Electronic restoration of vision in those with photoreceptor degenerations



Complete loss of vision is one of the most feared sequelae of retinal disease. Currently, there are few if any treatment options available to patients that may slow or prevent blindness in diseases caused by photoreceptor loss, such as retinitis pigmentosa and age-related macular degeneration. Electronic restoration of vision has emerged over recent years as a safe and viable option for those who have lost substantial numbers of photoreceptors and who are severely vision impaired. Indeed, there has been a dramatic increase in our understanding of what is required to restore vision using an electronic retinal prosthesis. Recent reports show that for some patients, restoration of vision to the point of reading large letters is possible. In this review, we examine the types of implants currently under investigation and the results these devices have achieved clinically. We then consider a range of engineering and biological factors that may need to be considered to improve the visual performance of newer-generation devices. With added research, it is hoped that the level of vision achieved with newer generation devices will steadily improve, resulting in enhanced quality of life for those with severe vision impairment.

Photoreceptor death accounts for approximately 50 per cent of all cases of blindness in Australia and the Western world, contributing to visual loss in both atrophic forms of age-related macular degeneration (AMD) as well as in inherited retinal degenerations such as retinitis pigmentosa (RP).[1] Currently, there are few methods that slow or prevent photoreceptor loss in these retinal diseases.

RP refers to a family of inherited retinal degenerations that are caused by a variety of different mutations affecting rod photoreceptors, resulting in gradual loss of rods followed by cones.[2] RP is associated with over 150 mutations in photoreceptor-associated proteins and as many as 50 per cent of patients have no known family history. Although gene therapy for those with Leber congenital amaurosis, the most severe form of inherited retinal degeneration, has been shown recently to be safe and effective,[3-5] it should be noted that RP can be inherited in an autosomal dominant fashion and the current gene therapy approaches are less viable for these conditions.[6]

Like RP, AMD is a disease that results in loss of rod and cone photoreceptors.[7] Both genetic and environmental factors influence the development of AMD, including non-modifiable risk factors such as age, gender and family history and infectious agents[8] or the inheritance of a range of single nucleotide polymorphisms in genes affecting components of the innate immune system, such as complement factor H, complement factor B or C3.[9-11] Modifiable risk factors, such as smoking, diet, higher body mass index, serum cholesterol levels, cataract surgery, cardiovascular disease, hypertension and sunlight exposure have also been implicated.[8] In AMD, deposits known as drusen develop between the retina and the underlying support tissues.[12] Vision is lost either from a slow process of photoreceptor death (atrophic AMD) or from a much more rapid and destructive process of blood vessel growth into the retina, which contributes to the speed of photoreceptor death (wet AMD).[7] Although there have been tremendous advances in the treatment of neovascular forms of AMD in recent years,[13] the dry form of AMD is more common and like RP, there are few treatment options. Therefore, methods that replace the function of lost photoreceptors are keenly sought and currently are being evaluated in research settings.

An important observation in patients with RP and atrophic AMD is that although photoreceptors gradually die, the ganglion cells and remaining visual pathway remain largely intact.[14, 15] Therefore, restoration of vision may be possible in these patients by targeting visual information to the remaining neurons of the retina, especially the ganglion cells. Since the 1960s, it has been known that an electrical current can induce a response in the retina in patients with no or limited sight.[16-18] Moreover, stimulation of retinal neurons is capable of generating a response in higher brain centres.[19]

Here, we provide an overview of the electronic devices currently under development for restoration of vision of those with photoreceptor degenerations like RP and AMD. We emphasise the patients most likely to benefit, the success reported and also the issues that need to be addressed in order to gain further visual improvements with these devices. Although devices that restore vision in those with optic nerve disease are under development, the design and information available on those devices is outside of the scope of this review.

Types of Retinal Prostheses Under Development

Retinal prostheses are among the most promising intervention techniques to assist people with late stage-retinal degeneration. All clinical approaches use electrical stimulation that ultimately leads to signalling from surviving ganglion cells to higher visual areas of the brain. The goal of these devices is to induce localised regions of visual perception in the visual field, known as phosphenes.

