University of Southern California Dissertations and Theses

Modeling retinal prosthesis mechanics

(USC Thesis Other)

Modeling retinal prosthesis mechanics

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2009 Brooke Christine Basinger

MODELING RETINAL PROSTHESIS MECHANICS

by
Brooke Christine Basinger

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

May 2009

ii
Acknowledgements

This work would not have been possible without the help and guidance of my
advisor, Dr. Jim Weiland. Dr. Adrian Rowley performed experimental measurements of
pressure under an array, contributed to tissue characterization tests, and offered insightful
and productive discussions throughout my time on this project. I’d also like to thank Dr.
Kinon Chen for his thorough work in mechanical tissue characterization, and Tim Nayar
for his help performing composite eye wall compression experiments.
My family and friends have supported me throughout this process. Particular
thanks are due to my rugby team, the Belmont Shore Women’s Rugby Club, which has
provided entertainment and friendship along with an athletic outlet for my occasional
intellectual frustrations. Additionally, I’d like to thank my parents, Tom and Carma
Basinger for their constant support. Without their encouragement and guidance I would
not have become the person I am today.

iii
Table of Contents

Acknowledgements ii

List of Tables vii

List of Figures viii

Abbreviations xiii

Abstract xiv

Chapter 1: Background and Motivation 1
1.1 Retinal Blindness 1
1.1.1 Degenerative Retinal Diseases 1
1.1.2 Treatment Options 2
1.2 Visual Prostheses 3
1.2.1 Cortical Stimulation 3
1.2.2 Retinal Stimulation 4
1.3 Retinal Prosthesis Mechanics 7
1.4 Problem Definition 11

Chapter 2: My Approach: Solid Modeling and Finite Element Analysis 12
2.1 Solid Modeling 12
2.2 Finite Element Analysis 14
2.3 Use in Product Development 16
2.4 Use in Biomechanics 17
2.5 Use in Modeling Retinal Prosthesis Mechanics 18

Chapter 3: Model Development 19
3.1 Analysis Type 20
3.1.1 Options 20
3.1.2 Discussion 20
3.1.3 My Choice 22
3.1.4 Advantages of this Choice 22
3.1.5 Model Limitations due to this Choice 23
3.2 Geometric Scope 23
3.2.1 Options 23
3.2.2 Discussion 24
3.2.3 My Choice 24
3.2.4 Advantages of this Choice 25
3.2.5 Model Limitations due to this Choice 25
3.3 Mesh Continuity 25
3.3.1 Options 25
iv
3.3.2 Discussion 25
3.3.3 My Choice 26
3.3.4 Advantages of this Choice 26
3.3.5 Model Limitations due to this Choice 26
3.4 Application of Tacking Force 26
3.4.1 Options 26
3.4.2 Discussion 27
3.4.3 My Choice 28
3.4.4 Advantages of this Choice 28
3.4.5 Model Limitations due to this Choice 28
3.5 Organization of the Eye Wall 29
3.5.1 Options 29
3.5.2 Discussion 29
3.5.3 My Choice 29
3.5.4 Advantages of this Choice 30
3.5.5 Model Limitations due to this Choice 30
3.6 Eye Wall Shape 30
3.6.1 Options 30
3.6.2 Discussion 30
3.6.3 My Choice 31
3.6.4 Advantages of this Choice 31
3.6.5 Model Limitations due to this Choice 31
3.7 Tissue Material Properties 32
3.7.1 Options 32
3.7.2 Discussion 32
3.7.3 My Choice 32
3.7.4 Advantages of this Choice 34
3.7.5 Model Limitations due to this Choice 34
3.8 Visualization of Results 34
3.8.1 Options 35
3.8.2 Discussion 36
3.8.3 My Choice 36
3.8.4 Advantages of this Choice 37
3.8.5 Model Limitations due to this Choice 37

Chapter 4: Final Model Description 39
4.1 Eye Wall 39
4.1.1 Porcine Eye Wall 40
4.1.1.1 Tissue Properties 40
4.1.1.2 Validation of Eye Wall Representation 44
4.1.1.3 Sensitivity 48
4.1.2 Human Eye Wall 50
4.2 Electrode Arrays 51
4.2.1 16-electrode Array 52
4.2.2 60-electrode Array 52
v
4.3 Constraints 53
4.3.1 Motion Constraints 54
4.3.2 Contact Constraints 55
4.4 Loading Conditions 55
4.4.1 Intraocular Pressure 56
4.4.2 Acute Tacking Force 56
4.4.3 Chronic Cable Forces 59
4.5 Mesh 61
4.6 FEA Solver 64

Chapter 5: Model Results 65
5.1 16-Electrode Array 66
5.1.1 Acute Tacking Force 66
5.1.2 Chronic Cable Force, Large Knee 67
5.1.3 Chronic Cable Force, No Knee 68
5.1.4 Chronic Cable Force, Large Knee with Twist 69
5.1.5 Chronic Cable Force, No Knee with Twist 70
5.2 60-Electrode Array 71
5.2.1 Acute Tacking Force 71
5.2.2 Chronic Cable Force, Large Knee 72
5.2.3 Chronic Cable Force, No Knee 73
5.2.4 Chronic Cable Force, Large Knee with Twist 74
5.2.5 Chronic Cable Force, No Knee with Twist 75

Chapter 6: Validation 76
6.1 Comparison to Eye Wall Compression 76
6.2 Comparison to Pressure Distribution Under an Array 77
6.2.1 Experimental Results 77
6.2.2 Model Predictions 81
6.3 Comparison to a Known Surgical Outcome 86

Chapter 7: Model Capabilities and Limitations 90
7.1 Capabilities 90
7.1.1 Specific Cases - Cut Cable 90
7.1.2 Evaluating Possible Design Features 93
7.1.2.1 Edge Features 94
7.1.2.2 Gap in Concentric Wings 94
7.1.3 Effect of Electrode Array Material Properties on the Retina 96
7.2 Limitations 97

Chapter 8: Conclusions and Future Work 100
8.1 Conclusions 100
8.1.1 Model Development 100
8.1.2 Results 101
8.1.3 Applications 102
vi
8.2 Future Work 103
8.2.1 Electrode Array Design 103
8.2.2 Increase Model Complexity 104
8.2.3 Determine Mechanism and Threshold for Mechanical Damage
to the Retina 104

Bibliography 105

Appendix A: Results from early flat tissue model 109

Appendix B: Results generated using literature tissue properties 110

Appendix C: Experimental evaluation of edge features 122

vii
List of Tables
Table 3.1 Summary of major choices made in model development. 20

Table 4.1 Mean and standard deviation of porcine ocular tissue properties
measured by Chen and Rowley. 42

Table 4.2 Mean and standard deviation of orbital fat properties measured
by K. Chen. 42

Table 4.3 Mean and standard deviation of porcine ocular tissue thickness
measured by Chen and Rowley. 43

Table 4.4 Tissue properties applied to model representation of a porcine
eye wall. 44

Table 4.5 Mean and standard deviation of human ocular tissue properties
measured by Chen and Rowley. 50

Table 4.6 Mean and standard deviation of human ocular tissue thickness
measured by Chen and Rowley. 51

Table 4.7 Tissue properties applied to model representation of a porcine
eye wall. 51

Table 4.8 Summary of five loading configurations representing acute tacking
and four chronic cable configurations. 56

Table 6.1 Pressure under an array measured experimentally by A. Rowley
in six approximate locations due to the weight of the array and
electrode array tacking. 80

viii
List of Figures
Figure 1.1 Illustration of epiretinal and subretinal approaches to electrical
stimulation of retinal cells. 5

Figure 1.2 Illustration of epiretinal prosthesis system. 6

Figure 1.3 Fundus photo of a 16-electrode array implanted in a human subject. 7

Figure 1.4 Illustration of a 16-electrode array, inserted through a sclerotomy,
being secured in place with a retinal tack. 8

Figure 1.5 Histological evidence of retina layer disruption due to focal
mechanical pressure. (Colodetti 2008) 9

Figure 1.6 Optical coherence tomography image showing an electrode array
whose edge has caused displacement of the retina. Courtesy of
H. Ameri. 10

Figure 3.1 Typical stress-strain curve for steel. 21

Figure 3.2 Model of entire human right eye with surgical markers noted and
eyewall portion intended for FEA. 24

Figure 3.3 Isometric view of an early model using a flat, layered eye wall (left)
and a cross-sectional view of a later model using a 4-layer, curved
eye wall with a curved electrode array (right). 31

Figure 3.4 Bose 3100 ELF uniaxial mechanical characterization machine with
temperature-controlled submersion chamber. 33

Figure 3.5 Four possible outcome plots for a FEA study. 35

Figure 3.6 A typical visualization of model results, showing the top view of
Von Mises stress using increasingly narrowed scales. 37

Figure 4.1 Cross-sectional view of 4-layer curved eye wall model. 40

Figure 4.2 The posterior region of an enucleated porcine eye wall, with Vertical
and Horizontal sample orientations marked. Courtesy of Chen
and Rowley. 41

Figure 4.3 A typical stress-strain curve for porcine choroid subjected to uniaxial
tensile testing. 41

ix
Figure 4.4 Schematic representation of experimental eye wall compression
test setup. 45

Figure 4.5 Cross-sectional view of modeled eye wall compression. 46

Figure 4.6 Comparison of modeled and experimental results for a composite
eye wall compression test. 47

Figure 4.7 Sensitivity of model results to tissue modulus. 49

Figure 4.8 Top view of a fabricated 16-electrode array meant for implantation
into a human or canine subject (left). Top view of my model
representation of the same electrode array (right). 52

Figure 4.9 Top view of 60-electrode array model. 53

Figure 4.10 Several views of final human eye wall model with a 16-electrode
array. The location of motion constraints, contact constraints, and
load sites are marked. 54

Figure 4.11 Screenshot of force recorded during automated tack insertion. 57

Figure 4.12 Maximum force required to insert a retinal tack into porcine eye
wall using either automated or manual insertion. 58

Figure 4.13 Tacking force increases with tack usage. 58

Figure 4.14 Cable model used to define chronic cable force boundary conditions
for a Large Knee and No Knee, both with and without a Twist in
the cable. 60

Figure 4.15 Von Mises stress at a single point (blue line) and numerical degrees
of freedom (green line) as mesh density increases. 62

Figure 4.16 Tetrahedral mesh with global element size = 0.75 mm and mesh
controls applied to the retina, choroid and orbital fat. 63

Figure 4.17 Element aspect ratios plotted for a mesh with global
element size = 0.75 mm. 64

Figure 5.1 Tacking force applied to a 16-electrode array. 66

Figure 5.2 Chronic cable force for a large knee applied to a 16-electrode array. 67

Figure 5.3 Chronic cable force with no knee applied to a 16-electrode array. 68
x

Figure 5.4 Chronic cable force with a large knee and a twist in the cable applied
to a 16-electrode array. 69

Figure 5.5 Chronic cable force with no knee and a twist in the cable applied to
a 16-electrode array. 70

Figure 5.6 Tacking force applied to a 60-electrode array. 71

Figure 5.7 Chronic cable force for a large knee applied to a 60-electrode array. 72

Figure 5.8 Chronic cable force with no knee applied to a 60-electrode array. 73

Figure 5.9 Chronic cable force with a large knee and a twist in the cable applied
to a 60-electrode array. 74

Figure 5.10 Chronic cable force with no knee and a twist in the cable applied
to a 60-electrode array. 75

Figure 6.1 Photograph of an enucleated porcine eye cup with a hole in the sclera
and choroid layers created using a trephine. 78

Figure 6.2 Screenshot of downward force (green line) recorded by the Bose
ELF 3100 load cell as an electrode array was tacked in place over
the force measurement site. 79

Figure 6.3 Photo of a 16-electrode array with approximate force measurement
sites marked for Mid-Heel, Mid-Mid, Mid-Toe, Left-Heel, Left-Mid
and Left-Toe. 80

Figure 6.4 Experimental measurements of pressure (mm Hg) at six locations under
a 16 electrode array during tacking overlaid on modeled prediction for
16-electrode array tacking stress distribution. 81

Figure 6.4 Top view of model with 16-electrode array, 6 force measurement
locations marked, and a hole in the sclera and choroid with a load cell
post visible in the Mid-Mid location. 82

Figure 6.5 Cross-section view of a 16-electrode array placed over a
force-measurement post in the Mid-Mid position. 82

Figure 6.6 Comparison of experimental results and model predictions for
pressure under an electrode array in six locations during tacking. 83

xi
Figure 6.7 Comparison of experimental results and model predictions for
pressure under an electrode array in six locations due solely to
the weight of the array. 85

Figure 6.8 A Fundus photo showing retinal damage caused by a 16-electrode
array chronically implanted in a canine model (left). Model
predictions of Von Mises stress distribution on the retina for a
16-electrode array subjected to tacking (top) and a cable with a
large knee (bottom) (right). 87

Figure 6.9 Shear stress in the XZ plane (the plane of the retina) caused by tacking
force applied to a 16-electrode array along the angled tacking axis. 88

Figure 6.10 Comparison of retinal damage in a surgical case with predicted Von
Mises stress due to tacking, shear stress due to tacking, and Von
Mises stress due to a cable with a large knee. 89

Figure 7.1 Fundus photo of a 16-electrode array that pivoted from its original
orientation (horizontal) to a new orientation (vertical) when the
ribbon cable anchoring it in place was cut. 91

Figure 7.2 Intensity of histological retinal damage mapped to the original and
eventual position of an electrode array that pivoted when the ribbon
cable anchoring it in place was cut. Courtesy of Y. Morales. 92

Figure 7.3 Modeled predictions of deformation, Von Mises stress, and shear stress
in three planes for a 16-electrode array rotated from one position to
another. 93

Figure 7.4 A proposed wide-field electrode array with concentric “wings”. 94

Figure 7.5 Top view of concentric electrode array profile when curved
(transparent image) and when flat (solid). Inset shows the
curved array being flattened. 95

Figure 7.6 Top view of Von Mises stress as a concentric electrode array is
subjected to a tacking force. 96

Figure 7.7 Top view of Von Mises stress for a 16-electrode array subjected to
tacking force and to a chronic cable force as array elastic modulus
is varied. 97

Figure C.1 Photograph and diagram of experimental setup. 123

xii
Figure C.2 Downward view of experimental setup. 123

Figure C.3 Raw and processed data for a representative force recording. 125

Figure C.4. Three designs tested dry and after soaking. 126

Figure C.5. Average slope for three designs tested dry and after soaking. 127

Figure C.6. One representative force recording for 14 different designs. 128

Figure C.7. Average slope for 14 different designs. 129
xiii
Abbreviations
AMD – age-related macular degeneration
CAD – computer aided drafting, later computer aided design
CAE – computer aided engineering
CAM – computer aided manufacturing
FEA – finite element analysis
RP – retinitis pigmentosa
xiv
Abstract

Degenerative retinal diseases such as Retinitis Pigmentosa and Age-Related
Macular Degeneration are, together, one of the leading causes of blindness in the United
States. Retinal prostheses bypass the degenerated cells and apply electrical stimulation
directly to the visual neural pathway, creating an artificial sensation of sight in subjects
with retinal blindness. Current prostheses utilize a polymer electrode array, which is
affixed to the retina using a retinal tack. The electrode array applies mechanical pressure
to the retina, which can cause mechanical damage to the very cells which are required to
transmit the applied signal. We can mitigate or relocate mechanical damage to the retina
through electrode array design changes, but to do so, we must thoroughly understand the
mechanical interface between the array and the retina.
I have generated a computerized model of the mechanical interaction of array and
eye wall using solid modeling and Finite Element Analysis techniques. I describe and
justify the choices made in the development of this model, define model inputs, provide
results from ten studies, and validate the accuracy of model results through comparison to
experimental data. This work is significant in that it combines a modifiable device and a
complex series of biomaterials in the same model, demonstrating that it is feasible to
satisfy the sometimes conflicting goals of product design and biomaterials modeling
simultaneously. This model can now be used by retinal prosthesis developers to optimize
electrode array design, reduce mechanical damage to the retina, and improve device
efficacy.

1
Chapter 1

Background and Motivation

1.1 Retinal Blindness

The retina is a complex, laminar organization of neural tissue lining the inside of
the eye. When light strikes the retina, a series of chemical and electrical events produces
a neural signal that is transmitted via the optic nerve to visual centers in the brain,
producing the sensation of sight. (webvision.med.utah.edu 2009)

1.1.1 Degenerative Retinal Diseases
Although disruptions can occur at any point in the visual pathway, degenerative
retinal diseases including Retinitis Pigmentosa (RP) and Age-Related Macular
Degeneration (AMD) together are among the leading causes of blindness in the United
States. (Klein 1997, Margalit 2003) These diseases cause a gradual, but generally
irreversible, deterioration of retinal cells that can eventually result in a complete loss of
natural vision.
AMD, which affects primarily the central portion of the visual field, can be
separated into two broad types: dry AMD and wet AMD. Dry AMD, which constitutes
80% of all cases, is caused by deposits and hypertropic and atropic changes in the retinal
pigment epithelium. Wet AMD is caused by neovascularization, which can cause rapid
vision loss as the fragile vessels near the macula rupture and leak. (Rakoczy 2006)
2
According to a 2000 U.S. census, AMD affects an estimated 1.75 million people over 40
and that population is expected to grow to 3 million people by the year 2020. (Klein
1997, Friedman 2004)
RP is a family of genetic disorders in which abnormalities in the photoreceptors
or retinal pigment epithelium lead to progressive vision loss across the entire visual field.
Approximately 1 in 4,000 people in the United States are affected by RP. (Berson 1993)

1.1.2 Treatment Options
Recently, progress has been made in a drug treatment for wet AMD (Genentech
2006). However, no treatment is currently available for RP or the more common dry
AMD. Several strategies for restoring vision to affected patients are currently under
investigation. Retinal fetal sheet transplantation attempts to replace a degenerate
photoreceptor layer, but its success is limited by the number of useful synaptic
connections that are established between the donor and host tissue. (Lund 2001) Gene
therapy has been shown to be effective in treating animal models of Leber’s disease,
which causes retinal ganglion cell and axonal degeneration, but each genetic mutation of
the disease requires an individualized therapy. (Preeising 2004) The expression of
channelrhodopsin-2 in ganglion cells in order to make those cells light sensitive is also
being investigated and has been successful in vivo in degenerate model rats, but the
channel is expressed using a viral vector which may introduce complications and the
cell’s low sensitivity requires an external system to amplify light. (Bi 2006) While these
research directions are all promising in some respects, each has disadvantages or
3
complicating factors which must be overcome before a clinical treatment option can be
established and approved.

