|
The Utah, cortically based visual neuroprosthesis system
With the successful
development of a penetrating microelectrode array for implantation in the
brain, artificial vision is ready to step beyond the original systems built
in the 1960’s. The development of the Utah artificial vision system is being
guided by four principles: 1) long-term safety and biocompatibility of the
implant, 2) vision capable of navigation without a guide dog, family member
or friend, 3) vision capable of reading printed text, and 4) a prosthesis
that is as unobtrusive as possible. As a general high-level description, the
Utah Artificial Vision system will consist of a micro-video camera hidden in
a pair of eyeglasses to transform light in the visual scene into electrical
signals, signal processing electronics to convert these signals into patterns
of electrical stimulation for the brain as well as a power source carried in
a shirt pocket, a totally implanted multichannel stimulator with power and
data to be delivered to the implant system via a radio-frequency telemetry
link, and an electrode array with 625 microelectrodes.
|
|
|
Figure 1:
Artist conception of an artificial vision system.
|
The Utah visual
neuroprosthesis program
Work is ongoing at the John Moran Laboratories at the
University of Utah to develop the cortically based visual prosthesis system
described above. The program conducts
research in the following broad areas:
|
|
Development
of improved electrode array architectures
|
|
|
Evaluation
of long term biocompatibility of implant systems
|
|
|
Basic
investigations of information processing in the retina and visual cortex
|
|
|
Development
of support instrumentation for a cortical neuroprosthesis
|
|
|
Conduct of
experimentation in human volunteers
|
History of visual
neuroprosthetics
The historical foundations for a cortically based
visual prosthesis. The concept
of cortically based artificial vision had its origins in studies of the
functional architecture of the cerebral cortex that were started in the early
twentieth century, and were pursued by Wilder Penfield. Penfield and Rasmussen observed the behavioral
consequences of electrically stimulating various regions of cerebral cortex
and noted that electrical stimulation of the surface of the visual cortex
generally evoked the perception of points of light (called phosphenes) at
specific regions in space. He also
observed that the location of the phosphenes in front of the observer
appeared to be deterministically related to the region of primary visual
cortex which was being electrically stimulated (a ‘visuotopic’
organization). These preliminary observations
on the visuotopic organization of visual cortex have been extended by many
others using electrical stimulation of human visual cortex, by recording of receptive fields in
primate visual cortex, and recently validated in humans with fMRI.
The observations
of the visuotopic organization of electrically evoked phosphenes have led a
number of investigators to propose that electrical stimulation of visual
cortex via arrays of electrodes might provide the profoundly blind with a
limited form of functional vision.
Subsequent experiments in the late sixties and early seventies by
Brindley, Dobelle, Pollen, and their co-workers demonstrated that a field of
individual phosphenes could be evoked by stimulating visual cortex with an
array of electrodes implanted subdurally over its surface. These studies confirmed the visuotopic
organization of visual cortex, and demonstrated that subjects could
assimilate information that was delivered to the visual cortex by electrical
currents passed via groups of electrodes.
Dobelle’s subjects were able to read phosphene evoked Braille
characters at a faster rate than they could using their tactile sense. However, it was also learned that currents
in the milliampere range were required to evoke individual phosphenes, and
that currents passed through groups of electrodes that were spaced too close
to neighboring electrodes produced highly non-linear interactions between the
location and character of the evoked phosphenes. It became clear from these experiments that stimulating visual
cortex via an array of surface electrodes would not be an effective means to
produce a useful visual sense in individuals with total blindness.
Recent experiments by Schmidt et al. have caused renewed interest in
cortically based visual neuroprosthetics.
In their most recent experiments, they chronically stimulated visual
cortex of a profoundly blind human volunteer with groups of microelectrodes
that were designed to penetrate the cortex to the level of its normal
thalamic input. The Schmidt et al.
study is the most complete to date that explores the psychophysical percepts
produced by intracortical microstimulation.
It has long been known that neural stimulation via penetrating
electrodes occurs with electrical currents that are much smaller than those
used to excite neurons via surface stimulation. Schmidt and his coworkers
validated this observation and demonstrated that phosphenes could be evoked
with currents that were orders of magnitude lower than those used with
surface stimulation (their lowest thresholds were in the 1 to 10 microampere
region). More importantly, they
showed that simple patterned perceptions could be evoked by current
stimulation via small groups of these microelectrodes (simple ‘lines’ were
evoked by simultaneously stimulating a set of six electrodes that were close
to each other). While their data was
anecdotal (in that it came
from a single subject), they also demonstrated that electrical
stimulation of pairs of electrodes that were separated by 250 microns often
evoked two discriminable percepts, while stimulation of electrodes separated
by 500 microns almost always evoked two discriminable phosphenes.