A key component of a retinal prosthesis is the means by which information about the visual world is converted to an electrical current. Currently, there are two main approaches to do this. The first is a device that incorporates an array of microphotodiodes, which are similar in concept to using miniature solar panels. The device is implanted in the location of the lost photoreceptors (subretinal) and the array of microphotodiodes convert light energy falling on the device into an electrical current that is delivered to the retinal neurons via attached microelectrodes. The second approach is to use an external device to capture images, such as a small camera fitted to a pair of spectacles, which sends signals to a microprocessor for conversion to electrical signals. The electrical signals are then targeted to retinal neurons via an array of electrodes within an implant that is positioned either on the surface of the retina (epiretinal) or under the retina (subretinal) or within the wall of the eyeball (suprachoroidal or episcleral) (Figure 1).

Figure 1.

A. Vertical section of a degenerated retina showing the locations of a retinal prosthesis. Currently, devices can be inserted between the sclera and choroid in a suprachoroidal location, in a subretinal location in the region where photoreceptors formerly resided or on the inner surface of the retina in an epiretinal position.

B. Stimulation configuration of electronic implants. A monopolar configuration consists of a stimulating electrode (black) and one return to ground (grey). A bipolar configuration consists of an active electrode surrounded by a neighbouring ground. Multipolar configurations consist of a single active electrode surrounded by multiple returns (Adapted from Cicione and colleagues[81]).

Retinal Prostheses can Restore Vision

The results from studies with blind patients provide the most useful information for determining the proof of concept that electronic restoration of vision is both safe and effective in restoring some level of vision. Currently, six groups worldwide have trialled devices in blind patients for varying lengths of time[20-25] (Table 1).

Table 1. Summary of visual outcomes for devices tested in patients with retinal degeneration
Group and referencesNumberRetinal conditionDevice typeDevice locationStudy durationVisual function
  1. RP: retinitis pigmentosa, CORD: cone-rod dystrophy, VA: visual acuity, ETDRS: Early Treatment Diabetic Retinopathy Study
Optobionics[12, 26] 105 RP, 1 Usher-type 2Artificial silicon retina: device is 2 mm diameter; 5,000 microphotodiodes, 20 x 20 μm in sizeSubretinal18 months

Subjective improvement in 6/6 subjects

VA (ETDRS) increase in 3/6 subjects

Second Sight Medical Products[22] 6RPArgus I (16 electrodes)EpiretinalUp to 5 yearsPatients identified the orientation of grating targets, could recognise objects (e.g. knife, plate, or cup); Maximal VA = 6/1,200; device shown to be safe
Second Sight Medical Products[20, 30] 27RPArgus II (60 electrodes; 200 μm2 in size)Epiretinal430 days26/27 subjects performed better in locating an image with device on than off.
Retinal Implant GmbH[29] 117 RP, 2 CORD, 1 choroideraemia16 electrodes over 0.28 mm; 50 x 50 μm in sizeSubretinalLess than 1 year8/11 subjects reported visual percepts; patterns distinguished; logMAR acuity was 1.78 (6/360)
Retinal Implant GmbH[23] 32 RP, 1 choroideraemiaMultiphotodiode array (1,500 electrodes)SubretinalLess than 1 year3/3 subjects had light perception; 2/3 spatial resolution 0.46 cycles/deg-1 1/3 subjects read letters 8.5cm high; 3/3 subjects localised objects on a table
EpiRet GmbH[25] 66 RPEPIRET3, 25 electrodes, 100 μm diameterEpiretinal4 weeksSix subjects could detect visual percepts.
Osaka University[24] 22 RP49 electrodesScleral pocket4 weeksPhosphenes detected; discrimination of two bars better with device turned on than off.
Intelligent Medical Implants GmbH[95] 20RP49 electrodesEpiretinal45 minutes vision testing19/20 reported seeing phosphenes.

Microphotodiode-based implants

Two different microphotodiode-based implants have been developed and tested in humans.[21, 23, 26] Optobionics based in Illinois, USA was the first company to conduct a government-funded clinical trial evaluating an electronic implant. The ‘artificial silicon retina’ (ASR) consists of 5,000 microphotodiodes with stimulating electrodes at each tip.[21] The implant is positioned in the subretinal space in the location where the photoreceptors have died. Consequently, light enters the eye as normal and is sensed by each individual microphotodiode, which generates a current that activates overlying neurons. It is the only type of device that is fully implantable without any external components. Because light is the sole source of current generated, it was originally thought that there would not be enough current to produce a phosphene; however, clinical results have been reported for 10 patients with RP, who were implanted with the artificial silicon retina in the mid-periphery. All had some improvement in light perception or vision, with four showing an improvement in visual acuity.[21, 26] In addition, improvements in light detection were observed in regions of the retina well away from where the device was located. This suggests that light detection in the artificial silicon retina is not solely determined by light-stimulated activity from the implant itself but rather indirect and beneficial effects were induced on the remaining retina. Follow-up studies in animal models indicate that implantation of an artificial silicon retina induces release of growth factors that have a beneficial effect on the surrounding retina.[21, 26-28]