1.2 Visual Prostheses

Visual prostheses rely on the principle of electrical activation of nerves, which is
central to other successful neural stimulators such as cochlear implants and deep brain
stimulators. In the case of visual prostheses, electrical stimulation is applied to different
points along the visual pathway within the central nervous system in an attempt to create
the sensation of visual perception in sight impaired patients.

1.2.1 Cortical Stimulation
Electrical stimulation of the cortex was first shown to produce visual percepts in
1929, when Forester exposed the occipital pole of a subject with normal vision, applied
an electrical stimulus, and produced local percepts termed phosphenes that were spatially
correlated to the visual field. (Forester 1929) Krause and Schum performed a similar
experiment in 1932 using a subject who had been blind for 8 years as a result of injury. It
was not known whether the cortex could create visual percepts after being deprived of
visual stimuli for an extended period of time, but Krause and Schum were also successful
in producing phosphenes. (Krause 1931) Almost 40 years later, Brindley and Lewin
were the first to use an implanted array of multiple individually-addressed surface
4
electrodes to stimulate the primary visual cortex, producing the first chronically
implanted visual prosthesis. (Brindley 1968)
Although surface stimulation provided significant early results, later attempts to
develop a cortical visual prosthesis have turned to intracortical stimulation using
penetrating probe electrodes. (Dobelle 1976, Schmidt 1992, Normann 1999)
Intracortical electrodes are able to generate phosphenes with a lower level of electrical
stimulation (Dobelle 1976) and can produce a higher spatial resolution of individually
distinguishable phosphenes. (Schmidt 1996) Important advances have been made in
cortical stimulation, but the difficulties of surgical implantation and the complexity of the
visual cortex have slowed the development of an FDA approved cortical visual
prosthesis.

1.2.2 Retinal Stimulation
In order to avoid some of these difficulties, and with the thought that intervening
as early as possible in the visual pathway would take advantage of some natural neural
coding and processing, researchers have begun developing retinal visual prostheses.
(Humayun 1996, Chow 1998, Zrenner 1999, Rizzo 2001) Although degenerative retinal
diseases cause considerable photoreceptor cell death, some viable inner retinal cells and
ganglion cells have been shown to survive (Stone 1992, Santos 1997, Kim 2002) and
direct stimulation of those cells has been shown to elicit phosphenes (Humayun 1996,
Humayun 1999).
Retinal stimulation has been attempted using two approaches: subretinal and
epiretinal, which are illustrated in Figure 1.1. In the subretinal approach, an active or
5
passive device is placed between the outer and inner retina layers. Zrenner et al have
implanted an active subretinal device with 1500 pixels with light controlled stimulation
and 16 hardwires electrodes accessible via a percutaneous connector. In early trials with
this device, RP subjects were able to distinguish between horizontal and vertical bars, but
long term efficacy has yet to be established. (Zrenner 1999, Kuttenkeuler 2006) Another
group has developed subretinal silicone arrays with 5000 microelectrode-tipped
microphotodiodes that use the photoelectric effect to convert incident photons to a
stimulating pulse. RP subjects with this device reported some improvements in their
detection of brightness, contrast and shape, but percepts were often not spatially
correlated with array placement and it is still uncertain whether the limited current
induced by this device can depolarize a cell. (Chow 1998, Chow 2004)

Figure 1.1 Illustration of epiretinal and subretinal approaches to electrical stimulation of retinal cells.

Epiretinal prostheses locate stimulating electrodes on the ganglion cell side of the
retina. Compared to subretinal devices, epiretinal prostheses have had a longer history of
successful chronic implantation and partial restoration of visual function. Humayun et al
6
have developed an epiretinal prosthesis that applies electrical stimuli to the retina through
a polymer microelectrode array affixed to the epiretinal surface with a retinal tack. This
system is illustrated in Figure 1.2. Six RP subjects implanted with a 16 electrode version
of the device have reported visual percepts that are spatially correlated with electrode
location and have been able to perform simple visual tasks. (Humayun 2003, Yanai
2007) Recently, an updated version with 60 electrodes has been approved for clinical
trials and implanted in human subjects. (Humayun 2009) Simulations performed with
normal-sighted subjects suggest that an increased number of electrodes would provide
subjects with greater spatial resolution and may eventually allow them to recognize faces
and read large text. (Hayes 2003)

Figure 1.2 Illustration of epiretinal prosthesis system.

7
1.3 Retinal Prosthesis Mechanics

The work in this thesis focuses on the intraocular mechanics of the epiretinal
prosthesis developed by Humayun et al. (Humayun 2003, Humayun 2009) In this
system, an image is captured by a head-mounted video camera, processed by an external
video processing unit, and transmitted inductively to a stimulation pulse generator
implanted either behind the ear or on an ocular belt. Stimulation pulses are passed via a
trans-scleral cable to a polymer microelectrode array affixed to the retina by means of a
retinal tack.
A typical epiretinal prothesis electrode array is essentially rectangular in shape,
with a multiwire cable attached to one side. A tack hole is located at the heel of the
array, near the cable insertion point, with metalized electrodes located near the center and
tip of the array. The shape of a 16-electrode array implanted in a human subject is shown
in Figure 1.3.

Figure 1.3 Fundus photo of a 16-electrode array implanted in a human subject.

The flexible, polymer electrode array of this epiretinal prosthesis is inserted into
the eye via a pars plana sclerotomy. The array is positioned over the macular region and
8
a retinal tack is inserted to anchor the array in place as shown in Figure 1.4. Retinal tacks
were originally developed to treat retinal detachments and have been adapted for the
fixation of retinal prostheses.

Figure 1.4 Illustration of a 16-electrode array, inserted through a sclerotomy, being secured in place with a
retinal tack.

The tip and shaft of the tack pass through a tack hole in the array and pierce the
retina, choroid and sclera. The tip of the tack anchors itself outside of the sclera, securing
the tack in place. Although this attachment method works relatively well to secure the
electrode array in place, (Majii 1999) it causes a clear insult to the tissue at the tack site
that may cause a fibrovascular reaction (Majii 1999) or retinal fold (Walter 1999). In
addition, portions of the electrode array may come into contact with, and apply pressure
to, other areas of the retina during the process of tacking.
Chronic pressure on the array may be caused by the tack head, which prevents the
array from sliding off the tack shaft and, in some configurations, applies constant
downward pressure on the array at the tack site. Chronic forces may also be transmitted
9
to the array through the cable which connects it to control hardware located outside of the
eye and is sutured in place in a chronic trans-scleral incision. (The intraocular portion of
the prosthesis is therefore fixed in place at two locations: the cable insertion site and the
tack hole.)
The retina is generally thought to be a very delicate tissue and retinal surgeons are
taught that any accidental instrument touch to the retina can cause mechanical damage,
possibly even a retinal tear that can lead to detachment. Colodetti et al. showed that an
unquantified amount of mechanical pressure from a probe held just in contact with a rat
retina for less than one hour caused significant disruption of the retinal layers. Although
his work focused on electrical damage to the retina, he found that retinal damage was
present when the retina was directly contacted by a probe electrode. (Colodetti 2007)

Figure 1.5 Histological evidence of retina layer disruption due to focal mechanical pressure. (Colodetti
2007)

Optical coherence tomography images of implanted electrode arrays in canine
subjects have shown retinal disruption associated with the presence of the array. In
Figure 1.6, the electrode array (the thin, white line on the right side of the OCT image),
has created a retinal fold where its edge contacts the retina. (Ameri, unpublished data)

10

Figure 1.6 Optical coherence tomography image showing an electrode array whose edge has caused
displacement of the retina. Courtesy of H. Ameri.

Long-term retinal implant studies have shown that damage to the retina occurs
where significant pressure is exerted by the stimulating array. (Majii 1999, Walter 1999)
Although the mechanism and threshold for mechanical retinal damage are not known, it
is presumed that mechanical pressure on the retina may injure the retinal ganglion cells
that are required to generate a neural impulse and transmit it to the optic nerve (i.e. those
cells immediately under the array and in the areas between the array and the optic nerve).
Any mechanical damage to the retina may, therefore, significantly interfere with the
efficacy of an epiretinal prosthesis.
Although it is important to prevent mechanical damage to the retina caused by
pressure from the electrode array, it is important to simultaneously achieve a near-
uniform close proximity between the array and the retina in order to improve electrical
coupling. The stimulation intensity required to elicit a percept from an epiretinal
electrode has been found to increase with distance from the retina. (deBalthasar 2008,
McMahon 2006) In addition, electric field theory suggests that the stimulus current from
an electrode further from the retina will activate a larger retinal area and thus result in a
more diffuse perception.
In order to make the most efficient use of limited power resources and to provide
the highest acuity vision, it is important that the retinal prosthesis electrode array be
11
positioned in close proximity to the retina without applying enough pressure to cause
mechanical damage. This balance is becoming more difficult to achieve as larger, higher
density electrode arrays are developed, electronic control components are moved near or
even inside the eye, and new materials are introduced to allow large-scale fabrication.

1.4 Problem Definition

Mechanical stress on the retina could be mitigated or redistributed away from
critical regions through mechanical design changes, stress relief features, or material
changes. However, early retinal prosthesis electrode array designs were based primarily
on electrical considerations, fabrication feasibility and surgical preferences. It is only
now that new iterations of the device have been suspected of causing mechanical damage
to the retina in animal models and an effort has been made to reconsider electrode array
design from a mechanical point of view. Thus far, any mechanically-motivated changes
in electrode array design have been guided by intuition and trial-and-error.
In order to improve the mechanical design of increasingly complex electrode
arrays and prevent mechanical damage to the retina while still securing the array in close
proximity, researchers needed a means of systematically evaluating the mechanical
behavior of a retinal prosthesis electrode array and its interaction with the retina.

12
Chapter 2
2 test
My Approach: Solid Modeling and Finite Element Analysis

2.1 Solid Modeling

In the 1950s and 1960s, the earliest computer programs meant to sketch or draft a
graphic representation of two dimensional engineering designs were being developed at
universities (particularly MIT’s Mathematical Laboratory, now the Department of
Computer Science) and government agencies. These early attempts laid the foundation
for what would become the fields of Computer-Aided Drafting
*
(CAD), Computer-Aided
Manufacturing (CAM) and Computer-Aided Engineering (CAE). Automobile and
aerospace manufacturers were among the first to see potential applications for CAD
techniques and many began internal development of proprietary programs to automate
repetitive drafting chores. By the start of the 1970s, several CAD software companies
had been established and 2D CAD began to migrate from research into commercial use.
These programs were still 2D replacements for drafting, but they helped manufacturers
reduce drawing errors and increased the reusability of drawings.
Meanwhile, research and commercial interest in 3D CAD software was gaining
momentum. Although 3D modeling was still too computer intensive for most practical
applications, research done in complex 3D curve and surface modeling at that time
formed the basis of the 3D modeling programs used today. One of the most influential

*
CAD has since come to mean Computer-Aided Design
13
pieces of research was Vesprille’s 1975 Syracuse University Ph.D. dissertation,
Computer-Aided Design Applications of the B-Spline Approximation Form, which
provides the foundational mathematical approach for representing 3D curves in CAD
software. (Vesprille 1975) In addition, the two different approaches to solid modeling
that form the basis for modern software packages were being developed by Charles
Lang’s group at Cambridge University and Herb Voelcker’s group at the University of
Rochester. Lang’s group developed the application of boundary representation data
structures (which had previously been used in finite element analysis applications) in
solid modeling, while Voelcker focused on constructive solid geometry methods. The
two approaches were fundamentally different, but each had its advantages and most
modern CAD software is now based on one or the other.
As computers became faster, more powerful and more common through the
1980s, both 2D and 3D CAD became more practical and more accessible. Incremental
improvements in software functionality were made as computing power progressed, but
models, once developed, were still very difficult or impossible to change. In 1989,
Parametric Technology Corp. launched Pro/ENGINEER, a UNIX-based parametric
feature-based solid modeler that allowed users to change the dimensions of individual
geometric features. Pro/ENGINEER drastically improved the ease of use and speed of
solid modeling, and has been one of the most commonly used modeling programs ever
since.
In 1995, Solidworks was released and quickly replicated Pro/ENGINEER’s
success in a more moderate price bracket. SolidWorks uses a similar parametric feature-
based approach on a Windows-based system and was soon known for excellent surface
14
modeling capabilities and a superior user interface. Several other 2D and 3D CAD
software packages are available, and there is a wide range of application-specific
programs, but Pro/ENGINEER and SolidWorks are still two of the most well-known 3D
solid modeling programs available. Both have implemented incremental improvements
over the years, but have retained the same fundamental parametric feature-based
approach since their introduction. That approach allows users to generate virtual
prototypes and explore design options without the expense or difficulty of generating and
testing physical prototypes.

2.2 Finite Element Analysis

Finite element analysis (FEA) (sometimes referred to as the finite element
method) is a numerical technique of solving field problems described by a set of partial
differential equations (PDEs). It is based on the work of Richard Courant, who
approximated the solution of second order elliptical partial differential equations
describing the torsion of a cylinder by discretizing the domain into finite triangular
subregions and reducing the governing PDE to a solvable ordinary differential equation
for each subregion. (Courant 1942) Combined, the solutions within each discretized
region approximate the global solution. Strang and Fix’s An Analysis of the Finite
Element Method provided a rigorous mathematical foundation for the method, and
readers interested in a derivation of the mathematics of FEA are referred to that work.
(Strang 1973)
15
Initially, FEA was used to solve complex elasticity and structural analysis
problems in civil and aerospace engineering, but the method has since been generalized
to provide numerical models of mechanical, electrical, fluid and thermal systems.
Because FEA requires the simultaneous solution of large numbers of ordinary differential
equations, its practicality is often limited by computing power. But as computers have
become larger and faster, the size and scope of the problems that can be solved by FEA
has increased dramatically.
To apply FEA to a problem, the user must generate a 2D or 3D geometric model
of the component to be analyzed (often from a CAD or solid modeling program), apply
material properties to each component of the model, and define all boundary conditions.
A meshing program then discretizes the model into finite elements with nodes at their
corners. Elements can be concentrated in areas of interest to provide a more detailed and
accurate solution in those areas. Each node has up to six degrees of freedom (3
translational, 3 rotational), some of which may be eliminated by the user-defined
boundary conditions. The user then decides what type of analysis to use (most FEA
software packages offer several types of linear analysis and a few offer nonlinear
analysis) and runs the FEA software. The software solves thousands of simultaneous
differential equations to obtain a displacement at each node, which can then be
extrapolated into a variety of different outcome variables including (for a mechanical
problem) strain, stress and pressure. A postprocessor is then used to digest the data and
display it superimposed over the model geometry.
The accuracy of a FEA solution depends, in part, on the mesh applied. A coarse
mesh will provide a good qualitative description of system behavior, but quantitative
16
results are likely inaccurate. As mesh density is increased, quantitative accuracy will
increase and results will approach an asymptote at the true numerical solution of the
problem. Creating and using a mesh with infinitely small nodes is impractical, but a
reasonable level of quantitative accuracy can be ensured by increasing mesh density only
until the mesh “converges” and further increases in mesh density produce only
insignificant changes in model results.
The accuracy of a FEA solution also depends on the accuracy and appropriateness
of the user-defined geometric model, material properties and boundary conditions as well
as the type of FEA equations applied. Complex model geometries must often be
defeatured and simplified before FEA analysis can be successfully applied. Other
simplifications in material properties and boundary conditions are often necessary as well
in order to make FEA analysis of a problem feasible within limited resources.
Simplifications must be chosen carefully to aid in solvability without distorting the true
behavior of the system. The relevance and viability of a simplified FEA model is best
evaluated by comparing the model’s behavior to known numerical solutions or
experimental data for a well-defined problem. Only after model behavior is determined
to be reasonable and relevant for problems with a known solution should it be applied to
more novel uses.

2.3 Use in Product Development

Solid modeling is now heavily used in industrial product design and development,
particularly in the automotive, aerospace and consumer product design industries. It
17
allows everyone involved in the design process (including the customer) to visualize the
end product. Solid modeling can shorten the design cycle, streamline the manufacturing
process and accelerate product introductions by improving the sharing of product design
information and communication throughout an organization and among suppliers and
customers, all of which can increase a manufacturer’s competitiveness.
In some cases, FEA is also included in the design process in order to validate a
virtual prototype or compare the behavior of several design options without having to
build and test a physical prototype.
The focus of solid modeling and FEA use in product development applications is
geometry; the components being analyzed may be changed at any time and users must be
able to easily and quickly modify components and institute design changes. The
materials used are generally well established engineering materials whose properties are
well-understood. Therefore, complex nonlinear FEA capabilities are generally
unnecessary and feature-based solid modeling programs with strong user interfaces and
excellent rendering abilities are generally chosen for these applications.

2.4 Use in Biomechanics

In recent years, solid modeling and FEA have also been used to better understand
biomechanical systems and analyze biological activities like cardiac contractions and
gait-related bone stress. In these applications, the geometry of the system is, if not
simple, at least static. Researchers are generally not interested in redesigning the heart;
rather, they are interested in accurately representing the behavior of a complex, non-
18
homogenous and often non-linear tissue. In this application, then, the usability and
changeability of feature-based solid modeling programs is generally unnecessary, and
advanced FEA programs that combine a complex, sometimes-cellular level,
representation of biological tissues with strong meshing capabilities and multiple analysis
options are generally preferred.

2.5 Use in Modeling Retinal Prosthesis Mechanics

In order to model a retinal prosthesis electrode array and its interaction with the
eye wall, we must combine the easily modifiable geometry of a feature-based solid
modeling program with a sufficiently accurate representation of complex and poorly
understood biological tissues.
There is little precedent for combining a device and a tissue in the same model
and analyzing the interactions between them. Commercially available solid modeling
and FEA programs are generally strong either in geometry rendering and modification or
in complex material representation, but not both. For our application, is important that
the software package chosen integrate well with industry partners who will eventually
fabricate commercial retinal prostheses and that the model is easily used and modified by
future users who may have little or no training in complex FEA techniques. We therefore
chose to use SolidWorks, which provides good surface rendering and an easy user
interface, and COSMOSWorks, a FEA pre-processing, meshing, analysis and post-
processing package that was developed specifically for user-friendly integration into
SolidWorks.
19
Chapter 3
3 test
Model Development

Modeling an in vivo system using a software package intended for engineering
design applications requires careful thought. A model should closely replicate the true
geometry and properties of the system in order to provide the most relevant results, but
such a replication is generally too large and complex to realistically solve within the
constraints of time and computing power. Simplifications and approximations are
necessary to make modeling feasible; the modeler’s challenge is to choose simplifications
that increase the practicality of a model while maintaining the relevance of its results.
This chapter discusses the choices I made in developing a model of the pressures applied
to the retina by a retinal prosthesis electrode array. A summary of those choices is
presented below in Table 3.1.

