Long-Term Histological and Electrophysiological Results of an Inactive Epiretinal Electrode Array Implantation in Dogs

Abstract

purpose. Short-term pattern electrical stimulation of the retina via multielectrode arrays in humans blind from photoreceptor loss has shown that ambulatory vision and limited character recognition is possible. To develop an implantable retinal prosthesis that would provide useful vision, these results need to be sustained over a prolonged period of retinal electrical stimulation. As a first step toward this goal, the biocompatibility and the feasibility of surgically implanting an electrically inactive electrode array onto the retinal surface was tested.

methods. A 5 × 5 electrode array (25 platinum disc–shaped electrodes in a silicone matrix) was implanted onto the retinal surface using retinal tacks in each of the 4 mixed-breed sighted dogs. Color fundus photography, fluorescein angiography, electroretinography, and visual evoked potentials were obtained preoperatively, at 1-week intervals for 2 weeks postoperatively, then at 2-week intervals up to 2 months postoperatively, and thereafter at 1-month intervals. One dog was killed at 2 months after implantation and a second dog after 3 months of implantation. Histologic evaluation of the retinas was performed. The remaining two dogs continue to be followed beyond 6 months after the implantation surgery.

results. No retinal detachment, infection, or uncontrolled intraocular bleeding occurred in any of the animals. Retinal tacks and the retinal array remained firmly affixed to the retina throughout the follow-up period. Hyperpigmentation of the retinal pigment epithelium was observed only around the site of retinal tack insertion. No fibrous encapsulation of the implant or intraocular inflammation was visible. A- and b-wave amplitudes of the electroretinogram were depressed at the first postoperative week testing but recovered over the ensuing 1 week and were not statistically different from the normal unoperated fellow eye throughout the postoperative period. N₁ and P₁ wave amplitudes of the visual evoked potentials were not significantly different from the normal fellow eyes at any of the postoperative test intervals. Fluorescein angiography showed that the entire retina including the area under the electrode array remained well perfused. Similarly, histologic evaluation revealed near total preservation of the retina underlying the electrode array.

conclusions. Implantation of an electrode array on the epiretinal side (i.e., side closest to the ganglion cell layer) is surgically feasible, with insignificant damage to the underlying retina. The platinum and silicone arrays as well as the metal tacks are biocompatible. With the success of implanting an electrically inactive device onto the retinal surface for prolonged periods, the effects of long-term retinal electrical stimulation are now ready to be tested as the next step toward developing a prototype retinal prosthesis for human use.

Photoreceptor loss from diseases such as retinitis pigmentosa (RP) and age-related macular degeneration is the leading cause of retinal blindness. ¹ In these disorders, despite a near total loss of photoreceptors, there is relative preservation of the inner retinal neurons. ^{2 3} A number of groups including ours are evaluating the possibility of restoring sight by developing a retinal prosthesis that would electrically stimulate the remaining retinal neurons. ^{4 5 6 7 8} The feasibility of this approach is supported by clinical studies that have shown that controlled electrical stimulation of retinal areas results in localized and retinotopically correct perceptions in patients blind from end-stage RP and age-related macular degeneration. ⁹ Pattern electrical stimulation using a multielectrode array resulted in multiple dots of light, which when viewed together were perceived by the patients as letters and geometric shapes. ¹⁰ However, these tests used hand-held electrodes and lasted for short periods, usually less than 60 minutes. To develop an implantable device, long-term implantation and electrical stimulation needs to be equally beneficial and safe. Because both mechanical stress and electrical current could be harmful to the retina, the mechanical and electrical biocompatibility issues must initially be studied separately, so as not to confound a negative result. The most straightforward way to do this is to first-test the biocompatibility of the surgery and chronic implantation without electrical stimulation, an approach taken in the development of other neural prostheses. ¹¹ Toward achieving this goal, in this study we address (a) techniques for surgical implantation of the electrode array on the retina, (b) biocompatibility of the implanted array, and (c) the mechanical effects of the electrode array on the retina. 