The use of microphotodiodes to generate light-induced current has been adopted successfully by a German group, Retina Implant AG.[23] Their device consists of 1,500 microphotodiodes arranged in a 38 by 40 grid. In contrast to the artificial silicon retina, each microphotodiode in this device has an amplifier and stimulating electrode (50 by 50 μm) located at the site. This device has been implanted in a total of 11 patients, including seven with RP, three with cone-rod dystrophy and one with choroideraemia.[23, 29] With direct stimulation, all patients were able to detect phosphenes that were round or elongated,[23, 29] confirming the proof of concept that direct electrical stimulation from the subretinal space results in phosphenes. Moreover, one patient was able to discern gratings, recognise objects and read letters five to eight centimetres in size.[23]

External image capture-based implants

Second Sight Medical Products (California, USA) has undertaken the most extensive chronic human trials to date. They have developed two devices, known as the Argus I and Argus II.[20, 22, 30] Argus I was primarily a proof of concept device consisting of 16 stimulating electrodes, while Argus II is a scaled-up version containing 60 stimulating electrodes. Both devices use an external processing unit that contains a camera to capture an image and a wireless transmitter to transfer information to the implant. Additionally, a video processor and battery are worn on the user's belt. Data are wirelessly transmitted to an extraocular electronic processing unit positioned in the temporal skull.[31] The unit connects to the intraocular stimulating electrode array via wires traversing the sclera to provide the stimulation currents and strategies for the electrode array. Both devices have microelectrode arrays that are positioned on the surface of the retina via custom retinal tacks (that is, positioned in an epiretinal location).[30]

All patients implanted with either device have had severe to profound RP.[20, 22, 30] Argus I was implanted over the macular region in a total of six RP patients.[32] It provided patients with the perception of phosphenes that were generally white or yellow in colour and round or oval in shape;[22, 31] however, their functional ability was minimal. Argus II has been implanted in a total of 32 patients either as part of a pilot study (n = 2) or as part of a worldwide safety and feasibility study at 10 clinical centres worldwide (n = 30) and followed from six months to 2.7 years. All subjects were able to perceive light and some were able to locate objects, identify direction of motion and distinguish orientations of black and white gratings.[20, 30, 33] Moreover, four of five patients with some residual vision showed improved accuracy in localising objects with the device turned on, suggesting that patients are capable of ‘combining’ prosthetic vision with any residual vision.[20]

In contrast to the Argus II, the EPIRET3 device, from The University of Marburg, is designed to be completely implanted within the eye without the need for any cabling to pass through the sclera.[25] The EPIRET3 consists of an extraocular part, comprising a computer system and transmitter unit, as well as an intraocular component that contains a receiver, electronics and microelectrode array. The electronics of the intraocular part are positioned in an artificial lens in the posterior chamber of the eye and are connected via a wire to a microelectrode array on the retina containing 25 stimulating electrodes of 100 μm diameter. This microelectrode array is fixed to the epiretinal surface of the retina using two retinal tacks.[25] A small safety trial of the EPIRET3 has recently been completed in six legally blind patients with RP.[25] All six patients reported visual percepts and could discriminate patterns of differing orientations.[25]

The devices described consisted of stimulating electrodes that are either placed subretinally or epiretinally. In contrast, a Japanese group based at Osaka University has developed a device and system similar to that used by Second Sight but that is placed in a scleral pocket in the wall of the eyeball.[24] This device contains 49 electrodes but their trials have included testing of only nine electrodes on this array.[24] They have currently implanted two patients with RP for a four-week study. One patient was able to detect reproducible phosphenes from stimulation of six of the nine stimulating electrodes as well as distinguish multiple phosphenes from simultaneous stimulation with two electrodes.[24] In contrast, another patient was able to detect reproducible phosphenes from four electrodes.[24] Further testing is necessary to determine the efficacy of this approach and whether it will provide patients with functional improvements.