20

My Choice Alternatives Reasoning
Analysis Type Linear Nonlinear
Simplicity, ease of use,
accessibility to others.
Geometric
Scope
Eye wall segment Entire eye
Reduce model size and
computation time.
Mesh
Continuity
Single mesh with
compatible elements
at part boundaries
Unique meshes for each
individual part.
Simplify contact
constraints, prevent
crashing.
Application of
Tacking Force
Directly to electrode
array surface
Including a retinal tack in
the model
Avoids chaotic results
near tack tip.
Organization of
the Eye Wall
Four layers
A single layer, or further
subdivision
Balances accuracy and
solvability.
Eye Wall Shape Curved Flat Most realistic.
Tissue Material
Properties
Measured
experimentally
Literature values
Most accurate and
complete
Visualization of
Results
Von Mises stress
Principle stress, strain,
deformation,
displacement, resultant
force, shear stress, etc.
Intuitive overview of
results
Table 3.1 Summary of major choices made in model development.

3.1 Analysis Type

3.1.1 Options
FEA solution algorithms can broadly be categorized into linear and nonlinear
analysis approaches.

3.1.2 Discussion
In linear analysis, materials are assumed to behave linearly in accordance with
Hooke’s Law. Most engineering materials maintain a linear relationship between stress
21
(applied force over a given area) and strain (change in length, or deformation) until they
reach the material’s yield point, after which its shape and properties are permanently
altered. Linear FEA is appropriate so long as the modeled materials remain within the
elastic, or linear, region of their stress-strain curve. This is generally assumed to be true
when deformations remain small in comparison to the size of the structure.

Figure 3.1 Typical stress-strain curve for steel.

Once a material is stressed past its yield point and begins plastic deformation, the
relationship between stress and strain is no longer linear, and nonlinear FEA is required
to fully understand the part’s behavior. Most nonlinear FEA algorithms approach the
problem in a series of time steps, solving it as a linear problem with only a small portion
of the defined load applied, then updating material properties and solving again with
slightly more of the load applied. This continues, with material properties updated at
each step, until the full load has been applied. One nonlinear FEA solution therefore
requires a long series of individual linear solutions, which is extremely slow and
22
computationally expensive. Nonlinear FEA is a powerful tool for analyzing plastic
deformations, fracture, and changing loads, but it is not always practical or necessary.
The material properties of biological tissues are not often well established and can
vary significantly from tissue to tissue. Many biomaterials are thought to have several
separate and distinguishable linear or near-linear regions and often do not have a clearly
defined yield point. Linear analysis can be accurately used if the material remains within
one of those linear or near-linear regions, but nonlinear analysis is needed to fully capture
the behavior of ocular tissues across the entire range of their stress-strain curves.

3.1.3 My Choice
I chose to develop this model using linear Finite Element Analysis algorithms.

3.1.4 Advantages of this Choice
Linear FEA reduces the computational resources necessary to model this system.
It eases the model development process and allows adjustments and corrections to the
model to be made and tested within a reasonable time frame. Linear FEA also makes this
model more intuitive and accessible to other retinal prosthesis researchers, who should be
able to use this model to quickly and easily design and evaluate multiple electrode array
design options without having to develop an expertise in FEA first. Also, this model
developed using linear analysis techniques provides a foundation for later expansion to
non-linear analysis.

23
3.1.5 Model Limitations due to this Choice
Typically, linear analysis is said to limit accuracy to situations with small
deformations, but this rule of thumb is slightly misleading. Linear analysis limits model
accuracy to situations where materials remain within a linear portion of their stress-strain
curve which, for engineering materials, is true when deformations are small. For
biomaterials, there may be several linear or near-linear regions within which linear
analysis is appropriate.
However, the use of linear FEA algorithms here limits the accuracy of my model
to material behavior within the single region used to define that material’s elastic
modulus. If that modulus is defined using a low-strain region, this model will
underestimate the material’s stiffness for larger deformations. If that modulus is defined
using a high-strain region, this model will overestimate the material’s stiffness for small
deformations. The choice to use linear analysis makes it extremely important to choose
elastic modulus values appropriate for the deformation range expected in a given case.

3.2 Geometric Scope

3.2.1 Options
For this project, I can model the entire eye and all intraocular device components
or reduce the model to include only a portion of the posterior eye wall and the electrode
array.

24
3.2.2 Discussion
Modeling the entire eye provides broader insight into effects that an intraocular
device may have on all parts of the eye. However, the retina immediately surrounding
the electrode array is the area whose health is most important to electrode array
designers. Modeling the entire eye increases the number of necessary nodes, elements
and computational time, spending valuable resources on areas that are not considered to
be a priority for this project.

3.2.3 My Choice
I did generate a model of the entire human eye purely to locate the physiological
and surgical markers important in electrode array placement, but I used only a portion of
the eye wall for all finite element analysis studies.

Figure 3.2 Model of entire human right eye with surgical markers noted and eyewall portion
intended for FEA.

25
3.2.4 Advantages of this Choice
The primary advantage of reducing the model to a portion of the eye wall is a
significant reduction in computation time.

3.2.5 Model Limitations due to this Choice
By modeling only a portion of the eye wall and electrode array, I introduce
boundaries at the cut edges of the eye wall and electrode array cable that can affect
results if not defined appropriately.

3.3 Mesh Continuity

3.3.1 Options
This model includes 5-10 individual parts or assemblies. The COSMOSWorks
meshing algorithm can apply a separate, unique mesh to each individual part, or it can
mesh all parts together in a single mesh with compatible elements at part boundaries.

3.3.2 Discussion
When a separate mesh is applied to each individual part, we can capture the
movement or sliding of one part against another. However, it is difficult to define how
one mesh interacts with another. Often, for models with several independent parts such
as this one, FEA algorithms cannot resolve multiple meshes with one another, fail to
converge on a solution, and crash.

26
3.3.3 My Choice
I have meshed this model using a single mesh with compatible elements at part
boundaries.

3.3.4 Advantages of this Choice
Applying a single mesh to this problem allows FEA studies to reliably converge
on a solution without risk of crashing.

3.3.5 Model Limitations due to this Choice
Meshing all parts together means that this model does not capture the effects of an
electrode array sliding against the retina well. So we must limit the application of this
model to loading conditions where significant sliding is not expected.
Additionally, in order to mesh all parts together, contact between parts must occur
over a relatively large surface area, rather than at single points or edges which would
cause numerical singularities. For my purposes, that means that electrode arrays must
match the curvature of the retina along part, or all, of the bottom surface of the array.

3.4 Application of Tacking Force

3.4.1 Options
During the insertion of a retinal tack, the surgeon applies downward force to the
retinal tack and the tack then pierces the eye wall and applies downward force to the
electrode array immediately under the tack head. I could include a retinal tack in this
27
model and apply tacking force to it, or I could eliminate the tack from the model and
apply tacking force directly to the electrode array.

3.4.2 Discussion
I believe that the acute force applied to an electrode array and the eye wall during
the initial insertion of the retinal tack, which secures the array in place, is very large
compared to chronic forces that may be applied afterwards. It is possible that some, or
even all, retinal damage occurs during the tacking process, so I felt it was important to
model a loading condition representative of tack insertion.
The tissue pierced by the tack is stressed well beyond its yield point and its
behavior during fracture cannot be accurately predicted using linear analysis techniques.
However, we don’t need a model to know that the tissue immediately underneath the tack
is stressed to the point of mechanical damage. That damage is unavoidable unless
another attachment method entirely is used to secure the electrode array in place. For the
purposes of electrode array design, it is more important that we understand smaller
stresses caused by contact between the electrode array and retina away from the tack site,
which may damage retinal cells that we hoped to leave intact. If a retinal tack were
included in this model, those smaller stresses would be masked by the large, chaotic and
inaccurate mesh deformations occurring around the tack tip.
Because the deformation of the metal tack is negligible compared to the
deformation of the polymer electrode array and ocular tissues, force applied to the tack
head can be estimated to transfer directly to the electrode array surface underneath it with
little change in magnitude or direction. We can therefore apply a tacking force directly to
28
the surface of the electrode array in the area contacted by the tack head, avoiding chaotic
model behavior where the tack tip pierces the eye wall without altering the intent of the
model. By applying a tacking force directly to the electrode array, we are most closely
modeling the instant when the tack has already pierced the eye wall, but the surgeon has
not yet release pressure on the insertion tool.

3.4.3 My Choice
I have eliminated the retinal tack from this model and instead applied tacking
forces directly to the surface of the electrode array underneath the tack head.

3.4.4 Advantages of this Choice
Applying tacking force directly to the electrode array eliminates potentially
chaotic model behavior at the point where the tack tip pierces the eye wall and instead
focuses on smaller stresses which occur where the electrode array contacts the retina
away from the tack site. Additionally, eliminating the tack from the model simplifies
model geometry and contact constraints considerably and facilitates the creation of the
single mesh discussed in section 3.3.

3.4.5 Model Limitations due to this Choice
With the retinal tack eliminated, this model does not capture the tissue rupture or
associated stresses at the point where the tack tip pierces the eye wall.

29
3.5 Organization of the Eye Wall

3.5.1 Options
The eye wall in vivo is comprised of three tissues: the retina, the choroid and the
sclera. Each tissue is made up of individual cells, which form a non-homogenous and
possibly anisotropic substance. I could model the entire eye wall as a single material or I
could subdivide it by layer or, even further, by cell or cell layer.

3.5.2 Discussion
It is important to provide enough detail and realism in the eye wall model to
adequately capture its behavior without creating a model that is too complex to solve.
Additionally, the material properties of each modeled part must be defined; subdivision
of the eye wall into smaller components is useful only if we know the material properties
for each component.

3.5.3 My Choice
I have modeled the eye wall as a composite of three individual homogeneous,
isotropic layers representing retina, choroid and sclera. Although it is not part of the eye
wall, I have also chosen to include a layer of orbital fat. I have not subdivided these
tissues down to a cellular level.

30
3.5.4 Advantages of this Choice
Modeling each tissue layer as a separate part provides a more realistic
representation than if I treated the entire eye wall as a single part, while maintaining a
solvable level of complexity. It allows me to define material properties independently for
each layer, and it allows me to visualize results specifically within the retina. The
inclusion of an orbital fat layer provides realistic cushioning to the model and provides
additional distance between the area of interest (the retina) and the model boundaries
(and any associated boundary effects).

3.5.5 Model Limitations due to this Choice
Modeling each layer of the eye wall as an individual part without further
subdivision treats each tissue as a homogenous material with aggregate isotropic
properties. It does not capture tissue anisotropy or effects due to non-homogenous
elements including blood vessels, Muller cells, or collagen fibers.

3.6 Eye Wall Shape

3.6.1 Options
The layered eye wall can be modeled as a flat or curved substrate.

3.6.2 Discussion
Modeling the eye wall as a flat substrate simplifies the contact constraints
between the electrode array and the retina considerably and facilitates the creation of the
31
single mesh discussed in section 3.3. Modeling the eye wall as a curved substrate is
clearly more realistic, though it requires careful arrangement of array-retina contact.

3.6.3 My Choice
In early development of this model, I used a flat representation of the eye wall in
order to simplify the problem. Select results generated using that model are presented in
Appendix A. Once initial challenges were solved, I moved to a curved eye wall model
with a radius of curvature of 12 mm, that of the average adult human eye. (Bron)

Figure 3.3 Isometric view of an early model using a flat, layered eye wall (left) and a cross-sectional view
of a later model using a 4-layer, curved eye wall with a curved electrode array (right).

3.6.4 Advantages of this Choice
Initial use of a flat model allowed me to understand and overcome some of the
challenges in modeling this problem. However, the eventual use of a curved eye wall
model was essential to the end relevance of my work.

3.6.5 Model Limitations due to this Choice
The use of a curved eye wall model complicates the contact constraints between
the electrode array and retina. The curvature of those parts must match along some or all
32
of their contacting surfaces to allow the FEA algorithm to converge on a solution without
singularities.

3.7 Tissue Material Properties

3.7.1 Options
To define the material properties of each tissue layer for the purposes of FEA
studies, the user must define mass density, Poisson’s ratio and elastic modulus. I could
define those values using estimates, data presented by others in previous publications, or
experimental testing.

3.7.2 Discussion
Most soft tissues contain more than 70% water, so they are often assumed to have
the mass density and Poisson’s ratio of water (Reichenblach 1991). No guidelines for
estimating elastic modulus values of biomaterials exist. Although a number of studies
have performed mechanical experiments on the retina, choroid and sclera, the reported
data is limited in scope and often contradictory. None of these studies tested tissues
within a body-temperature saline environment, which is most closely representative of
the environment inside the eye. However, experimental biomaterials testing, particularly
of ultrasoft biomaterials like retina, is extremely complex and requires specialized
equipment and knowledge.

3.7.3 My Choice
33
For this project, I have assumed all biological tissues to have the mass density of
water, 1 g/cm
3
and to be nearly incompressible (with a Poisson’s ratio of 0.49). Both are
common assumptions for soft biomaterials. (Reichenblach 1991)
In initial studies, I defined an elastic modulus value for each tissue using literature
values. Results generated using those values are presented in Appendix B. In order to
obtain the most relevant and accurate elastic modulus values, it was necessary to measure
each tissue’s elastic modulus experimentally. I worked with Dr. Adrian Rowley to select
and purchase appropriate equipment (a Bose 3100 ELF uniaxial mechanical
characterization machine equipped with a temperature-controlled submersion chamber,
shown in Figure 3.4).

Figure 3.4 Bose 3100 ELF uniaxial mechanical characterization machine with temperature-controlled
submersion chamber.

Drs. Kinon Chen and Adrian Rowley performed tensile testing of porcine and
human retina, choroid and sclera in room-temperature air, room-temperature saline and
body-temperature saline environments. A discussion of how I applied their results to my
model is presented in section 4.1.
34
3.7.4 Advantages of this Choice
The choice to measure elastic modulus values experimentally provided accurate
values for both porcine and human tissues in several environments, which had not yet
been attempted and could not be learned from literature. I discuss the sensitivity of
model results to errors in the elastic modulus value for each tissue in section 4.1.1.3. The
model is most sensitive to changes in retina and choroid values; Chen and Rowley’s work
in characterizing those tissues, in particular, was extremely important in creating an
accurate model.

3.7.5 Model Limitations due to this Choice
Characterizing ocular biomaterials is a significant undertaking that constituted a
separate dissertation in and of itself. That work was performed by Chen and Rowley,
who are referenced in the text where appropriate. I used literature modulus values
throughout the development of this model and only after Chen and Rowley completed
their study was I able to integrate experimentally measured tissue properties. Although
the model development process was limited somewhat by the choice to measure elastic
modulus values experimentally, model accuracy and relevance were certainly improved
with that choice.

3.8 Visualization of Results

I present model results for two electrode array designs subjected to 5 different
loading conditions in Chapter 5. Here I discuss how I chose to visualize those results.
35
3.8.1 Options
Finite element analysis algorithms calculate deformation at each node point.
From that, a variety of results can be calculated and displayed, including deformed shape,
contact pressure, principle stress (in 3 directions), shear stress (in 3 directions), Von
Mises stress, strain, and strain energy density. Each of these results can be viewed from
any angle. Several possible outcome variables are shown from different angles in Figure
3.5.

Figure 3.5 Four possible outcome plots for a FEA study. An isometric view of model deformation is
shown in the top left panel. A top view of contact pressure is shown in the top right panel. A cross-
sectional side view of strain and Von Mises stress are presented in the bottom left and right panels,
respectively.

In addition, the information presented depends on the scale applied to the plot.
For a 16-electrode array subjected to a tacking force, the largest stresses and strains occur
in the electrode array itself, but I am most interested in the smaller stresses and strains in
36
the retina layer. SolidWorks does not provide a logarithmic plotting option, so results
must be presented in several different scales to fully visualize both the behavior of the
electrode array and the behavior of the retina.

3.8.2 Discussion
No single outcome variable or view encapsulates all of the results of a FEA study.
Indeed, we do not know which outcome variables are related to mechanical damage of
the retina. The capability of SolidWorks to display a wide variety of outcome variables
offers us a great deal of flexibility in presenting model results. But to directly compare
multiple design options, I must visualize model results in a consistent manner that
provides an intuitive understanding of the total stress distribution in the retina.

3.8.3 My Choice
Although I sometimes include additional views or outcome variables for some
cases, I have chosen to consistently present results as a series of top-down views of Von
Mises stress with increasingly narrowed scales as shown in Figure 3.6 in order to allow
direct comparison between the results of different studies. I remove the electrode array
from the image in the final frame in order to provide a clear view of stress distribution in
the retina. Von Mises stress is a measure that accounts for all six stress components in a
three-dimensional system.

37
[0-0.02MPa] [0-0.2MPa] [0-2MPa] [0-20MPa]
[0 – 0.02MPa]
a
b
c
d
e

Figure 3.6 A typical visualization of model results, showing the top view of Von Mises stress using
increasingly narrowed scales.

3.8.4 Advantages of this Choice
Von Mises stress provides an intuitive understanding of the magnitude of total
stress in all directions. Presenting it along several scale ranges provides an understanding
of both the entire mechanical system and the specific stresses in the retina. A top down
view with the electrode array removed from the image provides the clearest view of stress
distribution in the retina.

3.8.5 Model Limitations due to this Choice
Von Mises stress does not capture all of the information produced by a FEA
study. Additionally, we do not know which, if any, outcome variable is the direct cause
of mechanical damage to the retina. Model results visualized as described here provide
38
an overview of stress distribution on the retina but should not be thought of as a direct
map of retinal damage.

39
Chapter 4
4 test
Final Model Description

Although this model has gone through several stages in the development process,
here I describe the input for the final version of this model and demonstrate how that
input was defined. Tissue property testing was carried out by others and their
contributions are noted in the text. A summary of their experimental method and relevant
data is included for completeness.
When this model has been used to predict the outcome of benchtop experiments, I
have altered the geometry of the model to match the experimental conditions (as in
section 4.1.1.2 and section 6.2.1).

4.1 Eye Wall

In general, my model of the posterior eye wall includes four curved, isotropic,
homogenous layers representing the retina, choroid, sclera and orbital fat.

40

Figure 4.1 Cross-sectional view of 4-layer curved eye wall model.

Human eye samples can be difficult to obtain, so more readily available porcine
eyes are often used as a substitute for experimental purposes. I have therefore generated
both porcine and human eye wall models.

4.1.1 Porcine Eye Wall
4.1.1.1 Tissue Properties
Chen and Rowley have performed experimental uniaxial tensile testing to
determine the properties of porcine ocular tissues. (Chen 2008) Rectangular tissue strips
were cut from enucleated porcine eyes. Samples were cut in two directions, defined as
Vertical and Horizontal, and subjected to tensile testing to yield while submerged in
room-temperature air, room-temperature saline and body-temperature saline
environments. The resulting stress-strain curves could be separated into two distinct
regions prior to tissue rupture, low strain and high strain, with a separate elastic modulus
reported for each region.
Sclera
Orbital Fat
Choroid
Retina
41

Figure 4.2 The posterior region of an enucleated porcine eye wall, with Vertical and Horizontal sample
orientations marked. Courtesy of Chen and Rowley.