Methods

Electrode Array Specifications

The electrode array was composed of 25 platinum disks arranged in a 5 × 5 square array (Fig. 1) . A single 25-μm diameter platinum wire was attached to each disk. The disks and wire were encapsulated in medical grade silicone, except for the surface of the platinum disks juxtaposed against the retina, which was not enclosed by the silicone. The exposed surface of the platinum disks formed an array of planar, stimulating electrodes in a silicone matrix. The disks were 400 μm in diameter and mounted on 600 μm centers. The side of the implant that was placed next to the retina measured 3 × 5 mm and was curved to match the retina. The implant was less than 1-mm thick. The 25 wires from the disks formed a cable 600 μm in diameter, extending from the electrode array. Each wire was individually insulated. The cable was 10-cm long, a sufficient length to allow the cable to exit the eye through the sclerotomy wound and to be sutured to the sclera, in the superotemporal quadrant, under the conjunctiva. 

Surgical Procedure

All procedures conformed to the ARVO Statement on Use of Animals in Ophthalmic and Vision Research. Four mixed-breed dogs of 15- to 20-kg body weight were used in this study. The dogs were anesthetized with halothane gas using a fluotec MK 3 vaporizer (Orchard Park, NY). Pupils were dilated using 10% phenylephrine hydrochloride and 1% tropicamide eye drops. A conjunctival peritomy was made, followed by three sclerotomies each 3.5-mm posterior to the limbus. The inferotemporal port was used for an infusion, and then via the superior two ports a vitrectomy was performed using techniques that are commonly used to strip the posterior hyaloid away from the retina in patients with macular holes. Specifically, at the beginning of the vitrectomy procedure, the aspiration pressure of the vitreous cutter was increased to 200 mm Hg without activating the guillotine cutter. The vitreous cutter was positioned close to the surface of the optic nerve head to engage the posterior hyaloid and yet not aspirate retinal tissue. Under the high-power magnification of the surgical microscope, the posterior hyaloid was observed as it was aspirated into the port of the vitreous cutter, and then the engaged vitreous was detached slowly from the macular region. The vitreous was then cut and removed. To ensure that no vitreous remained attached to the retinal surface, a silicone soft-tip cannula was swept across the surface of the macula while continuously aspirating through the cannula. If vitreous was present, the cannula would engage it and the silicone rubber tip would bend. This bending of the tip is akin to the bending of a fishing rod when a fish is hooked onto the line. No such “fish-strike” sign was seen in any of the dogs, indicating a thorough and complete stripping of the posterior hyaloid. After the vitrectomy, the superotemporal port was enlarged circumferentially to 3.5 mm to allow introduction of the electrode array into the eye. The electrode array was introduced through the superotemporal port and held in the mid-vitreous cavity with intravitreal end-gripping forceps (Grieshaber Co, Kennesaw, GA). Next, a retinal tack loaded in a retinal tack holder (Grieshaber Co) was introduced through the superonasal port. The electrode array was then tacked onto the area centralis. Usually, two tacks were used to affix the electrode array to the retina. During the procedure, to prevent any choroidal bleeding at the site of retinal tacks, the intraocular pressure was raised by adjusting the height of the infusion fluid bottle. The cable from the array exited the eyeball and was passed underneath the lateral rectus muscle and sutured at several points to the sclera. The cable was left subconjunctivally. Superotemporal and superonasal sclerotomies were closed with 7-0 vicryl (Ethicon, Somerville, NJ) mattress sutures, and the conjunctiva was closed with interrupted 6-0 vicryl sutures (Ethicon). Antibiotic and steroid ointment was applied at the end of the surgery. The dogs were then extubated and when fully alert returned to their kennels. Postoperatively, prednisolone tablets (1 mg/kg body weight once daily) were given for 1 month and then tapered over the next 2 months. Topical antibiotic and steroid ointment were also applied for the first 1 month. 

