (Fried etal., 2006; Sekirnjak etal., 2008). Previous in vitro work
64
examining retinal responses to electrical stimulation has shown that, for short pulses less than 0.1
ms, every current pulse can very consistently generate a ganglion cell spike for pulse frequencies up
to250Hz–whichisnearthenaturalmaximumforretinalfiring(Friedetal., 2006;Sekirnjak etal.,
2008; Jensen and Rizzo, 2007; Ryu etal., 2009a; O’Brien etal., 2002). Similarly, our stimuli of
short duration pulses (0.45 ms) with a frequency of 120 Hz or less were likely to be stimulating
primarily ganglion cell populations, rather than bipolar cells. Electrophysiology findings were
based on relatively small conical electrodes with a length of 125 μm and a base diameter of
30 μm or disk electrodes 9-15 μm, while our findings are based on large disk electrodes, with
stimulations varying in frequency and amplitude. However, it is likely that, ganglion cell spiking
within the population of cells underneath our electrodes follows the electrical stimulus pulse train
pattern with reasonable precision across the range of temporal frequencies and amplitudes that
we used (Fried etal., 2006; Sekirnjak etal., 2008; Field and Chichilnisky, 2007; Ahuja etal., 2008).
Perceptually, this may translate to consistent phosphenes across repeat stimulation trials.
Occasionally, subjectswouldinconsistently(variedfromtrialtotrial)seetwophospheneswhen
a single electrode was stimulated, or conversely see a single phosphene when two electrodes were
stimulated. This inconsistency would create a much higher variability in area, major and minor
axis shape descriptors, thus we also measured the variability in Number of Phosphenes across
trials. We found that subjects were more consistent in performing single electrode stimulation
experiments that were interleaved with other single electrode stimulation trials compared to sin-
gle electrode experiments interleaved with paired stimulation trials, as shown by the increase in
Number of Phosphenes variability between Single and Single w/Pairs experiments. One possible
explanation for additional phosphenes on single electrode experiments, is that the retina is sensi-
tized by prior stimulation. Thus, the subject perceived a phosphene at the location of the most
recent electrode stimulus and a previous paired electrode stimulus (where threshold levels had
decreased due to a prior stimulus in that region). There has been extensive literature examining
retinal adaptation to light stimulus suggesting that the retina has slow adaptation mechanisms
65
(on the order of tens of seconds) (Baccus and Meister, 2002) partially at the bipolar-ganglion cell
synapse, and partially at the ganglion cell level (Manookin and Demb, 2006; Zaghloul etal., 2007),
but in these studies, adaptation to light stimulus tended to desensitize the retina. More likely,
subjects perception of an additional phosphene may be due to an increase in noise during the
experimental task. In the paired electrode with single electrode stimulation, the set of possible
outcomes increases from a single phosphene to two. Thus, subjects could mistakenly presume
some single electrode trials yielded two phosphenes by associating the perception of phosphene
with another. In support of this claim, paired stimulation (stimulated with singles) also increased
in variability compared to Singles stimulated in isolation.
3.4.2 Phosphene bias
Three subjects overestimated the distance between phosphenes generated with paired stimulation
and one subject underestimated the phosphene pair distance. At first, the cause for discrepancy
between electrode distance and paired phosphene distance was hypothesized to originate from cor-
ticalplasticitybasedonnumerousstudiescitingtheexistenceoflongtermcorticalchangestocom-
pensateforvisualloss. Inanimalstudiesofretinallesions, corticalcellswithinthelesionprojection
zone (LPZ) increase in receptive field size and LPZ cells respond to retinal stimulation outside of
the lesion area (Gilbert, 1998; Gilbert etal., 1990; Gilbert and Wiesel, 1992; Gilbert etal., 2009).
This compensation may be the result of lateral connection modifications within V1 (Gilbert,
1998; Palagina etal., 2009). However, other studies have shown no blood oxygen level difference
(BOLD) response in LPZ with passive stimuli (Smirnakis etal., 2005). A recent review of previous
plasticity studies speculated that analyses are based on flawed assumptions and proof for plastic-
ity after visual deprivation remains inconclusive (Wandell and Smirnakis, 2009). Task-dependent
response in LPZ cells may not be due to plasticity, but top down functional connections (Ma-
suda etal., 2008; Masuda etal., 2010). Our tactile data comparing tactile shape size to drawings
reveals that tactile size biases correspond quite well with pair distance misjudgment. The three
66
subjects that overestimate tactile size also overestimate paired phosphene distance and the single
subject that underestimates tactile size also underestimated paired phosphene distance. Perform-
ing an absolute size judgment task for blind subjects with artificially induced vision is extremely
difficultsincetheylackareferencefor depthand electricallyelicited percepts donot have a strictly
correctlocationindepth. Indeed, subjectshavequalitativelyreportedthatphosphenescanappear
as close as ‘in front of their face’ to ‘an arms length’. In general, since each individual subjects
tends to see all phosphenes at a consistent depth, relative size between phosphenes is measurable.
Thus, paired phosphene distance discrepancies do not seem to suggest cortical plasticity, rather
our findings suggest a consistent bias in their representation of depth.
67
Chapter 4
Phosphenes elicited by single electrode stimulation
4.1 Introduction
The previous chapter established that electrically elicited percepts are stable, since phosphene
variability across several shape descriptors in minimal. In this chapter, we directly quantify
percept appearance in terms of the same shape descriptors to better understand the basic factors
mediating form perception.
There are several factors which may influence the shape of phosphenes. Firstly, the retina
is comprised of cells with a wide variation of structure (Masland, 2001) and function (Field and
Chichilnisky, 2007). Large epiretinal electrodes are likely to simultaneously activates 100’s to
1000’s of ganglion cells with different receptive fields sizes, simultaneously activating ON and
OFF pathways (ibid.) and potentially driving complex IPL adaptive mechanisms (Roska etal.,
2006). Across the retina, both the density of ganglion cells and the size of each cell’s receptive
fields changes with distance to the fovea. The fovea has higher density of cells compared to the
periphery; thereceptivefieldstendtogetlargerwithdistancefromthefovea. Thus, eachelectrode
activates a different number of cells, with differing receptive field sizes depending on its location
relative to the fovea. Moreover, changes in ganglion cell density due to RP, as described earlier
could influence macular and extramacular phosphenes.
68
It is likely that epiretinal stimulation with short duration pulses primarily activates ganglion
cells, rather than bipolar cell circuitry (Greenberg, 1998; Freeman etal., 2010a; Fried etal., 2006;
Sekirnjak etal., 2006; Sekirnjak etal., 2008; Sekirnjak etal., 2009). While this could allow for
high temporal precision, epiretinal electrodes might stimulate both the ganglion cell axon initial
segment (hillock) and passing axon fibers from peripheral locations. Such axonal fiber stimulation
could lead to phosphenes that are elongated in shape and poorly localized (Fried etal., 2009;
Rizzo etal., 2003; Freeman etal., 2010a; Rattay and Resatz, 2004).
Given the myriad of factors influencing retinal stimulation, there is a need to investigate
the intricacy of electrically elicited percepts generated by a retinal prosthesis. The goal of these
experimentswastodeterminehowsingleelectrodephospheneshapepropertieschangewithrespect
to stimulus electrode location.
4.2 Methods
4.2.1 Subjects
This set of experiments was performed on Subjects 1-4. Details about each of the subjects can be
found in sections 2.2.2.1and 2.2.2.2.
4.2.2 Psychophysical methods
4.2.2.1 Retinal stimulation
Stimuli were charge-balanced, 0.45 ms/phase cathodic-first biphasic pulse trains at 20 Hz fre-
quency, and 500 ms (Argus I subject) or 250 ms (Argus II subjects) in duration (see Figure 4.1).
Pulses were charge balanced across cathodic and anodic pulses for safety reasons. Each pulse
train was presented on a single electrode. All data were recorded under photopic conditions.
69
Figure 4.1: Retinal stimulation pulse train. Experiment used a biphasic, 0.45 ms cathodic
first charge balanced 20 Hz stimulation pulse train that varied in amplitude between 1.25 - 2X
threshold.
Datawascollectedonasubsetofsingleelectrodesacrosstheentirearrayatamplitudesbetween
1.25 - 2X threshold for each subject within 2-4 testing session and a few months period. In some
cases, data on a particular single electrode for a given amplitude condition was collected more
than once across different testing dates.
Shape data was recorded for Subject 1 using the methods explained in section 2.3.2.1 and for
Subjects 2-4 using methods explained in section 2.3.2.2.
4.2.3 Analysis
Only a subset of collected retinal stimulation data was used in the analysis. For each subject,
a single amplitude level (between 1.25 - 2X threshold condition experiments) was selected. The
selected amplitude was the lowest of 1.25 - 2X threshold containing the highest number of different
single electrode experiments. If the selected set of electrodes (at a single amplitude level) con-
tained the same test electrode across multiple days, the least consistent experiment (by measuring
phosphene area and orientation variability) was eliminated from the data set. The rational was to
haveasmanyconsistentelectrodestoanalyzepersubjectatthelowestamplitudelevelpossible(to
have minimal effects of current spread). In addition, for each subject, electrodes needed to be at
the same amplitude level (with respect to threshold) so phosphene size effects could be compared.
Final data analyzed for Subject 1 contained 13 different single electrodes at 1.25X threshold; data
70
for Subject 2 contained 29 electrodes at 1.5X threshold; data for Subject 3 contained 24 electrodes
at 2X threshold; data for Subject 4 contained 29 electrodes at 2X threshold.
Resulting binary images from each drawing trial were described using four shape descriptors:
area, major and minor axis length, and orientation (calculations of shape descriptors described in
Section 3.2.3). Shape elongation was calculated as the ratio of mean major axis to mean minor
axis. Using each subject’s fundus photograph, we first estimated the location of each electrode
relative to the optic disc (x-y coordinate system) and the location of the fovea relative to the optic
disc (see Section 2.4.1). The horizontal raphe (a line marking the transition between superior and
inferior retina) was also estimated by marking a line that passes through the center of the optic
disc and the fovea. Finally, for each shape experiment, we calculated the distance from the
stimulating electrode to the approximate fovea and shortest distance to the horizontal raphe.
4.3 Results
4.3.1 Phosphene descriptions
Percepts varied for each subject and electrode, but were relatively consistent from trial-to-trial.
The subjects reported that percepts appeared light gray, white or yellow in color. Examples of
typical phosphenes drawings for each subject (single electrode shown per subject) are shown in
Figure 4.2.
In Figures 4.3-4.6 single electrode phosphene shape data is overlaid on the electrode array
for each subject. Data from the entire data set of electrodes was split into groups and overlaid
on separate arrays to prevent overlap of adjacent phosphenes. Red circles in each plot mark the
stimulating electrode for each phosphene (located underneath the red circle). Estimated fovea
marked by a blue square. Note that data for both Subject 1 and Subject 4 was resized by a factor
of ~ 2 (decrease in size) than actual drawings, such that the data could be overlayed on the array
71
Figure 4.2: Single electrode example phosphenes. The first column shows an schematic
of the subject’s array (position adjusted to match actual position in fundus photograph) with
stimulating electrode marked by a red circle. The second column represents the individual trials
(each trial shown in a different color) aligned based on their position on the reference grid. The
third column shows the average drawing across five trials, plotted in a gray-scaled image. The
fourth column shows the calculated mean shape descriptors across all trials with standard error
values. Area given in degrees squared visual angle; Major and Minor axis given in visual angle
degrees; Orientation given in degrees. Each row represents an example for a different subject.
Although absolute position varied (marked in different colors, second column), phosphene shapes
were very consistent from trial to trial.
with minimal overlap. Data shown in Figures 4.3-4.6 are the same data set used in subsequent
analyses in this chapter.
Subject 1: Percepts on individual trials were drawn as curved and straight lines, wedges, or
relatively round spots (see Figure 4.3). Uncorrected for size bias, phosphenes subtended ~8-21°
along their major axes and 2-5° along their minor axes, and ranged in area from ~11-58° squared.
Subject 2: Percepts on individual trials were drawn as curved and straight lines of varying
thickness and ovals or relatively round spots (see Figure 4.4). Phosphenes subtended ~3-17°
72
Figure 4.3: Subject 1 single electrode phosphenes. Drawings from 13 single electrodes
overlayed on electrode array. Red circles indicate stimulating electrode, blue square marks the
approximate location of the fovea
along their major axes and 1-4° along their minor axes, and ranged in area from ~1-15° squared.
Noticeably, phosphenes in column 1 (i.e. A01-E01 were all extremely long, but thin arc-like
phosphenes.
Subject 3: Percepts were always slightly curved or straight thin lines. Unlike other subjects,
none of the phosphenes were round or oval-shaped, (see Figure 4.5). Phosphenes subtended ~2-11°
along their major axes and 0.5-1° along their minor axes, and ranged in area from ~1-5° squared.
Phosphenes in column 1 and 2 (i.e. - C01, E01 and F01) were extremely elongated.
Subject 4: Percepts were ovals, wedges and curved or straight lines (see Figure 4.6). Unad-
justed for size bias, phosphenes subtended ~2-48° along their major axes and 1-9° along their
minor axes, and ranged in area from ~2-99° squared. Phosphenes in the first 3 columns (A01,
E01, C03, F01 and F02) were extremely long and thin lines. Phosphenes along the last row(F04,
F06,F07 and F08) were triangular in shape. Phosphenes closer to the fovea were qualitatively
smaller and less elongated.
73
Figure 4.4: Subject 2 single electrode phosphenes. Drawing from 29 single electrodes
overlayed on electrode array. Red circles indicate stimulating electrode, blue square marks the
approximate location of the fovea
4.3.2 Phosphene shape analysis
After qualitatively assessing single electrode percepts, we quantitatively analyzed the data in
terms of shape descriptors and stimulation location relative to the fovea and horizontal raphe.
Phosphenes were considered elongated if their major axis was more than 2 times the length of the
minor axis. Figures 4.7-4.9 plot the change in phosphene area, major axis length and elongation
(major axis/minor axis) for each electrode with respect to its absolute location relative to the
fovea and horizontal raphe. Area and major axis length were normalized to the mean value across
all electrodes (for each subject), since subjects clearly differed in overall drawing size bias. The
straight dashed lines on each plot are linear regression best-fit lines of the dataset for each subject
with goodness of fit R
2
for each subject listed in the legend. Slope values with p-value (from
student’s t-test) are displayed in cases where slopes were significantly different than zero.
74
Figure 4.5: Subject 3 single electrode phosphenes. Drawing from 24 single electrodes
overlayed on electrode array. Red circles indicate stimulating electrode, blue square marks the
approximate location of the fovea
Figure 4.6: Subject 4 single electrode phosphenes. Drawing from 29 single electrodes
overlayed on electrode array. Red circles indicate stimulating electrode, blue square marks the
approximate location of the fovea
In terms of area (Figure 4.7), there did not seem to be any relationship between phosphene
size and stimulus location relative to either the fovea or horizontal raphe. All R
2
values were near
zero.
In terms of major axis length (Figure 4.8), Subject 2 phosphenes increased in length with
shallow slope, and Subject 4 phosphenes increased in length with steep slope as stimulation
distance to the fovea increase . Both Subjects 1 and 2 phosphene length increased with shallow
slope as distance to the horizontal raphe increased.
75
Figure 4.7: Single electrode phosphene area. Normalized phosphene area as a function of
distance to the fovea (A) and the horizontal raphe (B) for all subjects.
In terms of elongation (Figure 4.9), Subject 4 phosphenes tended to be more elongated with
electrode distance from the fovea. Subjects 1 and 2 phosphenes increased in elongation with
distance to the horizontal raphe. Furthermore, 76% of phosphenes were elongated for all subjects
(92% of phosphenes for Subject 1, 69% of phosphenes for Subject 2, 100% of phosphenes for
Subject 3 and 55% of phosphenes for Subject 4).
Figure 4.10 plots phosphene orientation (between -90° and 90°) for all subjects with respect to
the stimulating electrode location relative to the horizontal raphe (inferior retina marked in blue
withnegativelocations, superior retina in yellow with positive locations). Note that the horizontal
axis refers to the orientation of the phosphene and the vertical axis refers to the location of the
stimulating electrode relative to the horizontal raphe.
In most cases, phosphenes oriented at a positive angle were located in the inferior retina (36/49
phosphenes or ~ 73 %) and phosphene oriented at a negative angle were located in the superior
retina (35/46 cases or ~ 76%).
76
Figure 4.8: Single electrode phosphenes major axis length. Normalized phosphene length
as a function of distance to the fovea (A) and the horizontal raphe (B) for all subjects.
4.4 Discussion
Our goal was to explore the shapes of single electrode phosphenes both qualitatively and quantita-
tively across all 4 subjects. We did so by visually assessing single electrode percepts for different
subjects across the array and by calculating the change in phosphene shape with stimulation
location on the retina.
4.4.1 Shape descriptions
Our findings demonstrate that single electrode phosphenes shapes are distinctive from electrode
to electrode and qualitatively reproducible. Reported shapes partially coincide with previous
literature that claiming that epiretinal elicited percepts are generally oval or round in shape
(Horsager etal., 2009; De Balthasar etal., 2008; Horsager etal., 2010; Greenwald etal., 2009;
Horsager etal., 2011). Previous findings, based on subject’s description claim that phosphenes
were ’approximately 0.5 to 2 in. in diameter at arm’s length, corresponding to roughly 1° to 3° of
visual angle’ and that the major axis of oval phosphenes was 2 to 3 times greater than the minor
axis. Here we see that phosphenes are not only oval or round, but for some subjects phosphenes
were thin lines and arcs and for others, sharply drawn wedges. Electrode current spread is likely
77
Figure 4.9: Single electrode phosphene elongation. Phosphene elongation as a function of
distance to the fovea (A) and the horizontal raphe (B) for all subjects.
both symmetric and round, and can be assumed to activate a symmetric and circular region of
the retina. Interestingly, phosphene shapes, though consistent, did not match the supposed area
of activation. Likely, a perceived phosphene is result of response from a population of neural cells
(and passing axon fibers) located directly below the stimulating electrode. Unlike the normal
visual pathway, electrical stimulation simultaneously activates cells of differing function with a
series of pulses(Field and Chichilnisky, 2007; Masland, 2001). Furthermore, electrodes across the
array will each stimulate a population of cells that differ in cell density (Curcio and Allen, 1990)
and receptive field size . Thus, the variety of distinctive of phosphenes is likely a result of these
mentioned inhomogeneities across the retina.
4.4.2 Shape analysis
Quantitative analysis of phosphene shape showed no relationship between phosphene area and
electrode position for all subjects, an increase in phosphene length and elongation with either
distance to the fovea or raphe for 3/4 subjects, and distinct change in phosphene orientation
between the inferior and superior retina.
78
Figure 4.10: Single electrode phosphene orientation. Phosphene orientation as a function
of electrode location relative to the horizontal raphe for all subjects.
4.4.2.1 Phosphene area
There are two possible factors that would influencing phosphene area: cell density and receptive
field size. Receptive field size increases with distance from the fovea (Peichl and Wässle, 1979),
however cell density tends decreases with eccentricity (Curcio and Allen, 1990). Furthermore,
photoreceptor degeneration, and subsequently inner retina cell death tends to progress from ex-
tramacular to macular regions, thereby having a greater effect on ganglion cell density in the
periphery(Humayun etal., 1999b; Santos etal., 1997; Stone etal., 1992). While electrodes in the
periphery would stimulate cells with larger receptive field sizes, there would be a decrease in the
number of cells being stimulated compared to electrodes in the fovea. These two conflicting effects
may have contributed to the lack of relationship between phosphene area and electrode location.