Figure 4.3 A typical stress-strain curve for porcine choroid subjected to uniaxial tensile testing.
Two near-linear regions are identified and a separate linear elastic modulus value is defined for each.
Courtesy of K. Chen.

Relevant results of their work for porcine tissues are repeated in Table 4.1, with N
= 10 in each group.

Strain (mm/mm)
Stress (kPa)
Low-strain
High-strain
42

Table 4.1 Mean and standard deviation of porcine ocular tissue properties
measured by Chen and Rowley. N = 10. *25º C saline

A pilot study of porcine orbital fat was completed by Kinon Chen in a body-
temperature saline environment using the same methods, with N = 4 for each group.
Porcine Orbital Fat

Vertical Horizontal
37º C saline
Yield stress (kPa) 17.0 ± 21.7 5.4 ± 2.5
Yield strain (mm/mm) 0.015 ± 0.001 0.015 ± 0.001
Low modulus (kPa) 29.6 ± 37.6 17.8 ± 8.5
High modulus (kPa) 1703.7 ± 2131.4 552.3 ± 266.7
Table 4.2 Mean and standard deviation of orbital fat properties measured by K. Chen. N = 4

Because tissues in this model are considered to be isotropic, I have defined each
modeled tissue’s elastic modulus to be the average of the Vertical and Horizontal
orientation moduli reported by Chen and Rowley. I have maintained a distinction
Porcine Retina Porcine Choroid Porcine Sclera

Vertical Horizontal Vertical Horizontal Vertical Horizontal
25º C air
Yield stress
(kPa)
1.5 ± 1.3 1.3 ± 0.8 423 ± 246 626 ± 259
2630 ±
1203
3231 ±
1677
Yield strain
(mm/mm)
0.023 ±
0.010
0.028 ±
0.008
0.669 ±
0.116
0.746 ±
0.182
0.681 ±
0.188
0.724 ±
0.208
Low modulus
(kPa)
1.9 ± 1.7* 0.8 ± 1.7* 13 ± 14 22 ± 19 51 ± 53 70 ± 103
High modulus
(kPa)
82.3 ± 73.8 62.6 ± 48.6
1532 ±
1142
2179 ±
1026
7766 ±
3470
8829 ±
3823
37º C saline
Yield stress
(kPa)
0.8 ± 0.4 0.4 ± 0.3 568 ± 357 540 ± 229
1350 ±
886
964 ±
608
Yield strain
(mm/mm)
0.033 ±
0.003
0.034 ±
0.004
0.523 ±
0.175
0.547 ±
0.059
0.452 ±
0.160
0.409 ±
0.109
Low modulus
(kPa)
0.7 ± 0.5 0.3 ± 0.3 15 ± 22 14 ± 22 21 ± 35 16 ± 8
High modulus
(kPa)
39.2 ± 18.8 19.5 ± 12.2
2222 ±
1440
2542 ±
1462
5744 ±
3088
4358 ±
3232
43
between room-temperature air and body-temperature saline categories in order to apply
those moduli which best replicate live conditions for a given case. I have compared
model results using both Low modulus and High modulus values to experimental data as
described below in order to determine which is most appropriate for this project.
In the course of experimental tensile tissue testing, Chen and Rowley also
measured the thickness of each retina, choroid and sclera sample using an optical method.
Tissue thicknesses were measured for several experimental groups, so N = 60 in each
case. I have used the reported mean thickness of each tissue to define the geometry of
modeled tissue layers.
Porcine Retina Porcine Choroid Porcine Sclera
Thickness (mm) 0.21 ± 0.05 0.21 ± 0.04 1.19 ± 0.22
Table 4.3 Mean and standard deviation of porcine ocular tissue thickness measured by Chen and
Rowley. Thickness was measured optically prior to tensile testing. N = 60.

Although orbital fat generally does not appear as a single, smooth layer with a
consistent thickness in vivo, I have defined the thickness of an orbital fat layer to be 3.0
mm simply to provide sufficient distance between boundary conditions on the outside
surface of that layer and the area of interest, the retina. For the purposes of modeling, all
tissues were assumed to have the mass density of water, 1 g/cm3, and to be nearly
incompressible, with a Poisson’s ratio of 0.49. Both are common assumptions for soft
tissues in the field of biomaterials. (Reichenblach 1991)
The properties I ascribed to each tissue included in the porcine eye wall model are
summarized in Table 4.4, with assumed quantities shown in shaded cells and measured
quantities in unshaded cells.
44
Porcine
Retina
Porcine
Choroid
Porcine
Sclera
Porcine
Orbital Fat
Thickness (mm) 0.21 0.21 1.19 3.0
25º C
air
Low modulus
(Pa)
17500 60500 23700*
High modulus
(Pa)
72450 1855500 8297500 1128000*
37º C
saline
Low modulus
(Pa)
500 14500 18500 23700
High modulus
(Pa)
29000 2382000 5051000 1128000

Poisson’s ratio
0.49 0.49 0.49 0.49
Mass density
(g/cm³)
1 1 1 1
Table 4.4 Tissue properties applied to model representation of a porcine eye wall. Experimentally
measured quantities are shown in white cells with assumed quantities in shaded cells. * 37ºC saline data

4.1.1.2 Validation of Eye Wall Representation
The relevance of the porcine eye wall model described above can be validated
through comparisons to a benchtop composite eye wall compression test. Enucleated
porcine eyes were dissected and a rectangular sample (~16mm square) of posterior eye
wall including retina, choroid and sclera was taken. Samples were cut to exclude the
optic nerve region. Each sample was laid sclera side down on a glass slide and secured in
place with a small dot of commercial adhesive. Although this adhesive may affect the
results of any tissue mechanics experiment, it is thought that the movement of an
unsecured sample would add significantly more error.

45

Figure 4.4 Schematic representation of experimental eye wall compression test setup.

I placed glass slides with the tissue sample attached on top of a 50 g load cell in a
room-temperature air environment. Compression was applied using a cylindrical steel
probe, 1.0 mm in diameter, which was controlled by a BOSE ElectroForce 3100
mechanical characterization apparatus. I lowered the probe a total of 0.50 mm from its
initial position, at a rate of 0.1 mm/sec while force was recorded by the load cell. The
probe was initially placed very close to the retinal surface, but not in contact with it.
During later data analysis, the moment of contact was identified by a change in force-
displacement slope. Downward force was then recorded at several depths from the initial
contact point.
The porcine eye wall model was modified to represent the physical arrangement
of this experiment: the orbital fat layer was removed from the model, a central region of
the sclera’s exterior surface was fixed in all transitional and rotational directions
(simulating fixing the sample with an adhesive), and a proscribed downward
Load cell
Adhesive
Glass Slide
Eye wall
Probe
46
displacement was applied to the retina in a circular space with 1 mm diameter to cause
compression in the multi-layer tissue representation. The vertical reaction force along the
outside surface of the sclera (where the load cell was located in benchtop experiments)
was recorded. Because the experiment was conducted in a room-temperature air
environment, model tissue properties were defined using 25º C air data for this case.

Figure 4.5 Cross-sectional view of modeled eye wall compression.

Experimental compression test results are shown in Figure 4.6 by a dotted line,
with each data point representing the mean force recorded at a given indentation depth
over 10 samples and error bars representing standard deviation. This data shows a non-
linear trend similar in shape to the stress-strain curves reported for individual ocular
tissues by Chen and Rowley. In the field of nano-indentation testing, it is thought that
indentations greater than 10% of a sample’s overall thickness may be affected by
substrate effects (Doerner 1986). Here, the mean sample thickness was measured (using
calipers) to be 1.45 mm, so indentations of more than 0.145 mm are expected to have
Displacement applied
Vertical reaction force
recorded
47
(and are seen here to have) more variability and a higher slope than measurements at
smaller depths.

Figure 4.6 Comparison of modeled and experimental results for a composite eye wall compression test.

This composite compression experiment was modeled using both low modulus
and high modulus data. Modeled results are shown in the above plot by a solid colored
line, with results generated using low-strain modulus data in green, and results generated
using high-strain data in red. Modeled results are expected to be (and are) linear because
they have been generated using linear Finite Element Analysis algorithms that, by
definition, assume the stress-strain relationship of all materials to be linear.
Model results relying on low strain elastic modulus data (the green line in Figure
4.6) clearly do not approximate experimental results successfully in the displacement
x Experimental mean
Model – high strain moduli
Model – low strain moduli
48
range shown. It is likely that those results would match experimental data more closely
as displacement depth (and thus strain) approached zero.
Model results relying on high strain elastic modulus data (the red line in Figure
4.6) do match experimental results well within the displacement range shown. This
model slightly overestimates Von Mises stress for displacements under 0.1 mm but
otherwise produces results very similar to those obtained experimentally.
The similarity of experimental and modeled results here suggests that modeling
the eye wall using an arrangement of 4 isotropic, homogenous layers with elastic moduli
defined by the high-strain region of uniaxial tensile tests does produce an accurate and
relevant representation of the eye wall for our purpose in cases where non-zero
displacement occurs.

4.1.1.3 Sensitivity
The sensitivity of model results to tissue properties was demonstrated by running
the model of our composite eye wall compression test multiple times while varying the
applied elastic modulus of a given tissue from 1% - 10,000% of its original value. For
each test, a compression of 0.10 mm was applied, and the maximum Von Mises stress in
the model was recorded.

49

Figure 4.7 Sensitivity of model results to tissue modulus.

A steep line in such a sensitivity analysis indicates that a model’s results are very
sensitive to changes or errors in a given variable, while a shallow line indicates that the
model is not sensitive to changes or errors in that variable. Here, the model is most
sensitive to the elastic moduli of retina and choroid and is not sensitive to the moduli of
sclera and orbital fat. This suggests that, for the whole model to be accurate, it is most
important to accurately define the material properties of retina and sclera. The thorough
material property testing provided by Chen and Rowley for these two tissues, in
particular, is therefore critical in generating an accurate model of retinal prosthesis
mechanics.

50
4.1.2 Human Eye Wall
Chen and Rowley have also conducted experimental tensile testing of human
ocular tissues using the same methodology described in Section 4.1.1. (Rowley 2009)
Their relevant results for human retina, choroid and sclera tested in a body-temperature
saline environment are reproduced here in Table 4.5. Room-temperature air data is not
presented because I have not used the model to represent a human eye in a room-
temperature air environment and I do not expect such a model to be necessary in the
future.

Human Retina Human Choroid Human Sclera

Vertical Horizontal Vertical Horizontal Vertical Horizontal
37º C saline
Yield stress
(kPa)
0.4 ± 0.4 0.3 ± 0.2 96 ± 50 98 ± 56 1248 ± 814 1047 ± 872
Yield strain
(mm/mm)
0.040 ±
0.005
0.035 ±
0.003
0.474 ±
0.205
0.468 ±
0.176
0.524 ±
0.122
0.502 ±
0.154
Low modulus
(kPa)
0.3 ± 0.6 0.2 ± 0.2 4 ± 4 8 ± 13 24 ± 22 33 ± 29
High modulus
(kPa)
18.6 ± 16.3 12.5 ± 10.1 387 ± 127 362 ± 111
4395 ±
2720
4470 ±
4181
Table 4.5 Mean and standard deviation of human ocular tissue properties
measured by Chen and Rowley. N = 10.

I have again used the average of the Vertical and Horizontal modulus numbers
reported by Chen and Rowley for body-temperature saline environments to define the
elastic modulus of each tissue for this model. I have again maintained distinct Low and
High modulus categories in order to allow each to be applied separately for appropriate
cases. The mean thickness of the human tissue samples tested by Chen and Rowley is
shown in Table 4.6.

51
Human Retina Human Choroid Human Sclera
Thickness (mm) 0.28 ± 0.06 0.21 ± 0.03 0.97 ± 0.18
Table 4.6 Mean and standard deviation of human ocular tissue thickness measured by Chen and
Rowley. Thickness was measured optically prior to tensile testing. N = 60.

I have again assumed to have the mass density of water, 1 g/cm3, and to be nearly
incompressible, with a Poisson’s ratio of 0.49. (Reichenblach 1991) Chen and Rowley
did not test human ocular fat tissues, so here I apply the previously measured properties
of porcine orbital fat to human orbital fat. The properties ascribed to each tissue included
in the human eye wall model are summarized in Table 4.7, with assumed quantities
shown in shaded cells and measured quantities in unshaded cells.
Human
Retina
Human
Choroid
Human
Sclera
Human
Orbital Fat

Thickness (mm)
0.28 0.21 0.97 3.0
37º C
saline
Low modulus
(Pa)
250 6000 28500 23700*
High modulus
(Pa)
31100 374500 4432500 1128000*

Poisson’s ratio
0.49 0.49 0.49 0.49
Mass density
(g/cm³)
1 1 1 1
Table 4.7 Tissue properties applied to model representation of a porcine eye way. Experimentally
measured quantities are shown in white cells with assumed quantities in shaded cells. * porcine tissue data

4.2 Electrode Arrays

The overall goal of this project is to allow a computer based evaluation of a wide
variety of electrode array designs. However, any modifications or changes should be
guided by an analysis of currently used electrode arrays. Additionally, any model should
first be shown to accurately describe the behavior of existing devices before it is used to
52
explore novel ideas. I have therefore focused my work on the two specific electrode
array designs that are currently used in human subjects.

4.2.1 16-electrode Array
This 16-electrode array is a slightly simplified version of a first generation device
designed, manufactured, and implanted in both canine and human subjects by Second
Sight Medical Products. The array is approximately 5.5 mm by 7.0 mm and 0.5 mm
thick. It is pre-curved to more closely match the curvature of the retina. The array is
fabricated using a silicone formulation that we have tested (using uniaxial tensile testing
in a body-temperature saline environment) to have an elastic modulus of 21549 kPa. The
fabricated electrode array includes a thin layer of metalized lines and metalized disc
electrodes, but those components have been eliminated in the model for the sake of
simplicity.

Figure 4.8 Top view of a fabricated 16-electrode array meant for implantation into a human or canine
subject (left). Top view of my model representation of the same electrode array (right).

4.2.2 60-electrode Array
This 60-electrode array is a slightly simplified version of a later generation device
designed, manufactured, and implanted in both canine and human subjects by Second
53
Sight Medical Products. The array is approximately 5mm by 9mm and 0.25 mm thick. It
is pre-curved to more closely match the curvature of the retina. The array is fabricated
using a core of polyimide that we have tested (using uniaxial tensile testing in a body-
temperature saline environment) to have an elastic modulus of 439610 kPa surrounded by
a cushioning layer of the same silicone described above. This design iteration is
described as the “bathtub” design by Second Sight engineers. I have again eliminated
any metalized lines or disc electrodes from the array model for the sake of simplicity.

Figure 4.9 Top view of 60-electrode array model.

4.3 Constraints

To allow the finite element analysis algorithm to run, the user must constrain their
model in space in order to prevent rigid body motion and must define how parts interact
with one another. The location of motion constraints, contact constraints, and applied
loading conditions are shown in Figure 4.10.

54

Figure 4.10 Several views of final human eye wall model with a 16-electrode array. The location of
motion constraints, contact constraints, and load sites are marked.

4.3.1 Motion Constraints
The exterior surface of the orbital fat layer is fixed in all three translational and all
three rotational directions to prevent rigid body motion. The interior surface of the
electrode array tack hole is fixed using a Cylindrical Surface constraint to prevent
rotation while still allowing translation.

Fixed: orbital
fat surface
Fixed in rotational
direction: tack hole
Contact: no penetration
Tacking axis
Cable
55
4.3.2 Contact Constraints
The global contact for this model is defined to be Bonded (where touching
surfaces behave as though they are fixed together) using the Compatible Mesh option.
All tissue layers are therefore treated as though they are bonded together.
The boundary between the bottom surface of the electrode array and the retina is
defined as No Penetration (node to surface), which allows the parts to move away from or
against one another without intersecting. It is important to apply this contact constraint to
all fillets and surfaces on the electrode array that may come into contact with the retina
during analysis.

4.4 Loading Conditions

In general, there are three sources of loads: intraocular pressure, acute or chronic
pressure from a retinal tack, and forces transmitted to the array through the ribbon cable.
I have combined these loads to create 5 loading conditions representing acute tacking and
chronic forces for four different cable configurations. A summary of these loading
configurations is presented in Table 4.8.

56

Intraocular
Pressure
Acute or
Chronic Tack
Force
Chronic Cable Forces
Loading
Configuration
Array and
retina surface
Tack axis (30º
from vertical in x
and y directions)
Cut boundary of ribbon cable
Right Side Left Side
Tacking 2000 N/m² 1.25 N none
Large Knee 2000 N/m² 0.25 N
1 N Compressive
1 N Inward
No Knee 2000 N/m² 0.25 N
0.5 N Tension
0.1 N Inward
Large Knee +
Twist
2000 N/m² 0.25 N
0.5 N Compressive
0.5 N Inward
1.5 N Compressive
1.5 N Inward
No Knee + Twist 2000 N/m² 0.25N
1.0 N Tension
0.1 N Outward
0 N Tension
0.2 N Outward
Table 4.8 Summary of five loading configurations representing acute tacking and four chronic cable
configurations.

4.4.1 Intraocular Pressure
Normal intra-ocular pressure is 15 mmHg (2000 N/m²). For models representing
in vivo implants in human subjects, radial intra-ocular pressure is applied normal to the
top surface of the electrode array and retina.

4.4.2 Acute Tacking Force
The force used by the surgeon to insert a retinal tack is applied directly to the
surface of the electrode array tack well, where a ring has been marked and defined for
that purpose. Tacking force is applied along an axis 30º from vertical in both the x and y
directions, mimicking the approach angle taken during surgical implantation of the array.
I have measured tacking force magnitude using the Bose 3100 ELF system. The
vitreous was removed from enucleated porcine eyes and a large posterior sample of eye
wall was located above a load cell. The load cell recorded downward force either as a
57
surgeon inserted a retinal tack into the tissue sample using the same tack insertion tool
used in surgery or as the Bose was used to drive the tack downward at a constant speed.

Figure 4.11 Screenshot of force (blue line) recorded during automated tack insertion.

Initially, automated tack insertion appeared to require less downward force than
manual tack insertion. However, retinal tacks are difficult to obtain, and all of the tack
insertion force measurements shown below were conducted using the same tack.
Automated trials were conducted first, while the tack was still relatively sharp. Manual
trials were conducted only after the tack tip had been dulled by multiple uses and
therefore required higher forces to insert regardless of insertion method.

