The field of retinal prosthetics began about 20 years ago, largely because of advances in microelectronic technology which permitted the development of electronically sophisticated devices that were small enough to be implanted into the eyeball. The remarkable success of cochlear implants, which has been achieved without a large number of electrodes or very advanced electronic technology, served as a beacon for the emerging field of visual prosthetics. This chapter provides an overview of the challenges and achievements in the development of visual prostheses, which have the potential to provide vision to patients for whom there is otherwise little opportunity for significant rehabilitation. Potential sites for implementation of a visual prosthesis include the subretinal space, epiretinal surface, optic nerve, lateral geniculate nucleus (LGN), and visual cortex ( Figure 75.1 ).
Blindness is one of the most common forms of disability, and in industrialized countries retinal disease accounts for the majority of blind patients. Age-related macular degeneration (AMD) and retinitis pigmentosa (RP) are the two retinal diseases causing blindness that are considered to be potentially treatable with a retinal prosthesis. There are roughly 2 million Americans with AMD, and the percentage of affected individuals is expected to increase by 50% by the year 2020. RP is the leading cause of inherited blindness in the world, affecting roughly 1.7 million patients. The cost to the US government to provide support services for the blind is enormous, reaching $4 billion annually.
AMD and RP cause blindness because of grossly similar pathologies ( Figure 75.2 ), although the mechanisms of injury are different. Both diseases cause blindness because of a loss of the rods and cones (i.e., the photoreceptors), which are the only cells in the retina that can convert incoming light into neural signals that create conscious visual perception. The neural signals are propogated to retinal ganglion cells (RGC) in the inner retina, which connect the eye to the brain and remain relatively healthy in AMD and RP. A prosthesis can potentially restore vision to patients with these diseases by providing electrical stimulation to the RGCs, which will then conduct the visual information to the brain. Diseases that damage the inner retina or optic nerve, like diabetic retinopathy and glaucoma, would not be amenable to the use of a retinal prosthesis.
The belief that there was “sparing” of RGCs was based upon the interpretation of standard histopathology of the retinas of affected patients ( Figure 75.2 : upper right and lower). More careful study of such retinas, however, showed the “sparing” to be relative. For RP, there is a loss of 30–70% of the RGCs; there is greater loss of RGCs in the peripheral retina, and cell loss is greatest in more advanced cases, especially in the X-linked and recessive forms of RP. The loss of RGCs is the result of anterograde transsynaptic degeneration, invasion of inwardly migrating retinal pigment epithelial cells into the blood vessels of the inner retina, and perhaps compression of axons due to altered anatomy of the inner retina. In patients with severe RP, only about 300,000 of the average 1.2 million RGCs in normally sighted humans survive. This degree of survival would still seem to be adequate to support the delivery of a substantial amount of visual information to the brain. By comparison, the auditory nerve contains roughly 30,000 cells, and stimulating some fraction of these cells has been sufficient to provide useful hearing for deaf patients, although admittedly the more complex visual sense will require more detailed information transfer.
Robert Marc has described in detail a predictable and orderly “reorganization” of molecules, synapses, cells, and networks in retinas following degeneration of photoreceptors. The neural retina initially responds to loss of photoreceptors by showing subtle changes in neuronal structure, like neuronal swelling and disruption of microtubular structure (i.e., phase I changes). In phase II, there is death of photoreceptors – first rods, then cones. The loss of cones is followed by whole-scale reorganization of retinal cell layers and interconnections. Prior to their death, the metabolically stressed photoreceptors sprout neurites that extend, quite anomalously, up to the inner plexiform layer and ganglion cell layers. As the photoreceptors die, the bipolar cells retract their dendrites; the horizontal cells retract their dendrites within the outer plexiform layer while (anomalously) extending axonal processes and dendrites toward the inner plexiform layer. The Müller cells increase synthesis of their intermediate filament proteins and extend processes beneath the retina that form a dense fibrotic layer within the subretinal space, which would presumably complicate attempts to stimulate the surviving cells of the outer retina from the subretinal space.