Unfortunately, the electrode arrays used by Schmidt et al. were too sparse to
allow them to answer the key question upon which a cortical approach to
artificial vision must be based: does patterned electrical stimulation via a
high electrode count electrode array evoke discriminable patterned percepts,
or nondiscriminable blobs of light?
If this psychophysical experiment can be performed, and the former
result maintains, the physiological foundation of cortically based artificial
vision could be established.
In order to
answer this critical question (and many others associated with phosphene
psychophysics), researchers need a new class of tools: arrays of
microelectrodes that can be safely implanted into the visual pathways, and
that will allow periodic injections of electrical currents at many closely
spaced sites. Such arrays could
eventually also form the cornerstone of visual neuroprosthetic systems. It therefore is clear that progress in
cortically based artificial vision systems is directly linked to progress in
the development of high electrode count microelectrode arrays, and over the
past decade, such arrays have been developed.
Summary of Utah’s
accomplishments
Because of the
importance of the neural interface in a visual prosthesis, much of this
section will focus on a penetrating cortical electrode array of our own
design: the Utah Electrode Array, or UEA.
We will discuss considerations that were used in its design, how the
array can be implanted in the visual cortex, and its biocompatibility as
revealed by histological and electrophysiological experiments. Finally, we will summarize psychophysical
experiments we have performed that provide rough estimates about the number
of electrodes that may be required to restore some functional vision in an
individual with profound blindness.
The historical motivation behind
much of what we have developed at the University of Utah has been to create
experimental systems that will allow researchers to better understand the
vertebrate visual system though electrophysiological and behavioral
experiments. These systems have been
used for many years, and we now look forward to applying them to study human
psychophysical questions that directly relate to the development of a
cortically based vision neuroprosthetic system.
A. Electrode Arrays
The cornerstone of a visual neuroprosthesis is the
interface between the functioning neurons in the visual pathways and
implanted devices that can selectively excite these neurons. The remaining elements in such a neuroprosthesis
will require modifying existing technologies, and as such, are simply matters
of engineering development. This
neural interface must individually stimulate a very large number of neurons
that have retained function even though more distal neurons have been
irreparably damaged by the etiology of the blindness. This interface will bypass the
malfunctioning distal components in the visual pathway and directly excite
neural pathways that are proximal to the implant site. Recent work at the University of Utah (Jones, Campbell et al. 1992) ,
at the University of Michigan by Wise et al. (Hoogerwerf and Wise 1994) , and Stanford University (Kewley, Hills et al. 1997) has
focused on the use of silicon as an electrode material from which high electrode
count, penetrating electrode arrays can be fabricated. Silicon is highly biocompatible (Stensaas and Stensaas 1978; Yuen, Agnew
et al. 1987; Schmidt, Horch et al. 1993) , can be micromachined using
standard microfabrication technologies, and can incorporate integrated
electronics. The Michigan and
Stanford electrode arrays have been built to take advantage of the planar
photolithographic manufacturing techniques used in the semiconductor industry,
while the Utah arrays were designed ‘from the ground up’ to meet the needs of
a neural interface for the cerebral cortex.
As such, new manufacturing techniques had to be developed in order to
build this device. These techniques have been described elsewhere (Jones, Campbell et al. 1992) .
|
|
|
|
The Utah Electrode Array (UEA), shown above, provides a multichannel
interface to the visual cortex. It
has a large number of 1.5mm long electrodes (typically 100 in a 10 x 10
square grid) that project out from a very thin (0.2mm) substrate and that are
separated from each other by 0.4mm. The tips of the electrodes are metalized
with platinum to facilitate electronic to ionic transduction. As the array’s substrate must rest on the
cortical surface, minimizing its thickness (its super-cortical profile) was
an important design consideration: if it was too thin, it might break upon
insertion, if it was too thick, the dura and the skull would produce a
constant downward force on the array, tending to push it into the
cortex.