79
4.4.2.2 Phosphene length and elongation
All subjects tended to draw phosphenes that were elongated (major axis > twice the minor axis
in 75% of drawings). Though the electric field and current spread generated by a disk electrode is
likely symmetric, theses findings suggest that cells are not activated symmetrically. Subject 4 had
thelowestpercentageofelongatedshapes(55%)comparedtotheotherthreesubjects. ForSubject
4, phosphene elongation and major axis length tended to increase with distance from the fovea,
but not with horizontal raphe. The increase in phosphene elongation with eccentricity, suggests
that symmetric shapes are located closer fovea. These findings could suggest that activation
near the fovea creates more localized percepts than the periphery. Indeed, cell density is higher
in the macular region and contains more ganglion cell somas than the periphery (Curcio and
Allen, 1990) which would tend to have more axon fibers. For Subjects 1 and 2, major axis length
and elongation increased with distance to the horizontal raphe. These findings are consistent with
axonal stimulation since with increasing distance from the horizontal raphe, axon fibers of passage
tend to originate from more peripheral cells (Oyster, 1999). Though neither phosphene elongation
or major axis length significantly changed with either fovea location or major axis length, all
phosphenes drawn were elongated. While the shape of the phosphenes and high percentage
(100%) of elongated shapes suggests that axon fibers are being stimulated, the tendency for the
subject to draw single lines without a clear minor axis, seemed to confound the analysis.
4.4.2.3 Phosphene orientation
Combining data across all subject showed that phosphenes oriented at a positive angle were
located in the inferior retina (36/49 phosphenes or ~ 73 %) and phosphene oriented at a negative
angle were located in the superior retina (35/46 cases or ~ 76%). Based on fundus photos for
all subjects, arrays are located on the macular region. Looking at the pattern that axons fibers
traverse before bundling at the optic nerve head (shown in Figure 4.11), axons in the macular
region tend to follow arcuate paths around the fovea. Axons located in the inferior retina within
80
the macula are oriented at a positive angle and axons located in the superior retina are oriented
at a negative angle. The orientations of subjects phosphenes are strongly suggestive of axonal
stimulation and orientation wise, seem to match the ganglion cell fiber paths in the retina.
Figure 4.11: Ganglion cell axon pathways. In the macular region, nerve fiber trajectories
drastically shift in orientation from inferior to superior retina.
Taken together our findings regarding phosphene length, elongation and orientation all seem
to suggest that axonal stimulation is at least partially mediating the shape of percepts. For one
subject, the increase in elongated shapes in the periphery and the prevalence of more symmet-
ric/compact shapes near the fovea is consistent with decrease in ganglion cell soma density with
eccentricity. In the next chapter, based on the assumption of axonal stimulation, a model is cre-
ated to predict the shapes of percepts based on the anatomy of axon pathways and to postulate
the degree of interference caused by axonal stimulation for multi-electrode stimulation.
81
Chapter 5
Development of a predictive model
5.1 Introduction
One concern for epiretinal implants is that percepts might be elicited by stimulation of the both
axon fiber as well as the axon initial segment (hillock). Recently, electrophysiological studies have
determined that the site of the action potential is in a region located 0-40 um from the soma
containing a high density of sodium channels (Fried etal., 2009) and with electrical stimulation
using small electrodes, some studies have shown that cathodic stimulation produces thresholds
in the initial segment that are up to a tenth lower than the rest of the axon (Jensen etal.,
2003; Sekirnjak etal., 2008). However, calcium imaging studies with larger electrodes have shown
that passing axon fibers are activated by epiretinal stimulation and that thresholds of the axon
initial segment and the entire fiber are similar at medium length pulse durations of ~ 0.5 ms
(Behrend etal., 2011).
Based on the shape characteristics of single electrode phosphenes, it seems plausible that
axonal stimulation is mediating the shapes of percepts. Such axonal fiber stimulation could lead
to phosphenes that are elongated in shape and poorly localized. Here we present the first model
that quantitatively predicts the apparent spatial position and shape of percepts elicited by retinal
electrical stimulation in humans based on the known anatomy of the retina. This model can
82
predict the shape of percepts elicited by single electrode stimulation and the shape and relative
positions of percepts elicited by multiple electrode stimulation.
5.2 Methods
5.2.1 Subjects
This set of experiments was performed on Subjects 1-4. Details about each of the subjects can be
found in sections 2.2.2.1and 2.2.2.2.
5.2.2 Psychophysical methods
5.2.2.1 Retinal stimulation
Stimuli on each electrode were charge-balanced, 0.45 ms/phase cathodic-first biphasic pulse trains
at 20 Hz frequency, and 500 ms (Argus I subject) or 250 ms (Argus II subjects) in duration (see
Figure 5.1). Pulses were charge balanced across cathodic and anodic pulses for safety reasons.
All data were recorded under photopic conditions.
Figure 5.1: Retinal stimulation pulse train. Experiment used a biphasic, 0.45 ms cathodic
first charge balanced 20 Hz stimulation pulse train that varied in amplitude between 1.25 - 4X
threshold and was presented on single electrodes or pairs of electrodes.
Data was collected on a subset of electrode pairs that varied in spacing across the entire
array at amplitudes between 1.25 - 4X threshold. Collecting data for an electrode pair consisted
of 3 conditions: each of the two electrodes individually activated and both electrodes activated
83
asynchronously (separated by 25 ms). Synchronous paired stimulation data was also collected,
but not used in the subsequent analysis. From this, we separated the data into single electrodes
and electrode pairs. Electrode pairs were eliminated from the data set if the single electrode
phosphenes were not visible (i.e. - stimulation of either single electrode from a pair resulted in
percepts in less than 3 trials).
Data was collected within 2-4 testing sessions. While data from Argus II subjects was collected
over a few months period, data from Argus I Subject 1 was in collected in several sessions nearly
2 years apart. During this period, the position of the array changed as previously described
in Section 2.2.2.1, however the resulting change in electrode position was accounted for in the
analysis. In some cases, data on a particular single electrode for a given amplitude condition was
collected more than once across different testing dates.
After each stimulation, phosphene shape data was recorded for Subject 1 using the methods
explained in section 2.3.2.1 and for Subjects 2-4 using methods explained in section 2.3.2.2.
5.2.3 Computational modeling
A model of ganglion cell axon pathways was generated as contour lines on a 2D surface (Janso-
nius etal., 2009; Airaksinen etal., 2008) and is shown in Figure 5.2A. The model was based on a
mathematical model describing nerve fiber bundle trajectories in the human retina by Jansonuis
(Jansonius etal., 2009). As described by Jansonius, the model is only consistent within the mac-
ular region (within about 20 degrees visual angle eccentricity). In the model, each nerve fiber
trajectory/curve can be described by the equation:
whereΦ is the angle of a point on the curve at a distance r relative to the center of the optic
disc;Φ
o
is the initial angle of the curve at the distance r
o
from the center of the optic disc; b and
c are constants that change as a function ofΦ
o
. Each curve is differentiated by having a different
84
Φ
o
and b and c constants were obtained by Jansonius through fitting the parameters with data
from 55 human subjects.
The position of each stimulation electrode relative to the optic disc was mapped using each
subject’s photograph (see Figure 5.2B) in x-y coordinates of visual angle.
Figure5.2: Predictingperceptswithacomputationalmodel. Basedon (A)amathematical
model of nerve fiber trajectories in a human retina and (B) the electrode positions relative to the
optic disc (measured by looking at a subjects fundus photograph).
A predicted percept was generated for a stimulation pattern (single or paired electrodes) by
making two assumptions: 1. each electrode stimulates the neural tissue below it, 2. activating a
passing axon fiber produces a percept at the perceived location of that axon’s initial segment.
Thequalityofthemodelfitwasquantifiedasthepixel-by-pixelcorrelationbetweenthepercept
drawn by subjects with the predicted percept generated from the model.
5.3 Results
5.3.1 Model validation
Half of the single electrode percept drawings were used as a training set to optimize and validate
the position of the model. The model fits were optimized for each subject by changing the location
85
of the fovea (which in turn changed the position of the axon fiber trajectories) and calculating the
correlation between the percepts drawn by subjects with the predicted percept, averaged across
the set of electrodes. In the optimization process the only parameter that changed was the fovea
location that rotated freely about the optic disc at a constant radius. As an example Figure
5.3A demonstrates the optimization process for a single electrode. In the case where the model
placement is optimal, correlations are high (middle plot of 5.3A), while a poor model placement
causes the prediction to have a low correlation value with the phosphene drawing.
Figure 5.3: Optimizing the model. The model was rotated about the optic disc and predicted
percepts were correlated with subject drawings. (A) Subject drawing had high correlation at
an optimized model rotation and poor correlations otherwise. (B) Optimal orientation for each
subject placed the fovea (marked by colored dots) at a location that is anatomically realistic. (C)
Average model correlations dropped off as a function of rotation away from optimal orientation.
The optimized model placement for each subject is represented in Figure 5.3B by the location
of the fovea (shown as different colored dots). Anatomically, the location of the fovea in human
subjects has been approximated at an angular distance of 15.5±1.1° (temporal to the center of
86
the optic disc), and at a vertical angular distance of -1.5±0.9° (inferior to the center of the optic
disc) (Rohrschneider, 2004). In the axon model, fovea locations for each subject were very close
to this value; Subject 1 fovea located at (-15.6, -0.6); Subject 2 fovea located at (-15.5, -1.2);
Subject 3 fovea located at (-14.8, -4.7); Subject 4 fovea located at (-14.9, -4.3). Furthermore,
model correlation dropped off sharply as a function of rotation away from the optimal orientation
for both Subject 1 and Subject 3 (see Figure 5.3C). For Subject 2, model correlations dropped
off with a shallow slope as a function of rotation away from the optimal model orientation. For
Subject 4, model correlations dropped off with a shallow slope as a function of rotation away from
the optimal model orientation
for the optimal model orientation was not the orientation with the highest correlation, but
was higher than a correlations from model orientations in the superior retina.
5.3.2 Single electrode stimulation
5.3.2.1 Percepts vs. Predictions
One half of the single electrode stimulation data was used to train the model, while the other half
was used as a test set. Using test data set, the average correlation of the dataset was calculated
as a function of model orientation. For most subjects test data phosphene drawings matched
the predictions generated by the model. In Figures 5.4-5.7, phosphene percepts shown beside
predictions under conditions with stimulation of passing axon fibers or stimulation of just the
tissue under the electrode. Correlation values indicate similarity between predictions under either
condition to the phosphene data.
Subject 1 axonal stimulation predictions matched the experimental data better than no ax-
onal stimulation. Across the entire data set the average correlation of axonal stimulation model
with predictions was 0.67±0.03 which was statistically significantly larger than the correlation of
0.54±0.03 with no axonal stimulation (paired t-test of p< 0.001).
87
Figure 5.4: Subject 1 model prediction examples. Phosphene drawings (second column) vs.
predictions with axonal stimulation (third column) and no axonal stimulation (fourth column)
show that axonal predictions tend to match experimental data better than no axonal stimulation.
Subject 2 axonal stimulation predictions also matched the experimental data better than no
axonal stimulation. Across the entire data set the average correlation of axonal stimulation model
with predictions was 0.56±0.04 which was statistically significantly larger than the correlation of
0.36±0.01 with no axonal stimulation (p< 0.001).
Subject 3 axonal stimulation predictions also matched the experimental data better than
no axonal stimulation predictions. Across the entire data set the average correlation of axonal
stimulation model with predictions was 0.63±0.03 which was statistically significantly larger than
the correlation of 0.55±0.03 with no axonal stimulation (p< 0.05).
Subject 4 axonal stimulation predictions sometimes matched experimental data. Across the
entire data set the average correlation of axonal stimulation model with predictions was 0.62±0.05
which was statistically no different than the correlation with no axonal stimulation of 0.54±0.05
88
Figure5.5: Subject2modelpredictionexamples. Whilemostaxonalstimulationpredictions
matched experimental data better than non-axonal stimulation predictions (row 1 and 2), in some
cases (row 3), neither axonal nor non-axonal predictions could explain phosphene shape.
(p=0.14). This means that the axonal model did no better at predicting phosphene shape than
the assumption of no axonal stimulation.
Across all subjects (summarized in Table 5.1) the assumption that passing axons fibers were
activated during stimulation was a better predictor of phosphene shape than the no axonal stim-
ulation (average axonal stimulation correlation = 0.62 vs. no axonal stimulation correlation =
0.49).
5.3.2.2 Phosphene length
The major axis length of each phosphene was compared to the major axis length of the prediction
across the entire set of single electrodes (training and test data combined). Results shown in
Figure 5.8 indicate that for Subject 1 and Subject 3, if electrodes are located on axon fibers that
89
Figure 5.6: Subject 3 model prediction examples. Axonal stimulation predictions matched
experimental data better than non-axonal stimulation predictions (row 1 and 2) in most cases,
while in others, non-axonal predictions had higher correlations than the axon model prediction
(row 3).
predict percepts longer than 10
o
visual angle, the actual phosphene length tends to be less than
predictions. For short predictions < 10
o
visual angle, phosphene length was comparable or longer
than model prediction lengths.
5.3.3 Paired electrode stimulation
For paired electrode stimulation, when two electrodes were stimulated asynchronously, the model
was used to predict whether subjects saw one or two phosphenes. Model performance was eval-
uated in two ways. First, cross correlations between subjects reported percepts and model pre-
dictions were calculated. Next, the model was evaluated for its ability to predict the number of
individual phosphenes. For each prediction image, the number of phosphenes was calculated by
90
Figure 5.7: Subject 4 model prediction examples. Axonal stimulation predictions were not
always correlated to experimental data.
the number of distinct regions. This value was then compared to the number of distinct regions
in each drawing trial. For example, in Figure 5.9, the number of distinct regions in the top row is
2 (for both experimental data and Axonal model prediction) and in the bottom row this value is
1. While the assumption of no axon stimulation would predict that the subject sees 2 phosphenes
in both cases, the axon stimulation assumption is actually better at predicting the number of
phosphenes. In Figures 5.9-5.12, phosphene percepts are shown beside predictions under condi-
tions with stimulation of passing axon fibers or stimulation of just the tissue under the electrode.
Correlation values indicate similarity between predictions under either condition to the phosphene
data.
Subject 1 axonal stimulation predictions matched the experimental data better than no axonal
stimulation. Across the entire data set the average correlation of axonal stimulation model with
predictions was 0.68±0.02 compared to 0.54±0.02 with no axonal stimulation (paired t-test of p<
91
Table 5.1: Axonal stimulation average correlation. Average axonal stimulation correlations
for Subject 1, Subject 2 and Subject 3 tended to be significantly higher than no axon stimulation.
0.05). Additionally, the axon model was able to predict whether there would be 1 or 2 phosphenes
correctly 84% of the time compared to 44% assuming conventional electrode separation.
Subject 2 axonal stimulation predictions also matched the experimental data better than no
axonal stimulation. Across the entire data set the average correlation of axonal stimulation model
with predictions was 0.47±0.03 compared to 0.29±0.01 with no axonal stimulation (p< 0.01).
Furthermore, the axon model was able to predict whether there would be 1 or 2 phosphenes
correctly 70% of the time compared to 43% assuming conventional electrode separation.
Subject 3 axonal stimulation predictions also matched the experimental data better than no
axonal stimulation predictions. Across the entire data set the average correlation of axonal stim-
ulation model with predictions was 0.54±0.03 compared to 0.43±0.03 with no axonal stimulation
(p< 0.05). The axon model was only marginally better at predicting the number of phosphenes
at 60% compared to 58% with conventional electrode separation.
Subject 4 axonal stimulation predictions sometimes matched experimental data. Across the
entire data set the average correlation of axonal stimulation model with predictions was 0.51±0.03
compared to 0.37±0.02 with no axonal stimulation. The axon model seemed to be able to predict
whether there would be 1 or 2 phosphenes correctly 67% of the time compared to 56% assuming
conventional electrode separation.
92
Figure 5.8: Phosphene Length vs. Axon Model Length. For Subject 1 and Subject 3,
phosphene length was shorter than predictions when stimulated axon tracts were greater than 10
o
Across all subjects (summarized in Table 5.2) the assumption that passing axons fibers were
activated during stimulation was a better predictor of the number of phosphenes than no axonal
stimulation (axonal stimulation = 67% compared to 56%).
5.4 Discussion
Our goal was to develop a model based on the axonal anatomy of the retina that would predict
the shapes of single electrode phosphenes and the shapes and relative positions of paired electrode
phosphenes. With our model, we were able to predict single electrode phosphene shape with an
average correlation of 0.62 compared with 0.49 for no axon stimulation. For longer predictions,
93
Figure 5.9: Subject 1 paired electrode model prediction examples. Phosphene drawings
(secondcolumn)vs. predictionswith axonalstimulation (thirdcolumn) andno axonal stimulation
(fourth column) show that predictions with axon model are better at predicting both the shape
of phosphenes and the number of percepts.
Table 5.2: Summary of paired stimulation predictions with axon model.
results suggest that the phosphene length may be shorter than the entire axon. Our model was
able to predict the shapes of phosphenes from paired stimulation with an average correlation
of 0.55 compared to 0.41 with no axonal stimulation. Furthermore, we could predict whether
stimulating two electrodes was likely to result in two discrete phosphenes or a single percept with
an accuracy of 67%.
The high correlation of predictions with actual percept drawings seems to confirm the hy-
pothesis that axonal stimulation is at least partially mediating the shapes of these phosphenes.
94
Figure 5.10: Subject 2 paired electrode model prediction examples. Phosphene drawings
(secondcolumn)vs. predictionswith axonalstimulation (thirdcolumn) andno axonal stimulation
(fourth column) show that predictions with axon model can sometimes predict the shape and
number of percepts.
Furthermore, the results from paired stimulation suggest that axonal stimulation is likely to im-
pede the system resolution. For example, in Figure 5.9, the second row demonstrates that paired
stimulation can result in a single phosphene even when stimulating electrodes are greater than
2500μm apart. Resolution is not only limited by the current spread between adjacent electrodes,
but must also take into account the location of passing axon fibers. It is unlikely that we can
create a digital scoreboard image using the current electrodes and stimulation protocols, where
each pixel can be varied in brightness independently of others. By developing a model to cor-
rectly predict the interactions between non-adjacent electrodes, we may be able to optimize the
representation of the visual world.
95
Figure 5.11: Subject 3 paired electrode model prediction examples. Phosphene drawings
(secondcolumn)vs. predictionswith axonalstimulation (thirdcolumn) andno axonal stimulation
(fourth column) show that predictions with axon model predicts the shape of phosphenes and
occasionally predicts the number of percepts.
Figure 5.12: Subject 4 paired electrode model prediction examples. Phosphene drawings
(secondcolumn)vs. predictionswith axonalstimulation (thirdcolumn) andno axonal stimulation
(fourth column) show that axon model predictions do not always match the experimental data.
96
Chapter 6
Frequency and amplitude modulation
6.1 Introduction
To create form vision, multiple electrodes must be stimulated in concert. Present technology
uses light sensors, either in an external camera or as part of the implant, to capture an image.
Each electrode will stimulate the retina in accordance with the amount of light detected by the
corresponding area of the image sensor. The goal is to create an image much like a gray-scale
digital scoreboard, where each phosphene produced by an electrode can be thought of as a pixel
that varies in brightness. However, findings from previous chapters have shown that stimulation
of passing axon fibers are likely to impede resolution. Image quality could be partially optimized
by limiting the size of each phosphene (despite axonal stimulation) and correctly representing
relative brightness. Preferably, electrodes located on different axon tracts would be able to create
a phosphene that remains spatially distinct across a range of brightness levels.
Ideally, retinal stimulation would have the capacity to target individual ganglion cells (includ-
ing specific sub-types of the approximately 20 different types of ganglion cell (Masland, 2001)
and would be capable of producing activation patterns within these cells that match the spatio-
temporal activity of the normal retina. However such cellular resolution is far beyond the abilities
of current technology. Current chronic and semi-/chronic devices contain square electrodes of 50 x
97
50μm or quadruple square electrodes of 120 x 120μm (a semi-chronic array recently implanted by
Retina Implant AG), 220μm diameter disk electrodes (the Argus II device, Second Sight Medical
Products, Inc.) and 260-520 μm diameter disk electrodes (Argus I, Second Sight) described in
previous chapters. Thus, in all these devices, a single electrode simultaneously activates hundreds
tothousands ofcellswith awidevariation of structure (ibid.)and function (Field and Chichilnisky,
2007).