58

Figure 4.12 Maximum force required to insert a retinal tack into porcine eye wall using either automated or
manual insertion.

Conducting multiple manual insertion trials with the same tack confirms that
required tacking force increases with tack usage.

Figure 4.13 Tacking force increases with tack usage.

59
Retinal tacks used to secure retinal prosthesis electrode arrays in canine or human
subjects are always new. For the purposes of this model, I have therefore defined tacking
force magnitude to be that required to manually insert a tack on its first usage: 1.25 N.

4.4.3 Chronic Cable Forces
Chronic forces are transmitted to the electrode array through the attached ribbon
cable, which is sutured in place at the sclerotomy insertion point. The cable is therefore
fixed in place in two locations: at the insertion site and at the retinal tack site. If, when
the device is originally implanted, the cable is bent or twisted in any way between those
two points, the cable will apply chronic forces to the electrode array. Any number of
cable loading configurations is possible in vivo, and we cannot define a single definitive
chronic loading configuration. Instead, I have defined four configurations representing
possible cable geometries that could occur in vivo:
Large Knee
No Knee
Large Knee with Twist
No Knee with Twist
When extra cable length is introduced into the eye between the retinal tack and
fixed sclerotomy, the cable bends inward toward the center of the eye, creating a “knee”.
If the sclerotomy is not perfectly aligned with the retinal tack, the cable will twist in
between the fixed sclerotomy and the tacking point.
I have used COSMOSWorks to perform FEA studies on a model of an electrode
array cable and have replicated Large Knee and No Knee cases, both without and with a
60
Twist between the sclerotomy and the retinal tack site. Side and front views of those four
cases are shown in Figure 4.14.

Figure 4.14 Cable model used to define chronic cable force boundary conditions for a Large Knee
and No Knee, both with and without a Twist in the cable. Cables without Twists are shown in a
side view, cables with Twists are shown in a front view.

For each cable model, I recorded the stresses on the edge of the cable and used
that information to estimate boundary conditions to apply to the cut face of the electrode
array in my primary model. I applied tangential forces to the electrode array face either
Inward (towards the center of the eye) or Outward (away from the center of the eye) and
applied normal forces to the electrode array face to simulate either Compression or
Tension in the cable. Those forces are listed for each loading configuration in Table 4.8.
No Twist Twist
Large Knee No Knee
61
4.5 Mesh

I have used COSMOSWork’s Standard Mesher to generate a solid mesh using
high quality tetrahedral elements. Mesh controls are applied to increase mesh density in
the retina and choroid and to decrease mesh density in the orbital fat layer.
Typically, as mesh density increases (global element size decreases), model
results asymptotically approach the true numerical solution of the problem. However, as
mesh density increases, the number of nodes, elements and degrees of computational
freedom (and therefore the computation time) increase exponentially. Even on large,
modern, fast computers, complex three dimensional models with dense meshes can take
up to twelve hours to run.
The Von Mises stress at a specific point (the center of the retinal surface)
resulting from a tacking force applied to a 16-electrode array is shown in Figure 4.15 for
several mesh element sizes. Note that, as the absolute value of global element size
decreases, mesh density increases and Von Mises stress approaches an asymptote as
expected. I have also recorded the number of numerical degrees of freedom for each
mesh.
62

Figure 4.15 Von Mises stress at a single point (blue line) and numerical degrees of freedom (green line) as
mesh density increases.

At a global element size of 0.75 mm (circled in Figure 4.14), the solution is
approaching the asymptotic solution but the number of degrees of freedom (and thus
computation time) is still relatively small. I have therefore chosen this mesh size for
continued studies. This mesh uses 74215 total nodes and 50199 total elements.
63

Figure 4.16 Tetrahedral mesh with global element size = 0.75 mm and mesh controls applied to the retina,
choroid and orbital fat.

Mesh elements provide the most accurate solution when their aspect ratio is close
to 1. In this mesh, 94.9% of elements have an aspect ratio < 3, while 0.337% have an
aspect ratio > 10. The maximum aspect ratio is 30.731, located along a thin fillet edge on
the top surface of the electrode array. A top view of element aspect ratios for this mesh is
shown in Figure 4.17.

64

Figure 4.17 Element aspect ratios plotted for a mesh with global element size = 0.75 mm.

In this type of plot, we want the model to appear primarily blue (aspect ratios near
1). Here, there are only small deviations from that, with the largest deviations occurring
along the upper surface of the electrode array, where a sharp edge is hard to mesh
accurately. This is not the electrode array surface in contact with the retina; the lower
surface of the electrode array is curved with no sharp edges and all mesh elements on that
surface have aspect ratios near 1. This mesh is sufficiently dense and well-formed to
provide accurate results for this model.

4.6 FEA Solver

To produce the results presented in Chapter 5, I have used COMOSWork’s Linear
FFEPlus solver with no special effects, stabilizers or mesh adapters applied.
65
Chapter 5
5 5. placeholder
Model Results

In this chapter, I present model results for electrode arrays implanted in human
eyes based on the model input defined in Chapter 4. For each of two electrode arrays, I
present results from the five different loading conditions detailed in Figure 4.8.
These loading configurations are intended to represent acute tacking force,
chronic forces when the cable has a large knee, chronic forces from a cable with no knee,
chronic forces from a twisted cable with a large knee, and chronic forces from a twisted
cable with no knee. The chronic force configurations modeled here represent only a
selection of many possible in vivo chronic loading conditions.
In each study, I have defined ocular material properties to be High-strain, human
properties measured in a body-temperature saline environment. The electrode arrays,
constraints and mesh used are detailed in Chapter 4.
For each case, I provide a trimetric view of model deformation, a cross-sectional
view of Von Mises stress, and a top view of shear stress in the XZ plane (the plane of the
retina) as well as a series of top view Von Mises stress plots in progressively narrower
scales. The last image in that series is repeated in the largest panel with the electrode
array removed from view. Because these images are generated using screenshots, the
scale bar is often too small to read. The range of values for all plots is therefore noted in
brackets. Scales for each image remain the same from one study to the next.

66
5.1 16-Electrode Array

5.1.1 Acute Tacking Force

Figure 5.1 Tacking force applied to a 16-electrode array.

The soft array material causes the electrode array tip to lift off of the retina as the
tack site is driven downward. Stress is concentrated around the tack site and heel of the
array, with larger stresses in the sclera level in the same location. Shear stress reveals
stress asymmetry due to the angled tack approach.

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
67
5.1.2 Chronic Cable Force, Large Knee

Figure 5.2 Chronic cable force for a large knee applied to a 16-electrode array.

The cable pulls the heel of the electrode array upward, pushing the body of the
array downward with the tack as a lever point. Stresses are concentrated under the body
and, to a lesser extent, the toe of the array, particularly near the edges. Retinal stresses
are lower than seen during tacking, though high stresses still occur in underlying layers.

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
68
5.1.3 Chronic Cable Force, No Knee

Figure 5.3 Chronic cable force with no knee applied to a 16-electrode array.

Stress distribution is symmetrical, and fairly evenly distributed with some
concentration towards the heel of the array. Stress magnitude in the retina is consistently
low, with higher stresses visible in the sclera under the array heel.

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
69
5.1.4 Chronic Cable Force, Large Knee with Twist

Figure 5.4 Chronic cable force with a large knee and a twist in the cable applied to a 16-electrode array.

The large knee again pulls the heel of the array upwards, causing the body of the
array to push downward into the retina. More negative effects are due to the cable twist,
which causes some electrode array rotation and generates significant asymmetric stress
along the sides of the array and throughout the sclera and orbital fat layers.

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
70
5.1.5 Chronic Cable Force, No Knee with Twist

Figure 5.5 Chronic cable force with no knee and a twist in the cable applied to a 16-electrode array.

This study produces the lowest overall stress magnitude. The cable twist causes
slight array rotation, but stresses remain fairly evenly distributed. The twist does move
stress towards the toe of the array (when compared to the same cable with no twist).

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
71
5.2 60-Electrode Array

5.2.1 Acute Tacking Force

Figure 5.6 Tacking force applied to a 60-electrode array.

Large, somewhat asymmetrical, stresses are highly concentrated under the heel of
the array. Some high-magnitude stress also occurs at the tip, where slightly mismatched
array and retina curvatures create focal points. However, virtually no stress is applied to
the retina underneath the body of the array, where stimulating electrodes are located.

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
72
5.2.2 Chronic Cable Force, Large Knee

Figure 5.7 Chronic cable force for a large knee applied to a 60-electrode array.

The cable knee pulls the array heel up and pushes the tip down. Large magnitude
stresses remain concentrated near the heel and toe edges.

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
73
5.2.3 Chronic Cable Force, No Knee

Figure 5.8 Chronic cable force with no knee applied to a 60-electrode array.

Tension in a cable with no knee pulls the heel of the electrode array up, removing
the high stress that was present under the heel of the array in other loading
configurations. The end result is reduced stress around the edges of the array and
virtually no stress underneath the central region.

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
74
5.2.4 Chronic Cable Force, Large Knee with Twist

Figure 5.9 Chronic cable force with a large knee and a twist in the cable applied to a 60-electrode array.

Very high magnitude stresses are again generated underneath the heel and toe
edges of the array, while no stress is applied to the retina under the center of the array.

Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
75
5.2.5 Chronic Cable Force, No Knee with Twist

Figure 5.10 Chronic cable force with no knee and a twist in the cable applied to a 60-electrode array.

This loading configuration causes slight array rotation, concentrating
stress under the left heel corner of the array and along the right side.
Von Mises stress
[0-20 kPa]
Deformation Von Mises Stress [0-20kPa] Shear Stress [0-20kPa]
[0-20 kPa] [0-200 kPa] [0-2000 kPa] [0-20000 kPa]
76
Chapter 6
6 placeholder
Validation

I have validated the predictions of this model through direct comparison to two
experimental tests and through qualitative comparison to a known surgical outcome. In
each of these methods, model predictions correlate well with experimental data and the
conclusions we can draw from each validation method correlate well with one another.
Taken together, these three independent validations demonstrate my eye wall model
representation is viable, the predicted interaction of that eye wall with an electrode array
is accurate, and that model predictions can be correlated to in vivo retinal damage.

6.1 Comparison to Eye Wall Compression

The four-layer, homogeneous, isotropic representation of the eye wall I have used
in this project, and the process by which I have defined linear tissue properties for that
model, is validated through the comparison to experimental eye wall compression tests
presented in Section 4.1.1.2.
That experiment demonstrates that, when the model geometry is modified to
appropriately mimic the experimental setup and when high-strain elastic modulus values
are applied, modeled predictions of eye wall behavior match experimental results well.

77
6.2 Comparison to Pressure Distribution Under an Array

Further validation is necessary to demonstrate that the accurately captures the
interaction of my validated eye wall and an electrode array. Ideally, such a validation
would compare model predictions of stress or pressure distribution on the retina to in vivo
measurements of pressure underneath an implanted array at several different locations.
However, we have not yet been able to establish a reliable means of making those in vivo
pressure measurements. Instead, we have measured pressure under an array during
tacking in a benchtop arrangement. Though the conceptualization of this experiment was
collaborative, experimental work for this comparison was conducted by Dr. Adrian
Rowley and he is referenced in the text where appropriate. I have described his methods
and results in Section 6.2.1 for completeness and for comparison to model predictions.

6.2.1 Experimental Results
Dr. Rowley used a 3 mm diameter trephine to remove a circular portion of sclera
and choroid from an enucleated porcine eye cup while leaving the retina intact. He
positioned the sample over a 1.6 mm diameter cylindrical post such that the post was
centered in the scleral hole and flush with the lower surface of the retina. That post was
connected to the Bose 3100 ELF load cell.

78

Figure 6.1 Photograph of an enucleated porcine eye cup with a hole in the sclera and choroid layers created
using a trephine. The retina was left intact. The eye is located above a cylindrical load cell post such that
the post is flush with the lower surface of the retina.

Dr. Rowley positioned an electrode array above the hole so that the force-
measurement site would fall under a pre-determined portion of the array and then tacked
the array in place using a new retinal tack. During tacking, he recorded downward force
on the load cell post. A typical recording is shown in Figure 6.2.

Porcine eye
Load cell post
Trephined hole in
sclera and choroid
79

Figure 6.2 Screenshot of downward force (green line) recorded by the Bose ELF 3100 load cell as an
electrode array was tacked in place over the force measurement site. A = artifact; B = zero point; C – B =
weight of array; D – B = tacking force; E – B = chronic force; F = cable touch. Courtesy of A. Rowley.

The change in force in section A is an artifact of the measurement system.
Section B defines the zero point. At the end of section B, the array is placed on the eye
wall, and the force in section C minus the force in section B is equal to force created by
the weight of the array. At point D, the array is tacked in place and the maximum
downward force minus the zero point is equal to tacking force. In section E, the tacked
array applies chronic pressure to the retina but tissue creep produces a gradual reduction
in downward force. The two signal jumps labeled as F resulted from Dr. Rowley
touching the array cable slightly.
Dr. Rowley repeated this experiment in six locations, with the force measurement
site located near the heel, midpoint and toe of the array along both the midline and the
left side. Locations are referred to as Mid-Heel, Mid-Mid, Mid-Toe, Left-Heel, Left-Mid
80
and Left-Toe. Locating the array so that a predetermined region of it would lie on top of
the force measurement site was extremely difficult; we anticipate that repetition of the
experiment would show significant variability in force magnitude due largely to wide
variations in array location with respect to the force measurement site. However, due to
limited availability of retinal tacks, we could only run one trial at each location.

Figure 6.3 Photo of a 16-electrode array with approximate force measurement sites marked for Mid-Heel,
Mid-Mid, Mid-Toe, Left-Heel, Left-Mid and Left-Toe.

As one can see from the above figure, force measurement sites were quite large
with respect to the array itself and sites overlapped somewhat with one another. Dr.
Rowley’s results for pressure due to the weight of the array and tacking pressure are
provided in Table 6.1.

Heel Mid Toe
16-Electrode Array
Weight of
Array
Left
(mmHg)
55 88 55
Mid
(mmHg)
55 248 90
Tacking
Force
Left
(mmHg)
5108 1890 1733
Mid
(mmHg)
6660 2798 2085

Table 6.1 Pressure under an array measured experimentally by A. Rowley in six approximate locations due
to the weight of the array and electrode array tacking.
81

6.2.2 Model Predictions
When these experimental results are compared to the predicted stress distribution
for a 16-electrode array during tacking, shown in Chapter 5, the distribution of pressure
throughout the six locations qualitatively matches the distribution predicted by the model:
larger pressures occur near the heel and along the midline, with lower pressures at the toe
and along the left side.

Figure 6.4 Experimental measurements of pressure (mm Hg) at six locations under a 16 electrode array
during tacking overlaid on modeled prediction for 16-electrode array tacking stress distribution.

However, a closer comparison can be made by modifying the model to more
closely replicate the experimental setup. I have removed the orbital fat layer from the
model (as no orbital fat was included in the live experiment), and created a 3 mm
diameter hole in the sclera and choroid at each of six locations. I inserted a metal,
cylindrical post in the appropriate location, fixed it in space, and recorded vertical
reaction force on the post as gravity alone and gravity plus a tacking force were applied
to the electrode array. In this case, I used the ocular material properties of porcine tissues
5108 1890 1733
6660 2798 2085
82
measured in 25º air (instead of human properties measured in 37º saline, as used for
Chapter 5 results).

Figure 6.4 Top view of model with 16-electrode array, 6 force measurement locations marked, and a hole
in the sclera and choroid with a load cell post visible in the Mid-Mid location.

Figure 6.5 Cross-section view of a 16-electrode array placed over a force-measurement post in the Mid-
Mid position.

The tacking pressure predicted by the model using both High-strain and Low-
strain porcine ocular properties is shown below. With the exception of the Left-Toe
83
location, differences between experimental results and tacking pressure predicted by the
High-strain model ranged from 2-30%. This range is considered reasonable given the
lack of location precision in the live experiment (as the model shows, values can change
significantly with the location of the force-measurement site with respect to the array).

Figure 6.6 Comparison of experimental results and model predictions for pressure under an
electrode array in six locations during tacking.

Midline - Tacking Pressure
0
10000
20000
30000
40000
50000
Pressure (mmHg)
Experimental
High-strain model
Low -strain model
Experimental 6660 2798 2085
High-strain model 8796 2803 1327
Low -strain model 41898 16469 3060
Heel Mid Toe
Left Side - Tacking Pressure
-5000
5000
15000
25000
35000
45000
Pressure (mmHg)
Experimental
High-strain model
Low -strain model
Experimental 5108 1890 1733
High-strain model 4345 2038 179
Low -strain model 30178 10354 3365
Heel Mid Toe
84
Earlier comparisons to composite eyewall compression tests (presented in Chapter
4) indicated that High-strain elastic modulus values are most appropriate when modeling
situations with non-zero deformations and Figure 6.6 confirms that result. Predictions for
pressure under an array generated using Low-strain elastic modulus data are significantly
larger than experimental results in all six locations, further indicating that tissue behavior
falls into the High-strain region during tacking.
Model predictions for pressure due solely to the weight of the array in six
locations are shown in Figure 6.7. Note the much smaller y-axis scale compared to
Figure 6.6.

85

Figure 6.7 Comparison of experimental results and model predictions for pressure under an
electrode array in six locations due solely to the weight of the array.

Pressure due solely to the weight of a small, polymer electrode array is much
smaller in magnitude than the force needed to insert a retinal tack. Tissue deformation in
response to array weight will be correspondingly small when compared to tissue
deformation during tacking. With small deformations, differences between high-strain
and low-strain modulus predictions are not as apparent and both models’ predictions fall
close to experimental results.
Midline - Weight of Array
0
100
200
300
400
500
600
700
800
900
1000
Pressure (mmHg)
Experimental
High-strain model
Low -strain model
Experimental 55 248 90
High-strain model 86 85 54
Low -strain model 65 70 40
Heel Mid Toe
Left Side - Weight of Array
0
100
200
300
400
500
600
700
800
900
1000
Pressure (mmHg)
Experimental
High-strain model
Low -strain model
Experimental 55 88 55
High-strain model 42 36 26
Low -strain model 37 32 0
Heel Mid Toe
86
Although it is difficult to conduct a precise experiment that can be directly
compared to modeled predictions of stress distribution on the retina, the relatively
imprecise experiment described here matches model predictions relatively well when
appropriate elastic modulus values are applied to the model. The experiment needs to be
repeated several times to get an average value, but each trial would require a new retinal
tack.
Comparing model predictions to single-trial experimental results indicates that,
when the model is modified to appropriately mimic the experimental setup and when
High-strain modulus values are applied, modeled predictions of system behavior
generally fall within 30% of experimental results. This correlates well with the eye wall
compression validation presented in Section 4.1.1.2.