The large number of penetrating electrodes in the UEA presents a very
large surface area to the cortex and the implanted array tends to ‘self
anchor’ to the cortical tissues. Such
an array has the strong advantage that it ‘integrates’ with the cortical
tissues and, therefore, ‘floats’ with respect to the cranium. As the cortex moves due to respiration and
blood pumping, or to displacements in the skeletal musculature, the array
moves with it, thereby producing little or no relative motion between the
electrode tips and the neurons near its active tips. This design feature therefore, produces an
extremely stable interface with the surrounding neurons.
The penetrating electrodes in an implanted array must compromise as
little cortical volume as possible (ideally zero). Thus, each needle must be made as slender as possible yet
retain sufficient strength to withstand the implantation procedure. Further, consistent with concept of ‘blunt
dissection’ used by neurosurgeons, these penetrating structures should
displace the tissues they are inserted into rather than cut their way through
them. Thus, the needle should be conical
and have a very sharp tip rather than a planar or knife-like geometry. They must also be strong enough so that
they are not deflected by the tissues they are inserted into. The needle electrodes in the UEA meet
these criteria and are about 80 microns in diameter at their bases. They taper to a sharpened tip that has a
radius of curvature of two to three microns.
Shown below is an electron micrograph of the tips of these
electrodes. Electrodes with these
dimensions have been shown to be sufficiently strong to withstand insertion
into materials that are considerably less compliant than cortical tissue
(cork, balsa wood, even egg shell).
These electrodes do not bend during the insertion process, and they
only displace about 4% of the cortical volume into which they are inserted.
|
|
|
|
Each electrode is electrically isolated from its neighboring electrodes
with a moat of glass that surrounds each electrode’s base. Each electrode has a bonding pad on the
rear surface of the substrate.
Conduction of signals along the length of the silicon needles is
achieved by the doped silicon used in its fabrication. The entire electrode array (with the
exception of the platinum coated tips) is insulated with a 2 micron thick
coat of silicon nitride. An
electrical connection is made to each electrode by bonding an insulated 25
micron diameter wire to each bond pad, and connecting these wires to a
percutaneous connector. The rear of
the array (with bonded lead wires) is encapsulated with a silicone
elastomer. Electrode impedances
(measured with a 100 nanoamp, 1 kHz sine wave current) are typically in the
100 to 500 kW range.
We have used both chronic and acute arrays in our experiments. The acute array has all 100 electrodes
brought out to a small printed circuit board containing four, 26-pin IDC
connectors. Our original chronic
system had only eleven of the 100 potentially functional electrodes brought
out to a Microtech connector via eleven, 25 micron diameter, platinum-iridium
lead wires (the twelfth was a platinum-iridium reference wire). Our most recent design brings out 38 of
the 100 electrodes to a custom made connector. The connector is integrated into a titanium pedestal that is
mounted with titanium bone screws to the cranium. The figure below shows a
photograph of this ‘tulip’ version of our chronic electrode assembly.
|
|
|
|
When we
have received IRB approaval to perform human experimenation, the UEA will be used in most of
our acute and chronic work. However,
we will also conduct a series of acute experiments using our ‘Utah Slant
Array’, or USA. These experiments
will explore the relationship between phosphene thresholds and electrode
shank length. The USA is ideally
suited for these experiments, as it contains electrodes that vary in shank
length from 0.5 mm to 1.5 mm. An
electron micrograph of the USA is shown in the figure below. The USA is virtually identical in all
aspects to the UEA, except in the varying length of the electrode shanks.
B.
Surgical
Issues: implanting high count electrode arrays
A cortically based visual prosthesis will be a highly invasive system
that, eventually, must contain active integrated electronic circuitry. While relatively large electronic and
mechanical systems have been implanted in the body (pacemakers, artificial
joints, and the cochlear prosthesis), miniaturized devices of the mechanical
and electronic complexity of the UEA have yet to be implanted on a long term
basis in any part of the human body, much less the brain. Whether the implant site is intended to be
the cortex or the retina (or the optic nerve (Veraart, Raftopoulos et al. 1998) , safe and effective
surgical procedures that are cost-effective must be developed and validated
in animal models before they are attempted in human volunteers.
The Utah researchers have developed new surgical
techniques and tools that enable the UEA to be implanted in cortical
tissues. Even though the individual
electrodes of the UEA are extremely sharp, early attempts at implanting large
numbers of them into the visual cortex only deformed the cortical surface and
resulted in incomplete implantation.