The results of these clinical trials demonstrate that electrically stimulating the degenerated retina can produce phosphenes and in some cases improve vision to the point of reading large letters. The aim now is to design devices that provide the best possible visual function to each patient. This is driven heavily by the engineering decisions of each approach; however, it is also the underlying biological interface, which may begin to explain why some patients have remarkable restoration of vision, while others receive little benefit. Below, we discuss a range of factors that may need to be considered to optimise the performance of retinal prostheses.

Important Factors for Improving Retinal Implant Performance

Patient selection

Retinal prostheses solely target diseases of the retina, of which the most common types are RP and AMD. Understanding the specific changes that occur in the retina over time, especially the cellular changes that occur well after loss of vision, may be important for optimising the visual outcomes following implantation of a retinal prosthesis. Consideration of changes to the inner retina, especially ganglion cells, synaptic remodelling, glial scarring or vascular effects may all have an impact on the long-term function of the device.

An inherent assumption for the successful development of any photoreceptor restoring therapy is that the remaining layers of the retina, and especially ganglion cells, remain intact. Although, there is preservation of the inner retina at the late stages of geographic atrophy,[34] one report[35] noted a significant loss (47 per cent loss) of ganglion cells in neovascular AMD. With respect to inherited retinal degenerations, there has been a number of studies of the inner retina in animal models and humans with inherited retinal degenerations. These studies show that a substantial number of ganglion cells remain well after complete loss of photoreceptors.[14, 15, 36, 37] There may be losses of up to 40 per cent of ganglion cells at the macula, following the complete loss of photoreceptors.[36, 37] In addition, several studies[38, 39] have shown that the intrinsic function of ganglion cells is altered in animal models of retinal degeneration.

A second major change that has been characterised in animal models of retinal degeneration is the widespread inner retinal remodelling that occurs at late stages of degeneration.[40-42] Following loss of rods and cones, downstream neurons, including bipolar, amacrine and ganglion cells, change[40, 42] (Figure 2). Some neurons form new synaptic connections and/or migrate to new locations within the inner retina. In addition, glial cells form a glial scar that gradually entombs the retina. Thus, the conventional retinal circuits that underpin our ability to see may be corrupted with increasing time after photoreceptor death.

Figure 2.

Vertical sections of a normal (A) and degenerating (B) rat retina that has been labelled for the amacrine cell marker, glycine. Glycine labels about half of all amacrine cells in the inner nuclear layer (INL). In the degenerating retina, several glycine-immunoreactive neurons can be seen migrating through the distal retina (arrows).

ONL: outer nuclear layer, OPL: outer plexiform layer, INL: inner nuclear layer, IPL: inner plexiform layer, GCL: ganglion cell layer. Scale bar = 20 um.

Despite the concerns that inner retinal changes might have on the optimal visual outcome obtained with an implanted device, the clinical results suggest that these effects are not likely to be factors in the success of the device. Notably, all current chronically trialled devices have been implanted in patients with RP or choroideraemia,[20-24, 30] who had little or no light perception and who had been visually impaired for varying lengths of time, and thus likely to have developed some form of inner retinal change. Yet, detection of phosphenes and even letters was possible despite some of these patients having been visually impaired for a substantial time.[23, 30, 43] Similar positive visual outcomes have been obtained in recent gene therapy trials of the most severe form of inherited retinal degeneration, Leber congenital amaurosis.[3, 4, 44, 45] In particular, gene therapy restored cortical function in dogs with long-standing visual loss due to a mutation in RPE65[46] and has resulted in substantial improvements in vision for patients with Leber congenital amaurosis for up to three years.[47]

Overall, these studies demonstrate that provided some ganglion cells remain, restoration of vision is possible despite widespread changes to the inner retinal circuitry. It is not yet known whether inner retina remodelling will prevent more detailed vision being possible with an implant. Indeed, different approaches may be necessary to treat those with more recent visual impairment (for example, at a stage when very little inner retinal remodelling has occurred) compared with those with long-standing disease (when significant remodelling may be present).

Location of the device

The location of an implant is an important consideration in optimising the performance of the device. Consideration needs to be made with respect to pathological changes associated with the stage of disease, the position in relation to the macula and also the distance from the target neurons, the ganglion cells.