6.3 Comparison to a Known Surgical Outcome

Although we do not yet understand the threshold or mechanism for mechanical
damage to the retina, we can assume that retinal damage is most likely to occur in
locations where stress magnitude is high. We can qualitatively compare modeled
predictions for the location of high-stress regions to the location of retinal damage in
surgical cases.
A review of surgical images from 16-electrode retinal prostheses implanted in
canine models in combination with surgeons’ experience suggests that retina damage,
when present, is often located around the edge of the array as suggested by this model.
Figure 6.8 shows a Fundus photo of an implanted 16-electrode where visible retina
87
damage is apparent, marked by dark discoloration, at the tack site, under the toe of the
array, and at each of the heel corners. This correlates well with modeled predictions for
damage during tacking (damage near the heel) in combination with damage due to
chronic forces from a cable with a large knee (damage near the toe).

Figure 6.8 A Fundus photo showing retinal damage caused by a 16-electrode array chronically implanted in
a canine model (left). Model predictions of Von Mises stress distribution on the retina for a 16-electrode
array subjected to tacking (top) and a cable with a large knee (bottom) (right).

However, damage in this case appears to be asymmetrical near the heel of the
array, a result that seems to be unrelated to model predictions. In truth, though, the
model does predict this asymmetry; it simply is not visible when results are visualized
using Von Mises stress. If, instead, results are presented as shear stress in the XZ plane
(the plane of the retina) as shown in Figure 6.9, asymmetrical shear stress caused by the
supero-nasal tack approach becomes apparent.

[0-20 kPa]
[0-20 kPa]
88

Figure 6.9 Shear stress in the XZ plane (the plane of the retina) caused by tacking force applied to a 16-
electrode array along the angled tacking axis.

Shear stress magnitude is higher near the lower heel corner of the array, which
correlates well to the asymmetrically high level of damage seen near the lower heel
corner in the Fundus photo. Combined together, the locations of Von Mises stress due to
tacking, shear stress due to tacking, and Von Mises stress due to a large knee in the cable
correlate well with the locations of retinal damage shown in Figure 6.10.

[0-2 kPa]
89

Figure 6.10 Comparison of retinal damage in a surgical case with predicted Von Mises stress due to
tacking, shear stress due to tacking, and Von Mises stress due to a cable with a large knee.

90
Chapter 7
7 spaceholder
Model Capabilities and Limitations

7.1 Capabilities

This model succeeds in its intended purpose by allowing current and future users
to easily modify array geometry and loading conditions to evaluate different cases and
design options while still accurately capturing the behavior of a series of complex
biological tissues. With the validations presented in Chapter 6, we can use this model to
evaluate novel or hard-to-characterize cases and be confident in the accuracy of predicted
results.
This model can also be used to evaluate specific cases, design features and
electrode array properties. In this chapter, I present several examples of ways in which
this model has been used and could be used in the future to further electrode array design
efforts or interpret experimental and clinical results.

7.1.1 Specific Cases - Cut Cable
Electrode arrays subjected to specific, unique loading conditions in vivo can be
modeled to either predict the result of loading condition changes or to retroactively
reverse engineer what must have happened in a given case. In one such case, the ribbon
cable connected to a 16-electrode array implanted in a human subject was cut for surgical
reasons. Without the cable to help anchor it in place, the electrode array pivoted around
91
the tack and settled in a new orientation. The original orientation (horizontal) and the
new orientation (vertical) of this array are shown in Figure 7.1.

Figure 7.1 Fundus photo of a 16-electrode array that pivoted from its original orientation (horizontal) to a
new orientation (vertical) when the ribbon cable anchoring it in place was cut.

A post-mortem histological study of the subject’s retina was conducted by Dr.
Yael Morales, who categorized retinal damage into five intensity levels and mapped that
damage to the electrode array’s original and eventual locations as shown in Figure 7.2.
Significant damage immediately around the tack site was likely due to the retinal tack
itself, but other damage may have been caused by the movement of the electrode array.

Original location
Location after rotation
92

Figure 7.2 Intensity of histological retinal damage mapped to the original and eventual position of an
electrode array that pivoted when the ribbon cable anchoring it in place was cut. Courtesy of Y. Morales.

My model can be used to replicate this situation. To do so, gravitational force is
applied to the electrode array to cause it to rotate around the tack while the cable
boundary is left unconstrained. The Von Mises stress and shear stress predicted by my
model when a 16-electrode array is pivoted is shown in Figure 7.3. It suggests that stress
will be located primarily near the leading edge of the electrode array which is, in fact,
where moderate to high levels of damage did occur.
Modeling this specific situation can help us understand what loading conditions
this subject’s electrode array must have been subject to prior to the cutting of the ribbon
cable. It could also potentially be used to broadly correlate different types of histological
retinal damage with mechanical or electrical stresses.
2
3
4
5
Intensity
650
680
710
750
780
820
840
860
890
920
950
980
1010
1040
1070
1100
1130
1160
1190
1220
1250
1280
1300
93
Y dir on YZ plane
Z dir on YZ plane
Z dir on XZ plane
Von Mises stress
Deformation
[0 – 0.02MPa]

Figure 7.3 Modeled predictions of deformation, Von Mises stress, and shear stress in three planes for a 16-
electrode array rotated from one position to another.

7.1.2 Evaluating Possible Design Features
This model can also be used to evaluate the logistics of specific design features.

94
7.1.2.1 Edge Features
It has been suggested that patterned vias around the outside of an electrode array
would help make the array edge more flexible and hopefully less likely to cause
mechanical damage to the retina.
In order to evaluate a broad range of possible via patterns before this model was
in place, each prospective design was fabricated and I performed a long series of
experimental tests to determine edge stiffness for each design option. My report on these
experiments is presented in Appendix C. With this model now operational, a range of
possible via patterns could be evaluated much more quickly without having to physically
fabricate the designs.

7.1.2.2 Gap in Concentric Wings
Electrode arrays designs have been considered that provide wide field vision
through concentric “wings” that can be folded up for insertion through a sclerotomy and
then unfurled inside the eye.

Figure 7.4 A proposed wide-field electrode array with concentric “wings”.

95
This array would be fabricated flat with the hope that the gap between the wings
and the body would allow it to conform to the curvature of the retina. My model can be
used to determine exactly how big that gap should be for the array to conform smoothly.
I generated a curved model of what we would like the array to look like once inside the
eye, and then flattened the model to reverse engineer the geometry that should be used to
fabricate the flat original.

Figure 7.5 Top view of concentric electrode array profile when curved (transparent image) and when flat
(solid). Inset shows the curved array being flattened.

We can also use this model to predict the stresses in this array as it conforms to
the curvature of the retina. A top view of Von Mises stress for this concentric array
subjected to a tacking force is shown in Figure 7.6.

96

Figure 7.6 Top view of Von Mises stress as a concentric electrode array is subjected to a tacking force.

7.1.3 Effect of Electrode Array Material Properties on the Retina
In Chapter 5, we can see differences in the effects of a 16 and 60 electrode array
design on stress in the retina. This model can be used to predict the effect of other
specific designs, design features and design choices on the retina, including the material
properties of the electrode array.
Figure 7.7 shows a top view of Von Mises stress for a 16-electrode array subject
to a tacking force (top row) and a cable with a large knee (bottom row) as the elastic
modulus of the array material is varied.
0 - 5MPa
97

Figure 7.7 Top view of Von Mises stress for a 16-electrode array subjected to tacking force and to a
chronic cable force as array elastic modulus is varied.

The results in Figure 7.7 indicate that harder array materials distribute stress
somewhat more evenly across the retina, while softer materials concentrate in specific
locations, particularly during tacking.

7.2 Limitations

Like any model, this model is limited by the simplifications and choices made in
its development and the accuracy of model input. In general, all FEA models are likely
to produce reasonable qualitative descriptions of system behavior, but quantitative
Acute tacking force:
Chronic cable force (large knee):
E = 0.4 GPa (silicone) E = 2.4 GPa (parylene) E = 3.7 GPa (polyimide)
[0 – 0.02MPa] [0 – 0.02MPa] [0 – 0.02MPa]

[0 – 0.02MPa]

[0 – 0.02MPa] [0 – 0.02MPa]
98
accuracy is more sensitive to model simplifications, analysis algorithms and mesh
quality.
Based on the capabilities of this software package and the choices that I made in
developing it, this model:

a) Is limited in accuracy to the linear High-strain region of materials’ stress-strain
curves. Because it is based on High-strain elastic modulus values, it will
overestimate material stiffness and stress for very small deformations.

b) Is only appropriate for static loads. If a load changes over time (i.e. an acute
tacking force), each time step must be considered as a separate loading condition.

c) Does not provide insight into the rest of the eye or other intraocular device
components (though it could be extended to include those things). In the
meantime, users must be careful to apply appropriate boundary conditions and/or
keep boundaries away from regions of interest.

d) Requires that the entire assembly be meshed as a single part, with compatible
elements across part boundaries. To accomplish this, the number of parts moving
against one another should be minimized, and electrode arrays must match the
curvature of the retina through some or all of their contacting surface area.

99
e) Does not capture gross tissue damage that occurs immediately underneath the
retinal tack tip.

f) Does not capture anisotropic or nonhomogeneous tissue properties, including any
effects from individual blood vessels, Muller cells or collagen fibers.

g) Does not explicitly predict mechanical damage to the retina. It can only predict
stress and strain, which we currently can only presume are causes of retina
damage.
100
Chapter 8
8 placeholder
Conclusions and Future Work

8.1 Conclusions

8.1.1 Model Development
My work is significant in that it combines a modifiable device with complex
biomaterials in the same model and analyzes interactions between them. I have been
guided by the fields of product design and biomaterials modeling, two very different
fields with often conflicting goals.
To combine user friendly and easily modifiable device models with complex
biomaterials, I found it best to use a software package intended for product design. It is
difficult to adequately represent the behavior of ultrasoft nonlinear biomaterials in such a
package, but if a suitable modeling system can be developed, it can be used over and over
again by non-specialist users who want only to try new device geometries or loading
conditions.
The initial introduction of biomaterials to design modeling, then, requires a
thorough understanding of the biomaterials being used and careful thought about how to
replicate those materials within the confines of software not intended for that purpose.
Sensitivity analysis indicates that this model is most sensitive to changes or errors in the
representation of the retina and choroid layers. As with any model, it is best to start
simply and add complexity over time.
101

8.1.2 Results
Results presented in this dissertation indicate that ocular materials can be
represented by linear elastic modulus values defined by the high-strain (50-100%) region
of their stress-strain curves. Models using those values were able to successfully predict
experimental data for two benchtop experiments (Sections 4.1.1.2 and 6.2.1).
With those properties defined, this model can be used to compare the effect of
different electrode array designs and different loading conditions. I have presented
extensive work for two electrode arrays: a first-generation 16-electrode array and a later-
generation 60-electrode array. The 16-electrode array is softer, with simpler geometry.
In particular, the bottom of the array is smooth and curved to fit the retina closely along
its entire surface area. Although stresses in the retina underneath this array are somewhat
concentrated near the array edges (where modulus mismatches are most apparent), the
smooth contact between this array and the retina distributes stress throughout the entire
contacting surface area. This produces lower stress magnitudes distributed under the
entire electrode array.
The 60-electrode array is more complex. It combines a layer of comparably stiff
polyimide with a surrounding layer of softer silicone. The bottom surface of this array
utilizes both materials and does not provide smooth contact with the retina. This array is
also longer than the first array, and its length and stiffness make it less likely to smoothly
conform to the curvature of the retina. Instead, stress is disproportionately applied under
the heel and toe of the array, where the flatter electrode array comes into the most direct
contact with the more curved retina. Stress magnitudes at those locations are very high,
102
but virtually no stress is applied to the retina under the center of the array, where
stimulating electrodes are located.
When either array is subjected to a tacking force, stress is concentrated near the
tack site and the heel of the array. The introduction of a large cable “knee” pulls the heel
of the array away from the retina and drives the body and toe of the array downward,
using the tack as a fulcrum. A cable without a large “knee” may actually generate
tension along the cable. But with the retina tack inserted, the array remains in place and
stress underneath all portions of the array is somewhat reduced. Neither a large “knee” or
a small “knee” produces large magnitude stresses in the retina.
When, however, a twist is introduced to the cable between the insertion point at
the sclerotomy and the tack site, the array may rotate slightly and asymmetrical shear
stresses occur. For a cable with a large knee, this asymmetry can produce high
magnitude localized stresses. For a cable with a small knee, this asymmetry actually
balances out the slight tendency towards stress near the heel and results in consistently
low stress magnitudes across the entire array. The effects of chronic cable forces are
smaller in magnitude than those of tacking forces, and small variations in chronic force
effects are most visible underneath the 16-electrode array. For the 60-electrode array,
high stresses located near the heel and toe edges for every case overwhelm small
variations in cable force effects.

8.1.3 Applications
Now that this model has been developed and validated, one can easily modify it to
analyze different array geometries, model features, and special cases. Even without
103
specialized knowledge of the model, users can get a broad understanding of the behavior
of a given mechanical system and that understanding will help guide future design
efforts. With knowledge gained from modeling, developers of the retinal prosthesis can
reduce the number of design fabrications and animal trials necessary to test new design
ideas. This makes the design process more efficient, and will help developers more
quickly optimize electrode array design in order to minimize damage to the retina.
This model can produce a wide range of outcome variables and viewpoints. That
flexibility is a powerful tool, although at this point it is still difficult to know how to
apply or evaluate model results. At the moment, I think we can assume that regions of
high stress should be avoided if possible as past surgical results suggest that the general
location of retinal damage correlates with predicted regions of high stress. But as our
understanding of the mechanisms and thresholds for retinal damage increases, we will be
able to use model results to directly predict (and thereby avoid) specific retinal damage.

8.2 Future Work

8.2.1 Electrode Array Design
Even without further development or the integration of increased complexity, this
model can be used immediately to model, evaluate and compare novel design ideas and
improve electrode array design.

104
8.2.2 Increase Model Complexity
Options for introducing anisotropic, nonhomogeneous and nonlinear material
properties to the eye wall model should be explored, though potential gains in numerical
accuracy should be weighed against practicality.

8.2.3 Determine Mechanism and Threshold for Mechanical Damage to the Retina
It is important to determine how different types of focal mechanical stresses affect
the health of retinal cells so that the results of this model can be applied more directly.

105
Bibliography

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

Bi A, Cui J, Ma YP, Olshevskaya E, Pu M, Dizhoor AM & Pan ZH. (2006). Ectopic
expression of a microbial-type rhodopsin restores visual response in mice with
photoreceptor degeneration. Neuron 50, 23-33.

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

Bron A, Tripathi R and Tripathi B 1997 Wolff’s Anatomy of the Eye. (London: Chapman
& Hall).

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

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

Colodetti L, Weiland JD, Colodetti S, Ray A, Seiler MJ, Hinton DR and Humayun MS.
(2007) Pathology of damaging electrical stimulation in the retina. Exp Eye Res 85 23-33.

Courant, R. L. (1943). Variational methods for the solution of problems of equilibrium
and vibration. Bulletin of the American Mathematical Society 49, 1-23.

de Balthasar C et al Factors affecting perceptual thresholds in epiretinal prostheses.
(2008). Invest Ophthalmol Vis Sci 49(6) 2203-14.

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

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

Doerner MF & Nix WD. (1986). A method for interpreting the data from depth-sensing
indentation instruments. J Mater Res 1, 601-609.

Foerster O. (1929). Beitrage zur pathophysiologie der sehbahn und der spehsphare. J
Psychol Neurol 39, 435-465.

106
Friedman DS, O’Colmain BJ, Munoz B, Tomany SC, McMarty C, de jong PT, Nemesure
B, Mitchell P & Kempen J. (2004). Prevalence of age-related macular degeneration in
the United States. Arch Ophthalmol 122, 562-572.

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

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

Humayun MS (2009). Preliminary results from Argus II feasibility study: a 60 electrode
epiretinal prosthesis. ARVO 4744, Ft. Lauderdale, FL.

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

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

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

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

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

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

Klein R, Klein BE, Jensen SC, and Meuer SM. (1997). The five-year incidence and
progression of age-related maculopathy: the Beaver Dam Eye Study. Ophthalmology
104(1) 7-21.

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

107
Kuttenkeuler C, Wilke R, Volker M, Sachs H, Gabel VP, Gekeler F, Besch D, Bartz-
Schmidt K & Zrenner E. (2006). Subretinal implant study group. Subretinal chronic
active multi-electrode arrays in blind patients: fundus appearance, optical coherence
tomography and angiography. In ARVO. Fort Lauderdale, FL.

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

Majji AB, Humayun MS, Weiland JD, Sukuki S, D’Anna SA and de Juan E Jr. (1999).
Long-term histological and electrophysiological results of an inactive epiretinal electrode
array implantation in dogs. Invest Ophthalmol Vis Sci 40(9) 2073-81.

Margalit E and Sadda SR. (2003). Retina and optic nerve diseases. Artif Organs 27(11)
963–74.

McMahon MJ. (2006). Electrode impedance as a predictor of electrode-retina proximity
and perceptual threshold in a retinal prosthesis. Incest Ophthalmol Vis Sci 47:E-abstract
3185.

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

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

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

Reichenblach A, Eberhart W, Scheibe R, Deich C, Seifart B, & Reichart W. (1991).
Development of the rabbit retina IV. Tissue tensility and elasticity in dependence on
topographic specializations. Exp Eye Res 53, 241-51.

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

Rowley AP, Chen K, Basinger B, Weiland J & Humayun M. (2009). Mechanical
modeling of an epiretinal prosthesis. ARVO 4592, Ft. Lauderdale, Fl.

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

108
Schmidt EM, Bak MJ, Hambrecht FT, Kufta CV, O'Rourke DK & Vallabhanath P.
(1996). Feasibility of a visual prosthesis for the blind based on intracortical
microstimulation of the visual cortex. Brain 119 ( Pt 2), 507-522.

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

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

Vesprille K (1975). Computer aided design applications of the rational B-spline
approximation form. PhD thesis, Syracuse U., 1975.

Walter P, Szurman P, Vobig M, Berk H, Ludtke-Handjery HC, Richter H, Mittermayer
C, Heinmann K and Sellhaus B. (1999). Successful long-term implantation of electrically
inactive epiretinal microelectrode arrays in rabbits. Retina 19(6) 546-52

Yanai D, Weiland JD, Mahadevappa M, Greenberg RJ, Fine I and Humayun MS. (2007)
Visual performance using a retinal prosthesis in three subjects with retinitis pigmentosa.
Am J Ophthalmol 143(5) 820-7.