Further, the compression of the cortical surface produced by slow
mechanical insertion can injure blood vessels, causing intracranial
hemorrhage and cortical edema.
Because the brain is a viscoelastic material, it will behaves in a much
more rigid fashion if the electrodes can be inserted into the cortex at a
very high velocity. We have developed
a unique surgical instrument based upon this concept that appears to
circumvent the above mentioned problems: a system that rapidly inserts the
UEA into the cortex (Rousche and
Normann 1992) . A drawing of
the pneumatically actuated insertion tool we have developed is shown
below. Array insertion is achieved by
a transfer of momentum between an accelerated piston, and an “insertion mass”
that rests against the back-side of the electrode array which is to undergo
implantation. When the momentum
transfer takes place, the array is rapidly inserted into the cortical tissues
in about 200 microseconds. The
insertion is so rapid that the viscoelastic properties of the cortical
tissues cause the cortex to experience only slight mechanical dimpling and
the insertion is generally complete.
Occasionally implantation of the UEA through surface vasculature is
accompanied by a small amount of subpial bleeding, but this typically
resolves itself, and single unit recordings of neural activity can often be
made within hours after the surgical procedures are completed. The complete implantation procedure (from
anesthesia induction to recovery) takes about four hours from initial
incision to final closure.
|
|
|
|
C. Chronic histology
A neuroprosthetic system must be implanted into the nervous system and
remain fully functional for periods that will eventually extend to many
decades. This consideration places
unique constraints on the architecture, materials, and surgical techniques
used in the implementation of the neural interface. Inattention to any of these issues can result in chronic
inflammatory responses around the implant site and generate a thick capsule
surrounding each electrode. Because
the UEA is a unique implantable structure, the consequences of its presence
in the body for extended durations must be evaluated. We have studied its biocompatibility using
basic histological and electrophysiological techniques.
The materials of which the UEA is built: silicon, silicon nitride,
silicon dioxide, platinum, titanium, tungsten, and silicone are known to be
well tolerated by the CNS (Stensaas
and Stensaas 1978; Edell, Toi et al. 1992) . Our six month histological experiments support this notion (Schmidt, Horch et al. 1993) . The figures below show hematoxylin and
eosin stained sections of such tissue.
A thin capsule (2-5 microns thick) forms around each electrode track,
but neuronal cell bodies are typically seen in close apposition to the
electrode tracks (Turner, Shain et
al. 1999) . In fact, in
sections where the tracks are similar in diameter to blood vessels, it is
often difficult to tell a track from a vessel. These figures illustrate that the electrodes of the UEA are not
displaced laterally by the cortical tissues during implantation. The histological photographs below are
examples of a particularly benign tissue response. However, we also have histological samples showing gliosis,
buildup of fibrotic tissue between the array and the meninges, subtle array
displacement in the cortex, and the presence of a small number of red blood
cells (or hemosiderin) in some
tracks. While the histology we have
performed to date supports the use of the UEA in acute human experimentation,
before we can consider its use in chronic applications, more work will be
done to ensure that histological findings like those shown in the histology
below can be achieved on every implantation. Specifically, while the
hematoxylin and eosin stain has been useful in the gross cytological analyses
we have performed to date, it is wholly inappropriate as the sole means of assessing
biocompatibility. We must extend
these studies and use immunofluorescent techniques for evaluating structures
and specific cell types such as capillaries, astrocytes, macrophages,
microglia, meningeal cells and endothelia, all of which have been shown to
play a role in the brain host response to biomaterial implantation regardless
of the material employed. Our preliminary histological observations with
hematoxylin and eosin have been elaborated upon in a recent study by Maynard
on the long term consequences of UEA implantation (Maynard, Fernandez et al. 2000) .
|
|
|
|
D. Neuronal recordings
Acute recording capability
An excellent index of the biocompatibility of a cortical implant is its
ability to record single- and/or multi-unit activity from the neurons near
the electrode tip for prolonged periods of time. If the materials used in the array, or the implantation techniques
are not biocompatible, neuronal processes close to the electrode track will
degenerate and it will not be possible to record single-unit activity. However field potential activity located
far from the electrode tracks might still be recorded from functioning
neurons, and such neurons may still be stimulated effectively in a
neuroprosthetic application.