Current chronic prostheses have had a range of locations from peripheral implantation to a position closer to the macula.[21, 23, 32] As RP is a defect of rod photoreceptors, it affects the peripheral retina first, followed by central degeneration at a later stage.[48-50] As a consequence, the degenerative process in these patients will be more severe in the peripheral retina.[49, 51] This dictates that the best location of an implant is over the macula, where the degenerative process will be less severe and the density of ganglion cells the highest;[52] however, some forms of RP show a non-uniform loss across the retina with islands of retina showing some preservation until late in the disease.[53] Thus, careful attention to the location of the device with respect to any localised changes in the retina has been suggested. Recently, a scoring system has been developed to aid surgeons in positioning a device across the retina.[54] The scoring system takes into account the location of the macula, the presence and extent of pigment migration from the RPE and any scarring, vascular attenuation and thickness of the inner retina.

There are several locations in which a prosthesis could be positioned within the wall of the eye with increasing distance from ganglion cells, including epiretinal, subretinal, suprachoroidal and episcleral placement (Figure 1). The decision as to which location to choose is often based on a careful appraisal of the distance from the target neurons, the benefit of utilizing residual retinal circuitry to improve vision, and the surgical risk.

Epiretinal implants are positioned against the inner surface of the retina, the nerve fibre layer. This location ensures the implant is in close proximity to the target cells, the ganglion cells. This is beneficial as lower thresholds of activation are needed when the stimulating electrode is close to the target neurons.[22, 55, 56] An epiretinal design also provides a large amount of space in the vitreous cavity to mount electronics or components onto the back of the stimulating electrodes; however, the stability of the device could be affected by eye movements. Second Sight use a single retinal tack to fix their implant to the retina and have successfully reported that the Argus II device has been intact and stable in 28 of 30 patients.[20, 30] There is also some concern that electrical stimulation from this position will result in activation of passing ganglion cell axons, producing non-focal vision.[57]

Subretinal implants are placed distal to the inner nuclear layer (INL) in the region where photoreceptors have died. The stability of the device in this location is excellent, as the retina assists with anchoring the device in position and glial scarring may actually assist with stabilisation. For the current clinically tested subretinal implants, there have been no reports of failure due to positioning of the device[23] or from animal trials;[58] however, the surgery to place the device in this position is challenging and requires a localised retinal detachment.[23] With respect to the visual outcome, it has been suggested that subretinal placement may offer better vision because electrical stimulation uses the remaining retinal circuits to modulate a stimulus. Interestingly, a study using the rd1 mouse showed that there was no difference in the thresholds required to elicit a retinal response in the degenerated retina, when comparing epiretinal and subretinal stimulation.[59]

Suprachoroidal implants are placed between the choroid and sclera. Most surgical techniques for these implants involve slipping an electrode array between these layers, which protects the retina from any manipulation.[24, 60-62] Additionally, it may be easier to anchor these implants into the sclera for long-term stability and it may also be possible to implant multiple arrays, allowing vision across a greater extent of the visual field. The implants are located further from target neurons and in animal models require higher levels of current to generate cortical responses than an implant located subretinally.[63] Furthermore, the impact of glial scarring is unknown and may further increase the required stimulation level. On the other hand, a clinical trial with a device located in a scleral pocket revealed that, at the very least, percepts from a device in this position are possible.[24]

Episcleral prostheses are being designed and tested in animal models. These implants are positioned onto the outer section of the sclera. This position is the furthest from the ganglion cells and will potentially require larger currents for effective stimulation and may activate large subsets of neurons due to current spreading across the retina. Siu and Morley[64] tested this approach using a 21-electrode Cochlear Ltd array with electrode diameters of 700 μm. Their study showed electrically evoked potentials in the visual cortex following electrical stimulation of the rabbit retina.[64] A further study[65] showed electrically evoked potentials recorded in the cat visual cortex following episcleral stimulation with one, two and three millimetre diameter electrodes. These electrodes are quite large and even though the surgery offers minimal risks, this approach is unlikely to provide high-resolution vision.

Technological developments in engineering

To improve the performance of retinal prostheses, advances in engineering may improve implant technology, including how the visual image is captured and converted to electrical signals, the number and size of electrodes used to stimulate the retina and the manner in which retinal neurons are stimulated by these electrodes. We discuss each of these engineering factors in turn.