In phase 3, the reorganizing retina displays widespread sprouting of new neurites, migration of neurons into ectopic locations, development of new and aberrant synapses, creation of reciprocal synapses, “synaptic microneuromas” ( Figure 75.3 ), and widespread death of all neuronal cell types in the retina.
Furthermore, the degeneration of photoreceptors alters the physiology of the surviving RGCs, which develop an increased spontaneous firing rate. The increased spontaneous activity might produce “noise” within the signal transduction pathway, which could complicate the attempt to create a useful visual image. As such, a new body of research is being pursued within the field of retinal prosthetics to define the response properties of degenerating retinas, with special emphasis on the response characteristics of retinal neurons following electrical stimulation.
More specific rationale for treatment of retinal blindness with a retinal prosthesis
Guidelines for considering the use of a retinal prosthesis should require that a patient had normal vision at some point in life. This provides assurance that the complex series of interconnections between the photoreceptors and the primary visual cortex had at one time been properly established.
RP is widely considered to be the primary target for a retinal prosthesis because affected patients often become more severely blind than patients with AMD. Thus, the Food and Drug Administration would almost certainly require a greater demonstration of safety and efficacy for any proposal to use a retinal prosthesis for AMD, since these patients generally have better vision and therefore would be taking a greater risk than patients with RP.
The hope that vision with spatial detail can be achieved is based upon the presence of a predictable topographic order along the afferent visual pathway between the photoreceptors and the visual cortex. It therefore seems reasonable to assume that direct electrical stimulation of bipolar cells or RGCs (but not their overlying axons) might generate percepts at locations within the visual field similar to those which would have been obtained by photic stimulation in the same area ( Box 75.1 ). Indeed, human patients who have been severely blind for decades from RP have seen photopsias of varying degrees of detail (and more, see below) following electrical stimulation of the retina by numerous groups.
In patients with photoreceptor loss, a retinal prosthesis has the potential to improve vision by delivering electrical stimulation to retinal nerve cells that survive the degeneration of photoreceptors
The rationale for use of a retinal prosthesis is based on the premise of a predictable topographic order extending from cells of the retina, especially the retinal ganglion cells (RGCs), to the visual cortex, such that electrical stimulation of the retina in a specific geometric pattern could yield percepts of a similar geometry
Successful candidates for a prosthesis should have had normal vision at some point in their lives and a substantial survival of RGCs to make it possible to deliver electrical input to the brain by electrically stimulating the retina
There is relative preservation of RGCs in age-related macular degeneration, but in retinitis pigmentosa there is significant “reorganization” of the retinal neurons and glia of the middle and inner retina, including sprouting of new neuritis by neural cells, development of aberrant synapses, and significant cell death (including in the RGC layer), which complicates the goal of creating useful vision for blind patients with a retinal prosthesis
Diseases that cause significant damage to the optic nerve, like glaucoma, could not be treated with a retinal prosthesis
More detailed considerations of prosthetic intervention in AMD
Visual loss with AMD is limited to central vision, and although patients can be significantly handicapped for tasks like reading, retention of peripheral vision provides them with some degree of independence. A prosthesis for patients with AMD would only be helpful if it could improve central vision, which is a much more demanding goal than helping patients with RP by providing relatively coarse visual input to help with navigation.
The same rationale of relative preservation of RGCs for use of a retinal prosthesis has been offered for AMD as for RP, although the story is more complex. In the aging retina, photoreceptor loss begins in the parafoveal region (initially inferiorly) and is dominated by loss of rods. Early on, this cell loss can occur without RPE abnormalities, although in many cases RPE abnormalities, mostly drusen, are evident and constitute perhaps the most important clinical signs of dry AMD. For wet AMD, the pathology is much more severe, with marked loss of photoreceptors and moderately severe (50% loss) of RGCs overlying the areas of photoreceptor degeneration, although, in some cases, the inner retina can be relatively spared. For dry AMD, these surviving photoreceptors also may not have normal synaptic connections ; and unlike RP, there is only scant physiological evidence that RGCs can be stimulated to produce vision in patients with AMD. There are also a number of other anatomical features of the central macula that will likely complicate efforts to obtain vision in the range of 20/400 or better for patients with AMD. First, the anticipation that patients might see a geometrically similar (pattern to the stimulus) pattern is confounded by the fact that the RGCs are stacked upon one another within the parafovea. Thus, there may be a lack of topographical order for the centralmost RGCs such that adjacent RGCs might not correspond to adjacent points in the visual field.
For patients with wet AMD, the scarring that follows the hemorrhages can severely distort the retinal contour ( Figure 75.4 ), which to some extent would complicate surgical positioning or attachment of the prosthesis to the inner retinal surface. As such, the anatomy of each patient with AMD will have to be considered individually, and it must be assumed that some severely blind patients with AMD will not be good candidates for a prosthesis.
Attempts to assist patients with AMD or RP have included use of optical and electro-optical devices, such as telescopes, closed-circuit monitors and mobility training, primarily to learn how to use a cane for walking. These strategies can be beneficial but are limited in the functional gains that can be made. Newer rehabilitation strategies are also being explored, including sensory substitution; transplantation of stem cells, embryonic, or adult cells; and molecular genetic approaches, which may offer the best long-term treatment option. A prosthesis also has the significant advantage that restoration of function could be achieved by stimulation of nerve fibers that had been properly established during development – no new connections would have to be developed, as would be the case for transplanted cells.
The relative merits and disadvantages of various alternative strategies that could potentially be used in lieu of a retinal prosthesis are briefly compared below.
In sensory substitution therapy, a sensory modality is used to provide spatial information about the environment that cannot be appreciated by the compromised visual system. A customized device captures visual information and relays it to a nonvisual sense, like the tactile or the auditory system, to provide input about the spatial detail of the patient’s local environment. The late Paul Bach-y-Rita, a pioneer in this field, advocated the introduction of visual information (captured with a camera) through tactile input to the skin or tongue. His subjects were (to some extent) able to recreate an impression of their environment, experience a sense of objects in space, and perform “eye”–hand coordination tasks.
A more recent device uses a camera to capture visual images, which are then electronically modified (by configuring the loudness, frequency, and inter-ear disparity) to provide a “soundscape” that represents the visual landscape. Some completely blind patients have navigated through unfamiliar environments and even found their way through a maze on a computer screen using only the auditory cues provided by this vOICe device ( Figure 75.5 and Box 75.2 ).
Historically, patients with severe visual loss from RP have benefited from use of a white cane, which provides important assistance for navigation. Patients with severe loss of central vision, from either RP or AMD, can benefit from use of optical or electronic devices such as telescopes and closed-circuit monitors
Molecular genetic approaches for treatment of one form of RP have recently provided some benefit to blind humans
Transplantation of stem cells, retinal pigment epithelium (RPE), or retinal neurons has been generally well tolerated in animals and humans, and in some cases has provided some benefit to blind patients
Human trials with ciliary neurotrophic factor administration have shown enhancement of RPE survival and a delay in visual loss in some human patients
Sensory substitution recreates an impression of the environment using sensation to a nonvisual sense such as touch or hearing. This approach can assist in navigation but could not restore “sight”
The efforts to treat blindness by transferring healthy genes to repair genetic mutations achieved a major milestone in 2001 when Acland et al demonstrated that blind dogs (suffering from a retinal disease caused by the same mutation that causes Leber congenital amaurosis in humans) were able to regain lost sight. Within 3 months, the dogs were able to navigate within a dimly lit room. More recently, modifications of their packaging techniques have produced a 90% reduction of the amplitude of nystagmus. Additional studies on blind large animals and rodent models of retinal degeneration have provided substantial evidence that gene replacement therapy, or gene silencing therapy, represents a scientifically sound strategy potentially to treat certain forms of retinal blindness. Gene therapy trials to treat early-onset retinal degeneration have taken place in the USA and England, and have shown modest but encouraging visual results.
Human transplantation studies, following earlier animal studies, have been performed using RPE, retinal neurons, partial-thickness sections of retina, and stem cells. These studies have ranged from a single patient to 56 patients, and most report relatively short follow-up of the patients. Generally, the transplants have been well tolerated and immune rejections and surgical complications have been uncommon. Many studies have reported improvement in visual acuity in some patients, occasionally up to four lines on a Snellen chart. In general, patients with wet AMD generally fared worse than dry AMD or RP patients, and worse pretreatment vision correlated with decreased chance of benefit from transplantation.
Ciliary neurotrophic factor (CNTF), a chemical which may be released normally by the RPE in response to cellular injury, has demonstrated significant retinal protection in animal models of RP where a delay in photoreceptor cell loss and enhancement of RPE survival in a rat model were demonstrated. A company called Neurotech USA has created an encapsulated vehicle containing RPE cells that have been modified using viral vectors to secrete CNTF continuously. Neurotech has conducted a phase I trial on 10 RP patients, 7 of whom demonstrated improved visual acuity, which was maintained in 3 patients 6 months after the removal of implant. However, one patient suffered a complete choroidal detachment. Other neurotrophic factors, like brain-derived neurotrophic factor and nerve growth factor, have also shown some neuroprotective benefit in animal experiments and may be other options for future human tests.
What is a visual prosthesis?
A retinal prosthesis is a complex device that functions by: (1) capturing visual images; (2) communicating the images to electronic components that interface with the retina; and (3) selectively delivering electrical pulses to the retina to create vision ( Figure 75.6 ). The neurons can also be stimulated by some nonelectrical means using neurotransmitters or cations. Although such nonelectrical strategies offer the potential for being less toxic to the host tissue, development lags substantially behind the efforts to use electrical stimulation.
Attempts to develop visual cortical prostheses began in the 1970s, and have been advanced by Drs. Normann (University of Utah) and separately by Philip Troyk (Illinois Institute of Technology). Advancements in microtechnology roughly 20 years ago allowed our group and another at Duke University to begin efforts to develop a retinal prosthesis. Since then, the field has enjoyed enormous growth and now includes 22 retinal prosthetic research groups in six countries. There is not yet enough evidence to know which approach(es) might be preferable, and it is possible different diseases might be treated differently.
Comparison of different types of visual prosthetic devices
Each of the locations being considered as a site for a visual prosthesis ( Figure 75.1 ) has certain advantages and disadvantages, and each approach has merit ( Box 75.3 ). In general, the more central the placement of a prosthesis, the greater the range of diseases it can treat. For instance, a retinal prosthesis could not be used to treat glaucomatous blindness (because of damage to the optic nerve), but a LGN or cortical approach could potentially provide some vision. By comparison, the convoluted topography of the visual cortex and the need for a substantial neurosurgical procedure to implant a cortical device impose challenges on the widespread use of visual cortical prosthetic devices ( Figure 75.1 ). On the other hand, the eye is prone to develop chronic inflammations whereas the brain seems to respond fairly indolently to long-term implantation of foreign material. A cuff electrode array can be easily placed around the optic nerve in the orbit, but in this location the nerve is invested by all three meningeal sheaths and requires greater charge density for stimulation. Conversely, placing electrodes on the meninges-free, intracranial part of the optic nerve would require a large craniotomy. At the level of the LGN, it would be very challenging to implant a large number (i.e., hundreds) of electrodes that might ultimately be needed to create spatially detailed vision there. The recent use of small craniotomies for placement of deep brain implants in the treatment of patients with Parkinson’s and other neurological diseases has been well tolerated, and it is possible that some similar technique can be adapted for visual cortical implants in the future. Each potential site for a visual prosthesis has its advantages and drawbacks.