We have recorded responses from the UEA in both acute and chronically
implanted animals to better understand its short and long term
biocompatibility. Specimen responses
with high, medium and low signal-to-noise ratios that were evoked by a bar of
light moving across the receptive fields of visual cortical units are shown
below. To generalize these findings,
we have tabulated below the recording quality of our past 17 acute
implants. In this table, we
classified responses on each electrode using the scale in this figure.
|
|
High
SNR >3
|
Medium
3< SNR <1.5
|
Low
SNR <1.5
|
Inactive
|
Useful SNR>1.5
|
|
Minimum
|
1%
|
3%
|
0%
|
16%
|
10%
|
|
Maximum
|
53%
|
52%
|
35%
|
88%
|
68%
|
|
Mean +/- S.D.
|
20+/-15%
|
16+/-13%
|
8+/-9%
|
56+/-23%
|
36+/-19%
|
Table
legend. Recording statistics from our
past 17 acute implants illustrating the mean and standard deviation of the number of electrodes with high, medium
and low signal-to-noise ratios for each of the arrays. Minimum and maximum are the lowest and highest
number of electrodes with the indicated quality of recording on each of the
arrays. Some arrays had a few broken
electrodes, so data is presented as percent of potentially functional
electrodes.
In our best acute implantations, we have been able to record good
single- and multi-unit responses from 68% of the electrodes (because of the
curvature of the gyri in the cortex, it is not possible to implant all 100
electrodes in a given gyrus and some electrodes end up in sulci). More typically, however, we are able to
record good quality single- and multi-unit responses from 36% of the
electrodes in a given array. Global
field potentials, non-discriminable multi-unit responses, and an absence of
responses are recorded from the remaining electrodes.
Chronic recording capability
We have used chronic implantations of the UEA in visual and auditory
cortex and motor cortex to better monitor the stability and long term
biocompatibility of the UEA. We have recorded multi- and single-unit evoked
activity and identified single-units (based on response kinetics) on many of
our electrodes. We have been able to
record single- and multi-unit responses for over three years (the longest
intervals studied, see examples below).
The presence of single- and multi-units on many electrodes, and the
stability of these identified units over periods of months provides the most
compelling evidence for the biocompatibility of the UEA.
E. Behavioral Experiments
Useful function will be achieved in a visual neuroprostheses by
injection of electricalof electrical
currents into the visual pathways through large numbers of electrodes. Current injections can produce short term
and long term complications depending upon the levels of the currents that
are injected (McCreery, Yuen et
al. 1994) (or see Agnew, Ch6 for a review (Agnew and McCreery 1990) ). In order to determine the levels of current injections via the
UEA that are required to evoke sensory percepts, we have conducted a series
of behavioral experiments (Rousche
and Normann 1998) . Ideally, these experiments would be performed with
implants in visual cortex, but, because of the greater ease of providing
auditory stimuli to a behaving animal, we targeted the auditory cortex as our
implant site and used auditory stimulation rather than visual stimulation.
Three animals
were trained over a one to two month period to lever press as a result of
auditory stimulation, and auditory thresholds were measured. Trained animals that performed this task
at 90% correct were chronically implanted with the UEA. Following implantation of the UEA, we
interspersed current injections via the UEA and conventional auditory
stimulation. Current injections that
evoked auditory percepts should have resulted in a ‘positive’ lever press in
the trained animals. The chronic
percutaneous connectors used in these experiments had limited pin counts and
permitted access to a total of only 22 of the active electrodes in these
animals. Behavioral thresholds,
measured on different electrodes as the amount of charge injected per phase
of stimulation, ranged from 1.5 nC/phase up to 26 nC/phase. The average threshold, measured in 71
sessions was 8.9 nC/phase. The
stability of four of these threshold measurements was recorded in one animal
over a three month period. Over the 100 days monitoring interval, thresholds
varied by no more than 50%. This provides an additional index of the
biocompatibility of the “floating” array design.
F. Human Psychophysical Experiments:
visual performance versus number of pixels
Before one can consider conducting experiments on human
subjects that relate to a cortically based visual prosthesis, one should have
some rough idea about how many electrodes would be required to restore a
useful visual sense in a profoundly blind subject. We have performed a number of human psychophysical experiments
with a pixelized vision simulator to get rough estimates of this number (Cha, Horch et al. 1992; Cha, Horch et
al. 1992; Cha, Horch et al. 1992) .