Image Capture

Retinal implants based around microphotodiode arrays have individual diodes that detect light and generate a stimulus based on the amount of light detected.[21, 23] In this manner, they are similar to the responses of native photoreceptors. One unusual and potential advantage of this technique is that the microphotodiodes detect light wavelengths across a greater spectrum of colours than can naturally be perceived by the human eye. This means that patients with these implants should have the ability to perceive light wavelengths outside the normal spectrum, such as infrared.[66, 67]

A second point of difference between implants comprising microphotodiodes compared with those using an external camera and direct stimulation, such as the Argus II, is image fading. Following stimulation, nine out of ten patients implanted with the Argus II device reported fading of a white/yellow visual percept within ten seconds and some as quickly as two seconds.[33] Changing stimulus parameters led to individual variations in visual perception but did not eliminate the rapid fading of the percept.[33] This means that patients would need to interpret image ‘flashes’, before they fade. Similar observations were made in patients implanted with a subretinal array that directly stimulated overlying neurons.[23] It is not clear why patients experienced this rapid fading of vision. One possible explanation is a reduction in response following repetitive stimulation of ganglion cells, especially in the absence of involuntary eye movements. Direct repetitive stimulation of rabbit ganglion cells with pulse trains or stroboscopic light was associated with a reduction in the amplitude of the ganglion cell response when the inter-stimulus interval was less than 15 ms.[68] It is notable that a microphotodiode-based array placed in the subretinal space was not associated with image fading, possibly because this type of device moves with the eyeball, including with microsaccades.[23]

The main disadvantage of the microphotodiode-based arrays is that there is no opportunity for external, complex image processing. In contrast, retinal implants that rely on an external camera to capture an image have the greatest potential for modification of the signal that is used to stimulate the retina.[20, 24, 30, 31] This is because the image capture and processing occurs in an external electronic unit that can easily be replaced and updated. Image processing strategies that enhance the representation of depth, edges and contrasts are possible and could be tuned for the individual needs of patients. This becomes important when considering the need to adapt to changing background light levels. In the future, camera-based image capture and external processing will have the greatest potential to allow for providing vision over a large range of background light levels, while optimising dynamic range.

Design of the Electrode Array

Electrode array design and stimulation capabilities are some of the most important aspects for eliciting neural activity. The number, size and material of electrodes are all likely to influence the visual outcomes.

The number of electrodes on an array has become a major distinguishing feature of a device. Current devices range from 16 to 5,000 electrodes.[20, 21, 23, 24, 30, 31] Argus I, with 16 electrodes, allowed patients to perceive phosphenes and discriminate between simultaneous and sequential stimulation.[22, 31, 32, 69] In contrast, Argus II, with 60 electrodes, allowed patients to perform spatio-motor tasks, distinguish black and white gratings and direction of motion.[20, 30] The results with these two devices show that an increase in electrode number improved visual performance.

The number of pixels necessary to achieve different levels of vision can be calculated theoretically and tested in patients with simulated visual impairment. Cha and colleagues[70] and Cha, Horch and Normann[71] used psychophysical studies to determine the minimum number of electrodes required to be able to read letters. They showed that 625 electrodes in a 25 by 25 array with a centre-to-centre spacing of 400 μm all contained within a 1.0 cm by 1.0 cm grid[71] allowed resolution of images in a tiny central field. For navigational tasks, a lower resolution but larger field of view was necessary. In this case, 500 sampling points over a 10° field (equates to an electrode array of approximately 10 mm diameter) was sufficient.[72] With reference to the currently available devices, Fornos, Sommerhalder and Pelizzone[73] measured the maximum reading performance from the 60 electrode Argus II device in simulated visually impaired patients. They showed that if reading material was presented with a text size that corresponded to 4.5 pixels per character (that is, one to two characters imaged on the whole surface of the array), a reading rate of 20 to 35 words per minute was achievable. This rate is considerably slower than the reading rate of a normally sighted individual (approximately 250 words per minute) or a Braille reader (approximately 100 words per minute), emphasising the need for further improvement in the electrode array.

The resolution achieved with any device is likely to depend on the size of the electrodes and their density. For example, many neurons of diameter 10 to 20 μm will be excited simultaneously with 500 μm diameter electrodes. To achieve 6/24 acuity, approximately 100 electrodes, 20 μm in diameter are required in a device placed within the macula[74] (Figure 3). To restore a greater visual field with the same level of vision, a larger implant is required; for a 10° field of view 18,000 electrodes are required over a 3.0 mm diameter device (that is, a density of 2,500 electrodes per mm2). When manufacturing devices with many smaller electrodes, there are safety issues, namely, the potential for electrical cross talk and power requirements that need consideration.

Figure 3.

A. Graph showing the number of electrodes required to obtain theoretical levels of vision over visual fields of increasing size.

B. Graph showing the relationship between visual acuity and density of electrodes. It is possible to achieve near normal vision with 1,000 electrodes, provided the pixel size obtained with the electrodes is small and the density high.