Color fundus photography, fluorescein angiography, electroretinography (ERG), and visual evoked potentials (VEP) were obtained preoperatively and at postoperative weeks 1 and 2. These tests were then repeated every 2 weeks up to 2 months and then at every 1-month interval. 

Fundus Photography and Fluorescein Angiography

The animals were anesthetized and dilated as described above. Color fundus photography was performed using 100 ASA color film with a Topcon TRC-50FT retinal camera (Tokyo Optical, Tokyo, Japan). A No. 0.9 Wratten neutral density filter was used to reduce the tapetal reflex from the flash of the camera. Two- and one-half milliliters of 10% sodium fluorescein solution was injected intravenously, immediately before taking the fluorescein pictures. Fluorescein angiography was done using 400 ASA black-and-white film with a matched pair of DeLori fluorescein exciter and barrier filters in place. 

Data Acquisition System for Electrophysiology

Our electrophysiological setup consisted of a LKC MGS-1 bright flash stimulator (LKC Technologies, Gaithersburg, MD) and a computer for stimulus triggering and waveform capture. The light intensities (given below for each measurement) are set and calibrated at the factory and conform to International Society for Clinical Electrophysiology of Vision Standards for the particular measurement. The evoked potential was amplified (up to ×1000) and filtered with a P15 Grass amplifier (Grass Instruments, Warwick, RI). Additional amplification of up to ×50 was available on the A/D board (model ATMIO64-F5; National Instruments, Austin, TX) in the desktop computer (IBM–compatible). The signal was acquired at a 10-kHz sampling rate (Labview software; National Instruments). 

ERG

Animals were anesthetized by inhalation of halothane and oxygen and kept under stage 3 and degree 2 to 3 of anesthesia (loss of palpebral and flexor reflexes and eyes in ventral position). The animals were maintained in a sternally recumbent position, and their body temperatures monitored. All animals were dark-adapted for 45 minutes before the beginning of the electrophysiological testing. During the testing, the eyelids of the stimulated eye were kept open with an eyelid speculum. The cornea was kept lubricated by the administration of artificial tears. Three electrodes were used: a corneal recording electrode, reference electrode at the outer canthus, and ground electrode in the mouth. For the dark-adapted bright flash ERG, a single full-field flash stimulus (flash intensity of 2.09 candela/m² [cd/m²]) was used. The signal was filtered and amplified (0.3–300 Hz bandwidth,× 20,000 amplification). A flicker ERG response was measured by presenting the same bright flash at 31 Hz in the presence of a dim (25.2 cd/m²) background light. The same amplification was used, but the filter was narrowed (1–100 Hz bandwidth, ×20,000 amplification). Fifty waveforms were averaged for each recording of the flicker ERG. The averaged 50 waveforms were digitally processed with Fourier analysis (1024 point fast Fourier transformation subroutine supplied with MATLAB 5.0; Mathworks, Natick, MA). Three recordings of the dark-adapted and flicker ERG types were obtained for each eye. Recordings of the operated eye were compared with those from the unoperated eye of the same animal at each follow-up examination. The data were analyzed by measuring the amplitude and latency of the a-wave and the b-wave. The b-wave amplitude was measured from the trough of the a-wave. The averaged recordings of unoperated eye were taken as 100%, and then the averaged recordings of the operated eye were calculated as a relative percentage of unoperated eye. Statistical analysis was performed, using a paired Student’s t-test. P ≤ 0.05 was considered statistically significant. 

Visual Evoked Potentials

Animal anesthesia and positioning were the same as has been described for the ERG recordings. Two subcutaneous platinum iridium electrodes were used, one over the visual cortex and the other over the frontal region. VEPs were recorded as the average of 100 bright flash stimuli presented at 1 Hz (flash intensity of 2.09 cd/m², no background illumination, 0.3–300 Hz bandwidth, ×20,000 amplification). To exclude the chance that a far-field ERG was contaminating our VEP responses, we sequentially moved the recording electrode closer to the eye and showed that the VEP waveforms quickly diminished as we moved away from the scalp overlying the visual cortex. Three recordings were done in both of the eyes at each follow-up. The averaged recordings of the operated eye were compared with those from the unoperated eye and statistically analyzed as described for the ERG recordings. 