The simulator, shown to the right, is portable, and consists of a
battery powered video camera and video monitor. The monitor is masked with one of a number of perforated foil
masks that have from 100 to 4000 perforations. Using this simulator, six student volunteers were asked to read
text out loud from a computer monitor, and, when trained, to navigate through
a complex, programmable maze. After
this task had been performed, our subjects were challenged to navigate
through normal living environments.
|
|
We learned that reading and navigation
performance improved as the number of pixels increased, but that it began to
plateau at about 625 pixels (simulating a 25 x 25 array of phosphenes) (Cha, Horch et al. 1992; Cha, Horch et
al. 1992; Cha, Horch et al. 1992) . The study also suggested that with
as few as 100 pixels, trained subjects could navigate our maze without error,
and that they could navigate in normal living environments, avoiding objects
in their paths. While we fully
appreciate the limitations of this simulation, it indicates that significant
visual task performance could be achieved with modest amounts of visual
input.
Overview of human
engineered and biological approaches to sight restoration
This past decade
has seen a gradual shift in vision research from basic studies of the cell
biology and neuroscience of the visual pathways to application of this
knowledge base to restoring sight to the profoundly blind. This shift has taken two broad directions:
human engineered or neuroprosthetic solutions, and biologically inspired
solutions.
|
New approaches to site
restoration for the profoundly blind
|
|
Human engineered / Neuroprosthetic approaches
|
Biological approaches
|
|
Cortically Based (Utah’s
approach)
|
Retinal Transplantation
|
|
Retinally Based
|
RPE transplantation
|
|
Optic Nerve Based
|
Growth Factors
|
|
|
Gene Therapy
|
As this informational web-site is focused
on the human engineered approach, recent progress in the biological
approaches will only be described briefly below. Further, as Utah’s interests are focused on a cortical
neuroprosthesis, this review of the retinal and optic nerve approaches will
also be intentionally brief.
Retinal based
visual neuroprosthetic research.
Over the past few years, the research efforts of an increasing number
of laboratories around the world have been committed to the development of a
retinal based visual neuroprosthesis. The lay media has recently featured this work on television and
in the press, and has created a great deal of public interest in the
potential of this neuroprosthetic technology. This retinal approach has
taken two basic directions: an epiretinal approach where an electrode array
will be placed on the vitreal surface in an effort to stimulate ensembles of
ganglion cells (or possibly bipolar cells), and a subretinal approach where
an electrode array is intended to be implanted between the retina and the
pigment epithelium and extrinsic currents are intended to stimulate either
remnant photoreceptor inner segments, or bipolar cells.
The epiretinal approach is being
aggressively pursued in America by teams at Harvard/MIT and at Johns
Hopkins/the Mann Foundation. This
approach is also being aggressively pursued in Germany with substantial
financial support from the German government. While these teams have yet to publish their efforts to develop
a long lasting and efficacious electrode array that could realistically be
used as a chronic retinal implant, the American researchers have performed
pilot retinal stimulation experiments in human volunteers. The Harvard/MIT and the Hopkins groups
have shown that passing transretinal electrical currents can evoke visual
percepts in individuals with retinitis pigmentosa, and both have shown that
very simple patterned stimulation evokes elementary patterned percepts in
these volunteers. The Harvard/MIT and
the German epiretinal projects have also focused on developing surgical
procedures that will ensure that the implant will be closely apposed to the
retina (i.e., removal of all remnant vitreous humor without damaging the
delicate retinal tissue). The German
counterparts have also directed much of their efforts and research dollars on
a ‘retinal encoder’ system that will be used to remap visual input into
appropriate patterns of electrical retinal stimulation. Because of various non-visuotopic mapping
problems, such a video encoding / neural stimulator system is likely to be
needed in any visual neuroprosthesis system (retinal or cortical).