Over the last several years our group has carried out a series of experiments quantifying the
relationship between stimulation and percept for a single electrode or pair of electrodes. This work
hasshownthatperceptualthresholdsareinfluencedbytheproximityoftheelectrodestotheretina
surface, (De Balthasar etal., 2008; Mahadevappa etal., 2005) and that both threshold and bright-
ness for a single electrode or pairs of electrodes can be predicted across a variety of parameters
such as frequency, pulse duration and amplitude (Horsager etal., 2009). In particular, it has been
shown that increases in both current amplitude and stimulation frequency result in an increase
in percept brightness (Greenwald etal., 2009; Horsager etal., 2009). A cortical visual prosthesis
group has also reported an increase in percept brightness with an increase in either stimulation
frequency or current amplitude (Schmidt etal., 1996). However, these studies relied primarily
on either threshold or brightness matching judgments and did little to examine how the shape of
elicited percepts varies as a function of stimulus amplitude and frequency. A preliminary study by
our group showed that size of phosphenes increased non-uniformly with amplitude (Nanduri etal.,
2008). Previous clinical visual prosthesis studies reporting on phosphene shape were largely anec-
dotal, did not systematically repeat multiple trials for a given stimulus and did not compare
the shapes of percepts produced across a variety of pulse trains (Rizzo etal., 2003; Evans etal.,
1979; Dobelle and Mladejovsky, 1974; Brindley and Lewin, 1968b; Brindley etal., 1972).
Here we show that changes in stimulus amplitude and frequency have separable effects on
the shape of elicited percepts. As a consequence, it may be possible to develop stimulation
protocols for encoding visual images that use a combination of frequency and amplitude coding to
98
independently manipulate the size and brightness of phosphenes, thereby providing an increased
flexibility that may improve the ability of these prostheses to represent the visual world.
6.2 Methods
6.2.1 Subjects
Experiments were performed on Subject 1 (Argus I Subject) see section 2.2.2.1. Argus II Subjects
2, 3 and 4, were not available for extensive testing during the time of this study (limited shape
data was collected for a few electrodes; brightness experiments were not performed). Thus,
findings from these subjects are not included in this chapter. Experiments with Argus II Subject
4 exploring a different effect of frequency on percept shape are reported in Section 7.2.
6.2.2 System
The Argus I system (discussed in Section 2.1) was used for experiments .
6.2.3 Psychophysical methods
Inadditiontoexploringtheeffectsoffrequencyandamplitudeonphospheneshapeandbrightness,
phosphene data was also analyzed to confirm shape consistency by analysis methods given in
Section 3.2.3.
6.2.3.1 Control task - tactile drawing
Control experiment performed with tactile drawing is previously reported in Section 3.2.2.1.
6.2.3.2 Retinal stimulation
In the retinal stimulation experiments, stimuli were charge-balanced, 0.45 ms/phase cathodic-first
biphasic pulse trains that were always 500 ms in duration. Pulses were charge balanced across
99
cathodicandanodicpulsesforsafetyreasons. Eachpulsetrainwaspresentedonasingleelectrode,
and 9 electrodes were tested in total. All data were recorded under photopic conditions.
Phosphene shape and brightness were manipulated in two ways. In the Modulated Amplitude
condition we modulated current amplitude between 1.25 – 6X threshold (threshold was defined
in a separate experimental run using a 20Hz pulse train using a method of adjustment procedure
described previously (De Balthasar etal., 2008) while holding frequency constant at 20Hz. In the
Modulated Frequency condition the frequency of the pulse train was varied between 13 Hz - 120
Hz, while holding current amplitude constant at 1.25X threshold, Figure 6.1.
Figure6.1: SchematicoftheBiphasicpulsetrain. Pulsetrainswerevariedbyeitherchanging
pulse amplitude (gray dashed arrows) or pulse frequency (black arrows).
Phosphene shape (for both Modulated Amplitude and Modulated Frequency conditions) was
measured using methods analogous to those described above for the tactile felt shapes. Phosphene
brightness was measured using a brightness rating procedure where the subject compared the
brightness of the phosphene to a reference stimulus(Stevens, 1957). The subject was explicitly
instructed to rate the apparent brightness independently from the apparent size of the phosphene.
For a given electrode, the reference stimulus was the same for both Modulated Amplitude and
Modulated Frequency conditions.
Shape and brightness judgments were conducted in separate runs. Within a run each fre-
quency/amplitude was presented 5-10 times in random order amongst other test stimuli that
varied in either amplitude or frequency. For each electrode we measured responses for 10 different
pulse trains (6 frequencies at 1.25X threshold, and 5 amplitudes at 20Hz. Modulated Amplitude
100
and Modulated Frequency conditions therefore contained a single pulse train (1.25X threshold at
20Hz) that was common to both conditions. A total of 970 trials of data were collected. On a
small percentage of trials (1.33% for brightness runs, 5.38% for shape runs) when stimulation was
near threshold no phosphene was seen. On brightness runs we recorded a brightness of zero. On
shape runs these trials were excluded. In total we collected 450 trials of brightness data (since all
trials were included) and 520 trials of shape data (where 28 trials were excluded).
6.3 Results
6.3.1 Control tactile drawing experiment
Results with tactile control data are previously reported in Section 3.3.1 and summarized in Table
6.1.
6.3.2 Retinal stimulation experiment
Here, we report phosphene variability for experiments exclusively performed in this chapter and
compare our findings to tactile control variability from Chapter 3, Section 3.3.1.
Nearly all phosphenes appeared as elongated ellipses (~93%) with their minor axis length
being less than 50% of the major axis length. We therefore compared phosphene variability to
control experiment results for elongated shapes. Phosphene drawing was surprisingly consistent
compared to performance for tactile shapes.
Area: For phosphenes, the mean (across electrodes) variability in phosphene area across trials
was 16±1.4% (as compared to ~34±2% for elongated tactile shapes). This difference in trial-by-
trial variability between drawings of phosphenes and drawings of elongated tactile shapes was
significant (p<0.05, Students 2-tailed t-test).
101
Table 6.1: Phosphene shape variability for frequency and amplitude modulation ex-
periments. Tactile control drawing variability with standard error (row 1 & 2) and phosphene
data variability (row 3) across area, major and minor axes, and orientation shape descriptors.
Tactile variability was separated into two categories: compact shapes (minor axis length > 50%
major axis length) and elongated shapes (minor axis length < 50% major axis length). Phosphene
drawing variability (row 3) was equal to (orientation, major and minor axes) or significantly less
than (area) tactile drawing variability. Asterisks represent significantly better performance for
phosphenes than for elongated tactile shapes, * p<0.05, Students 2-tailed t-test).
Orientation: For phosphenes, the mean orientation variability was ~6.9±0.5°, (as compared
to ~8±2° for elongated tactile shapes), see Table 6.1, and there was no significant difference in
drawing variability between phosphenes and tactile shapes (p = 0.799, Students 2-tailed t-test).
These results suggest that our subject was able to accurately and consistently report the size
and orientation of phosphenes elicited by retinal stimulation for the amplitude and frequency
conditions used in experiments. As stated previously, our results suggest that a major part of
the variability across drawing of phosphenes may in fact be due to drawing error rather than
variability in the elicited percept.
6.3.2.1 Phosphene descriptions
Percepts on individual trials were drawn as curved and straight lines, wedges, or relatively round
spots. Phosphenes subtended ~5-24° along their major axes and 1-8° along their minor axes, and
ranged in area from ~7-150° squared. The subject reported that percepts appeared light gray,
white or yellow in color. As either the stimulation amplitude or frequency increased, the subject
reported that phosphenes were perceived as brighter and as had sharper contours. Examples of
typical individual drawings of phosphenes for three different electrodes are shown in figure 6.2.
102
Figure 6.2: Phosphene drawings from 3 different electrodes. Example phosphenes from
3 different electrodes (a single trial for each electrode) show (A) Schematic of array highlighting
the example electrodes shown in B-D (D2, C4, B3). For all three electrodes stimulation 0.45 ms
biphasic, 20 Hz pulse train for a duration of 500 ms. (B,C) For electrodes D2 and C4 the pulse
train was at 1.25X threshold. (D) For electrode B3 the pulse train was at 3X threshold.
6.3.2.2 Phosphene size and brightness
Figure 6.3 shows, for a typical electrode, D2, how phosphene shape changes with an increase in
either stimulation amplitude or frequency. The left column represents the individual trials (each
trial shown in a different color) aligned based on their position on the reference grid. The right
column shows the average drawing across five trials, plotted in a gray-scaled image. Because
there was some variability in the positioning of the drawing of phosphenes across individual trials
we aligned the drawings from each trial shown in the first column along their mean centroid.
The first row (Panels A & B) represents phosphene drawings for stimulation using the baseline
parameters of 1.25X threshold and 20 Hz, the second row (Panels C & D) shows drawings when
103
the amplitude was increased to 4X threshold, but frequency was kept constant at 20 Hz. The
third row represents phosphene drawings when stimulation frequency was increased to 80 Hz, but
amplitude was kept constant at 1.25X threshold. Comparison of the different rows demonstrates
that phosphene size increases with amplitude (comparing rows 1 and 2), but size/shape does not
increase substantially with frequency (comparing rows 1 and 3).
Figure 6.3: The effects of current amplitude and frequency for electrode D2. Row
1: phosphene drawings at baseline stimulation, Row 2: phosphene drawings after an increase in
amplitude, Row 3: phosphene drawings after an increase in frequency
104
As described above, as well as measuring the size of the elicited percepts, we also measured
theirbrightnessusingaratingprocedure. Forthiselectrode, phosphenebrightnessdidnotincrease
significantly (p=0.37, 1-tailed Students t-test) with increases in stimulation amplitude, but did
increase significantly (p<0.01, 1-tailed Students t-test) with an increase in stimulation frequency
(see Table 6.2 row 7).
Figure 6.4 shows analogous brightness (Panels A & B) and size (Panels C & D) data for
all 9 electrodes (each in a different color) for both Modulated Amplitude (Panels A & C) and
Modulated Frequency (Panels B & D) conditions. In amplitude plots, x-axes are normalized with
respect to threshold. In size plots, y-axes are normalized with respect to the apparent size of a
standard stimulus of 1.25X threshold and a frequency of 20 Hz. The straight lines on each plot
are linear regression best-fit lines of the datasets.
Brightness increased as a function of both amplitude and frequency. Panel 6.4A shows bright-
ness ratings as a function of increasing amplitude in the Modulated Amplitude condition: bright-
ness increased with amplitude, as indicated by slope of the best-fit line being significantly larger
than zero in seven out of nine electrodes (p<0.01, 1-tailed Students t-test for 7 electrodes, p>0.05
in the remaining 2). Panel B shows brightness ratings as a function of increasing frequency in the
Modulated Frequency condition: apparent brightness increased as a function of frequency for all
9 electrodes (p<0.01, 1-tailed Students t-test).
Apparent size always increased with increasing amplitude, but generally did not increase with
frequency. Panel C shows drawing size (mean area) as a function of amplitude in the Modulated
Amplitude condition: for all 9 electrodes, the size of the phosphenes increased as a function of
amplitude (p<0.01, 1-tailed Students t-test). Panel D shows size as a function of frequency in the
Modulated Frequency condition; in 6/9 electrodes, size did not vary with frequency (p>0.05; in
the remaining three electrodes slopes were significantly larger than zero p<0.01, 1-tailed Students
t-test). Although phosphene size statistically increased in three electrodes, in all cases the slopes
were relatively shallow (s=0.3-4.2 % increase/Hz). In other words, a doubling of the amplitude
105
Figure 6.4: The effects of amplitude and frequency on apparent brightness and size.
Panels A and C show brightness and apparent size as a function of normalized (relative to thresh-
old) amplitude for 9 electrodes. Panels B and D show brightness and apparent size as a function
of normalized (relative to threshold) frequency for the same 9 electrodes. Each electrode’s data
are fit with the best fit linear regression.
resulted in a 1.7-2.9 times (mean of 2.2 times) increase in size, whereas a doubling of the frequency
only resulted in a 1-1.8 times (mean of 1.2 times) increase in size (normalized by frequency). Slope
values and corresponding statistics for Figure 6.4 plots are also shown in Table 6.2.
Figure 6.5 shows data for normalized brightness and normalized size averaged across all elec-
trodes. Best fit lines are again based on a linear regression model. As for the individual electrode
data, phosphene brightness increases with either amplitude or frequency, as indicated by a best
fittingslopethatissignificantlygreaterthanzero(PanelsA&B,p<0.01, 1-tailedStudentst-test).
Percept size increases with increasing amplitude (Panel C, p<0.01, 1-tailed Students t-test) but
does not change with increasing frequency (Panel D, p>0.05, 1-tailed Students t-test). It is also
worth noting that larger increases in apparent brightness can be elicited by changes in frequency
than changes in current amplitude – doubling the frequency results in an 1.4 times increase in
106
Table 6.2: Slope values for best fit of ’Brightness and Size vs. Amplitude and Fre-
quency’ plots. P values specify whether slope is significantly greater than zero slope. The last
row is mean slopes across all 9 electrodes (* p<0.05, ** p<0.01).
apparent brightness, whereas doubling amplitude only results in an 1.2 times increase in apparent
brightness. These data are also summarized in the last row of Table 1.
A previously published study of perceptual brightness vs. amplitude used single pulses (Green-
wald etal., 2009) (instead of pulse trains as carried out here). In that study it was found that
brightness rating data were best fit with a power function with an exponent close to 0.4. Our
results were comparable to the single pulse published results within the range of amplitudes that
we tested – similarly to the single pulse data we found that brightness roughly doubles across a
five times increase in amplitude.
6.3.3 Computational modeling
These results can be qualitatively described using a simple model based on estimates of the spread
of current from a metal disk in a semi-infinite medium (based on electrophysiological spatial
threshold data) (Ahuja etal., 2008) and the Horsager et al. (Horsager etal., 2009) ‘Perceptual
Sensitivity’ model previously used by our group to predict the perceptual sensitivity of the retina
to electrical stimulation in human subjects. As described previously, this model bears a strong
resemblance to models used to describe both temporal sensitivity for normal vision and retinal
sensitivity to electrical stimulation as measured neurophysiologically (Shannon, 1989; Chander
107
Figure 6.5: Normalized size and brightness as a function of amplitude and frequency.
Averaged across 9 electrodes; data are fit using linear regression.
and Chichilnisky, 2001; Rieke, 2001; Watson, 1986), and can predict both absolute thresholds and
suprathreshold brightness matching across a wide variety of pulse trains. A model schematic is
shown in Figure 6.6 .
We began by applying a spatial attenuation function to the temporal input stimulus pulse
train to produce b
1
(r,t), a spatiotemporal stimulus profile:
where f(t) is the electrical stimulation input pattern, t is the time (in milliseconds), r is the
distance from the center of the stimulating electrode in μm and I(r) is the current attenuation
from a disc electrode. The function used to model the spatial attenuation of current is given by:
108
Figure 6.6: Model schematic. BOX 1: The time stimulus, f(t), was transformed into a spatio-
temporal representation, based on the measured electrophysiological thresholds from a disk elec-
trode. BOXES 2-5: The output of this was passed through a modified version of the ‘Perceptual
Sensitivity Model’ incorporating threshold and suprathreshold parameters. The resulting output
corresponds to a spatial brightness response, B(r).
where r is the distance from the center of the stimulating disc electrode and a is the radius
of the electrode (BOX 1). The I(r) function was obtained by inverting the relationship between
threshold versus distance from the edge of a 200μm diameter platinum disk electrode (previously
reportedin(Ahuja etal., 2008). IntheAhujapaper, thresholdsfromsalamanderretinawereshown
to increase with distance r from the stimulating electrode. We assumed that current attenuation
at distance r was inversely proportional to the increase in threshold at r.
We then passed this spatiotemporal stimulus through the ‘Perceptual Sensitivity model’. In
brief, the stimulus was convolved with a temporal low-pass filter with a one-stage gamma function
with a time constant τ
1
= 0.42 ms as its impulse response (BOX 2). We then assumed that the
system became less sensitive as a function of accumulated charge by calculating the amount of
accumulated cathodic charge over time, and convolving this accumulation with a second one-stage
gamma function with time constant τ
2
= 45.25 ms (BOX 3). The output of this convolution was
scaled (by a factor= 8.73) and subtracted from the output of the first convolution. The resulting
time course was half-rectified.
109
The previous instantiation of this model modeled the amplitude or frequency required to
reach threshold or a fixed brightness level. To do this, the half-rectified output, b
3
(r, t), was
passed through a power non-linearity ofβ=3.4 at threshold andβ=0.8 at suprathreshold. For our
purposes a continuous mapping of amplitude/frequency to brightness was required. We therefore
replaced β with a continuous function that nonlinearly rescaled b
3
(r, t), across space and time,
based on a sigmoidal function dependent on the maximum value of b
3
(r, t) (BOX 4):
The parameter values of a (asymptote) = 14, s (slope) = 3 and i (shift) = 16 were chosen to
match the observed psychophysical data. Interestingly, for parameter values that reproduced the
observed behavior, we found that the sigmoidal function had accelerating slope near threshold
and a compressive slope when amplitude values were at suprathreshold levels: properties very
similar to those demonstrated by parameter beta in the original Horsager model. These were the
only free parameters used to develop this model – all other parameters were based on those of the
original Horsager data and model.
Finally, as in the Horsager model, the output, b
4
(r, t) was convolved with a low-pass filter
described using a three-stage gamma function with time constant τ
3
= 26.25 ms (BOX 5). The
maximum value of the output from this slow integration over time was used to represent the
brightness response for each location in space, B(r). For a given stimulus, the brightness of a
phosphene was assumed to be linearly related to the maximum brightness of the B(r) plot. We
estimated the size of the phosphene by calculating the area where B(r) >Θ, whereΘ was fixed as
the maximum brightness elicited by a threshold stimulus.
Thus, most of the parameters in this model (τ
1
, τ
2
, τ
3
and ) were fixed based on previous
work or separate measurements of threshold. Only the parameters of the sigmoid function (a, s
and i) were varied to match our psychophysical results.
110
Phosphene predictions generated with this model are shown in Figure 6.7. The top row shows
the effect of increasing amplitude and the bottom row shows the effect of increasing frequency.
Note that the 1.25X threshold at 20Hz stimulus (outlined with a blue box) is common to both
the Amplitude Modulation and the Frequency Modulation conditions. The maximum value of
B(r), representing the brightness of the phosphene, is reported below each simulated phosphene.
Analogous to our psychophysical data, the model replicates the finding that increasing current
amplitude results in increases in both brightness and size, whereas increasing frequency results
in an increase in apparent brightness, but little increase in size. The model also replicates the
finding that larger increases in apparent brightness can be elicited by changes in frequency than
changes in current amplitude.
Because our model assumes uniform current spread from a disk electrode and equal sensitivity
acrosstheretina, allpredictionsareofaroundandsymmetricalpercept. Althoughoursubjectdid
occasionally draw circular percepts, the majority of his percepts resembled elongated ellipses, such
as those shown in Figure 6.3. This is probably due to unequal sensitivity across the retinal surface.
It is likely that retinal stimulation not only activates neural tissue directly below the electrode
but also passing axon fibers tracts (Rizzo etal., 2003; Rattay and Resatz, 2004; Fried etal.,
2009; Jensen etal., 2005b) . For further details see Chapter 5.