Electrode Impedance Measurement

Impedance of several electrodes was evaluated before implantation and after explantation in phosphate-buffered saline. The impedance of the electrode was evaluated using a custom measurement system. The DC potential of the electrode was held at 300 mV (versus saturated calomel reference electrode), and a 5 mV sine wave (1 kHz) was used as the excitation signal. A constant DC bias is used, because platinum impedance is sensitive to DC potential. ¹²  

Light and Electron Microscopy

One animal was killed at 2 months and one at the 3-month postoperative period. Two animals remain under follow-up which now is more than 6 months in duration. Eyes of the killed animals were enucleated and subjected to light and electron microscopic examination. The unoperated eye served as the control. After hemisecting the eyes, the posterior halves were immersed overnight in 2% glutaraldehyde and 2% paraformaldehyde in 0.1 molar phosphate buffer at 4°C. The area under the retinal prosthesis was sectioned and post-fixed for 2 hours in 2% osmium tetroxide in phosphate buffer, alcohol dehydrated, and embedded in epoxy resin. Thick sections (1 μg) stained with toluidine blue were used for light microscopy, and thin sections (60–90 nm) stained with lead citrate and uranyl acetate and were examined with a JOEL JEM-100 CX2 transmission electron microscope (Hitachi, Tokyo, Japan). 

Results

Ophthalmoscopic Examination and Fluorescein Angiography

Anterior segment and lens remained normal in all dogs despite a transscleral cable that remained implanted in the subconjunctival space. No retinal detachment, infection, or excessive intraocular bleeding occurred in any of the animals. There was minimal vitreous hemorrhage in dog 1, which cleared within 2 to 3 weeks. Similarly, there were a few vitreous cells and minimal vitreous haze in dog 3, which cleared within the first week. A stable array position is critical to maintaining steady image resolution and stimulation current levels. The retinal tacks and the electrode array remained firmly affixed to the retina of all dogs throughout the follow-up period. A mild retinal fold was observed at one edge of the electrode array intraoperatively in all dogs. There was no progression of the retinal fold with time. Around the site where the tack pierced the retinal pigment epithelium (RPE), hyperpigmentation and hypopigmentation of the RPE was noted. Hyperpigmentation was ophthalmoscopically visible within the first week and progressively enlarged, but remained confined to the tack area and did not involve the entire implanted area in any of the dogs (Figs. 2 A, 2B). This hyperpigmentation of the RPE blocked the reflex from the underlying tapetum. 

Pre- and postoperative fluorescein angiography showed good retinal vascular perfusion for the entire follow-up period (Figs. 3A 3B) . Metal sites of the electrode array appeared as blocked fluorescence. Hyperpigmented areas around the tack insertion site showed initial blocked fluorescence followed by hyperfluorescence (Figs. 3B 3C 3D) . The area of hyperfluorescence associated with the hyperpigmented areas did not increase in size in the late phases of the angiogram and if anything faded in the later frames (10 minutes after injection), indicating localized staining contained within the RPE. There were no areas of fluorescein leakage indicative of choroidal neovascularization. Also, no retinal vascular leakage was seen either under the electrode array or around the retinal tack. 