In creating a device that optimises vision, the size and density of the electrodes must be weighed up against the ability of the device to apply current in a manner that does not impose deleterious effects on the tissue. In animal models, the threshold current needed to elicit an action potential in ganglion cells is proportional to the electrode area;[75, 76] however, electrode material and the characteristics of how the charge is injected into the tissue (for example, pulse duration and waveform) are also likely to be important considerations in optimising the operation of a device. There is only limited information relating electrode size to visual perception. In three patients implanted with an epiretinal four by four array, where the electrode size alternated between 260 μm and 520 μm in diameter, there was no difference in threshold response between the two different electrode sizes.[22] Similar results have been demonstrated in sighted cats implanted with a suprachoroidal device, containing electrodes of either 395 or 160 μm diameter.[60] These results emphasise that for the devices currently being evaluated in clinical trials, the electrode size plays only a minor role if any, in determining whether phosphenes are perceived. It remains to be determined whether smaller and more densely packed electrodes are needed to improve vision beyond that already demonstrated. The challenge now is to develop technology that optimises image capture and processing in such a way as to provide useful information to patients with as small a number of electrodes as possible.

The Stimulation Strategy

To see objects, multiple electrodes will need to be activated across a device. This requires consideration of how much current is being applied to the tissue, whether there is a dynamic range of stimulation, such that brighter regions of an object elicit more current and whether there is retinotopic mapping of the applied current across the retina. Thus, the way current stimulates the remaining neurons in a degenerate retina could influence the spatial resolution achieved with a device. Several stimulation strategies are being trialled, including monopolar, bipolar or multipolar[24, 77, 78] and examples of these configurations are shown in Figure 1B. Monopolar stimulation uses a single return electrode that is positioned on the opposite side of the neural tissue from where the stimulating electrode is to ensure that current flows through the tissue. The same return electrode is used regardless of the stimulating electrode. This approach requires that the return electrode be connected to the stimulating circuit, which is typically achieved with a flexible wire. In comparison to bipolar or multipolar returns, monopolar stimulation could result in more current diffusing across the retina, perhaps limiting spatial resolution.

Bipolar stimulation uses a single return electrode for each active electrode. The return electrode is often adjacent to the active electrode; however, the location of the return electrode can vary. Provided the electrodes are in close contact with the target cells, this stimulation should reduce thresholds and the spread of current in the retina. Interestingly, studies have shown that in the cochlear implant, monopolar stimulation produced lower thresholds than bipolar stimulation.[79, 80]

Multipolar stimulation is similar to that of bipolar stimulation; however, multiple electrodes near the stimulating electrode are used for the return path. This stimulation should contain the stimulating current within particular regions of the retina and theoretically improve resolution. Indeed, recently, the response of the visual cortex to suprachoroidal stimulation from electrodes with different returns was compared in sighted cats.[81] Monopolar stimulation was found to elicit a cortical response with the lowest level of current. However, a multipolar configuration where the surrounding electrodes acted as ground, improved spatial resolution.[81]

By choosing different return electrodes, multipolar stimulation also offers the ability to steer the current to particular locations in the retina. This could be useful for implants with a minimal number of electrodes with large electrode spacing, as the current could be directed into locations where an electrode is not present but neural activation is necessary to form a useful image. Current steering could also be used with high-density arrays to create various patterns. Bipolar and multipolar stimulation contain current more effectively and thus assist with spatial resolution. They are the methods most likely to be used for simultaneous stimulation with a large number of electrodes; however, it does require sophisticated electronics and the ability to control each electrode individually.

Design of the Electronics

For implants with an external image capturing unit and external processing, the electronics that drive the electrodes need to be carefully designed.

The method of transmitting data and power from the implanted electronics to the implanted electrode array is often achieved with wires; however, due to the size of the electronics, they are usually positioned away from the microelectrode array and the retina, which often requires cables to exit the eye at some location. Although this has not been shown to be problematic with current long-term implants,[30] the potential for damage to the eye due to infection is much higher. Ideally, devices of the future would be designed to have an entirely intraocular system to minimise the number and size of implanted components; however, this can complicate the design of the electronics. Currently, the only implants that have power and data processing contained within the intraocular component of the device are the microphotodiode array-based implant designed by Optobionics[21] and the EPIRET3 device.[25]