6.3.4 Comparing experimental data with modeling predictions
Figure 8 re-plots the data from Figure 6.5 so as to compare size and brightness ratings for the
same stimuli. As mentioned above, both the Modulated Amplitude and Modulated Frequency
conditions contained a stimulus at 1.25X threshold at 20Hz. This stimulus can therefore be used
as a “standard reference” – i.e. both brightness ratings and size ratings were normalized so that
this “standard reference” had a brightness rating of 10 and a size of 1. For each stimulus we
plotted the brightness rating assigned to that stimulus normalized by the “standard reference”
along the x-axis and the apparent size of that stimulus normalized by the “standard reference”
111
Figure 6.7: Predicted percepts with model. Increasing amplitude shown in top row and
increasing frequency in bottom row.
along the y-axis. The stimuli presented as part of the Amplitude Modulation condition are shown
in light gray and the stimuli presented as part of the Frequency Modulation condition are shown in
dark gray. Dotted and solid lines show model predictions for amplitude and frequency modulation
respectively. Asdescribedabove, the6-foldincreaseinamplitudetestedinourexperimentresulted
in large changes (~6 fold increase) in area but only moderate changes in brightness (~2 fold
increase). In contrast, the 6-fold frequency range tested in our experiment resulted in relatively
little (~2 fold) change in area but a larger change in brightness (~3.5 fold). Our model produced
the same qualitative behavior. Note that, despite its good performance, this should be considered
a descriptive rather than a predictive model given that it was not used to predict an independent
data set.
112
Figure 6.8: Comparing experimental findings to model. Comparing brightness rating and
apparent size psychophysical data to model predictions for the effects of changing amplitude and
frequency. Experimental data from Amplitude Modulation are shown in dark gray and data from
Frequency Modulation are in light gray. Dotted solid lines show model predictions.
6.4 Discussion
Ourmaingoalwastoseewhetheramplitudeandfrequencyhadseparableeffectsonphosphenesize
and brightness. We also confirmed that shape data from experiments performed in this chapter
were reproducible based on earlier criteria from Chapter 3.
6.4.1 Phosphene brightness
Manipulating current amplitude and frequency have different effects on phosphene size and bright-
ness. Changing current pulse amplitude effects both phosphene area and, to a lesser degree,
phosphene brightness. In contrast, changing current pulse frequency results in relatively little
change in area, but a much larger change in brightness.
Our finding that increasing either frequency or amplitude results in an increase in brightness
is consistent with the previous literature for both epiretinal prostheses (Greenwald etal., 2009;
113
Horsager etal., 2009) and cortical prostheses (Schmidt etal., 1996), as well as neurophysiological
studies examining the responses of the retina to electrical stimulation (Hensley etal., 1993; Field
and Chichilnisky, 2007).
Ganglion cell spike rates increase with contrast (Troy and Enroth-Cugell, 1993). Perceptually,
brightness is a relative measure, so something considered bright by a typical observer has high
contrast with its surroundings. Previous in vitro work examining retinal responses to electrical
stimulation has shown that, for pulses less than 0.1 ms, every current pulse can generate a spike
for pulse frequencies up to 250 Hz – which is near the natural maximum for retinal firing (Sekirn-
jak etal., 2008; Jensen and Rizzo, 2007; Ryu etal., 2009a; Fried etal., 2006; O’Brien etal., 2002).
These findings were based on relatively small conical electrodes with a length of 125 μm and a
base diameter of 30 μm or disk electrodes 9-15 μm, while our findings are based on large disk
electrodes. However, it is likely that, ganglion cell spiking within the population of cells under-
neath our electrodes follows the electrical stimulus pulse train pattern with reasonable precision
across the range of temporal frequencies that we used (Fried etal., 2006; Field and Chichilnisky,
2007; Sekirnjak etal., 2008; Ahuja etal., 2008).
We also found that increasing amplitude has limited effects on the maximum brightness. This
effect of saturation of brightness at higher amplitudes have been noted in previous experiments
by our group: for example we found in a previous study that brightness ratings elicited using
single (0.975 ms) pulse stimulation were best fit using a power function (Greenwald etal., 2009).
In our model the saturation of brightness with current amplitude was modeled by a sigmoidal
nonlinearity that assumes asymptotic responses as a function of current amplitude at higher
amplitudes. One possible explanation for this observation is that increasing amplitude increases
brightness by producing multiple spikes per pulse (Ryu etal., 2009b), which is interpreted by
the visual system as increased brightness. However, cell refractory properties will eventually
impose an upper limit on the number of spikes elicited by a single pulse and at that point further
increases in amplitude beyond that amplitude will not increase the rate of neural response. Our
114
data supports this explanation, since brightness increases with amplitude at near threshold, but
by 4X above threshold the increase in brightness asymptotes.
6.4.2 Phosphene size
Manipulating amplitude results in large changes in area but only moderate changes in brightness.
Manipulating frequency results in relatively little change in area and a larger change in brightness.
Once again, this can be easily understood based on current understanding of retinal integration of
electrical stimuli and basic electrostatics. When amplitudes are increased, supra-threshold current
will be applied to a wider area of retina(Ahuja etal., 2008).
In frequency coding, the current spread remains the same across the range of frequencies. The
electrical current dissipates virtually immediately at the end of the current pulse. Although cell
depolarization may decay at a slower rate, the use of biphasic pulses reduces summation across
pulses. Neurons lying at a distance from the electrode at which current is well below the spiking
threshold will not fire, regardless of increasing frequency. This behavior is captured by the early
part of the sigmoidal non-linearity in our model, whereby low amplitudes on the retina result in
no response. While increasing frequency does reduce perceptual threshold (Horsager etal., 2009)
this attenuation in threshold is small compared to the decrease in current with distance from the
stimulating electrode. Thus, threshold attenuation with increasing frequency is likely to result in
only a small expansion of phosphene size.
6.4.3 Caveats
Our model was able to successfully replicate our general finding that increasing frequency has
greater effects on phosphene brightness than on phosphene size, whereas increasing current am-
plitude has greater effects on phosphene size than phosphene brightness. However this model
should be considered descriptive rather than predictive given the number of free parameters and
115
the fact that it was not used to predict an independent data set. Moreover, the model is indis-
putably a major simplification of the complex retinal processes involved.
First, our model was not designed to successfully predict percept shape. The model assumed
stimulation of cells below the electrode with a uniform current spread and equal sensitivity of
the retina in all directions. As a consequence, our model predicts perfectly circular phosphenes.
In contrast, our experimental data generally found that phosphenes were elongated and angled
lines or arcs. This finding suggests activation of axon fiber tracts as well as ganglion cell bodies.
Activation of fiber tracts would be expected to produce a phosphene elongated in the direction of
the axon tract, since the percept would include the receptive field locations of all (or a proportion
of) the ganglion cells whose axon fibers lie under the electrode (Rizzo etal., 2003; Rattay and
Resatz, 2004; Fried etal., 2009; Jensen etal., 2005b).Consistent with this hypothesis, increases in
phosphene size with increasing amplitude tended to be mainly along the minor axis dimension.
Phosphene size does not increase along the major axis, since the spread in stimulation area along
an axon tract simply continues to stimulate the same collection of fiber tracts though along more
of the tract length. Along the minor axis the spread in stimulation area would recruit additional
axon tracts. Incorporating biophysical models of activation may allow us to predict shape as well
as brightness and size.
Second, these data are from a single subject, for whom the array was implanted in the macular
region with electrodes in contact with the retina. Results may vary in other prosthesis subjects
with different implant locations and threshold values.
6.4.4 Implications for future retinal prostheses
A variety of engineering constraints are likely to limit the capabilities of retinal prostheses, such
as electrode size, charge density limit and electrode-to-retina distance (De Balthasar etal., 2008).
Working within these limitations, a successful retinal prosthesis will use a stimulation paradigm
that optimizes available resolution and contrast to present the visual world to patients. The
116
work described here makes several advances towards this goal. We find that phosphene size and
brightness can to a certain degree be dissociated using amplitude or frequency coding. Based on
these findings, we believe that frequency coding may offer a reasonable solution to maintaining
highresolutionwhileproducingperceptsthatrangewidelyinbrightness. Finally, theseresultscan
be plausibly explained with a simple model based on electric field spread from a disk electrode and
a previous model describing perceptual sensitivity to retinal stimulation. Together, these findings
demonstrate further progress towards developing the optimized stimulation patterns that will be
needed to create a visual prosthesis with form perception capabilities.
117
Chapter 7
Additional Experiments
7.1 Sinusoid stimulation
7.1.1 Introduction
Over the past 15 years, electrophysiological experiments have extensively explored means to
achieve greater stimulation selectivity by changing stimulus parameters. It is thought that elec-
trical stimulation with shorter duration pulses (<0.15 ms) activates ganglion cells, as they have
very short integration time constants (Fried etal., 2006), while longer pulse durations activate
bipolar cells since they have long integration time constants (Greenberg, 1998; Jensen etal.,
2005b; Shah etal., 2006). Furthermore, higher frequency pulse trains (>10 Hz) have been shown
to selectively stimulate ganglion cells and suppress bipolar cell layer activation (Ahuja etal.,
2008; Sekirnjak etal., 2006). A recent electrophysiology study has shown that extremely long
pulse duration pulse trains (~25-50 ms/phase), expressed in the form of 10 to 20 Hz sinusoid
waveforms, can be used to preferentially stimulate bipolar cells and avoid stimulation of ganglion
cells axons (Freeman etal., 2010a).
Thus far, epiretinal stimulation perceptual findings (from previous chapters) have concurred
with electrophysiological literature for ganglion cell stimulation. In these previous chapters, stim-
ulus parameters were cathode-first biphasic pulse trains of varying frequencies and amplitudes,
118
but with pulse duration constrained to 0.45 ms. Though a pulse duration of 0.45 ms is not as
short as the 0.15 ms reported in literature, the shapes of the resulting phosphenes resembled what
would be expected with ganglion cell axon stimulation, rather than bipolar cells. Specifically,
in Chapter 3, single electrodes phosphenes appeared as ‘arcs’, ‘ovals’ or ‘lines’ oriented in along
the direction of passing ganglion axon tracts and in the case of paired stimulation, electrodes
stimulating the same axon fibers yielded a single percept instead of two separable phosphenes. In
Chapter 4, while we demonstrated the capability to maintain resolution across a dynamic range
of brightness levels, it was clear that frequency modulation did not reduce axonal stimulation and
phosphenes remained elongated. Our findings expose a severe limitation of the device in terms
of system resolution that could potentially affect form perception capabilities of the epiretinal
implants. Ideally future high resolution arrays would need to improve stimulation selectivity by
decreasing ganglion cell axon stimulation.
Though electrophysiological findings have shown promise towards attaining stimulation se-
lectivity (i.e. - long pulses could selectively stimulate bipolar cells while concurrently avoiding
ganglion cell axon activation), from a practical standpoint, long pulses are extremely inefficient.
As pulse duration increases, perceptual (and electrophysiological) threshold amplitude tends to
asymptotes, such that the overall threshold charge (proportional to amplitude multiplied by pulse
duration) increases substantially. The maximum stimulation amplitude output is driven by the
charge density limit since high charge densities could potentially damage either stimulated neural
cells (Shannon, 1992; McCreery etal., 1988; McCreery DB, 1990) or compromise the electrode
material (Brummer etal., 1983; Brummer and Turner, 1977). With an increase pulse duration,
threshold charge approaches the device safety limit and the operable range of current amplitude
levels decreases. As a result, in the clinic, epiretinal prosthesis devices currently use short pulse
duration pulse trains. Short duration pulses maximize the dynamic range of amplitude levels
and allow for pulses from an electrode to be temporally separate from an adjacent electrode (to
minimize electric field overlap).
119
While electrical stimulation animal studies with long pulses has been explored, findings have
focused on normal retina. Morphological studies in a degenerate animal models have shown that
there are significant changes in connectivity within the bipolar and amacrine cells (Jones etal.,
2003; Marc etal., 2007; Strettoi and Pignatelli, 2000). The perceptual consequences of long pulse
duration stimulation is uncertain. On the one hand, if ganglion cell axonal stimulation is avoided,
exclusive bipolar stimulation could result in more punctate phosphenes, thereby improving intra-
electrodediscriminability. Ontheotherhand,connectivitychanges(increaseddendriticsprouting)
as a result of remodeling, could cause larger less reliable phosphenes.
In this chapter we explore the perceptual experience of stimulation with long pulses through
sinusoid stimulation at different frequencies. Due to hardware constraints and safety limitations,
these experiments could only be carried out in a single subject for an extremely limited set of
parameters. Though only partial in scope, we felt that our findings were necessary and compelling
to report.
7.1.2 Methods
7.1.2.1 Subjects
Data reported in this study was collected on a single subject (Subject 1) chronically implanted
with a 16-electrode retinal prosthesis (see section 2.1). Argus II subjects, who did meet selec-
tion criteria (see section 2.2.1), were also excluded from testing because hardware limitations in
Argus II devices do not allow for alternative types of stimulus waveforms such as sinusoids or
pseudosinusoids.
7.1.2.2 Retinal stimulation
Three separate types of stimulation were used: true sinusoids, pseudo-sinusoids, and the conven-
tional biphasic pulse train (see Section 5.1). In order to stimulate with true sinusoids (Figure 7.1
A), each stimulation trial was generated and sent individually through custom-built software with
120
a single trial per run (no repeated-trials were possible). The total stimulation duration of true
sinusoid waveforms was difficult to control, since it was manually set by pressing a start and stop
button, but was in the order of 1-2 seconds. In order to match stimulation paradigms from earlier
chapters in which multiple repeats were tested in a single experimental run and stimulus duration
controlled, a pseudo-sinusoid waveform was constructed using 14 evenly spaced square-wave steps
to produce a charge- balanced stimulation that closely resembled a true sinusoid (Figure 7.1B).
Software limited the maximum number of time steps to 14 steps per stimulus period, thereby pre-
venting a smoother curve . Step length was changed to create waveforms of different frequencies
and step height adjusted to match amplitude values of a true-sinusoid. Conventional biphasic
pulse trains (Figure 7.1C) were used as a comparison. Although, all waveforms in Figure 7.1 show
a single period, stimulation was repeated for multiple cycles for a total stimulus duration of 500
ms (except for true sinusoid).
7.1.2.3 Psychophysical methods
Phosphene shape was measured using methods analogous to those described in Section 3.2.2.1 for
three separate experiments: true sinusoids, pseudosinusoids and paired stimulation experiments.
Experiment 1: In the true sinusoid experiments, we stimulated 5 single electrodes for two trials
in total with 20 Hz sinusoid waveforms at 24 μA. The same 5 electrodes were also stimulated
with a biphasic pulse train at 20 Hz with an amplitude much greater than threshold (100-200 μA)
to confirm that results were similar to shapes measured in Chapter 4. Experiment 2: Pseudo-
sinusoid experiments were conducted on 6 single electrodes. Stimulation conditions tested were
500 ms pseudo-sinusoid waveform at 20 Hz, 40 Hz and 100 Hz and a biphasic pulse train at
20 Hz. Within a run each, shape data for a given electrode was collected for 5-10 stimulus
presentations which were randomly interspersed amongst other test electrodes (with the same
stimulus conditions). Experiment 3: Electrodes selected for this experiment were carefully chosen
aspairsthatthatarelocatedonthesameaxonpathandlikelytoproducedasinglephosphenewith
121
Figure 7.1: Schematic of types of waveforms for experiments. (A) True sinusoids. (B)
Pseudosinusoids. (C) Conventional biphasic pulse train. (D) Paired stimulation with electrode
1 activated with a pseudosinusoid waveform and electrode 2 with a biphasic pulse train (top) or
with electrode 1 and 2 as asynchronously activated with biphasic pulse trains (bottom).
conventional biphasic pulse train stimulation (as in Chapter 5). A complete experiment consisted
of 4 stimulus conditions, electrode 1 individually activated with a 20 Hz pseudo-sinusoid waveform
(500 ms duration), electrode 2 individually activated with a 20 Hz biphasic pulse train, electrode 1
stimulated with a 20 Hz pseudo-sinusoid stimulation while electrode 2 is activated with a biphasic
pulse train, and both electrodes activated with asynchronously (separated by 25 ms) with biphasic
pulse trains. Stimulus for paired stimulation was interleaved such that the cathodic phase of the
sine wave matched the cathodic pulse in the biphasic waveform (see Figure 7.1D), so there was an
overlap in the ON phase of pulses from different electrodes. We collected shape data for paired
electrode stimulation for 2 pairs of electrodes.
122
7.1.2.4 Analysis
Findings from the sinusoid experiment were analyzed by qualitatively comparing shape draw-
ings from sinusoid stimulation to biphasic pulse train stimulation. Findings from the pseudo-
sinusoid experiments were analyzed by comparing axon model predictions (using methods similar
to Chapter 5) to shape drawings. For paired experiments we qualitatively assessed whether 1 or
2 phosphenes resulted when two electrodes were activated.
7.1.3 Results
7.1.3.1 Qualitative assessments
Theseweresomeofthefirstexperimentsmeasuringvisualexperienceofepiretinalstimulationwith
extremely long pulse widths. The subject found the drawing task more difficult than conventional
biphasic stimulation. Repeated trials of the same stimulus not only yielded different shape results
(see Section 7.1.3.2), but the subject often reported different levels of brightness from trial-to-trial.
At times the repeat of a stimulus amplitude that was initially too dim to draw a shape would
appear extremely bright. As a result, trying to brightness match paired electrode amplitudes
when two electrodes were stimulated at the same time (see Section 7.1.3.4) proved to be very
difficult. Thus, experiments were carried out on a trial basis and relevant data was reported.
7.1.3.2 Experiment 1: True sinusoids
Percepts resulting from 20 Hz sinusoid stimulation were drawn as a single round spot or several
separate and distinct round phosphenes. Resulting phosphenes for all 5 electrodes are shown in
Figure 7.2 (each electrode in a different row). The first column labels the stimulating electrode
on the array with axonal prediction; the second and third columns show subjects drawing for
two trials of the same stimulus; the fourth column shows a shape drawing from single trial with
a biphasic pulse train stimulus. Biphasic stimulus phosphenes were arcs or long lines similar
123
to shapes shown in Chapter 4 and resembled predictions with axon model. Phosphenes from
sinusoid stimulation were more punctate rather than elongated, but tended to result in multiple
phosphenes for single electrode stimulation. The order of stimulation follows the rows in the
figures, so it seems that the subject tended to draw more phosphenes for sinusoid stimulation
with successive stimulations.
7.1.3.3 Experiment 2: Pseudosinusoids
The subject was asked to draw the shape of the phosphene when presented with a pseudo-sinusoid
stimulus with a single period resembling Figure 7.1B (ranging in frequency from 20 Hz to 100 Hz)
or a biphasic pulse train (at 20 Hz). Results are shown in Figure 7.3 with a different electrode
in each row. In the first column, we show the stimulating electrode with a predicted phosphene
based on the axon model from Chapter 5. In the second through fourth columns we show the
average phosphene drawn (across 5-10 repeat trials) for 20, 40 and 100 Hz pseudo-sinusoid stimu-
lation respectively. At times, for the 20 Hz stimulation, many of the repeat trials were not visible.
Unfortunately, the stimulus amplitude could not be increased past the charge density limit. With
20 Hz pseudo-sinusoid stimulation, phosphenes did not resemble axonal stimulation predictions.
Pseudosinusoids 40 Hz and 100 Hz more closely resembled axon stimulation predictions and con-
ventional biphasic stimulation percepts.
7.1.3.4 Experiment 3: Paired stimulation
Paired stimulation parameters needed to be adjusted during the experiments so that phosphene
brightness from each electrode was comparable. The electrode stimulated with a pseudo-sinusoid
stimulus was chosen as the electrode located closer to the optic disc (when looking at the subject’s
fundus photograph). Results are shown in Figure 7.4. Experiment pairs are shown in the first
column; phosphenes from single electrode activation (pseudo-sinusoid and biphasic) shown in col-
umn two; phosphene spatial summation with pseudo-sinusoid shown in column three; phosphene
124
summation with both electrodes activated with biphasic pulse trains is shown in the last column.
On most trials, phosphenes generated with pseudo-sinusoid/biphasic paired stimulation consisted
of two distinct spatially separated phosphenes, resembling the sum of the individual electrode
percepts. On most trials with both electrodes activated with a biphasic pulse train produced a
single percept resembling the axon pathways located underneath the electrodes.