ERG

Although the electrophysiological setup and depth of the anesthesia were regulated closely, the ERG and VEP measurements in the dogs varied between animals and also to a lesser extent from day to day for each animal. We were able to overcome this variability by normalizing the data and reporting it as a percentage of the unoperated eye in the same animal. Throughout the postoperative period, in all dogs, the A-wave amplitudes and latencies were not statistically different from those of the unoperated normal eye (i.e., remaining within 88%–120% and 90%–135%, respectively, of the fellow eye). For the first 2 weeks postoperatively, the b-wave amplitudes were significantly less in one dog (dog 3). But after 2 weeks, in all dogs, the b-wave amplitudes recovered to within 80% to 155% of the control eye and remained stable throughout the follow-up period. B-wave latencies varied from 95% to 146% of the normal fellow eye and were also not significantly different between the eyes in any dog (Figs. 4 A, 4B). The 31-Hz flicker ERG similarly showed no difference in amplitude or latency between the operated and unoperated eyes 6 months or more after implantation (Figs. 4C 4D) . 

Visual Evoked Potentials

For the first 2 weeks postoperatively, N₁- and P₁-wave amplitudes were significantly lower in one dog (dog 3). But after 2 weeks, in all dogs, the N₁- and P₁-wave amplitudes were within 81% and 127% and 87% to 136%, respectively, of the unoperated eye. These values were not significantly different from the normal fellow eye throughout the postoperative follow-up. The average N₁- and P₁-wave latencies varied from 87% to 114% and 95% to 149%, respectively, of the normal fellow eye. The N₁ and P₁ latencies were not significantly different from the normal fellow eye throughout the postoperative period in all dogs (Figs. 5 A, 5B). 

Electrode Impedance Testing

Because no chronic electrical stimulation was performed with these arrays, these implants were not instrumented with an external connector, precluding in vivo impedance measurement. However, the preimplant and postexplant impedance of several arrays was measured in phosphate-buffered saline. Preimplant impedance averaged 8.4 kΩ (1 array, 8 electrode sites). Postexplant impedance averaged 8.0 kΩ (2 arrays, 14 electrode sites). The difference was not statistically significant (P > 0.05). 

Light and Electron Microscopy

Dissection of the enucleated eye confirmed that the array was in close apposition to the retina, as was indicated by the fundus photography. Light microscopy showed that the photoreceptor layer, the outer nuclear layer, the inner nuclear layer, and the ganglion cell layer were all of normal thickness and density under the electrode array in both dogs that were killed (Figs. 6 A, 6B). The nerve fiber layer appeared undisturbed. RPE, tapetum, and choriocapillaris appeared normal on light and electron microscopy. Electron microscopy also showed intact RPE with good intercellular connections with photoreceptor outer segments (Fig. 6C) . The area through which the tack was inserted showed loss of tapetum, RPE, and retinal layers; however, this damage was limited to the tack insertion site, and no pathology was observed in retinal layers under the electrode array compared with the normal eye. Last, close inspection with a high-power magnification surgical microscope revealed no mechanical degradation of the silicone rubber matrix, the platinum electrode wires, or disk electrodes of the explanted array. The structural integrity of the implanted arrays was also confirmed by measuring electrical impedance of the electrode sites before and after implantation (see results of electrode impedance testing). 

Discussion

Our results show that a multielectrode array can be secured to the inner retinal surface, is mechanically stable, and biologically tolerated well over a 6-month period (i.e., a length of time after which the outcome is unlikely to change at least as far as vitreoretinal surgical procedures are concerned). Although complex vitreoretinal procedures are commonly performed, affixing an implant to the retina is rare. The retinal tissue, with a thickness of 0.3 mm, is delicate and tears easily. The delicate nature of the retina also predisposes it to pressure necrosis from the weight of an implanted epiretinal device. The abundant vascular supply of the retina and the underlying choroid complicates surgical manipulation because of the threat of excessive bleeding. Despite these hurdles, we were able to place the electrode array epiretinally without inducing significant retinal damage. Before tack placement, a complete vitrectomy with posterior hyaloid separation was necessary to reduce postoperative vitreous traction and displacement of the electrode array. The posterior hyaloid separation in dogs was relatively easy to perform provided that careful technique as described in the Methods section was followed. We found the application of the first retinal tack to be critical because if it pulled out not only would this result in the displacement of the electrode array but also result in immediate bleeding and retinal detachment. The second retinal tack placement was easier, and we found the second tack necessary to keep the array affixed in close proximity to the retina. Retinal tacks were placed so as to avoid damaging retinal ganglion cell axons from the areas underlying the electrode sites. We believe that the minor retinal fold, which appeared intraoperatively after applying the tacks, was due to the thickness of silicone matrix of the electrode array, which displaced the retina toward the edge of the implant. Although this effect was minor and did not progress, the fold can be reduced further by making thinner electrode arrays. 