The subretinal approach was originally
proposed by Chow in America, and it has been adopted by another German team
led by Zrenner. The original idea
suggested that a suspension of non-powered microphotodiodes could be inserted
into the subretinal space, and that these photodiodes could passively replace
the function of lost rods and cones by directly stimulating second order
neurons in the retinal pathway. Chow
and Zrenner have shown that flat silicon discs can be implanted subretinally,
and have developed surgical techniques to achieve this. Recently, both groups seem to have come to
understand that silicon photodiodes are many orders of magnitude less
sensitive than rods and cones, exclusive of the difficulty of transducing
electronic currents into ionic ones and of stimulating the remnant
neurons. Both groups seem to be
acknowledging that the subretinal implant must be externally powered. All
players in the retinal game have begun working with a variety of in vivo and in vitro retinal models in their efforts to develop electrode
arrays that can effectively and selectively stimulate retinal pathways.
An optic nerve
based visual neuroprosthesis. A recent study has described phosphene
generation via optic nerve stimulation.
These experiments were conducted in a blind human volunteer who was
chronically implanted with an extracellular ‘cuff’ electrode that surrounded
the optic nerve, and which contained four surface electrodes. Stimulation with various current levels
through different combinations of electrodes evoked a variety of spatially
distributed phosphenes. However, the
researchers were unable to obtain a rational visuotopically organized
relation between stimulation parameters and evoked phosphenes. This approach may show promise if an
ultra-high electrode count array of penetrating electrodes were to be
implanted in the optic nerve. We have
implanted an array of 100 electrodes with varying
length shanks in the sciatic nerve (the ‘Utah Slant Array’, or ‘USA’). Because this array architecture
distributes the active tips of the electrodes uniformly throughout the entire
nerve, we were able to selectively drive individual muscle groups. Although principally
engineered as a peripheral nerve interface, such an array might form
the basis for an optic nerve based visual prosthesis.
Pro’s and Con’s of
different approaches to sight restoration
The Pros and Cons of these human
engineered approaches: Research
in visual neuroprosthetics falls into the four general camps outlined
above. Each has advantages and
disadvantages that are summarized on the following page. It is clear that all
four implant sites have common problems and advantages, and some that are
unique. Arguments can be advanced in
support of each approach. One common
problem is that all approaches are expected to be most beneficial to those
with acquired rather than congenital, complete blindness. However, the observation that a cortical
approach might provide an effective therapy for most sources of blindness
(and the only therapeutic approach for many) is a compelling argument to
support this approach to a cortically based visual prosthesis.
|
Visual
cortex approach:
|
|
|
Pros
|
Cons
|
|
|
The
only therapeutic approach for individuals with non-functional retinas
and/or optic nerves.
|
|
|
Implant
site is robust and protected by skull
|
|
|
Easy
surgical access (demonstrated)
|
|
|
High
density electrode implantation has been demonstrated
|
|
|
Phosphene
thresholds are low (in 1-10 microamp range)
|
|
|
Electrode
array architecture is likely to be well suited for application in other
sensory or motor regions of the cerebral cortex.
|
|
|
|
Stimulation
site far from photoreceptors (no retinal or thalmic processing)
|
|
|
Possibly
poor visuotopic organization
|
|
|
Problems
of multiple feature representations in V1 (color, lines, motion, ocular
dominance)
|
|
|
Societal
phobias about "brain implant"
|
|
|
Significant
consequences of surgical complications.
|
|
|
Epiretinal
Approach
|
|
|
Pros
|
Cons
|
|
|
Stimulating
close to photoreceptors so one can take advantage of native processing
power in thalamus and cortex.
|
|
|
Surgical
complications not necessarily as significant as cortical approach.
|
|
|
|
Requires
functional optic nerve pathway.
|
|
|
May
stimulate optic nerve fibers rather than cell bodies: this will greatly
complicate visuotopic organization.
|
|
|
Hard to
imagine how sacadic eye motions will not cause very high sheer loads on
implanted arrays (and eventual dislodging of array).
|
|
|
Difficult
surgical access.
|
|
|
Difficult
to adhere electrode array to retina.
|
|
|
Subretinal
approach:
|
|
|
Pros
|
Cons
|
|
|
Stimulating
closest to photoreceptors so one can take advantage of retinal, thalmic
and cortical signal processing.
|
|
|
If
bipolar cells can be directly stimulated, retinotopic organization
should be preserved.
|
|
|
Surgical
complications not necessarily as significant as cortical approach.
|
|
|
|
Requires
functional retina and optic nerve pathway to convey signals to cortex.
|
|
|
Blockage
of nutrients from choroid to remnant retina by the implant
|
|
|
Very
complex surgical access
|
|
|
Can't
| | | |