Zrenner E, Stett A, Weiss S, Aramant RB, Guenther E, Kohler K, Miliczek KD, Seiler
MJ & Haemmerle H. (1999). Can subretinal microphotodiodes successfully replace
degenerated photoreceptors? Vision Res 39, 2555-2567.
109
Appendix A: Results from early flat tissue model
0.313mm 0.425mm 0.171mm 0.174mm Array tip liftoff
1.38mm 1.00mm 1.43mm 1.23mm “crush width from hole
3.57mm 3.46mm 3.15mm 3.10mm “crush” area width
0.355mm 0.369mm 0.309mm 0.284mm Downward array disp.
16.49MPa 16.32MPa 13.75MPa 11.28MPa Max contact pressure
90.68MPa 174MPa 150MPa 65.79MPa Max stress
Front edge tack hole Smiley corner Front edge tack hole Front edge tack hole Location of max stress
Stress at retina/sclera
boundary
Contact pressure
Deformed shape
Silicone array With “reverse smiley” With “smiley” Polyimide array
0.313mm 0.425mm 0.171mm 0.174mm Array tip liftoff
1.38mm 1.00mm 1.43mm 1.23mm “crush width from hole
3.57mm 3.46mm 3.15mm 3.10mm “crush” area width
0.355mm 0.369mm 0.309mm 0.284mm Downward array disp.
16.49MPa 16.32MPa 13.75MPa 11.28MPa Max contact pressure
90.68MPa 174MPa 150MPa 65.79MPa Max stress
Front edge tack hole Smiley corner Front edge tack hole Front edge tack hole Location of max stress
Stress at retina/sclera
boundary
Contact pressure
Deformed shape
Silicone array With “reverse smiley” With “smiley” Polyimide array
110
Appendix B: Results generated using literature tissue properties

Finite element modeling of retinal prosthesis mechanics

B C Basinger
1,4
, A P Rowley
2,4
, M S Humayun
1,3,4
, and J D Weiland
1,3,4

Department of Biomedical Engineering,
Viterbi School of Engineering,
University of Southern California

Department of Cell and Neurobiology,
Keck School of Medicine,
University of Southern California

Department of Ophthalmology,
Keck School of Medicine,
University of Southern California

Intraocular Retinal Prosthesis Group,
Doheny Retina Institute,
Doheny Eye Institute,
Keck School of Medicine,
University of Southern California
1355 San Pablo Avenue, Suite 100
Los Angeles, CA 90033

Communicating author: J D Weiland
Phone: 323-442-6670
FAX: 323-442-6755
Email: jweiland@usc.edu

Subject classification numbers: PACS 87.19.R-, 87.85.E-, 87.85.em, 87.85.Tu

Submitted to: Journal of Neural Engineering

Abstract: Epiretinal prostheses used to treat degenerative retina diseases apply stimulus via an electrode
array fixed to the ganglion cell side of the retina. Mechanical pressure applied by these arrays to the retina,
both during initial insertion and throughout chronic use, could cause sufficient retinal damage to interfere
with the effectiveness of the device. In order to understand and minimize potential mechanical damage, we
have used finite element analysis to model the mechanical interaction between an electrode array and the
retina in both an acute and a chronic loading configuration. Modeling indicates that an acute tacking force
distributes stress primarily around the edges of the electrode array, particularly the edge nearest the cable.
Model predictions for one potential chronic loading configuration, where excess cable is included between
the incision in the sclera and the electrode array, suggest that stress in that case is concentrated along the
edge of the array furthest from the cable. Retinal damage is seen in both locations in a canine model
chronically implanted with a similar 16-electrode array.

111

1. Introduction
Degenerative retina diseases such as retinitis pigmentosa and age-related macular degeneration
affect nearly 2 million people in the United States [1-4]. These diseases eventually lead to
blindness that can negatively affect an individual’s quality of life. Under development are
experimental devices that bypass the degenerated photoreceptors and apply electrical stimulation
directly to the inner retina [5-15]. These retinal prosthetic devices rely on the principle of
electrical activation of nerves, which is central to other successful neural stimulators such as the
cochlear implant and deep brain stimulator.

A number of research groups around the world have implanted prototype retinal prosthetic
devices in blind humans. These implants are often categorized by the position of the stimulating
electrodes: epiretinal implants locate an electrode array on the ganglion cell side of the retina,
while subretinal implants stimulate from electrodes located underneath the retina. Studies of both
implant types have shown that humans with retinitis pigmentosa can detect light in response to
electrical stimulation [16-20]. The electrode arrays used in these studies are generally fabricated
using polymers such as polyimide or silicone, which are much stiffer than soft tissue, thus
creating a mismatch in material properties where the device interfaces with ocular tissues.

An epiretinal prosthesis has been developed at the University of Southern California and Second
Sight Medical Products, Inc. It utilizes a head-mounted video camera to capture images, which
are processed by an external video processing unit and transmitted inductively to a stimulation
pulse generator implanted behind the ear. Stimulation pulses are passed via a trans-scleral cable
to a flexible, polymer microelectrode array affixed to the retina by means of a retinal tack [21-
23]. The electrode array has 16 platinum disk electrodes arranged in a 4x4 pattern.

The stimulation intensity required to elicit a percept from an epiretinal electrode has been found
to increase with distance from the retina [24, 25]. In addition, electric field theory suggests that
the stimulus current from an electrode further from the retina will activate a larger retinal area and
thus result in a more diffuse perception. In order to make the most efficient use of limited power
resources and to provide the highest acuity vision, it is important that the retinal prosthesis
electrode array be positioned in close proximity to the retina.

Currently, that proximity is achieved with the use of a single metal retinal tack that passes
through a hole in the polymer array substrate and penetrates the retina, choroid and sclera. Retinal
tacks were developed to treat retinal detachments and have been adapted for fixation of retinal
prostheses. Although this attachment method works relatively well to maintain the electrode
array position [26], retinal tacks cause a clear insult to the tissue and tacks have been associated
with fibrovascular reaction [26] and retinal folds [27] immediately around the tack insertion site.
Furthermore, mechanical pressure has been shown to cause damage to retinal ganglion cells
independent of tacking [28].

With the electrode array positioned close to the retina, the potential exists for mechanical damage
caused by the attachment method or by contact between the device and the retina. However, it is
important to maintain the health and functionality of those cells needed to transmit an applied
electrical signal on to the optic nerve. Mechanical damage to the retina could be mitigated or
redistributed away from critical regions through mechanical design changes, stress relief features,
or material changes but mechanical design thus far has been guided primarily by intuition and
trial-and-error. No means of systematically evaluating the mechanical behavior of a retinal
prosthesis electrode array is currently in place. In order to better understand the mechanics of
112
retinal prosthesis electrode arrays and to guide mechanical design efforts, we have used solid
modeling and linear finite element analysis to model mechanical interactions of the electrode
array and eye wall. We have used this model to predict mechanical stress distribution in the
retina in both an acute and chronic loading case and compared the results qualitatively to a
surgical outcome.

2. Methods
We have used SolidWorks Office Professional 2008 (Concord, MA) and its integrated finite
element analysis package, COSMOSWorks Advanced 2008 (Concord, MA), to model an
epiretinal electrode array and a portion of the eye wall. Modeling was performed using a 2.13
GHz Intel® Core™2 Asus desktop computer.

2.1. Model Geometry
Model geometry is shown in figure 1. The electrode array model is a simplified version of the
16-electrode arrays used in an on-going clinical trial [18]. That electrode array design was chosen
because we have extensive preclinical testing data from long-term implants that can serve as
qualitative validation for our model. The modeled array is essentially rectangular in shape (with
rounded corners), 0.5 mm thick at its maximum thickness, with a cable protruding from the “heel
edge” and a tack hole located near the cable insertion point. The array model is curved to allow it
to contact the retina across the array’s entire lower surface. The electrode array rests on an eye
wall model composed of 4 layers: retina (0.25mm thick), choroid (0.2mm), sclera (0.5mm) and
orbital fat (3mm) [29]. The eye wall substrate has a radius of curvature of 12mm to represent an
average human adult eye [29]. The electrode array and eyewall are combined in an assembly,
with the bottom surface of the electrode array in initial contact with the retina and contact
constraints defined to prevent penetration between the parts.

113

Figure 1: Four views of model geometry. The electrode array model is similar to a 16-electrode
array currently in use and includes a partial cable and a via used for tack insertion. The eye wall
model consists of four layers: retina, choroid, sclera and orbital fat. The eye wall has a radius of
curvature of 12mm and the array model is curved to contact the retina along its entire bottom
surface.

2.2. Material Properties
The electrode array polymer is defined to have an elastic modulus of 3.7 GPa and mass density of
1.61 g/cm³. Retina, choroid, sclera and orbital fat are defined to have elas tic modulii of 0.1 MPa
[30], 0.6 MPa [31], 1.8 MPa [31] and 0.05 MPa [29], respectively. The mass density of all
tissues was assumed to be that of water, 1.0 g/cm³. All materials were as sumed to be nearly
incompressible, with a Poisson’s ratio of 0.45.

2.3. Boundary Conditions and Mesh
Two loading conditions have been modeled: 1) An acute loading condition representing the initial
tacking of the electrode array during surgical implantation 2) A chronic loading condition
representing one likely combination of chronic forces exerted by the cable and tack together. In
both cases, the eye wall segment is fixed in all 6 translational and rotational directions along the
outside edge of the orbital fat layer. In addition, the interior face of the electrode array tack hole
is fixed in the rotational direction to prevent rotation of the device. The applied mesh uses 37,304
tetrahedral elements, with element size reduced to the thickness of the retina in the retinal layer.
The mesh and boundary conditions for both the chronic and acute loading condition are shown in
figure 2.

114
2.3.1. Acute Tacking Force. The first loading condition represents acute forces applied to the
electrode array during the initial implantation and tacking of the device. A tacking force of 1.25
N is applied along an axis 30º from vertical in both the x and y directions, approximating the axis
of tack insertion during surgery. The tacking force required to insert a retinal tack into the
eyewall was measured experimentally with a load cell as a surgeon inserted a tack into an
enucleated porcine eye. In this model the retinal tack has been neglected and the tacking force
applied directly to the top face of the electrode array tack hole. Because the deformation of the
metal tack is negligible compared to the deformation of the polymer array and ocular tissues,
force applied to the tack head can be estimated to transfer directly to the electrode array surface
underneath it with little change in magnitude or direction. Eliminating the tack from the model
simplifies computation by allowing a single mesh to be used for the entire system. By applying a
tacking force directly to the electrode array, we are most closely modeling the instant when the
tack has already pierced the eyewall, but the surgeon has not yet released pressure on the
insertion tool.

2.3.2. Chronic Cable Forces. The second loading condition represents one likely arrangement of
chronic forces applied to the electrode array after surgery. The electrode array is subject to a
residual tacking force and to forces and torques transmitted through the trans-scleral ribbon cable
attached to the array. Many possible combinations of chronic forces are possible in vivo and it is
difficult to predict which loading conditions are most likely or to know the precise cable force
magnitude or direction for a given case. In order to define a likely chronic loading condition, we
generated a larger, less detailed model of the entire eye and the entire intraocular portion of the
device, including a cable extending from an incision in the sclera (sclerotomy) to the heel edge of
the electrode array. During surgery, the length of cable between the array and sclerotomy is
somewhat variable and excess cable is sometimes introduced into the eye to create a bend in the
cable. Thus we simulated the case where extra cable length is introduced between the sutured
cable insertion site and the electrode array. Modeling that geometry predicted an inward band of
the cable with compressive forces 0.5-1.5 N in magnitude (depending on exact cable length)
along its length.

Thus, to create the chronic loading configuration presented here, a 2 N compressive force was
applied to the cut face of the cable in the normal direction, and 1 N was applied tangent to that
face in a direction pointing towards the center of the eye to create an inward bend in the cable.
The cable force direction is intended to represent forces caused by excess cable and force
magnitudes are defined to be slightly higher than those predicted by a larger model in order to
represent a “worst-case” condition. In addition, a chronic force of 0.25 N to the tack hole along
the previously described tacking axis to represent chronic pressure from the tack head.

3. Results
Model deformations resulting from both loading conditions are shown in figure 2. In the acute
case, the downward tacking force pushes the electrode array into the eye wall, causing the array
to flex while the eye wall deforms and the orbital fat layer bulges outward. The chronic cable
force modeled here pulls the heel of the electrode array upward and causes the toe of the array to
push downward into the retina. The electrode array flexes somewhat in both cases, but because
of the modulus mismatch between the retina and the electrode array polymer, larger
displacements are seen in the retina layer.
115

Figure 2: (a) Model mesh and loading condition for the acute case, where a 1.25 N downward
tacking force is applied to the material around the tack hole. The interior of the tack hole is fixed
to prevent rotation, and the outside of the orbital fat layer is fixed in all 6 rotational and
translational directions. (b) Cutaway side view of model deformation resulting from the acute
loading condition. The heel of the array is pushed downwards while the orbital fat bulges
outward. (c) Model mesh and loading condition for the chronic case, where a residual tack force
of 0.25 N is applied, along with 2 N of compressive cable force and a 1 N force pushing the cable
inward. The model is constrained as described above. (d) Cutaway side view of model
deformation resulting from the chronic loading condition. The heel of the array is pulled upward,
while the toe moves downward into the eye wall.

3.1. Acute Tacking Force
A top view of the Von Mises stress produced by the acute tacking force is shown in figure 3. The
initial plot (a), suggests that the highest stresses present are in the electrode array itself, near the
tack hole. However, we are more concerned with potential damage to the retina than to the
electrode array. Narrowing the scale by several orders of magnitude in panels (b), (c) and (d)
allows us to resolve stresses at the retinal level. In panel (e), the electrode array has been
removed from the image to provide a clear view of stresses on the surface of the retina. Stress
magnitude is indicated by color, with red representing the highest stresses. In this case, the
highest stresses in the retina are located under the edges of the array, particularly the heel edge.

The maximum Von Mises stress predicted in the retinal layer is 5.92 MPa, located at the heel
edge of the array. (For comparison, the yield stress of structural steel (ASTM A36) is 145 MPa
[32] and the yield stress of rubber is 1-7 MPa [33]). The exact location of that stress can be more
clearly seen in an alternate view. Figure 3(f) shows a side view of Von Mises stress. The
116
electrode array cable would be on the left of the image and the array tip on the right, but in order
to isolate the retinal layer both the electrode array and the choroid have been hidden from view.
High stresses are present in the sclera but, more importantly, a small area of high stress is located
in the retina under the heel of the electrode array. Close inspection of this view also reveals that
the retinal layer appears thinner in this location. That thinning increases strain, as shown in figure
3(g), which shows a side view of strain energy density with the electrode array hidden.

Figure 3: Predicted results of the acute loading condition modeled here. (a)-(e) show a top view
of Von Mises stress. Panels (a)-(d) show identical results with an increasingly narrowed scale:
[0-20MPa], [0-2MPa], [0-0.2MPa] and [0-0.02MPa]. The largest stresses occur in the array
itself, and it is only with a smaller scale that the stresses in the retina can be resolved. Panel (e)
shows the same image as (d), but with the electrode array hidden from view to provide a clear
image of stress distribution on the retina surface. A side view of Von Mises stress is shown in (f)
and a side view of strain energy density is shown in (g), both with the array and choroid hidden
from view.

3.2. Chronic Cable Force
A top view of Von Mises stress produced by the chronic cable force loading condition is shown
in figure 4. Again, panels (a), (b), (c) and (d) show increasingly narrow scales and the electrode
array has been removed from the image in panel (e). The highest stresses in the retina layer in
117
this case occur under the toe edge of the electrode array, though the magnitude of those stresses is
64.8% lower than in the previous case. The maximum Von Mises stress in the retinal layer here
is 2.08 MPa. Figures 4(f) and 4(g) show a side view of Von Mises stress and strain energy
density for this case, again with the array and choroid hidden from view to isolate the retina layer.

Figure 4: Predicted results of the chronic loading condition modeled here. (a)-(e) show a top
view of Von Mises stress. Panels (a)-(d) show identical results with an increasingly narrowed
scale: [0-20MPa], [0-2MPa], [0-0.2MPa] and [0-0.02MPa]. The largest stresses occur in the
array itself, and it is only with a smaller scale that the stresses in the retina can be resolved. Panel
(e) shows the same image as (d), but with the electrode array hidden from view to provide a clear
image of stress distribution on the retina surface. A side view of Von Mises stress is shown in (f)
and a side view of strain energy density is shown in (g), both with the array and choroid hidden
from view.

3.3. Comparison to Surgical Results
A review of surgical results from 16-electrode retinal prosthesis arrays chronically implanted in
canine models suggests that retina damage, when present, is often located around the edges of the
array. Figure 5 shows a Fundus photo of a 16-electrode array where visible retina damage is
apparent, marked by dark discoloration, at the tack site, under the toe of the array, and at each of
the heel corners. Though we cannot determine the specific mechanism of retinal damage in this
118
case, the locations of this damage correspond well with the locations of high stress regions
predicted by our model.

Figure 5: A Fundus photo showing a 16-electrode array implanted chronically in a canine model.
The dark pigmentation around the tack, at the toe of the array, and at both heel corners indicates
retina damage.

4. Discussion
In order to better understand, predict and prevent such damage, we have developed the first
model of mechanical interactions between the eyewall (retina, choroid, and sclera) and electrode
array. This model can be used to predict stress distribution in the retina due to both acute and
chronic forces associated with an epiretinal prosthesis. Preliminary results correlate well with
surgical observations in animal studies. The quantitative accuracy of any model is limited both by
the assumptions and simplifications used in building the model and by mesh density. It is
assumed that geometric and material simplifications are most limiting here, so mesh density was
not optimized or increased in these initial studies.

The geometries of both the electrode array and the eye wall have been simplified. The electrode
array has been represented as a single, polymer layer without metalized electrodes or traces,
which may provide additional stiffness to the device. In addition, no retinal tack has been
included in the model and tacking forces have instead been applied directly to the electrode array
surface. This allows model contact constraints to be simplified and eliminates any results of tack
tip penetration from the model. The stresses generated directly by the insertion of a retinal tack,
though significant, are contained in the immediate vicinity of the tack site and are not thought to
interfere in device efficacy. Furthermore, including them in the model could mask smaller
stresses generated under the remainder of the electrode array, where retinal ganglion cells must be
healthy to respond to applied electrical stimuli.
119

In addition, the material properties used to represent ocular tissues in this model, most notably
retina, have not been completely established. Biological tissues are generally viscoelastic and
behave nonlinearly, but here each tissue has been approximated by the single, linear elastic
modulus values reported above in order to allow this problem to be solved through linear finite
element analysis.
For linear FEA to be appropriate, materials should remain within the linear, elastic region of their
stress-strain curves. It is not currently known whether that condition is satisfied in this case, but
further tissue characterization work is being pursued that may indicate whether non-linear finite
element analysis is necessary.