The distance from the stimulating electrode to the retina is critical in terms of electrical stimulation threshold and image resolution. An electrode should be positioned as close as possible to the cells that it will stimulate without harming those cells. Such positioning will maximize resolution and minimize the stimulating current. Although no electrical stimulation was performed in these tests, data from other experiments suggest that positioning the electrodes as was done in these experiments will allow safe and effective stimulation of underlying retinal neurons. Specifically, short-term (<60 minutes) electrical stimulation tests in blind humans have used a similar stimulating array positioned over the retina. ¹⁰ Test subjects, individuals with end-stage outer retinal degeneration, have reported focal visual percepts corresponding to individual electrodes and simple shapes created by multiple electrodes activated simultaneously. The stimulating current density, in a majority of the cases, was within the acceptable limit for long-term stimulation of neural tissue with either platinum or iridium electrodes. ¹² Based on these results, we believe that surface electrode positioning is capable of creating useful visual percepts. In the current experiments, we sought to show that the array could be maintained in this position over a long period. Evidence that the array position was maintained (i.e., the array is not floating in the vitreous cavity) comes from fundus photography. The retinal fold near one side of the array (caused by the array pressing on the retina, Fig. 3B ) is consistent in appearance throughout the implant period. The array contacted the retina during the surgery to produce the fold. If the array had later moved away from the retina, the fold would have changed in appearance. Finally, during the initial histologic evaluation, the electrode array was observed to be firmly attached to the retina. Great care was necessary when removing the array from its close apposition to the underlying retina to allow the retina and the array to be evaluated by microscopy. 

We chose epiretinal positioning for our electrode array because of the ease of surgical implantation. A subretinal location has the potential benefit of obviating the need to affix an implant to the retina by securing the implant between the retina and the RPE. However, placing the implant in the subretinal space will require either a retinotomy or an external approach that would cross the choroid after a partial thickness scleral flap. The retinotomy may not seal especially if the implant is thick and therefore carries the risk of retinal detachment. The thickness of the implant, even if perforated, reduces the choroidal blood flow to the retina and may result in ischemia of the overlying retina. ^{5 6} Last, crossing the choroid to gain entry into the subretinal space can result in excessive bleeding from the highly vascularized choroidal bed. 

We selected dogs for electrode array implantation, because of the well-established and readily available National Eye Institute colony of dogs with photoreceptor degeneration. ^{13 14 15} In these dogs, photoreceptors fail to differentiate normally and subsequently degenerate, due to abnormal cyclic nucleotide metabolism caused by the same genetic defect found in some forms of human RP. These dogs form a good model of outer retinal degeneration and can be used subsequently for implantation of an active electrode array to test the electrical stimulation. Sighted mixed-breed dogs were used for the mechanical biocompatibility tests, however. The retinal degeneration dogs are expensive because of the need for prolonged housing to allow the time for them to develop severe photoreceptor degeneration. Moreover, ERGs and VEPs have been used to study toxic retinal reactions, and only sighted dogs provide us with the opportunity to use those tests to quantify the extent of any damage. ⁴ Even the cost of housing sighted dogs for 6 months or more is high, which limited the number of animals used in this study. On the other hand, both electrophysiological and histologic data were consistent among all four dogs in showing little damage due to long-term epiretinal implants. Therefore, we did not see the need to duplicate these results above and beyond shown in here. 