The Effect of Threshold on Device Performance

The condition of the retina heavily impacts on the ability of a stimulating current to elicit a response. Studies in humans and animals have found that the degenerated retina requires more current or higher thresholds to generate a neural response.[59, 68, 82, 83] This may be due to the documented changes in the diseased retina, such as changes in cellular intrinsic properties, synaptic activity, glial scarring or a combination of all of these factors.[39, 42, 59, 83-85]

The position of the electrodes relative to their target cells is another aspect that could affect device performance. Electrodes that are close to their target cells require less current for stimulation.[56, 86] The distance of the stimulating electrode from the target cell, rather than the electrode size or tissue impedance, is the most important factor in designing an electrode that uses less current to generate a response.[22] Designing an implant that uses less current to elicit a response is important as high current injection could be problematic in terms of tissue heating and damage over time. To address this distance issue some groups are testing devices with spine electrodes that penetrate into the retina[74, 87-90] This technique may be most useful for implants placed in the subretinal, suprachoroidal or episcleral space, so that they can deliver enough current to stimulate the ganglion cells. Manufacturing capability will be a determining factor for the development of electrodes that penetrate into the retina, as these electrodes would need to be much smaller (tips approximately one to 10 μm in diameter) than currently available surface electrodes (100 to 400 μm).

In summary, a range of engineering design features is being developed that collectively impacts on the performance of a device. Notably, enhanced visual performance is not just associated with a greater number of stimulating electrodes but rather the manner in which pre-processing of the visual image occurs, the manner in which the electrodes are stimulated and the positioning of the implant in relation to the target neurons may all play a role in visual enhancement. Having considered the optimal design and placement of an implant, possibly the most important factor to consider in its long-term performance is how the tissue reacts to the device and whether this alters performance over time. These implant–tissue interactions are considered below.

Integrity of the Neural Retina Following Stimulation

The integrity of the degenerated retina following electrical stimulation is a major aspect for long-term implantation of any device. Recently, Eng and colleagues[91] reported a detailed histological analysis of a patient who had used an Argus I device at the macula for five years. Retinal changes in the implanted eye were compared to retinae from un-implanted RP patients and also age-matched control patients. They showed that the implant had no impact on long-term ganglion cell viability and there was no neuronal loss in the region, where the implant was located. The region where the surgical tack was positioned showed fibrosis and a statistically significant loss of retinal cells. This tack traversed the entire retina and choroid, as it was positioned from the epiretinal device into the sclera. While this study is encouraging, it does highlight the delicacy of the retina, especially for devices that use penetrating electrodes.

Retinal morphology in response to electrical stimulation of the retina has also been examined in animal studies.[92, 93] Epiretinal stimulation of the healthy rat retina revealed a preservation of the inner retinal layers despite the close location of the stimulating electrode;[92, 93] however, the delicacy of the retina to any contact was again shown in these studies,[92, 93] as there was a significant decrease in outer retinal thickness measurements when an electrode was in contact with the retina regardless of stimulation. Furthermore, electrical stimulation combined with retina–electrode contact increased the area of retinal damage.[92] This highlights that epiretinal electrode arrays may not be able to be positioned in tight contact with the nerve fibre layer, as it may exacerbate cell loss. There will need to be a trade-off between obtaining the smallest distance from the target neurons to reduce thresholds and the minimisation of cellular changes attributed to the presence of the implant. Furthermore, excitability changes occurred in a study[94] that revealed an increase in cellular thresholds following electrical stimulation. It is important to note that these studies have been in healthy retinae and further investigation is needed to determine whether the response of the degenerated retina will be similar.

Neurotrophic effects on the degenerated retina following implantation of a prosthesis have been observed in the retina in both animal models of retinal degeneration and patients with RP.[21, 28, 66] Interestingly, histological analysis showed a significant increase in photoreceptors at the implant location when compared with sham-operated or non-operated eyes.[28] Analysis of the response at the superior colliculus revealed activity that matched where the implant was positioned in the retina[66] highlighting the visual pathway response from this retinal circuitry.


Over the last five years, there has been a dramatic increase in the number of groups developing electronic retinal prostheses designed to restore vision in patients who are severely visually impaired. With further engineering improvements and a better understanding of how the retina responds to a device, it is likely that improvements in vision will be possible with these devices in the future.

Grants and Financial Assistance

This work was supported by the National Health and Medical Research Council of Australia (Numbers 566814 and 566815 to ELF), Retina Australia and the Australian Research Council, through its Special Research Initiative in Bionic Vision Science and Technology grant to Bionic Vision Australia.