7.1.4 Discussion
We found that the percepts resulting from stimulation with sinusoids and pseudo-sinusoids were
remarkably different from stimulations with biphasic pulse trains. Though experiments were pre-
liminary in nature, there were three main findings. Firstly, stimulation with sinusoids in which the
total stimulus duration was uncontrolled resulted in multiple round phosphenes. Secondly, stim-
ulation with 20 Hz pseudo-sinusoids, with a controlled stimulus duration of 500 ms, resulted in a
single phosphene that was more round than the percept resulting from biphasic pulse train stimu-
lation. Furthermore, qualitatively, unlike biphasic pulse trains, 20 Hz pseudo-sinusoid phosphenes
did not match the axon model predictions. Lastly, for electrode pairs located on the same axon
pathway, wewereabletousea20Hzpseudo-sinusoidstimulationtogeneratetwospatiallydistinct
phosphenes, while with conventional biphasic stimulation, these same electrode pairs resulted in
a single phosphene.
7.1.4.1 Single electrode stimulation
Single electrode phosphenes generated with a 20 Hz Pseudo-sinusoid waveform stimulation re-
sulted in a single punctate phosphene. These results seem to closely match neurophysiology liter-
ature exploring stimulation with sinusoid waveforms and long pulse stimulations (Freeman etal.,
2010a; Greenberg, 1998; Shah etal., 2006; Jensen etal., 2003). Electrophysiological studies have
suggested that ganglion cells integrate current with a short time constant duration < 0.15 ms
(Lipton and Tauck, 1987; Fried etal., 2006). In animal studies, long latency ganglion cell spiking
125
is presumed to originate from presynaptic bipolar cell responses occurring > 8-60 ms after the
initial electrical stimulation. Furthermore, time constants of these bipolar cells have ranged be-
tween 20-26 ms (Jensen etal., 2005a; Fried etal., 2006) depending on electrode size. The drastic
change in phosphene shape between stimulation with low frequency sinusoid waveforms and the
conventional biphasic pulse trains suggests that low frequency sinusoids activate the presynaptic
cells of the retina (consisting of bipolar and amacrine cells), while avoiding the stimulation of
passing ganglion cell axon fibers. Thus, 20 Hz sinusoids having a cathodic phase of 25 ms tend to
generate phosphenes that are punctate instead of being elongated. Furthermore, as frequency was
increased to 40 Hz (with a cathodic phase of 12.5 ms) phosphenes closely resembled conventional
axonal stimulation phosphenes suggesting that pulse durations of 12.5 ms still activate ganglion
cells and passing axon fibers.
Single electrode stimulation with a 20 Hz true sinusoid waveform (uncontrolled total stim-
ulus duration) resulted in multiple punctate round phosphenes (that increased in number with
successive trials). Assuming that presynaptic activation is the primary determinant of these
phosphene shape, these results may be indicative of more complex mechanisms. Indeed, amacrine
cell are known to be involve in lateral inhibition and can cause wider field inhibition (>600 μm)
(Roska etal., 2006). It is also possible that eye movements during the longer stimulation resulted
in the perception of multiple phosphenes.
7.1.4.2 Paired electrode stimulation
Paired electrode stimulation experiment in which the electrode closest to the optic disc was stim-
ulated using a pseudo-sinusoid and the electrode farthest from the optic disc was stimulated with
a biphasic pulse train was compared to a conventional stimulation paradigm with both electrodes
stimulated with biphasic pulse trains. The pair containing a pseudo-sinusoid stimulating electrode
yielded 2 separate and distinct phosphenes (for two different pairs of electrodes), while biphasic
stimulation yielded a single phosphene. These findings suggest that pseudo-sinusoid stimulation,
126
that likely activates presynaptic cells, is able to avoid axonal stimulation and better the resolution
of an epiretinal prosthesis device.
7.1.4.3 Limitations
Our findings had several limitations. Firstly, these data are from a single subject, for whom the
array was implanted in the macular region with electrodes in contact with the retina. Results
mayvaryinotherprosthesissubjectswithdifferentimplantlocations, thresholdvaluesanddisease
progression in retina. Secondly, in order to yield round phosphenes, pseudo-sinusoid frequency
needed to be as low as 20 Hz, but the amplitude required for perceptual thresholds were at or
beyond the safe charge density limit. This is likely, because thresholds for longer pulses are
extremely inefficient. Interestingly, the subject also reported that a slight increase in amplitude
(andsometimesnoincreaseinamplitude)resultedinalargeincreaseinbrightness, suggestingthat
the mechanisms driving perceptual brightness with pseudo-sinusoid stimulation is very different
than with biphasic pulse trains. Thus, pseudosinusoid stimulation could limit the dynamic range
of bright levels and furthermore could compromise conventional amplitude-brightness mapping
strategies.
7.2 High frequency stimulation
7.2.1 Introduction
In order to confirm results from Chapter 6 with other subjects, preliminary experiments were
conducted on single electrodes using biphasic pulse trains (cathodic first pulse duration = 0.45
ms) at frequencies ranging from 6 -120 Hz (constant amplitude) on Argus II subjects . For Subject
4 results were strikingly different. At frequencies less than 40 Hz (6 Hz and 20 Hz), phosphenes
were comparable to results from previous chapters that seemed to indicate stimulation of passing
axon fibers temporal from the site of stimulation. However, at frequencies greater than 40 Hz
127
a unique effect was observed. At 60 Hz and higher phosphene length more than double from
the length observed at 20 Hz stimulation. Furthermore percepts resembled the combination axon
fibers pathways passing under the electrode in both directions. Based on these preliminary results,
the effects of high frequency stimulation were investigated further for large subset of electrodes
across the entire array for a single subject (Subject 4).
7.2.2 Methods
7.2.2.1 Subjects
Data reported in this study was collected on a Subject 4 chronically implanted with a 60-electrode
retinal prosthesis (see section 2.1). For further information about this subject see section 2.2.2.2.
Other subjects did not exhibit the same results for high frequency stimulation (see Chapter 6).
7.2.2.2 Retinal stimulation
In the retinal stimulation experiments, stimuli were charge-balanced, 0.45 ms/phase cathodic-first
biphasic 60 Hz pulse trains that were always 250 ms in duration. Pulses were charge balanced
across cathodic and anodic pulses for safety reasons. Stimulation amplitude was 1.25 X threshold
(threshold was defined in a separate experimental run using a 20Hz pulse train using a method
of adjustment procedure described previously (De Balthasar etal., 2008)). Each pulse train was
presented on a single electrode, and 15 electrodes were tested in total. All electrodes chosen
were ones that had already been tested at 20 Hz stimulation in 4. All data were recorded under
photopic conditions.
7.2.2.3 Psychophysical methods
Phosphene shape for was measured using methods analogous to those described in Section 3.2.2.1.
Each stimulation was repeated for 5 trials randomized amongst other stimuli.
128
7.2.2.4 Analysis
For each electrode, axon model predictions (from optimized fiber pathways generated in Chapter
5) were generated for two different conditions: 1. Assuming a unidirectional conventional ax-
onal stimulation pattern (temporal and away from the optic disc) , 2. Assuming a bidirectional
activation of axon patterns in both directions from the point of stimulation. Predictions were
then compared to phosphenes generated at 60 Hz and to phosphenes collected earlier in 4 at the
conventional 20 Hz stimulation.
7.2.3 Results
At60Hzstimulation, thesubjecttendedtodrawphosphenesthatwerecurvedarcs, smalldotsand
lines. Based on the axon model from Chapter 5, phosphenes were divided into 3 groups based on
thelocationofthestimulatingelectrode: inferiorelectrodes, superiorphosphenesandfoveal/raphe
electrodes. Figures 7.5, 7.6 and 7.7 plot average phosphene drawing results for inferior, superior
and foveal electrodes. The first column of the figure shows the schematic of the electrode array
with the stimulating electrode marked in red; the second column plots the phosphene drawings
at 60 Hz stimulation (biphasic pulse train 0.45 ms/cathodic phase); the third column plots the
phosphene predictions with an altered axon model that assumes bidirectional activation of the
axon pathways; the fourth column plots the phosphene drawings at 20 Hz stimulation (biphasic
pulse train 0.45 ms/cathodic phase, data collected originally in experiments of chapter 4); the
last column plots the phosphene predictions with the original axon model assuming unidirectional
activation of the axon pathways.
Using 60 Hz biphasic pulse train stimulation, electrodes that were clearly in the inferior or
superior retina produced phosphene arcs that resembled the entire axon pathway patterns passing
beneath the stimulating electrode (see Figures 7.5and 7.6). Furthermore, phosphenes at 60 Hz
129
stimulation did not (column 2) did resemble phosphenes at 20 Hz stimulation (column 4) or axon
prediction phosphenes assuming unidirectional activation.
However, 60Hzstimulationwithelectrodesthatwerenearthefoveaorhorizontalraphetended
to generate phosphenes that were even smaller in size than 20 Hz stimulation. Phosphenes were
smalldotsthatdidnotresemblepredictionswithaxonmodelassumingbidirectionalaxonpatterns
of activation.
7.2.4 Discussion
Phosphenes drawn with 60 Hz stimulation were strikingly different than phosphenes generated
with 20 Hz stimulation. In fact, at 60 Hz, phosphenes resembled axon model predictions generated
with bidirectional activation of axon tracts.
In the conventional axon model proposal, we assume that activation of a passing axon fiber
produces a percept at the perceived location of that axon’s initial segment. This scheme would
only generate phosphene streaks that look like the same shape of axon fiber pathways temporal to
the location of the electrode. Thus, phosphenes that resemble the entire axon pathway suggests
that fibers towards the optic disc are also being activated, though they do not pass under the
electrode.
One possible mechanism for activating fibers towards the optic disc is ephaptic coupling due
to action potentials from adjacent axons within the same bundle. Indeed, there have been some
limited studies that have mentioned ephaptic coupling in unmyelinated axon bundles within the
cortex (Anastassiou etal., 2011). The hypothesis for ephaptic coupling proposes that an action
potential from one fiber will create an eflux in concentration of K+ ions in the extracellular space,
that will then in turn cause a membrane potential difference in an adjacent fiber that is able to
elicit an action potential (Jefferys, 1995).
If we are to assume that ephaptic coupling can account for the phosphene shapes observed at
60 Hz, two potential questions are raised. Firstly, ’why does the effect only take place at higher
130
frequencies?’ and secondly, ’why do we only see this effect in Subject 4, but not others?’. It is
possible that the effect is only observed at high frequencies because the high extracellular K+
concentration is maintained with successive passing action potentials. However, at low frequencies
<20 Hz, extracellular K+ concentration diminishes rapidly and does not cause a potential differ-
ence across adjacent cell membranes. The observation of this effect in a single subject, but not
others may be due to a difference in presynaptic activation. Subject 4 is reported to have residual
vision that is better than other subjects. Indeed, it is possible that this subject has more remain-
ing photoreceptor cells in the macular region, since RP first affects the peripheral photoreceptors.
In this subject, hypothetically ganglion cells are not only activated by electrical stimulation, but
through the conventional network of photoreceptor to horizontal to bipolar/amacrine to ganglion
cells. Thus ganglion cells have a higher probability of being activated in comparison to other
subjects where presynaptic input is minimal.
ArclikeeffectssimilartothesehavebeenobservedwithvisualstimuliexperimentsbyMoreland
in 1960’s and described by Purkyne in 1825 as ’blue arcs of the retina’(Moreland, 1968; Moreland,
1969b; Moreland, 1969a). Unfortunately, exploring the exact cause of these phosphenes is highly
speculative and to confirm mechanisms through clinical experiments is difficult to say the least.
131
Figure 7.2: Phosphene drawings with true sinusoid stimulation. Stimulating electrode
shown in column 1; sinusoid stimulation phosphene trials in columns 2 and 3; biphasic stimulation
phosphenes in column 4.
132
Figure 7.3: Phosphene drawings with pseudosinusoid stimulation. Stimulating electrode
shown in column 1; average phosphene drawings for 20, 40 and 100 Hz pseudosinusoid stimulation
shown in columns 2-4; average phosphene drawings for 20 Hz biphasic pulse train in column 5.
133
Figure 7.4: Paired pseudosinusoid stimulation. Phosphene drawings for two different pairs
of electrodes (A & B). On electrode array schematic, pseudosinusoid stimulated electrode 1 is
indicated with a red circle, biphasic electrode 2 with a blue circle; (i) electrode 1 pseudosinusoid
stimulation phosphene drawing; (ii) electrode 2 biphasic stimulation phosphene drawing; (iii)
electrode 1 stimulated with pseudosinusoid and electrode 2 stimulated with biphasic pulse train;
(iv) electrode 1 and electrode 2 stimulated with biphasic pulse trains.
134
Figure 7.5: Inferior retina phosphenes. Phosphenes produced from electrodes located in the
inferior retina stimulated with high frequency biphasic pulse trains.
135
Figure 7.6: Superior retina phosphenes. Phosphenes produced from electrodes located in the
superior retina stimulated with high frequency biphasic pulse trains.
136
Figure 7.7: Foveal retina phosphenes. Phosphenes produced from electrodes located in near
the fovea or horizontal raphe stimulated with high frequency biphasic pulse trains.
137
Chapter 8
Conclusions
Experimental therapies can hope to restore sight to blind people with photoreceptor diseases such
as AMD and RP with differing degrees of success. Optimally, visual function would be completely
restored to its normal functioning state by repairing or replacing photoreceptors. An equally ideal
outcome would be to develop a therapy that interfaces with the visual system such that the final
output response, the spatio-temporal pattern of ganglion cell spikes, are identical to normal vision,
though the intermediary stage functioning could be rather different. Unfortunately, current gene
therapy solutions are specific to a single gene mutation and thus cannot treat a broad spectrum
of retinal disease. Additionally, stem cells, optogenetics and single cell electrical stimulation
approaches face technical hurdles and are still in developing stages.
Instead, the solution closest to clinical implementation, a retinal prosthesis with relatively
large electrodes, is an intermediary solution that offers partial visual restoration. While elicited
artificial vision is imperfect, the success of these devices will be determined by its ability to
generate useful vision in blind patients. For example, the cochlear implant, another implantable
neurostimulation device, is considered a success, not because implant patients hear exactly the
samesoundsaswithnormalhearing, butbecausetheyareabletousethesedevicesintheeveryday
sense to engage in similar hearing activities as normal hearing individuals. Previous studies have
shown that the visual experience of these subjects is not like a digital scoreboard image, with each
138
electrode acting as an independent pixel (Horsager, 2009). However, if a visual prosthesis device
can create images that contain predictable spatial and contrast information, patients may be able
to use this along with other contextual sensory information to understand the visual world. This
thesis characterizes the spatial properties of single and paired electrode phosphenes as a function
various stimulation parameters and stimulus location. Firstly, findings show that, within a limited
set of possible responses that change with stimulus parameters, percept size and brightness are
predicatble. Secondly, spatial resolution is limited by current spread and stimulation of axon
fibers. Lastly, the perceptual experience from particular stimulation parameters differs across
subjects. These key findings have implications for the development of future sight restoration
technologies and are detailed below.
8.1 Controlling spatial properties of percepts
One critical factor determining the quality of the vision produced by a prosthesis is its ability to
control the appearance of its percepts. The first step in generating predictable perception is a
stable image, meaning percepts from individual electrodes should be consistent and reliable. In
Chapter 3, we show that stimulation with short duration pulses (0.45 ms) at a variety of frequen-
cies with single electrode and paired electrodes produce phosphenes that are repeatable across
different shape descriptors such as size, orientation and axes length. Indeed, electrophysiological
literature also supports the reliability of retinal responses with short duration pulse trains (<250
Hz) that are likely primarily stimulating ganglion cells rather than presynaptic cells (Fried etal.,
2006). On the other hand, in Chapter 7, when the stimulus was drastically changed to a low fre-
quency sinusoid waveform, the subject occasionally saw multiple percepts that were less reliable
across repeated trials. If we are to assume that the low frequency sinusoid phosphenes correspond
to the stimulation of bipolar cells, as suggested in electrophysiology literature (Freeman etal.,
2010a), then these findings seem to demonstrate that bipolar and amacrine cell stimulation might
139
produce less reliable perceptual responses. Considering morphological studies have shown sig-
nificant remodeling of the bipolar and amacrine cell layers in a degenerated retina (Jones and
Marc, 2005), primary activation of presynaptic cells rather than ganglion cells may lead to more
complex percepts, which would be undesirable when multiple electrodes are activated simultane-
ously. Furthermore, although in Chapter 7 single electrode pseudosinusoid stimulation did seem
to reliably produce a single phosphene rather than multiple percepts, the subject qualitatively re-
ported striking differences in brightness from trial to trial (for a given amplitude level the percept
was not visible on some trials and for other trials the percept was reported as extremely bright).
Thus, short duration pulse trains still seem the most feasible in generating consistent percepts.
By demonstrating that the basic building blocks of form vision can be consistent, our findings
show promise that more complex stimulation patterns will also be reliable,
8.2 Factors affecting resolution
Ideally, a retinal prosthesis would be able to control the size and brightness of individual percepts
with each electrode/pixel independent of one another to maximize resolution and contrast. Our
findings have shown that though we have some control over spatial and brightness properties of
phosphenes by changing stimulation parameters, the system resolution may be limited by some
factors.
In Chapters 4 we show that the phosphenes are generally elongated shapes whose orientation
changes depending on the retinal location of the stimulating electrode. In Chapter 5, we demon-
strate that the shapes of these single electrode phosphenes are likely controlled by the stimulation
of axon fiber tracts. Furthermore, in 2/4 subjects, when pairs of electrodes on the same fiber tract
are stimulated, they do not produce two distinct phosphenes. Unfortunately, the activation of
passing axon fibers produces percepts that are elongated and in some cases poorly localized. As
we move towards higher resolution arrays, axonal stimulation will greatly hinder the resolution
140
of the system. Unless stimulation of axons can be avoided, subjects would be unlikely to resolve
very fine detail, low contrast images, such as facial features. Thus, the benefits of very high reso-
lution arrays will only be fully achievable with an increase in stimulation selectivity. There have
been several solutions proposed to limit axonal stimulation. Firstly, electrophysiology studies
have found that the site of the action potential and low threshold is at the initial segment of the
axon. Conceivably, if a stimulation amplitude was low enough, one would be able to stimulate
the initial segment at above threshold, while being below threshold of passing axon fibers. The
problem with this proposal is that electrical stimulation of groups of cells is not that precise and
perceptual threshold is not likely due to the activation of a single cell, but the integration across a
population of cells. A second solution is to design an array containing long electrodes places along
the fiber tract and activating them as a dipole. This solution is clearly not feasible considering
nerve fibers are not straight and take an arcuate path towards the optic disc. Furthermore, this
solution relies upon the surgeon’s ability to exactly place the array in line with the nerve fibers.
Another solution would be to preferentially activate the IPL rather than ganglion cells by using
long duration pulses. In Chapter 7, we use low frequency pseudosinusoid stimulation to create
smaller phosphenes (in a single subject) that likely avoid the stimulation of axon fibers. Unfor-
tunately, threshold amplitudes for low frequency sinusoids are just at the maximum safe charge
density which makes this type of stimulation less practical, unless new electrode materials with
higher charge density limits can be used. Furthermore with suprathreshold stimulation above
maximum charge density, an improvement in resolution is accompanied by a sharp increase in the
nonlinearity relating stimulation intensity to contrast.
Another factor affecting system resolution is the stimulus amplitude. In Chapter 6, we show
that coding brightness with amplitude results in larger phosphenes that may reduce the resolution
of the system. Conversely, coding brightness with frequency, we can limit the size of phosphenes,
while having a larger dynamic range of brightness levels than amplitude coding. However, the
141
control over phosphene size is limited to the minor axes and phosphenes remain elongated due to
axonal stimulation.
To span a greater field of view, future electrode arrays would need to be larger. Based on our
findingswithaxonalstimulation,largerarrayswouldcreateelongated,poorlylocalizedphosphenes
in the periphery that are likely to distract the subject from more relevant visual information.
Furthermore, in the macular region, axon fibers meet the horizontal raphe at a short distance and
tend to produce less elongated phosphenes. It is possible that electrodes in the macular region are
providing better form vision information, while peripheral phosphenes are complicating the final
image. Future experiments should attempt to omit electrodes in the periphery to see whether
performance improves or stays the same. If so, it might be much more important to focus on
decreasing electrode size and spacing to increase the system resolution rather than increasing the
size of the array.