In both loading conditions explored here, the model behaves predictably and realistically. In both
cases, we see relatively high levels of stress in the sclera and in the electrode array itself, but we
have focused on stresses in the retinal layer. The predicted stress distributions clearly indicate
that the retina is subject to the most severe trauma along the heel edge of the array during tacking
and along the toe edge when cable forces pull the heel of the array upwards.

The forces that an electrode array is subject to in vivo are highly variable and likely include cable
forces beyond those modeled here. At present, it is difficult to directly measure those forces or
the pressure under an electrode array in vivo in order to provide a direct validation of model
results. Yet the difficulty of accurately measuring the pressure under an implanted electrode
array is precisely why this computerized model of retinal prosthesis mechanics is needed. In the
future, this model will be validated through comparison to benchtop testing and a more thorough
correlation to clinical outcomes.

5. Conclusions
We have implemented a mechanical model that predicts stresses generated during surgical
attachment (tacking) and chronic implantation (cable forces). The above model indicates that
retinal stress is concentrated under the heel edge of the array during tacking, while the cable force
modeled here generates smaller amounts of stress under the toe edge. Initial comparisons suggest
that these results correlate well with clinical observations. With further refinement and validation,
it may be possible to use models of this type to design electrode arrays with stress relief features
that decrease mechanical stress on the retina.

Acknowledgements
This work was supported by the Department of Energy's Office of Science (grant no. DE-FC02-
04ER63735), W.M. Keck Foundation, and Clarence and Estelle Albaugh Trust.

References
[1] Heckenlively JR, Boughman J, and Friedman L 1988 Diagnosis and
classification of retinitis pigmentosa. In Heckenlively JR Retinitis Pigmentosa
(Philadelphia, PA: JB Lippincott) p 21
[2] Klein R, Klein BE, Jensen SC, and Meuer SM 1997 The five-year incidence and
progression of age-related maculopathy: the beaver dam eye study. Ophthalmol
104(1) 7-21
[3] Klein R, Klein BE, and Linton KL 1992 Prevalence of are-related maculopathy.
The beaver dam eye study. Ophthalmol 99(6) 933-43
[4] Margalit E and Sadda SR 2003 Retina and optic nerve diseases. Artif Organs
27(11) 963–74
120
[5] Chow AY and Chow VY 1997 Subretinal electrical stimulation of the rabbit
retina. Neurosci Let 225(1) 13-6
[6] Chow AY and Peachey NS 1998 The subretinal microphotodiode array retinal
prosthesis. Ophthalmic Res 30(3) 195–8
[7] Dawson WW and Radtke ND 1977 The electrical stimulation of the retina by
indwelling electrodes. Invest Ophthalmol Vis Sci 16(3) 249–52
[8] Eckmiller R 1997 Learning retina implants with epiretinal contacts. Ophthalmic
Res 29(5) 281-9[9] Humayun MS, de Juan E Jr, Weiland JD, Dagnelie G,
Katona S, Greenberg RJ and Suzuki S 1999
[9] Pattern electrical stimulation of the human retina. Vision Res 39(15) 2569–76
[10] Humayun MS 2001 Intraocular retinal prosthesis. Trans Am Ophthalmol Soc 99
271-300
[11] Margalit E, Maia M, Weiland JD, Greenberg RJ, Fujii G, Torres G, Piyathaisere
D, O’Hearn T, Liu W, and Lazzi G 2002 Retinal prosthesis for the blind. Surv
Ophthalmol 47(4) 335-56
[12] Rizzo JF 3
rd
, Wyatt J, Humayun M, de Juan E, Liu W, Chow A, Eckmiller R,
Zrenner E, Yagi T and Abrams G 2001 Retinal prosthesis: an encouraging first
decade with major challenges ahead. Ophthalmology 108(1) 13-4
[13] Wyatt J and Rizzo JF 1996 Ocular implants for the blind. IEEE Spectrum 33(5)
47–53
[14] Zrenner E et al 2001 Subretinal microphotodiode array as replacement for
degenerated photoreceptors? Ophthalmologe 98(4) 357-63
[15] Zrenner E et al 1997 The development of subretinal microphotodiodes for
replacement of degenerated photoreceptors. Ophthalm Res 29(5) 269-80
[16] Richard G, Hornig R, Keseru M and Feucht M 2007 Chronic epiretinal chip
implant in blind patients with retinitis pigmentosa: long-term clinical results.
Assoc Res Vis Ophth abstract 666/B290
[17] Walter P, Mokwa W, and Messner A 2008 The EPIRET3 wireless intraocular
retina implant system: design of the EPIRET3 Prospective clinical trial and
overview. Assoc Res Vis Ophth abstract 3023/D619
[18] Yanai D, Weiland JD, Mahadevappa M, Greenberg RJ, Fine I and Humayun MS
2007 Visual performance using a retinal prosthesis in three subjects with retinitis
pigmentosa. Am J Ophthalmol 143(5) 820-7
[19] Weiland JD, Yanai D and Mahadevappa M 2004 Visual task performance in blind
humans with retinal prosthetic implants. Conf Proc IEEE Eng Med Biol Soc 6
4172-3
[20] Sachs HG and Gabel V 2004 Retinal replacement - the development of
microelectronic retinal prostheses - experience with subretinal implants and new
aspects Graefe Arch Clin Exp Ophthalmol 242(8) 717-23
[21] Humayun MS et al 2003 Visual perception in a blind subject with a chronic
microelectronic retinal prosthesis. Vision Res 43(24) 2573-81
[22] Weiland JD and Humayun MS 2005 A biomimetic retinal stimulating array.
IEEE Eng Med Biol Mag 24(5) 14-21
[23] Weiland JD, Liu W and Humayun MS 2005 Retinal prosthesis. Annu Rev Biomed
Eng 7 361-401[24]
121
[24] de Balthasar C et al Factors affecting perceptual thresholds in epiretinal
prostheses. Invest OphthalmolVis Sci 49(6) 2203-14
[25] McMahon MJ et al 2006 Electrode impedance as a predictor of electrode-retina
proximity and perceptual threshold in a retinal prosthesis. Incest Ophthalmol Vis
Sci 47:E-abstract 3185
[26] Majji AB, Humayun MS, Weiland JD, Sukuki S, D’Anna SA and de Juan E Jr
1999 Long-term histological and electrophysiological results of an inactive
epiretinal electrode array implantation in dogs. Invest Ophthalmol Vis Sci 40(9)
2073-81
[27] Walter P, Szurman P, Vobig M, Berk H, Ludtke-Handjery HC, Richter H,
Mittermayer C, Heinmann K and Sellhaus B 1999 Successful long-term
implantation of electrically inactive epiretinal microelectrode arrays in rabbits.
Retina 19(6) 546-52
[28] Colodetti L, Weiland JD, Colodetti S, Ray A, Seiler MJ, Hinton DR and
Humayun MS 2007 Pathology of damaging electrical stimulation in the retina.
Exp Eye Res 85(1) 23-33
[29] Bron A, Tripathi R and Tripathi B 1997 Wolff’s Anatomy of the Eye. (London:
Chapman & Hall)
[30] Wollensack G and Spoeri E 2004 Biomechanical characteristics of retina. Retina
24(6) 967-70
[31] Friberg TR and Lace J 1988 A comparison of the elastic properties of human
choroid and sclera. ExpEye Res 47 429–36
[32] Ugural AC 2004 Mechanical Design: An Integrated Approach (New York:
McGraw-Hill)
[33] www.efunda.com

122
Appendix C: Experimental evaluation of edge features
Premitec
Edge feature evaluation
Brooke Basinger
July 3, 2008

Introduction
Epiretinal prostheses utilize polymer microelectrode arrays that are typically affixed to
the retina by means of a retinal tack. Mechanical damage to the retina near the edges of
the array has been observed in animal models, and finite element analysis of the array’s
mechanical interaction with the retina indicates that the retina is subject to the highest
stress levels near the edge of the array.

It is thought that stress relief features placed near the edge of the array would reduce the
device’s stiffness and therefore help reduce mechanical damage to the retina near the
edges of the array.

In this study, we have evaluated the effect of 14 possible edge designs.

Method

Experimental Setup
A variety of edge features utilizing thinned edges and patterns of vias of varying size and
arrangements were devised and produced.

Each disc was placed into the bottom of a plastic hemisphere with a radius of curvature of
1 inch (similar to the human eye), which was mounted on a 50g load cell. The disc was
visually lined up with the marked center of the hemisphere. A blunt pin was mounted
above the disc in the Bose 3100 ELF as shown in Figures 1 and 2. The load cell–
hemisphere–disc arrangement was manually raised until the disc was near the tip of the
pin. The Bose 3100 ELF was then used to move the pin slowly downward until the load
cell registered a downward force reading, indicating that the pin had come in contact with
the center of the disc, and then the pin was moved back up to the point just before it had
come in contact with the disc. Lastly, the pin was moved downward at a rate of
0.025mm/sec for a total of 0.15-0.20 mm. The downward force necessary to displace the
center of the disc into the hemispherical mount was recorded using the load cell. 2048
data points were recorded over 15 seconds and a 50Hz lowpass filter was applied to the
data at the recording site.

123

Figure C.1 Photograph and diagram of experimental setup.

A blunt pin was used to press the center of a test sample down into a hemispherical
mount while the transmitted force was recorded from a load cell below.

Figure C.2. Downward view of experimental setup.

The test sample can be more clearly seen in this downward view of the hemispherical
mount and load cell.
Pin
Test sample
Hemispherical
mount

Load Cell
124

Although discs with both 4mm and 5mm diameters were produced in order to explore the
effect of edge width, we found that the smaller vertical clearance between the center of a
4mm disc and the hemispherical mount limited the amount of data that could be recorded
before the pin came into contact with the bottom of the mount and overloaded the load
cell. Therefore, only discs with 5mm diameters are evaluated here.

For each of 14 different edge designs, 3 samples were tested, and tests were repeated 4
times consecutively on each sample. Three designs were also tested after being soaked
for 24 hours in 37º C phosphate buffered saline.

Data Processing

As the pin pushed the disc into the hemispherical mount, downward force was recorded
by the load cell. This force-displacement plot could be converted into a stress-strain plot
whose slope would be the device’s elastic modulus. Thus the slope of the force-
displacement line is related to the stiffness of the device, with steeper downward slopes
indicating stiffer devices and shallower slopes indicating more flexible devices. That
slope changes over time in each test, becoming steeper as the disc becomes more difficult
to deform downward.

In most cases, two distinct slopes were observed in each test: a shallower slope as the
more flexible edge deformed, and then a steeper slope as the center of the disc deformed.
Visual observations confirmed that the disc edge did deform first. In order to evaluate
the stiffness of the edge then, we have evaluated the slope of the initial portion of the
force-displacement line in each test.

The relevant data begins at a slightly different point in time for each test because,
although we began each test with the pin located close to the polyimide disc, it was
impossible to ensure that the pin was exactly the same distance from the disc in every
test. Data was trimmed automatically using Matlab 6.1 (The Mathworks Inc., Natick,
MA), to ensure consistency. The point where the slope dropped dramatically was
identified automatically and then data was trimmed to include only a series of 300 data
points beginning 600 points before the identified change in slope. In the few cases where
a change in slope was not identified automatically for any reason, that point was defined
manually. The trimmed data was normalized to begin at zero, filtered using a 10-point
moving average filter, and fitted with a linear regression.

Results

A representative force recording is shown below in Figure 3a. Force is increasingly
negative (downward), first with a shallow slope and then with a much steeper slope.
When downward displacement is stopped, noise increases as the Bose motor attempts to
maintain a constant displacement despite the reaction force of the disc. The pin is then
returned to its original starting position and the downward force on the load cell returns
125
to its original level. The “dropoff point”, starting point and ending point used to trim the
data are marked. The trimmed data is recorded over a region where the displacement rate
and direction are constant, so trimmed data can be plotted vs. Displacement as shown in
Figure 3b. Raw data, filtered and normalized data, and a linear regression are shown.

Figure C.3. Raw and processed data for a representative force recording.
(a) Downward force increases slowly as the edge of the disc deforms and then quickly once the center of
the disc begins to deform. Downward displacement of the pin stops and then the pin is moved upward to
its original position and pressure on the disc is released. (b) Data has been trimmed using the beginning
and end points shown. Raw data is filtered and normalized, then fit with a linear regression.

All 12 runs for each of 3 edge designs are shown in Figure 4, along with tests of the same
designs after a 24 hour soak in 37 º phosphate buffered saline. Different colors represent
a
b
Downward
displacement stops
Pin is returned to
original position
“Dropoff point”
Beginning/end points
for trimmed data
126
each of 3 different test samples for each design. Data is very consistent between multiple
repetitions of the same sample (indicating that the test is not altering the mechanics of the
device) and generally consistent between multiple samples of the same device. There are
slight variations in results between individual samples, but that is likely due to slight
variations in the alignment and setup of each sample rather than variations among the
physical devices.

The average slope for each device, both dry and after soaking are shown in Figure 5. N =
12 (3 samples, 4 repetitions each) for each, and error bars represent standard deviation.

Figure C.4. Three designs tested dry and after soaking.

Trimmed, filtered, normalized data for three designs (edge only, small via outer, and
wings 1-layer) tested dry and after a 24 hr soak in 37 deg. phosphate buffered saline.
Blue, red and green lines represent different samples.

127

Figure C.5. Average slope for three designs tested dry and after soaking.
N = 12 (3 samples, 4 repetitions each) for each. Standard deviation is represented by error bars.

In each case, the soaked device produced a shallower slope and could be described as
more flexible than its dry counterpart, but the relationship between the different designs
remained the same.

One representative force recording for each of 14 designs, tested dry, is shown below in
Figure 6.

….. ….. …..
..
..
..
128

Figure C.6. One representative force recording for 14 different designs.
Data has been trimmed, filtered and normalized.

The average slope for each design is shown in Figure 7. Data has been grouped into three
broad categories:
(yellow) Designs with “wings” cut through all layers,
(blue) Designs with any size or arrangement of vias on a thinned edge,
(green) Designs with a constant thickness throughout most of the device
Standard deviation is represented by error bars.

129

Figure C.7. Average slope for 14 different designs.
N = 12 (3 samples, 4 repetitions each). Standard deviation is shown with error bars.

Conclusions

Visual observations confirms that, particularly in devices with stress relief features of any
type along the edge, the edge of the disc deformed early in the experiment and the center
of the disc did not deform until several seconds later.

Data was consistent and repeatable across multiple repetitions of the same sample and
between multiple samples of the same design.

Soaking in 37º phosphate buffered saline softened the devices, but did not significantly
change their behavior or the relative stiffness between edge designs.

Designs can be grouped into 3 general groups: designs with a constant thickness
throughout most of the device (including devices where “wings” were cut only through 1
polyimide layer while a second polyimide layer remained whole), designs with some
arrangement of circular or elliptical vias on a thinned edge, and designs with “wings” cut
through all layers of the device. In general, devices with a constant thickness were the
stiffest. All designs with a thinned edge were slightly softer. Most arrangements of vias
on that thinned edge made the edge slightly more flexible, but differences between via
…. …. :::: --- == ++ ** /// ##
…. :::::
…..
:::::
…..

…..
:::::
---
===
+++
***
////
###

130
arrangements were generally not significant (with the exception of a double row of large
vias, which was softer than other via arrangements). Designs with wings cut through all
layers of polyimide were, by far, the softest.

Discussion

Using a thinner layer of material with any arrangement of vias softens the edge of the
device somewhat and cutting “wings” around the edge of device softens the edge
significantly.

We assume that increasing the flexibility of the edge will help to reduce mechanical
damage around the edge of the array, which suggests that retinal prosthesis electrode
arrays should include “wings” or, less preferably, a thinned edge to provide stress relief
around the edge of the device.

However, it is important to remember that the mechanism of mechanical damage to the
retina is not well understood and this study does not consider shear stress, slicing, or
other potentially negative effects that could be introduced inadvertently with these, or
any, stress relief features.

Abstract (if available)

Abstract Degenerative retinal diseases such as Retinitis Pigmentosa and Age-Related Macular Degeneration are, together, one of the leading causes of blindness in the United States. Retinal prostheses bypass the degenerated cells and apply electrical stimulation directly to the visual neural pathway, creating an artificial sensation of sight in subjects with retinal blindness. Current prostheses utilize a polymer electrode array, which is affixed to the retina using a retinal tack. The electrode array applies mechanical pressure to the retina, which can cause mechanical damage to the very cells which are required to transmit the applied signal. We can mitigate or relocate mechanical damage to the retina through electrode array design changes, but to do so, we must thoroughly understand the mechanical interface between the array and the retina.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Saliency based image processing to aid retinal prosthesis recipients

PDF

Towards a high resolution retinal implant

PDF

Manipulation of RGCs response using different stimulation strategies for retinal prosthesis

PDF

An in vitro model of a retinal prosthesis

PDF

PET study of retinal prosthesis functionality

PDF

Effect of continuous electrical stimulation on retinal structure and function

PDF

Understanding the degenerate retina's response to electrical stimulation: an in vitro approach

PDF

Investigation of the electrode-tissue interface of retinal prostheses

PDF

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

PDF

Cortical and subcortical responses to electrical stimulation of rat retina

PDF

Viscoelastic characterization of soft biological materials

PDF

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

PDF

Prosthetic visual perception: retinal electrical stimulation in blind human patients

PDF

Improving stimulation strategies for epiretinal prostheses

PDF

Protective mechanisms against oxidative stress in retinal pathogenesis

PDF

Oxygen therapy for the treatment of retinal ischemia

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

PDF

Intraocular camera for retinal prostheses: refractive and diffractive lens systems

PDF

Multi-scale modeling of functionally graded materials (FGMs) using finite element methods

PDF

Elements of photoreceptor homeostasis: investigating phenotypic manifestations and susceptibility to photoreceptor degeneration in genetic knockout models for retinal disease

PDF

Electrical stimulation approaches to restoring sight and slowing down the progression of retinal blindness

Asset Metadata

Creator Basinger, Brooke Christine (author)

Core Title Modeling retinal prosthesis mechanics

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 05/11/2009

Defense Date 03/18/2009

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

Tag finite element analysis,medical device design,OAI-PMH Harvest,retinal prosthesis,solid modeling

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Weiland, James D. (committee chair), Humayun, Mark S. (committee member), Meng, Ellis F. (committee member), Shiflett, Geoffrey R. (committee member)

Creator Email basinger@usc.edu,Brooke_Basinger@yahoo.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2231

Unique identifier UC1124809

Identifier etd-Basinger-2833 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-242613 (legacy record id),usctheses-m2231 (legacy record id)

Legacy Identifier etd-Basinger-2833.pdf

Dmrecord 242613

Document Type Dissertation

Rights Basinger, Brooke Christine

Type texts

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

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu