Retinal neuroprosthesis: science fact or science fiction?

When, over a quarter of a century ago, Steve Austin was equipped with prosthetic legs and bionic vision (and attendant sound effects) for a mere $6 million, there was no doubt that we were in the realm of ‘science fiction’. Cue the early 21st Century with advances in microelectronics, hermetic encapsulation, biomaterials suitable for long-term implantation and associated surgical techniques, and we have crossed the divide from ‘science fiction’ to the possibility of ‘science fact’, albeit with capabilities somewhat less impressive than the clinicians from Universal Studios, USA.

With vision being the most feature-rich and complex of the senses and visual cues being critical to tasks of everyday living, vision impairment results in momentous personal and economic burdens to both individuals and society. In this editorial we will examine a number of technical, scientific and clinical barriers to turning laboratory testing and clinical trialling of a retinal neuroprosthesis into a successful therapeutic device that provides patterned vision to the profoundly vision impaired.

The scope of this discussion will not extend to all visual prostheses, but will focus on the devices that, to date, have shown the most promise in chronic clinical trials in humans, namely those of the epiretinal neurostimulators from American and German teams. In other approaches, researchers have placed devices directly on the visual cortex, employed nerve cuffs surrounding the optic nerve and implanted unencapsulated microelectronic devices intimately contacting the subretinal surface.

For an epiretinal implant, electrodes are introduced into the vitreous space in intimate contact with the inner retinal layer. This approach is typically embodied as a two-unit device, wherein extraocular and intraocular devices communicate by way of transcutaneous radio-frequency (RF) telemetry or, in some proposals, a transcorneal laser [1–3]. The extraocular device comprises a camera and electronics for encoding and transmitting a stimulation paradigm. The intraocular device receives and decodes the transmission, usually by way of a microelectronic package known as an application-specific integrated circuit (ASIC). The ASIC connects to the electrode array, which is positioned at the vitreoretinal interface, providing controllable charge injection to the inner retina at each of several electrode sites (16 or 49, respectively, from the aforementioned teams). The stimulation paradigm is used to determine which of a select number of electrodes are actively delivering electrical stimuli at any given time.

In an epiretinal device, the tissue targeted for stimulation is the inner retinal layer, which contains the retinal ganglion cells (RGCs) that have survived the progression of diseases, such as retinitis pigmentosa and age-related macular degeneration. However, it is likely, with electrodes in close apposition to the inner retina, that bipolar cells, as well as the axons of the RGCs, will also be depolarized under certain conditions. This concern aside, one obvious advantage of an epiretinal implant is the well-defined topographic mapping of the visual space and established surgical techniques for introducing the electrode array [4].

The disadvantage of such an implantation approach is that it can only be used to treat diseases where there is a viable optic nerve. In response to such diseases, there can be large-scale reorganization of the retina with substantial gliosis. Despite this, studies have shown that human RGCs maintain their viability after the onset of these degenerative diseases [5], that the surviving retinal neurons are capable of being electrically stimulated [6] and that rudimentary phosphene vision is achievable in humans with advanced retinal degeneration.

Effectively replacing vision with the same resolution as that of normally sighted humans would require an image-capturing device to replace the function of the photoreceptor cells, of which there are more than 100 million in each eye, converging to some one million RGCs. Current epiretinal implant technology is capable of supporting, at best, up to 100 electrodes. The discord between these two numbers is where intense research is being conducted and where the divide between ‘science fiction’ and ‘science fact’ lies. The transition from tens to hundreds or even thousands of electrodes will require the resolution of major outstanding technical, scientific and clinical issues. These are discussed in the following sections. The list is by no means exhaustive; however, if significant advances are made in the following areas, then the path to successful long-term implantation of a device of ‘hundreds-order’ will be considerably clearer.

A question of numbers: psychophysics

Early studies in Utah, USA, by Cha and coworkers aimed to quantify the number of phosphenes – spots of visual percepts, patterned within the visual space, analogous to pixels on a television screen – necessary to read and navigate, and to achieve various levels of visual acuity. To a first-order approximation and, indeed, as logic would predict, the case of ‘more is better’ was found – but, significantly for the present discussion, low numbers of electrodes were found to be not without merit. Indeed, the ability to read text characters was arguably ‘useful’ at 100 phosphenes, with reading speeds of up to 70 words per minute [7]. Navigation was possible with a few hundred phosphenes with no marked improvement observed beyond 625 phosphenes. Once the 100-phosphene level was reached, visual acuity was affected to a greater extent by phosphene density than by phosphene quantity. Mobility was most significantly influenced by the size of the visual field [8]. So, if a one-to-one correspondence of phosphenes to electrodes can be assumed, how many electrodes are enough?

To address this question, a rather complex balance must be struck – a balance that considers not only patient benefit and safety, but also the limitations of our engineering and surgical expertise, and the commercial forces that bring such things to the patients for whom benefit may be derived.

To strike the balance described previously, is it ethical to go to market with a device with low numbers of electrodes and invest profits from those devices towards research and innovation of better devices? Conversely, is it ethical to wait for a magic number of electrodes that may be decades away and deprive today’s blind community of a chance at even rudimentary vision? If the latter is true, should the vision impaired be made to wait for 10, 100, 1000, 10,000 or 100,000 electrodes? To draw upon the cochlear implant analogy, recipients are able to hear – remarkably well in many cases – with just 22 electrodes. This, it is safe to say, has surprised many researchers in the field, not only because 22 electrodes have provided such remarkable benefits, but also because these benefits have, for the better part of two decades, continued to substantially improve in isolation from the electrode numbers [9]. This is due, in large part, to improved speech-processing strategies, that is, the way in which the electrodes are used. Will the same be true for visual prostheses? Will cortical plasticity, as it is known, allow the vision system to reorganize such that extraordinary sense can be made of the world through low numbers of electrodes? Will this vision continue to improve in unison with continued improvements in stimulation strategies?

Parallelization of stimulus encoding & delivery

In the field of neurostimulation to date, no end organ has required stimulation rates that would necessitate stimulus strategies and data throughput rates that would require a deviation from a serial mode of stimulation whereby a single current source is switched between multiple electrodes in a time-division multiplexed manner. Initial research in the area of epiretinal vision prosthetics has adopted the traditional serial approach. More recently, our group [10,11] and others [12] have been investigating alternate simulation paradigms that could be incorporated into an intraocular implant for epiretinal stimulation. One approach is to use multiple concurrent sources to activate electrodes in a parallel fashion. The need for parallelization has been driven by the increasing recognition that a device capable of effectively simulating prosthetic vision will require electrode numbers and transcutaneous communications protocols that exceed what is possible with a serial mode of electrode stimulation, in which an image is built up by sequentially activating electrodes in a rasterized approach, much like the electron beam in a cathode-ray tube television set draws an image. The difficulty with the rasterized approach is that there is a minimum amount of time (typically 100 µs to 1 ms) needed to inject charge in order to activate the neural tissue and to recover that charge to avoid damage to the same tissue, thus making the number of electrodes capable of being ‘serviced’ by a single stimulation source not only finite, but quite small indeed.

There are complex design issues associated with these revised stimulation strategies. The microelectronic design needed for providing multiple sites of charge injection and recovery at safe levels to ensure long-term viability of the tissue is one technological challenge. A major unknown is the interaction between multiple current sources in terms of current leakage between adjacent electrodes producing unpredicted or unwanted neural activation [13] and charge imbalances, which can affect electrode longevity and tissue health.

Neural–electrode interface

Possibly one of the most crucial yet least researched and understood areas of the retinal neuroprosthesis is the electrode–tissue interface. A suboptimal interface will probably cause electrode or tissue damage by exceeding predefined charge density limits. A concomitant factor is that the stimulus thresholds for visual perception are closely linked to the separation distance between electrode and tissue, as well as the amount of gliosis and retinal remodeling that occurs as part of the pathological process. To date, both these factors have limited implant designs to low electrode counts. With new materials and better methods to create a more intimate electrode–tissue contact, it should be possible to lower stimulation thresholds and also reduce electrode size without exceeding the safe limit for charge injection. However, many questions are yet to be answered. Will a chronically implanted neurostimulation device have a rescue effect on the neural retina and possibly arrest further degeneration? Will it be possible to create a biostable electrode through polymeric coating that is capable of high levels of charge injection but will also have drug-eluting capabilities – for example, to release neurotrophic factors to encourage RGC neurite ingrowth and reduce the electrode tissue impedance?

Device biocompatibility

The inherent attraction of creating a device with hundreds, if not thousands, of electrodes has provided impetus for the movement away from commonly used biomaterials, such as alumina-based ceramics, silicone elastomer and platinum. Hermetically encapsulated devices using these materials do not currently exist for such high electrode counts. Current auditory neuroprosthesis with, at most, a few tens of electrodes are probably at the limit of the current technological approaches. Our group and collaborators have been developing new microtechnologies and manufacturing processes to use materials with substantial histories in biostability and compatibility to create hermetic feedthroughs with the potential of interfacing hundreds of electrodes, constructed from silicone and platinum, to neurostimulator microelectronics, which are hermetically encapsulated in a ceramic [14].

Other groups have been exploring new materials for potential lifetime implantation within the ocular anatomy [15,16]. Such materials, in the absence of historical record in chronic implantations, will likely require protracted and detailed clinical trials and evaluation before they find widespread adoption in any therapeutic device. This said, once the benefit can be demonstrated to outweigh the risk, such devices and their materials may quickly find their place in therapeutic devices.

Summary

If future success for a clinically useful therapeutic device is linked to the research activity within a field, then there is great promise for a retinal neuroprosthesis. The number of research groups contributing in important ways to the vision prosthesis knowledge base is growing at an exponential rate. However, as detailed previously, there are many technological challenges and biological unknowns that stand in the way of creating such a device. In this editorial, we have touched on a number of critical issues, including parallelization of stimulus encoding and delivery, the neural–electrode interface, device biocompatibility and the plasticity at the visual system in relation to the psychophysics of visual perception. This is by no means a comprehensive list.

The research push is for more densely packed electrode grids with larger quantities of electrodes. If we are attempting to restore vision to the profoundly visually impaired in a manner analogous to normal vision, then this remains ‘science fiction’. However, the ‘science fact’ is that sufficient evidence exists to suggest that the provision of some degree of visual perception will occur in the near to medium future. The research, development and commercial effort will cost orders of magnitude more than the $6 million it took Steve Austin to run and see. But it is hard to put a monetary value on restoring some form of patterned vision to a person – giving them independence and mobility and providing them with some of the visual cues that most of us take for granted. The vexing question remains how many electrodes is sufficient for a ‘useful’ therapeutic device? As the scientific jury is still out on this issue, a look into the distant past may provide some wisdom. The 15th Century philosopher, Desiderius Erasmus, claimed, “In the kingdom of the blind, the one-eyed man is king”. In the vision prosthesis landscape, perhaps the equivalent question is, “In the kingdom of the blind, will the man/woman with 100 electrodes be king/queen?”