8.3 Subject variability
One main advantage of a retinal prosthesis is that the therapy is not limited to a specific gene
mutationofRPandthatelectricalstimulationofretinalneuralcellswillcausesimilareffectsacross
all subjects. While this is true for the most part, findings in this thesis seem to suggest that form
visiondoesvaryacrossdifferentimplantsubjects. InChapter4,singleelectrodephosphenesclearly
varied across each of the subjects. Subject 3 phosphenes were always very narrow and elongated
and Subject 4 tended to draw compact wedge shapes. Furthermore, Chapter 6 demonstrated that
frequency coding had little effect on the size and shape of phosphenes for Subject 1, but Chapter
7 showed very different effects of high frequency stimulation on Subject 4. Results from this thesis
suggest that subject’s perceptual experiences may differ. Hence, it may be important to maintain
the flexibility in the types of stimulation possible when designing implants.
142
Overall, this thesis has improved our understanding of the basic building blocks of pros-
thetic form vision through an understanding of the factors mediating single and paired electrode
phosphenes. Using this information, we can now make some inferences about the needs of future
visual prosthesis designs including the flexibility to change stimulation parameters and the ability
to alter the design of the electrode array, both in terms of electrode positioning and electrode
material. With these future directions, we may be able to improve stimulation selectivity and
realize the benefits of high resolution prosthetic vision.
143
Bibliography
Ahuja, A. K. etal., ‘Blind subjects implanted with the Argus II retinal prosthesis are able to
improve performance in a spatial-motor task.’, Br J Ophthalmol, 95:4 (Apr 2011), pp.539–
543.
Ahuja, Ashish K etal., ‘An in vitro model of a retinal prosthesis.’, IEEE Trans Biomed Eng, 55:6
(Jun 2008), pp.1744–1753.
Airaksinen, P. Juhani, Doro, Stephen and Veijola, Jukka, ‘Conformal geometry of the retinal
nerve fiber layer.’, Proc Natl Acad Sci U S A, 105:50 (Dec 2008), hURL: http://dx.doi.
org/10.1073/pnas.0801621105i, pp.19690–19695.
Anastassiou, Costas A etal., ‘Ephaptic coupling of cortical neurons.’, Nat Neurosci, 14:2 (Feb
2011), hURL: http://dx.doi.org/10.1038/nn.2727i, pp.217–223.
Ayuso, Carmen and Millan, Jose M, ‘Retinitis pigmentosa and allied conditions today: a paradigm
of translational research.’, Genome Med, 2:5 (2010), hURL: http://dx.doi.org/10.1186/
gm155i, p.34.
Baccus, Stephen A and Meister, Markus, ‘Fast and slow contrast adaptation in retinal circuitry.’,
Neuron, 36:5 (Dec 2002), pp.909–919.
Bak, M. etal., ‘Visual sensations produced by intracortical microstimulation of the human occip-
ital cortex.’, Med Biol Eng Comput, 28:3 (May 1990), pp.257–259.
Balthasar, Chloé de etal., ‘Factors affecting perceptual thresholds in epiretinal prostheses.’, Invest
OphthalmolVisSci,49:6(Jun2008), hURL:http://dx.doi.org/10.1167/iovs.07-0696i,
pp.2303–2314.
Baratta, R. etal., ‘Orderly stimulation of skeletal muscle motor units with tripolar nerve cuff
electrode.’, IEEE Trans Biomed Eng, 36:8 (Aug 1989), hURL: http://dx.doi.org/10.
1109/10.30809i, pp.836–843.
Barhoum, R. etal., ‘Functional and structural modifications during retinal degeneration in the
rd10 mouse.’, Neuroscience, 155:3 (Aug 2008), hURL: http://dx.doi.org/10.1016/j.
neuroscience.2008.06.042i, pp.698–713.
Barlow, H. B., HILL, R. M. and LEVICK, W. R., ‘Retinal Ganglion Cells Responding Selectively
to Direction and Speed of Image Motion in the Rabbit’, J Physiol, 173 (Oct 1964), pp.377–
407.
Behrend, Matthew R etal., ‘Resolution of the epiretinal prosthesis is not limited by electrode
size.’, IEEE Trans Neural Syst Rehabil Eng, 19:4 (Aug 2011), hURL: http://dx.doi.org/
10.1109/TNSRE.2011.2140132i, pp.436–442.
144
Behrend, Matthew R etal., ‘Selective labeling of retinal ganglion cells with calcium indicators
by retrograde loading in vitro.’, J Neurosci Methods, 179:2 (May 2009), hURL: http://dx.
doi.org/10.1016/j.jneumeth.2009.01.019i, pp.166–172.
Behrend, Matthew R, Ahuja, Ashish K and Weiland, James D, ‘Dynamic current density of
the disk electrode double-layer.’, IEEE Trans Biomed Eng, 55:3 (Mar 2008), hURL: http:
//dx.doi.org/10.1109/TBME.2008.915723i, pp.1056–1062.
Benav, Heval etal., ‘Restoration of useful vision up to letter recognition capabilities using
subretinal microphotodiodes.’, Conf Proc IEEE Eng Med Biol Soc, 2010 (2010), hURL:
http://dx.doi.org/10.1109/IEMBS.2010.5627549i, pp.5919–5922.
Bi, Anding etal., ‘Ectopic expression of a microbial-type rhodopsin restores visual responses in
mice with photoreceptor degeneration.’, Neuron, 50:1 (Apr 2006), hURL: http://dx.doi.
org/10.1016/j.neuron.2006.02.026i, pp.23–33.
Bradley, D. C. etal., ‘Visuotopic mapping through a multichannel stimulating implant in primate
V1.’, J Neurophysiol, 93:3 (Mar 2005), hURL: http://dx.doi.org/10.1152/jn.01213.
2003i, pp.1659–1670.
Brelén, M. E. etal., ‘Creating a meaningful visual perception in blind volunteers by optic
nerve stimulation.’, J Neural Eng, 2:1 (Mar 2005), hURL: http://dx.doi.org/10.1088/
1741-2560/2/1/004i, pp.S22–S28.
Brindley, G. S., ‘Beats produced by simultaneous stimulation of the human eye with intermittent
light and intermittent or alternating electric current.’, J Physiol, 164 (Oct 1962), pp.157–
167.
Brindley, G. S., ‘A new interaction of light and electricity in stimulating the human retina’, J
Physiol, 171 (Jun 1964), pp.514–520.
Brindley, G. S., ‘Sensations produced by electrical stimulation of the occipital poles of the cerebral
hemispheres, and their use in constructing visual prostheses.’, Ann R Coll Surg Engl, 47:2
(Aug 1970), pp.106–108.
Brindley, G. S. etal., ‘The extent of the region of occipital cortex that when stimulated gives
phosphenes fixed in the visual field.’, J Physiol, 225:2 (Sep 1972), pp.57P–58P.
Brindley, G. S. and Lewin, W. S., ‘The sensations produced by electrical stimulation of the visual
cortex.’, J Physiol, 196:2 (May 1968), pp.479–493.
Brindley, G. S. and Lewin, W. S., ‘The visual sensations produced by electrical stimulation of the
medial occipital cortex.’, J Physiol, 194:2 (Feb 1968), pp.54–5P.
Brummer, S. B., Robblee, L. S. and Hambrecht, F. T., ‘Criteria for selecting electrodes for elec-
trical stimulation: theoretical and practical considerations.’, Ann N Y Acad Sci, 405 (1983),
pp.159–171.
Brummer, S. B. and Turner, M. J., ‘Electrochemical considerations for safe electrical stimulation
of the nervous system with platinum electrodes.’, IEEE Trans Biomed Eng, 24:1 (Jan 1977),
hURL: http://dx.doi.org/10.1109/TBME.1977.326218i, pp.59–63.
Cai, Changsi etal., ‘Response variability to high rates of electric stimulation in retinal ganglion
cells.’, J Neurophysiol, 106:1 (Jul 2011), hURL: http://dx.doi.org/10.1152/jn.00956.
2010i, pp.153–162.
145
Calford, M.B.etal., ‘Plasticityinadultcatvisualcortex(area17)followingcircumscribedmonoc-
ular lesions of all retinal layers.’, J Physiol, 524 Pt 2 (Apr 2000), pp.587–602.
Carpenter, R. H., ‘Electrical stimulation of the human eye in different adaptational states.’, J
Physiol, 221:1 (Feb 1972), pp.137–148.
Carpenter, R. H., ‘Contour-like phosphenes from electrical stimulation of the human eye: some
new observations.’, J Physiol, 229:3 (Mar 1973), pp.767–785.
Carras, P. L., Coleman, P. A. and Miller, R. F., ‘Site of action potential initiation in amphibian
retinal ganglion cells.’, J Neurophysiol, 67:2 (Feb 1992), pp.292–304.
Caspi, Avi etal., ‘Feasibility study of a retinal prosthesis: spatial vision with a 16-electrode
implant.’, Arch Ophthalmol, 127:4 (Apr 2009), hURL: http://dx.doi.org/10.1001/
archophthalmol.2009.20i, pp.398–401.
Chader, Gerald J, ‘Beyond basic research for inherited and orphan retinal diseases: successes and
challenges.’, Retina, 25:8 Suppl (Dec 2005), pp.S15–S17.
Chader, Gerald J, Weiland, James and Humayun, Mark S, ‘Artificial vision: needs, functioning,
and testing of a retinal electronic prosthesis.’, Prog Brain Res, 175 (2009), hURL: http:
//dx.doi.org/10.1016/S0079-6123(09)17522-2i, pp.317–332.
Chan, Leanne L H etal., ‘Both electrical stimulation thresholds and SMI-32-immunoreactive reti-
nalganglioncelldensitycorrelatewithageinS334terline3ratretina.’, JNeurophysiol,105:6
(Jun 2011), hURL: http://dx.doi.org/10.1152/jn.00619.2010i, pp.2687–2697.
Chander, D. and Chichilnisky, E. J., ‘Adaptation to temporal contrast in primate and salamander
retina.’, J Neurosci, 21:24 (Dec 2001), pp.9904–9916.
Chang, B. etal., ‘Retinal degeneration mutants in the mouse.’, Vision Res, 42:4 (Feb 2002),
pp.517–525.
Chow, A. Y. and Chow, V. Y., ‘Subretinal electrical stimulation of the rabbit retina.’, Neurosci
Lett, 225:1 (Mar 1997), pp.13–16.
Chow, Alan Y etal., ‘The artificial silicon retina microchip for the treatment of vision loss from
retinitis pigmentosa.’, Arch Ophthalmol, 122:4 (Apr 2004), hURL: http://dx.doi.org/10.
1001/archopht.122.4.460i, pp.460–469.
Chowdhury, Vivek, Morley, John W and Coroneo, Minas T, ‘Stimulation of the retina with a
multielectrode extraocular visual prosthesis.’, ANZ J Surg, 75:8 (Aug 2005), hURL: http:
//dx.doi.org/10.1111/j.1445-2197.2005.03498.xi, pp.697–704.
Cottaris, Nicolas P and Elfar, Sylvia D, ‘How the retinal network reacts to epiretinal stimulation
to form the prosthetic visual input to the cortex.’, J Neural Eng, 2:1 (Mar 2005), hURL:
http://dx.doi.org/10.1088/1741-2560/2/1/010i, pp.S74–S90.
Curcio, C. A. and Allen, K. A., ‘Topography of ganglion cells in human retina.’, J Comp Neu-
rol, 300:1 (Oct 1990), hURL: http://dx.doi.org/10.1002/cne.903000103i, pp.5–25.
Dacey, D. M., ‘Physiology, morphology and spatial densities of identified ganglion cell types in
primate retina.’, Ciba Found Symp, 184 (1994), pp.12–28; discussion 28–34, 63–70.
Daiger, Stephen P, Bowne, Sara J and Sullivan, Lori S, ‘Perspective on genes and mutations
causing retinitis pigmentosa.’, Arch Ophthalmol, 125:2 (Feb 2007), hURL: http://dx.doi.
org/10.1001/archopht.125.2.151i, pp.151–158.
146
Dan, Y., Atick, J. J. and Reid, R. C., ‘Efficient coding of natural scenes in the lateral genicu-
late nucleus: experimental test of a computational theory.’, J Neurosci, 16:10 (May 1996),
pp.3351–3362.
Dobelle, W. H., ‘Introduction to sensory prostheses for the blind and deaf.’, Trans Am Soc Artif
Intern Organs, 20 B (1974), pp.761–764.
Dobelle, W. H. and Mladejovsky, M. G., ‘Phosphenes produced by electrical stimulation of human
occipital cortex, and their application to the development of a prosthesis for the blind.’, J
Physiol, 243:2 (Dec 1974), pp.553–576.
Dobelle, W. H., Mladejovsky, M. G. and Girvin, J. P., ‘Artifical vision for the blind: electrical
stimulation of visual cortex offers hope for a functional prosthesis.’, Science, 183:123 (Feb
1974), pp.440–444.
Dobelle, W. H. etal., ‘Mapping the representation of the visual field by electrical stimulation of
human visual cortex.’, Am J Ophthalmol, 88:4 (Oct 1979), pp.727–735.
Doroudchi, M. Mehdi etal., ‘Virally delivered Channelrhodopsin-2 Safely and Effectively Restores
Visual Function in Multiple Mouse Models of Blindness.’, Mol Ther, (Apr 2011), hURL:
http://dx.doi.org/10.1038/mt.2011.69i.
Drasdo,N.andFowler,C.W.,‘Non-linearprojectionoftheretinalimageinawide-angleschematic
eye.’, Br J Ophthalmol, 58:8 (Aug 1974), pp.709–714.
Dreher, B., Burke, W. and Calford, M. B., ‘Cortical plasticity revealed by circumscribed retinal
lesions or artificial scotomas.’, Prog Brain Res, 134 (2001), pp.217–246.
Evans, J. R. etal., ‘Brightness of phosphenes elicited by electrical stimulation of human visual
cortex.’, Sens Processes, 3:1 (Mar 1979), pp.82–94.
Fang, Z. P. and Mortimer, J. T., ‘Alternate excitation of large and small axons with different
stimulation waveforms: an application to muscle activation.’, Med Biol Eng Comput, 29:5
(Sep 1991), pp.543–547.
Fernández, E. etal., ‘Development of a cortical visual neuroprosthesis for the blind: the relevance
of neuroplasticity.’, J Neural Eng, 2:4 (Dec 2005), hURL: http://dx.doi.org/10.1088/
1741-2560/2/4/R01i, pp.R1–12.
Feucht, M. etal., ‘[Development of an epiretinal prosthesis for stimulation of the hu-
man retina].’, Ophthalmologe, 102:7 (Jul 2005), hURL: http://dx.doi.org/10.1007/
s00347-005-1186-6i, pp.688–691.
Field, G. D. and Chichilnisky, E. J., ‘Information processing in the primate retina: circuitry and
coding.’, Annu Rev Neurosci, 30 (2007), hURL: http://dx.doi.org/10.1146/annurev.
neuro.30.051606.094252i, pp.1–30.
Fitzgibbon, T. and Reese, B. E., ‘Organization of retinal ganglion cell axons in the optic fiber
layer and nerve of fetal ferrets.’, Vis Neurosci, 13:5 (1996), pp.847–861.
Freeman, Daniel K etal., ‘Selective activation of neuronal targets with sinusoidal electric stimula-
tion.’, J Neurophysiol, 104:5 (Nov 2010), hURL: http://dx.doi.org/10.1152/jn.00551.
2010i, pp.2778–2791.
Freeman, Daniel K and Fried, Shelley I, ‘Multiple components of ganglion cell desensitization in
response to prosthetic stimulation.’, J Neural Eng, 8:1 (Feb 2011), hURL: http://dx.doi.
org/10.1088/1741-2560/8/1/016008i, p.016008.
147
Freeman, Daniel K, Rizzo, Joseph F and Fried, Shelley I, ‘Electric stimulation with sinusoids and
white noise for neural prostheses.’, Front Neurosci, 4 (2010), hURL: http://dx.doi.org/
10.3389/neuro.20.001.2010i, p.28.
Freeman, Daniel K, Rizzo, Joseph F and Fried, Shelley I, ‘Encoding visual information in retinal
ganglion cells with prosthetic stimulation.’, J Neural Eng, 8:3 (Jun 2011), hURL: http:
//dx.doi.org/10.1088/1741-2560/8/3/035005i, p.035005.
Fried, S. I., Hsueh, H. A. and Werblin, F. S., ‘A method for generating precise temporal patterns
of retinal spiking using prosthetic stimulation.’, J Neurophysiol, 95:2 (Feb 2006), hURL:
http://dx.doi.org/10.1152/jn.00849.2005i, pp.970–978.
Fried, Shelley I etal., ‘Axonal sodium-channel bands shape the response to electric stimulation in
retinal ganglion cells.’, J Neurophysiol, 101:4 (Apr 2009), hURL: http://dx.doi.org/10.
1152/jn.91081.2008i, pp.1972–1987.
Fujikado, Takashi etal., ‘Evaluation of phosphenes elicited by extraocular stimulation in normals
and by suprachoroidal-transretinal stimulation in patients with retinitis pigmentosa.’, Grae-
fes Arch Clin Exp Ophthalmol, 245:10 (Oct 2007), hURL: http://dx.doi.org/10.1007/
s00417-007-0563-zi, pp.1411–1419.
Gargini, Claudia etal., ‘Retinal organization in the retinal degeneration 10 (rd10) mutant mouse:
a morphological and ERG study.’, J Comp Neurol, 500:2 (Jan 2007), hURL: http://dx.
doi.org/10.1002/cne.21144i, pp.222–238.
Gerding, H., Benner, F.P.andTaneri, S., ‘Experimentalimplantationofepiretinalretinaimplants
(EPI-RET) with an IOL-type receiver unit.’, J Neural Eng, 4:1 (Mar 2007), hURL: http:
//dx.doi.org/10.1088/1741-2560/4/1/S06i, pp.S38–S49.
Ghosh, Krishna K etal., ‘Types of bipolar cells in the mouse retina.’, J Comp Neurol, 469:1 (Jan
2004), hURL: http://dx.doi.org/10.1002/cne.10985i, pp.70–82.
Giannikopoulos, Dimitrios V and Eysel, Ulf T, ‘Dynamics and specificity of cortical map re-
organization after retinal lesions.’, Proc Natl Acad Sci U S A, 103:28 (Jul 2006), hURL:
http://dx.doi.org/10.1073/pnas.0604539103i, pp.10805–10810.
Gilbert, C. D., ‘Adult cortical dynamics.’, Physiol Rev, 78:2 (Apr 1998), pp.467–485.
Gilbert, C. D., Hirsch, J. A. and Wiesel, T. N., ‘Lateral interactions in visual cortex.’, Cold Spring
Harb Symp Quant Biol, 55 (1990), pp.663–677.
Gilbert, C. D. and Wiesel, T. N., ‘Receptive field dynamics in adult primary visual cortex.’,
Nature, 356:6365 (Mar 1992), hURL: http://dx.doi.org/10.1038/356150a0i, pp.150–
152.
Gilbert, Charles D, Li, Wu and Piech, Valentin, ‘Perceptual learning and adult cortical plasticity.’,
J Physiol, 587:Pt 12 (Jun 2009), hURL: http://dx.doi.org/10.1113/jphysiol.2009.
171488i, pp.2743–2751.
Girvin, J. P., ‘Current status of artificial vision by electrocortical stimulation.’, Can J Neurol
Sci, 15:1 (Feb 1988), pp.58–62.
Goo, Yong Sook etal., ‘Retinal ganglion cell responses to voltage and current stimulation in wild-
type and rd1 mouse retinas.’, J Neural Eng, 8:3 (Jun 2011), hURL: http://dx.doi.org/
10.1088/1741-2560/8/3/035003i, p.035003.
148
Greenberg, R. J. etal., ‘A computational model of electrical stimulation of the retinal ganglion
cell.’, IEEE Trans Biomed Eng, 46:5 (May 1999), pp.505–514.
Greenberg, R.J., ‘Analysis of electrical stimulation of the vertebrate retina- work towards a retinal
prosthesis’, Ph.D thesis, John Hopkins University (1998).
Greenwald, ScottHetal., ‘Brightnessasafunctionofcurrentamplitudeinhumanretinalelectrical
stimulation.’, Invest Ophthalmol Vis Sci, 50:11 (Nov 2009), hURL: http://dx.doi.org/10.
1167/iovs.08-2897i, pp.5017–5025.
Hauswirth, William W etal., ‘Treatment of leber congenital amaurosis due to RPE65 mutations
by ocular subretinal injection of adeno-associated virus gene vector: short-term results of
a phase I trial.’, Hum Gene Ther, 19:10 (Oct 2008), hURL: http://dx.doi.org/10.1089/
hum.2008.107i, pp.979–990.
Henderson, D.C., Evans, J.R.andDobelle, W.H., ‘Therelationshipbetweenstimulusparameters
and phosphene threshold/brightness, during stimulation of human visual cortex.’, Trans Am
Soc Artif Intern Organs, 25 (1979), pp.367–371.
Hensley, S. H., Yang, X. L. and Wu, S. M., ‘Relative contribution of rod and cone inputs to
bipolar cells and ganglion cells in the tiger salamander retina.’, J Neurophysiol, 69:6 (Jun
1993), pp.2086–2098.
Hoffer, J. A., Loeb, G. E. and Pratt, C. A., ‘Single unit conduction velocities from averaged
nerve cuff electrode records in freely moving cats.’, J Neurosci Methods, 4:3 (Oct 1981),
pp.211–225.
Horsager, A., ‘Prosthetic Visual Perception: Retinal Electrical Stimulation in Blind Human Pa-
tients’, Ph.D thesis, University of Southern California (2009).
Horsager, Alan etal., ‘Temporal interactions during paired-electrode stimulation in two retinal
prosthesis subjects.’, Invest Ophthalmol Vis Sci, 52:1 (Jan 2011), hURL: http://dx.doi.
org/10.1167/iovs.10-5282i, pp.549–557.
Horsager, Alan, Greenberg, Robert J and Fine, Ione, ‘Spatiotemporal interactions in retinal pros-
thesis subjects.’, Invest Ophthalmol Vis Sci, 51:2 (Feb 2010), hURL: http://dx.doi.org/
10.1167/iovs.09-3746i, pp.1223–1233.
Horsager, Alan etal., ‘Predicting visual sensitivity in retinal prosthesis patients.’, Invest Oph-
thalmol Vis Sci, 50:4 (Apr 2009), hURL: http://dx.doi.org/10.1167/iovs.08-2595i,
pp.1483–1491.
Hubel, D.H.andWiesel, T.N., ‘Receptivefields, binocularinteractionandfunctionalarchitecture
in the cat’s visual cortex.’, J Physiol, 160 (Jan 1962), pp.106–154.
Hubel, D. H. and Wiesel, T. N., ‘Receptive fields and functional architecture of monkey striate
cortex.’, J Physiol, 195:1 (Mar 1968), pp.215–243.
Humayun, M. S. etal., ‘Pattern electrical stimulation of the human retina.’, Vision Res, 39:15 (Jul
1999), pp.2569–2576.
Humayun, M. S. etal., ‘Morphometric analysis of the extramacular retina from postmortem eyes
with retinitis pigmentosa.’, Invest Ophthalmol Vis Sci, 40:1 (Jan 1999), pp.143–148.
Humayun, Mark S etal., ‘Visual perception in a blind subject with a chronic microelectronic
retinal prosthesis.’, Vision Res, 43:24 (Nov 2003), pp.2573–2581.
149
Jansonius, N. M. etal., ‘A mathematical description of nerve fiber bundle trajectories and their
variability in the human retina.’, Vision Res, 49:17 (Aug 2009), hURL:http://dx.doi.org/
10.1016/j.visres.2009.04.029i, pp.2157–2163.
Jefferys, J. G., ‘Nonsynaptic modulation of neuronal activity in the brain: electric currents and
extracellular ions.’, Physiol Rev, 75:4 (Oct 1995), pp.689–723.
Jensen, Ralph J and Rizzo, Joseph F, ‘Thresholds for activation of rabbit retinal ganglion cells
with a subretinal electrode.’, Exp Eye Res, 83:2 (Aug 2006), hURL: http://dx.doi.org/
10.1016/j.exer.2006.01.012i, pp.367–373.
Jensen, Ralph J and Rizzo, Joseph F, ‘Responses of ganglion cells to repetitive electrical stimu-
lation of the retina.’, J Neural Eng, 4:1 (Mar 2007), hURL: http://dx.doi.org/10.1088/
1741-2560/4/1/S01i, pp.S1–S6.
Jensen, Ralph J and Rizzo, Joseph F, ‘Activation of retinal ganglion cells in wild-type and rd1
mice through electrical stimulation of the retinal neural network.’, Vision Res, 48:14 (Jun
2008), hURL: http://dx.doi.org/10.1016/j.visres.2008.04.016i, pp.1562–1568.
Jensen, Ralph J and Rizzo, Joseph F, ‘Activation of ganglion cells in wild-type and rd1 mouse
retinas with monophasic and biphasic current pulses.’, J Neural Eng, 6:3 (Jun 2009), hURL:
http://dx.doi.org/10.1088/1741-2560/6/3/035004i, p.035004.
Jensen, Ralph J etal., ‘Thresholds for activation of rabbit retinal ganglion cells with an ultrafine,
extracellular microelectrode.’, Invest Ophthalmol Vis Sci, 44:8 (Aug 2003), pp.3533–3543.
Jensen, Ralph J, Ziv, Ofer R and Rizzo, Joseph F, ‘Responses of rabbit retinal ganglion cells to
electrical stimulation with an epiretinal electrode.’, J Neural Eng, 2:1 (Mar 2005), hURL:
http://dx.doi.org/10.1088/1741-2560/2/1/003i, pp.S16–S21.
Jensen, Ralph J., Ziv, Ofer R and Rizzo, Joseph F, ‘Thresholds for activation of rabbit retinal
ganglion cells with relatively large, extracellular microelectrodes.’, Invest Ophthalmol Vis
Sci, 46:4 (Apr 2005), hURL: http://dx.doi.org/10.1167/iovs.04-1018i, pp.1486–1496.
Jones, Bryan W and Marc, Robert E, ‘Retinal remodeling during retinal degeneration.’, Exp
Eye Res, 81:2 (Aug 2005), hURL: http://dx.doi.org/10.1016/j.exer.2005.03.006i,
pp.123–137.
Jones, Bryan W etal., ‘Retinal remodeling triggered by photoreceptor degenerations.’, J Comp
Neurol, 464:1 (Sep 2003), hURL: http://dx.doi.org/10.1002/cne.10703i, pp.1–16.
Jones, Bryan W, Watt, Carl B and Marc, Robert E, ‘Retinal remodelling.’, Clin Exp Optom, 88:5
(Sep 2005), pp.282–291.
Kanda,Hiroyukietal.,‘Electrophysiologicalstudiesofthefeasibilityofsuprachoroidal-transretinal
stimulation for artificial vision in normal and RCS rats.’, Invest Ophthalmol Vis Sci, 45:2
(Feb 2004), pp.560–566.
Keck, Tara etal., ‘Massive restructuring of neuronal circuits during functional reorganization of
adult visual cortex.’, Nat Neurosci, 11:10 (Oct 2008), hURL: http://dx.doi.org/10.1038/
nn.2181i, pp.1162–1167.
Kolomiets, Bogdan etal., ‘Late histological and functional changes in the P23H rat retina after
photoreceptor loss.’, Neurobiol Dis, 38:1 (Apr 2010), hURL: http://dx.doi.org/10.1016/
j.nbd.2009.12.025i, pp.47–58.
150
Kotler, S., ‘Vision Quest’, Wired, (2002).
Krause, F, ‘Die Sehbahnen in chirurgischer Beziehung und die faradische Reizung des Sehzen-
trums’, Klin Wschr, 3 (1924), pp.1260–1265.
Lagali, Pamela S etal., ‘Light-activated channels targeted to ON bipolar cells restore visual func-
tion in retinal degeneration.’, Nat Neurosci, 11:6 (Jun 2008), hURL: http://dx.doi.org/
10.1038/nn.2117i, pp.667–675.
LeRoy, Charles, ‘Ou ’lon rend compte de quelques tentatives que l’on a faites pour guerir plusieurs
maladies par l’electricite’, Hist Acad Roy Sciences (Paris), 60 (1755), pp.87–95.
Li, Z. Y. etal., ‘Rhodopsin transgenic pigs as a model for human retinitis pigmentosa.’, Invest
Ophthalmol Vis Sci, 39:5 (Apr 1998), pp.808–819.
Lin, Bin etal., ‘Restoration of visual function in retinal degeneration mice by ectopic expression
of melanopsin.’, Proc Natl Acad Sci U S A, 105:41 (Oct 2008), hURL: http://dx.doi.org/
10.1073/pnas.0806114105i, pp.16009–16014.
Lipton, S. A. and Tauck, D. L., ‘Voltage-dependent conductances of solitary ganglion cells disso-
ciated from the rat retina.’, J Physiol, 385 (Apr 1987), pp.361–391.
Liu, Tingting etal., ‘Incomplete cortical reorganization in macular degeneration.’, Invest Oph-
thalmol Vis Sci, 51:12 (Dec 2010), hURL: http://dx.doi.org/10.1167/iovs.09-4926i,
pp.6826–6834.
Loeb, G. E. and Peck, R. A., ‘Cuff electrodes for chronic stimulation and recording of peripheral
nerve activity.’, J Neurosci Methods, 64:1 (Jan 1996), pp.95–103.
Lowenstein, K and Borchardt, M., Dtsch Z Nervenheilk, 58 (1918), pp.264–292.
MacNeil, M. A. and Masland, R. H., ‘Extreme diversity among amacrine cells: implications for
function.’, Neuron, 20:5 (May 1998), pp.971–982.
Mahadevappa, Manjunatha etal., ‘Perceptual thresholds and electrode impedance in three retinal
prosthesis subjects.’, IEEE Trans Neural Syst Rehabil Eng, 13:2 (Jun 2005), hURL: http:
//dx.doi.org/10.1109/TNSRE.2005.848687i, pp.201–206.
Manookin, Michael B and Demb, Jonathan B, ‘Presynaptic mechanism for slow contrast adap-
tation in mammalian retinal ganglion cells.’, Neuron, 50:3 (May 2006), hURL: http:
//dx.doi.org/10.1016/j.neuron.2006.03.039i, pp.453–464.
Marc, Robert E and Jones, Bryan W, ‘Retinal remodeling in inherited photoreceptor degener-
ations.’, Mol Neurobiol, 28:2 (Oct 2003), hURL: http://dx.doi.org/10.1385/MN:28:2:
139i, pp.139–147.
Marc, Robert E etal., ‘Neural reprogramming in retinal degeneration.’, Invest Ophthalmol Vis
Sci, 48:7 (Jul 2007), hURL: http://dx.doi.org/10.1167/iovs.07-0032i, pp.3364–3371.
Marc, Robert E etal., ‘Neural remodeling in retinal degeneration.’, Prog Retin Eye Res, 22:5 (Sep
2003), pp.607–655.
Marg, E., ‘Magnetostimulation of vision: direct noninvasive stimulation of the retina and the
visual brain.’, Optom Vis Sci, 68:6 (Jun 1991), pp.427–440.
Margalit, Eyal etal., ‘Retinal prosthesis for the blind.’, Surv Ophthalmol, 47:4 (2002), pp.335–356.
151
Margolis, David J and Detwiler, Peter B, ‘Cellular origin of spontaneous ganglion cell spike
activity in animal models of retinitis pigmentosa.’, J Ophthalmol, 2011 (2011), hURL: http:
//dx.doi.org/10.1155/2011/507037i.
Margolis, David J etal., ‘Functional stability of retinal ganglion cells after degeneration-induced
changes in synaptic input.’, J Neurosci, 28:25 (Jun 2008), hURL: http://dx.doi.org/10.
1523/JNEUROSCI.1533-08.2008i, pp.6526–6536.
Masland, R.H., ‘Neuronaldiversityintheretina.’, Curr Opin Neurobiol,11:4(Aug2001), pp.431–
436.
Masuda, Yoichiro etal., ‘V1 projection zone signals in human macular degeneration depend on
task, not stimulus.’, Cereb Cortex, 18:11 (Nov 2008), hURL: http://dx.doi.org/10.1093/
cercor/bhm256i, pp.2483–2493.
Masuda,Yoichiroetal.,‘Task-dependentV1responsesinhumanretinitispigmentosa.’, InvestOph-
thalmol Vis Sci, 51:10 (Oct 2010), hURL: http://dx.doi.org/10.1167/iovs.09-4775i,
pp.5356–5364.
Mazzoni, Francesca, Novelli, Elena and Strettoi, Enrica, ‘Retinal ganglion cells survive and main-
tain normal dendritic morphology in a mouse model of inherited photoreceptor degener-
ation.’, J Neurosci, 28:52 (Dec 2008), hURL: http://dx.doi.org/10.1523/JNEUROSCI.
4968-08.2008i, pp.14282–14292.
McCreery, D. B. etal., ‘Comparison of neural damage induced by electrical stimulation with
faradaic and capacitor electrodes.’, Ann Biomed Eng, 16:5 (1988), pp.463–481.
McCreery DB, Agnew WF., Neural Prostheses: Fundamental Studies (NJ: Prentice Hall, 1990),
chap. Mechanisms of stimulation-induced neural damage and their relation to guidelines for
safe stimulation., pp.297–317.
Moreland, J. D., ‘Threshold measurements of the blue arcs phenomenon.’, Vision Res, 8:8 (Aug
1968), pp.1093–1106.
Moreland, J. D., ‘Possible mechanisms of the blue arcs of the retina.’, J Physiol, 201:2 (Apr 1969),
pp.60P–61P.
Moreland, J. D., ‘Retinal topography and the blue-arcs phenomenon.’, Vision Res, 9:8 (Aug 1969),
pp.965–976.
Morimoto, Takeshi etal., ‘Chronic implantation of newly developed suprachoroidal-transretinal
stimulation (STS) prosthesis in dogs.’, Invest Ophthalmol Vis Sci, (Jul 2011), hURL: http:
//dx.doi.org/10.1167/iovs.10-6971i.
Murakami, I., Komatsu, H. and Kinoshita, M., ‘Perceptual filling-in at the scotoma following a
monocular retinal lesion in the monkey.’, Vis Neurosci, 14:1 (1997), pp.89–101.
Naito, J., ‘Retinogeniculate projection fibers in the monkey optic nerve: a demonstration of the
fiber pathways by retrograde axonal transport of WGA-HRP.’, J Comp Neurol, 284:2 (Jun
1989), hURL: http://dx.doi.org/10.1002/cne.902840203i, pp.174–186.
Nakauchi, Kazuaki etal., ‘Threshold suprachoroidal-transretinal stimulation current resulting in
retinal damage in rabbits.’, J Neural Eng, 4:1 (Mar 2007), hURL: http://dx.doi.org/10.
1088/1741-2560/4/1/S07i, pp.S50–S57.
152
Nakauchi, Kazuaki etal., ‘Transretinal electrical stimulation by an intrascleral multichannel elec-
trode array in rabbit eyes.’, Graefes Arch Clin Exp Ophthalmol, 243:2 (Feb 2005), hURL:
http://dx.doi.org/10.1007/s00417-004-1060-2i, pp.169–174.
Nanduri, D. etal., ‘Retinal prosthesis phosphene shape analysis.’, Conf Proc IEEE Eng Med Biol
Soc, 2008 (2008), hURL: http://dx.doi.org/10.1109/IEMBS.2008.4649524i, pp.1785–
1788.
Nishida, Kentaro etal., ‘Efficacy of suprachoroidal-transretinal stimulation in a rabbit model of
retinal degeneration.’, Invest Ophthalmol Vis Sci, 51:4 (Apr 2010), hURL: http://dx.doi.
org/10.1167/iovs.09-4120i, pp.2263–2268.
Normann, R. A. etal., ‘A neural interface for a cortical vision prosthesis.’, Vision Res, 39:15 (Jul
1999), pp.2577–2587.
Normann, Richard A etal., ‘Toward the development of a cortically based visual neuroprosthe-
sis.’, J Neural Eng, 6:3 (Jun 2009), hURL: http://dx.doi.org/10.1088/1741-2560/6/3/
035001i, p.035001.
Obeid, Iyad, Veraart, Claude and Delbeke, Jean, ‘Estimation of phosphene spatial variability for
visual prosthesis applications.’, Artif Organs, 34:5 (May 2010), hURL: http://dx.doi.org/
10.1111/j.1525-1594.2009.00878.xi, pp.358–365.
O’Brien, Brendan J etal., ‘Intrinsic physiological properties of cat retinal ganglion cells.’, J Phys-
iol, 538:Pt 3 (Feb 2002), pp.787–802.
Oyster, Clyde W., The human eye : structure and function (Sinauer Associates, 1999).
Palagina, Ganna, Eysel, Ulf T and Jancke, Dirk, ‘Strengthening of lateral activation in adult
rat visual cortex after retinal lesions captured with voltage-sensitive dye imaging in vivo.’,
Proc Natl Acad Sci U S A, 106:21 (May 2009), hURL: http://dx.doi.org/10.1073/pnas.
0900068106i, pp.8743–8747.
Palanker, Daniel etal., ‘Design of a high-resolution optoelectronic retinal prosthesis.’, J Neural
Eng,2:1(Mar2005), hURL:http://dx.doi.org/10.1088/1741-2560/2/1/012i, pp.S105–
S120.
Peichl, L. and Wässle, H., ‘Size, scatter and coverage of ganglion cell receptive field centres in the
cat retina.’, J Physiol, 291 (Jun 1979), pp.117–141.
Penfield, W., ‘Some observations on the cerebral cortex of man.’, Proc R Soc Lond B Biol
Sci, 134:876 (Jul 1947), pp.329–347.
Penfield, W. and Jasper, H., Epilepsy and the functional anatomy of the human brain.Epilepsy
and the functional anatomy of the human brain (London: Churchill, 1954).
Penfield, W. and Rasmussen, T., The cerebral cortex of man. (NewYork: Macmillan, 1952).
Peng, Y. W. etal., ‘Ectopic synaptogenesis in the mammalian retina caused by rod photoreceptor-
specific mutations.’, Nat Neurosci, 3:11 (Nov 2000), hURL: http://dx.doi.org/10.1038/
80639i, pp.1121–1127.
Pezaris, John S and Eskandar, Emad N, ‘Getting signals into the brain: visual prosthetics through
thalamic microstimulation.’, Neurosurg Focus, 27:1 (Jul 2009), hURL: http://dx.doi.org/
10.3171/2009.4.FOCUS0986i, p.E6.
153
Pezaris, John S and Reid, R. Clay, ‘Demonstration of artificial visual percepts generated through
thalamic microstimulation.’, Proc Natl Acad Sci U S A, 104:18 (May 2007), hURL: http:
//dx.doi.org/10.1073/pnas.0608563104i, pp.7670–7675.
Pezaris, John S and Reid, R. Clay, ‘Simulations of electrode placement for a thalamic visual
prosthesis.’, IEEE Trans Biomed Eng, 56:1 (Jan 2009), hURL: http://dx.doi.org/10.
1109/TBME.2008.2005973i, pp.172–178.
Phillips,M.Joseph,Otteson,DeborahCandSherry,DavidM,‘Progressionofneuronalandsynap-
tic remodeling in the rd10 mouse model of retinitis pigmentosa.’, J Comp Neurol, 518:11
(Jun 2010), hURL: http://dx.doi.org/10.1002/cne.22322i, pp.2071–2089.
Polyak, S.L., The retina; the anatomy and the histology of the retina in man, ape, and monkey,
including the consideration of visual functions, the history of physiological optics, and the
histological laboratory technique (The University of Chicago press, 1941).
Rasmussen, T. and Penfield, W., ‘The human sensorimotor cortex as studied by electrical stimu-
lation.’, Fed Proc, 6:1 Pt 2 (1947), p.184.
Rattay, F., ‘Analysis of models for extracellular fiber stimulation.’, IEEE Trans Biomed Eng, 36:7
(Jul 1989), hURL: http://dx.doi.org/10.1109/10.32099i, pp.676–682.
Rattay, F., ‘The basic mechanism for the electrical stimulation of the nervous system.’, Neuro-
science, 89:2 (Mar 1999), pp.335–346.
Rattay, Frank and Resatz, Susanne, ‘Effective electrode configuration for selective stimulation
with inner eye prostheses.’, IEEE Trans Biomed Eng, 51:9 (Sep 2004), hURL: http://dx.
doi.org/10.1109/TBME.2004.828044i, pp.1659–1664.
Rieke, F., ‘Temporal contrast adaptation in salamander bipolar cells.’, J Neurosci, 21:23 (Dec
2001), pp.9445–9454.
Rizzo, Joseph F etal., ‘Perceptual efficacy of electrical stimulation of human retina with a mi-
croelectrode array during short-term surgical trials.’, Invest Ophthalmol Vis Sci, 44:12 (Dec
2003), pp.5362–5369.
Rodieck, R.W., Neurosciences (New York: Liss, 1988), chap. The primate retina, pp.203–278.
Rohrschneider, K., ‘Determination of the location of the fovea on the fundus.’, Invest Ophthalmol
Vis Sci, 45:9 (Sep 2004), hURL: http://dx.doi.org/10.1167/iovs.03-1157i, pp.3257–
3258.
Roska, B. and Werblin, F., ‘Vertical interactions across ten parallel, stacked representations in the
mammalian retina.’, Nature, 410:6828 (Mar 2001), hURL: http://dx.doi.org/10.1038/
35069068i, pp.583–587.
Roska, Botond, Molnar, Alyosha and Werblin, Frank S, ‘Parallel processing in retinal ganglion
cells: how integration of space-time patterns of excitation and inhibition form the spiking
output.’, J Neurophysiol, 95:6 (Jun 2006), hURL: http://dx.doi.org/10.1152/jn.00113.
2006i, pp.3810–3822.
Ryu, Sang Baek etal., ‘Decoding of retinal ganglion cell spike trains evoked by temporally pat-
terned electrical stimulation.’, Brain Res, 1348 (Aug 2010), hURL: http://dx.doi.org/10.
1016/j.brainres.2010.06.044i, pp.71–83.
154
Ryu, Sang Baek etal., ‘Electrically-evoked Neural Activities of rd1 Mice Retinal Ganglion Cells
by Repetitive Pulse Stimulation.’, Korean J Physiol Pharmacol, 13:6 (Dec 2009), hURL:
http://dx.doi.org/10.4196/kjpp.2009.13.6.443i, pp.443–448.
Ryu, Sang Baek etal., ‘Characterization of retinal ganglion cell activities evoked by tempo-
rally patterned electrical stimulation for the development of stimulus encoding strategies
for retinal implants.’, Brain Res, 1275 (Jun 2009), hURL: http://dx.doi.org/10.1016/j.
brainres.2009.03.064i, pp.33–42.
Sahel, Jose A. etal., ‘Subjects Blind From Outer Retinal Dystrophies Are Able To Consistently
Read Short Sentences Using The ArgusII Retinal Prosthesis System’, in: ARVO (2011).
Santos, A. etal., ‘Preservation of the inner retina in retinitis pigmentosa. A morphometric anal-
ysis.’, Arch Ophthalmol, 115:4 (Apr 1997), pp.511–515.
Schiefer, Matthew A and Grill, Warren M, ‘Sites of neuronal excitation by epiretinal electrical
stimulation.’, IEEE Trans Neural Syst Rehabil Eng, 14:1 (Mar 2006), hURL: http://dx.
doi.org/10.1109/TNSRE.2006.870488i, pp.5–13.
Schmidt, E. M. etal., ‘Feasibility of a visual prosthesis for the blind based on intracortical mi-
crostimulation of the visual cortex.’, Brain, 119 ( Pt 2) (Apr 1996), pp.507–522.
Schnapf, J. L., Kraft, T. W. and Baylor, D. A., ‘Spectral sensitivity of human cone photorecep-
tors.’, Nature, 325:6103 (1987), hURL: http://dx.doi.org/10.1038/325439a0i, pp.439–
441.
Schneider, Keith A, Richter, Marlene C and Kastner, Sabine, ‘Retinotopic organization and
functional subdivisions of the human lateral geniculate nucleus: a high-resolution func-
tional magnetic resonance imaging study.’, J Neurosci, 24:41 (Oct 2004), hURL: http:
//dx.doi.org/10.1523/JNEUROSCI.2413-04.2004i, pp.8975–8985.
Sekirnjak, Chris etal., ‘Electrical stimulation of mammalian retinal ganglion cells with multielec-
trode arrays.’, J Neurophysiol, 95:6 (Jun 2006), hURL: http://dx.doi.org/10.1152/jn.
01168.2005i, pp.3311–3327.
Sekirnjak, Chris etal., ‘High-resolution electrical stimulation of primate retina for epiretinal
implant design.’, J Neurosci, 28:17 (Apr 2008), hURL: http://dx.doi.org/10.1523/
JNEUROSCI.5138-07.2008i, pp.4446–4456.
Sekirnjak, Chris etal., ‘Loss of responses to visual but not electrical stimulation in ganglion cells
of rats with severe photoreceptor degeneration.’, J Neurophysiol, 102:6 (Dec 2009), hURL:
http://dx.doi.org/10.1152/jn.00663.2009i, pp.3260–3269.
Shah, Hassan A., Montezuma, Sandra R and Rizzo, Joseph F, ‘In vivo electrical stimulation of
rabbit retina: effect of stimulus duration and electrical field orientation.’, Exp Eye Res, 83:2
(Aug 2006), hURL: http://dx.doi.org/10.1016/j.exer.2005.11.023i, pp.247–254.
Shannon, R. V., ‘A model of threshold for pulsatile electrical stimulation of cochlear implants.’,
Hear Res, 40:3 (Jul 1989), pp.197–204.
Shannon, R. V., ‘A model of safe levels for electrical stimulation.’, IEEE Trans Biomed Eng, 39:4
(Apr 1992), hURL: http://dx.doi.org/10.1109/10.126616i, pp.424–426.
Shivdasani, Mohit N etal., ‘Evaluation of stimulus parameters and electrode geometry for an
effective suprachoroidal retinal prosthesis.’, J Neural Eng, 7:3 (Jun 2010), hURL: http:
//dx.doi.org/10.1088/1741-2560/7/3/036008i, p.036008.
155
Smirnakis, Stelios M etal., ‘Lack of long-term cortical reorganization after macaque retinal le-
sions.’, Nature, 435:7040 (May 2005), hURL: http://dx.doi.org/10.1038/nature03495i,
pp.300–307.
Stasheff,StevenF,‘Emergenceofsustainedspontaneoushyperactivityandtemporarypreservation
of OFF responses in ganglion cells of the retinal degeneration (rd1) mouse.’, J Neurophys-
iol,99:3(Mar2008), hURL:http://dx.doi.org/10.1152/jn.00144.2007i, pp.1408–1421.
Stevens, S. S., ‘On the psychophysical law.’, Psychol Rev, 64:3 (May 1957), pp.153–181.
Stone, J. L. etal., ‘Morphometric analysis of macular photoreceptors and ganglion cells in retinas
with retinitis pigmentosa.’, Arch Ophthalmol, 110:11 (Nov 1992), pp.1634–1639.
Strettoi, E. and Pignatelli, V., ‘Modifications of retinal neurons in a mouse model of retinitis
pigmentosa.’, Proc Natl Acad Sci U S A, 97:20 (Sep 2000), hURL: http://dx.doi.org/10.
1073/pnas.190291097i, pp.11020–11025.
Strettoi, Enrica etal., ‘Remodeling of second-order neurons in the retina of rd/rd mutant mice.’,
Vision Res, 43:8 (Apr 2003), pp.867–877.
Strettoi, Enrica etal., ‘Morphological and functional abnormalities in the inner retina of the rd/rd
mouse.’, J Neurosci, 22:13 (Jul 2002), hURL: http://dx.doi.org/20026533i, pp.5492–
5504.
Sullivan, Robert K P, Woldemussie, Elizabeth and Pow, David V, ‘Dendritic and synaptic plastic-
ity of neurons in the human age-related macular degeneration retina.’, Invest Ophthalmol Vis
Sci, 48:6 (Jun 2007), hURL: http://dx.doi.org/10.1167/iovs.06-1283i, pp.2782–2791.
Thyagarajan, Senthil etal., ‘Visual function in mice with photoreceptor degeneration and trans-
genic expression of channelrhodopsin 2 in ganglion cells.’, J Neurosci, 30:26 (Jun 2010),
hURL: http://dx.doi.org/10.1523/JNEUROSCI.4417-09.2010i, pp.8745–8758.
Tomita, Hiroshi etal., ‘Visual properties of transgenic rats harboring the channelrhodopsin-2 gene
regulated by the thy-1.2 promoter.’, PLoS One, 4:11 (2009), hURL: http://dx.doi.org/
10.1371/journal.pone.0007679i, p.e7679.
Tomita, Hiroshi etal., ‘Channelrhodopsin-2 gene transduced into retinal ganglion cells restores
functional vision in genetically blind rats.’, Exp Eye Res, 90:3 (Mar 2010), hURL: http:
//dx.doi.org/10.1016/j.exer.2009.12.006i, pp.429–436.
Troy, J. B. and Enroth-Cugell, C., ‘X and Y ganglion cells inform the cat’s brain about contrast
in the retinal image.’, Exp Brain Res, 93:3 (1993), pp.383–390.
Troyk, Philip etal., ‘A model for intracortical visual prosthesis research.’, Artif Organs, 27:11
(Nov 2003), pp.1005–1015.
Varela, Carolina etal., ‘Functional modifications in rod bipolar cells in a mouse model of retinitis
pigmentosa.’, Vision Res, 43:8 (Apr 2003), pp.879–885.
Veraart, C. etal., ‘Visual sensations produced by optic nerve stimulation using an implanted
self-sizing spiral cuff electrode.’, Brain Res, 813:1 (Nov 1998), pp.181–186.
Veraart, Claude etal., ‘Pattern recognition with the optic nerve visual prosthesis.’, Artif Or-
gans, 27:11 (Nov 2003), pp.996–1004.
156
Walia, Saloni etal., ‘Retinal nerve fiber layer defects in RP patients.’, Invest Ophthalmol Vis
Sci, 48:10 (Oct 2007), hURL: http://dx.doi.org/10.1167/iovs.07-0404i, pp.4748–
4752.
Walter, Peter etal., ‘Cortical activation via an implanted wireless retinal prosthesis.’, Invest Oph-
thalmol Vis Sci, 46:5 (May 2005), hURL: http://dx.doi.org/10.1167/iovs.04-0924i,
pp.1780–1785.
Wandell, Brian A and Smirnakis, Stelios M, ‘Plasticity and stability of visual field maps in adult
primary visual cortex.’, Nat Rev Neurosci, 10:12 (Dec 2009), hURL: http://dx.doi.org/
10.1038/nrn2741i, pp.873–884.
Wang, H. etal., ‘High-speed mapping of synaptic connectivity using photostimulation in
Channelrhodopsin-2 transgenic mice.’, Proc Natl Acad Sci U S A, 104:19 (May 2007), hURL:
http://dx.doi.org/10.1073/pnas.0700384104i, pp.8143–8148.
Wang, Xin etal., ‘Feedforward excitation and inhibition evoke dual modes of firing in the cat’s
visual thalamus during naturalistic viewing.’, Neuron, 55:3 (Aug 2007), hURL: http://dx.
doi.org/10.1016/j.neuron.2007.06.039i, pp.465–478.
Watson, A. B., Handbook of Perception and Human Performance (Wiley, 1986), chap. Temporal
Sensitivity, pp.–.
Wichmann, F. A. and Hill, N. J., ‘The psychometric function: II. Bootstrap-based confidence
intervals and sampling.’, Percept Psychophys, 63:8 (Nov 2001), pp.1314–1329.
Wiley, J. D. and Webster, J. G., ‘Analysis and control of the current distribution under circular
dispersive electrodes.’, IEEE Trans Biomed Eng, 29:5 (May 1982), hURL: http://dx.doi.
org/10.1109/TBME.1982.324910i, pp.381–385.
Wiley, J. D. and Webster, J. G., ‘Distributed equivalent-circuit models for circular dispersive
electrodes.’, IEEE Trans Biomed Eng, 29:5 (May 1982), hURL: http://dx.doi.org/10.
1109/TBME.1982.324911i, pp.385–389.
Wilke, Robert etal., ‘Spatial Resolution and Perception of Patterns Mediated by a Subretinal16-
Electrode Array in Patients Blinded by Hereditary Retinal Dystrophies.’, Invest Ophthalmol
Vis Sci, (Jun 2011), hURL: http://dx.doi.org/10.1167/iovs.10-6946i.
Wollner, D. A. and Catterall, W. A., ‘Localization of sodium channels in axon hillocks and initial
segments of retinal ganglion cells.’, Proc Natl Acad Sci U S A, 83:21 (Nov 1986), pp.8424–
8428.
Woodbury, J. W. and Woodbury, D. M., ‘Vagal stimulation reduces the severity of maximal
electroshock seizures in intact rats: use of a cuff electrode for stimulating and recording.’,
Pacing Clin Electrophysiol, 14:1 (Jan 1991), pp.94–107.
Wässle, H., Peichl, L. and Boycott, B. B., ‘Morphology and topography of on- and off-alpha cells
in the cat retina.’, Proc R Soc Lond B Biol Sci, 212:1187 (May 1981), pp.157–175.
Yanai, Douglas etal., ‘Visual performance using a retinal prosthesis in three subjects with retinitis
pigmentosa.’, Am J Ophthalmol, 143:5 (May 2007), hURL: http://dx.doi.org/10.1016/
j.ajo.2007.01.027i, pp.820–827.
Ye, J. H. and Goo, Y. S., ‘The slow wave component of retinal activity in rd/rd mice recorded
with a multi-electrode array.’, Physiol Meas, 28:9 (Sep 2007), hURL: http://dx.doi.org/
10.1088/0967-3334/28/9/009i, pp.1079–1088.
157
Zaghloul, Kareem A, Boahen, Kwabena and Demb, Jonathan B, ‘Contrast adaptation in sub-
threshold and spiking responses of mammalian Y-type retinal ganglion cells.’, J Neu-
rosci, 25:4 (Jan 2005), hURL: http://dx.doi.org/10.1523/JNEUROSCI.2782-04.2005i,
pp.860–868.
Zaghloul, Kareem A etal., ‘Functional circuitry for peripheral suppression in Mammalian Y-type
retinal ganglion cells.’, J Neurophysiol, 97:6 (Jun 2007), hURL: http://dx.doi.org/10.
1152/jn.01091.2006i, pp.4327–4340.
Zhang, Yi etal., ‘Ectopic expression of multiple microbial rhodopsins restores ON and OFF light
responses in retinas with photoreceptor degeneration.’, J Neurosci, 29:29 (Jul 2009), hURL:
http://dx.doi.org/10.1523/JNEUROSCI.0184-09.2009i, pp.9186–9196.
Zrenner, E. etal., ‘The development of subretinal microphotodiodes for replacement of degener-
ated photoreceptors.’, Ophthalmic Res, 29:5 (1997), pp.269–280.
Zrenner, E. etal., ‘Can subretinal microphotodiodes successfully replace degenerated photorecep-
tors?’, Vision Res, 39:15 (Jul 1999), pp.2555–2567.
Zrenner, Eberhart, ‘Will retinal implants restore vision?’, Science, 295:5557 (Feb 2002), hURL:
http://dx.doi.org/10.1126/science.1067996i, pp.1022–1025.
Zrenner, Eberhart etal., ‘Subretinal electronic chips allow blind patients to read letters and com-
bine them to words.’, Proc Biol Sci, 278:1711 (May 2011), hURL: http://dx.doi.org/10.
1098/rspb.2010.1747i, pp.1489–1497.
158
Abstract (if available)
Abstract
Over the last 8 years it has been shown that an implantable retinal prosthesis can partially restore visual capabilities to blind humans. With current electrode arrays it is only possible to stimulate groups of cells rather than individual single cells with spatio-temporal precision. While artificial vision from a retinal prosthesis is unable to completely replicate the neural response patterns of normal vision, by stimulating groups of cells with electrodes patients see electrically elicited visual percepts. Thus this work will focus mainly on the form of percepts created with single electrodes in a prosthesis given certain varying stimulation parameters, and the development of a model to predict how stimuli applied produce certain percepts with single and paired prosthesis electrodes. ❧ Ideally we would want this model to resemble a digital display with independent pixels in order to individually control each percept seen by the subject in relation to the stimuli. We find however that the model that best fits the data does not resemble a digital display but instead the nerve fiber bundle trajectories in the human retina. ❧ The work presented here gives insight into the factors affecting form perception with a microelectronic retinal prosthesis. Specifically, by directly measuring the shapes of visual percepts from single and paired electrodes at different stimulation parameters, the building blocks of prosthetic vision are understood. The incentive is that we can use this information to develop a strategy that can control the percepts from each individual prosthesis electrode and piece them together in an organized way to best represent the visual world.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Prosthetic visual perception: retinal electrical stimulation in blind human patients
PDF
Manipulation of RGCs response using different stimulation strategies for retinal prosthesis
PDF
Towards a high resolution retinal implant
PDF
Understanding the degenerate retina's response to electrical stimulation: an in vitro approach
PDF
Effect of continuous electrical stimulation on retinal structure and function
PDF
Improving stimulation strategies for epiretinal prostheses
PDF
Saliency based image processing to aid retinal prosthesis recipients
PDF
Characterization of visual cortex function in late-blind individuals with retinitis pigmentosa and Argus II patients
PDF
Investigation of the electrode-tissue interface of retinal prostheses
PDF
Stimulation strategies to improve efficiency and temporal resolution of epiretinal prostheses
PDF
Viewing the picture we paint
PDF
Modeling retinal prosthesis mechanics
PDF
Ultrasound neuromodulation and its applications for noninvasive vision restoration
PDF
Intraocular and extraocular cameras for retinal prostheses: effects of foveation by means of visual prosthesis simulation
PDF
Synaptic second wind: phosphene fading of the Argus II and glutamate secretion in the inner retina
PDF
PET study of retinal prosthesis functionality
PDF
Cortical and subcortical responses to electrical stimulation of rat retina
PDF
Robot vision for the visually impaired
PDF
An in vitro model of a retinal prosthesis
PDF
Oxygen therapy for the treatment of retinal ischemia
Asset Metadata
Creator
Nanduri, Devyani
(author)
Core Title
Prosthetic vision in blind human patients: Predicting the percepts of epiretinal stimulation
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Biomedical Engineering
Degree Conferral Date
2011-12
Publication Date
12/06/2012
Defense Date
08/24/2011
Publisher
Los Angeles, California
(original),
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
artificial vision,form vision,neural prostheses,OAI-PMH Harvest,psychophysics,retinal prosthesis,retintis pigmentosa
Format
theses
(aat)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Weiland, James D. (
committee chair
), Fine, Ione (
committee member
), Humayun, Mark S. (
committee member
), Loeb, Gerald E. (
committee member
), Medioni, Gérard G. (
committee member
)
Creator Email
devyani.nanduri@gmail.com,nanduri@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC11292336
Unique identifier
UC11292336
Identifier
etd-NanduriDev-452.pdf (filename)
Legacy Identifier
etd-NanduriDev-452
Dmrecord
196626
Document Type
Dissertation
Format
theses (aat)
Rights
Nanduri, Devyani
Internet Media Type
application/pdf
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
artificial vision
form vision
neural prostheses
psychophysics
retinal prosthesis
retintis pigmentosa