Both platinum and silicone are proven biocompatible materials. Silicone has been approved by the FDA for intraocular use (silicon foldable intraocular lens) and has a proven capability of being well tolerated in the cochlea as part of the cochlear implant electrode array. ^{16 17 18} Similarly, platinum iris clip lenses have been successfully used in ophthalmology. ¹⁹ Platinum metal also has a proven performance record as an electrode material (in cochlear implants) for long-term electrical stimulation because of its high charge injection limits before irreversible reactions occur at the electrode tissue interface. ¹² Undoubtedly, if the need for smaller closely packed electrode arrays should arise, the current means of hand-fabricating silicone/platinum electrodes will not suffice, and microfabricated arrays will become a necessity. ^{20 21} However, hurdles remain in microfabricating thin film electrode arrays. ²¹ Silicon has been used as a substrate to fabricate densely packed electrode sites, but to attempt to make a silicon electrode array thin enough to be both flexible and lightweight results in silicon that is unacceptably brittle. Polyimide (Kapton) is another substrate for electrode array fabrication, but it is prone to saline leaking as well as a weak material interface, resulting in delamination of the metal at the electrode site. ²² The electrodes we used were modified cochlear implant electrodes and have been tested extensively in saline baths with and without passing current and are part of FDA-approved cochlear implants. We did not see any fibrous encapsulation histologically, and visibly the wires to each electrode remained intact. The silicone rubber also showed no discoloration or any structural changes. Similarly, no significant difference was measured between the postexplant impedance and preimplant impedance. Finally, because it is unclear whether a completely wireless method of data and power transmission can be developed for a retinal prosthesis, we elected to evaluate the worst case scenario of testing an electrode array with an attached transscleral cable. Transscleral tubes are tolerated well as part of certain glaucoma procedures. ²³ We similarly found a transscleral cable to be well tolerated. We believe that as long as the cable is positioned subconjunctivally and the scleral wound is secure, the risk of infection and any other untoward effect remains low. Thus, if needed, an electrode array can be partially powered via such a cable attachment. 

Metal alloy tacks were chosen to secure the implants because of their known biocompatibility (FDA-approved for intraocular use). ²⁴ The retinal tack worked well to maintain the electrode array position. Tacks have been associated with fibrovascular reaction when placed at the edge of a retinotomy site. ^{24 25 26} However, when placed in uncut retina, this reaction was only limited to the insertion site and did not affect the area below the implant. There were minimal hyperpigmentary changes at the level of the RPE around the tack insertion site, obscuring the underlying tapetal reflex. But this proliferation of RPE was confined around the site of the tack and did not extend much beyond that site. We believe that this limited proliferation is a mild response for the injury caused by the tack. The mild nature of this change is confirmed by the fact that RPE proliferation did not adversely affect the overlying retina as determined by electrophysiology or histologic evaluations. Similar RPE changes are seen after submacular surgery in cases of choroidal neovascular membrane removal in presumed ocular histoplasmosis cases. In these patients the visual acuity often recovers to an excellent level despite these changes, again suggesting that RPE hyperpigmentary changes are not mutually exclusive with good vision. ²⁷ We believe that in lieu of the limited proliferation, an intact underlying choriocapillaris, and overlying retina, such an RPE response can be tolerated. The successful use of retinal tacks does not preclude the search for other even less traumatic techniques of fixation. Some of these include cyanoacrylate glue, magnets, using autologous thrombin, ²⁸ and even a preshaped electrode that would conform to the curvature of the back of the eye (cantilever design). ⁴ These techniques all have the advantage of not piercing the RPE and choroid and, therefore, minimizing the damage associated with fixation, but each has its own disadvantages. The magnetic attachment would hinder the future use of MRI but has the distinct advantage of making the implant easy to replace. The ease of replacement is also a feature of the cantilever design, but this approach may be limited by the disadvantage of having to build a different device for varying eye curvatures, as might be encountered in different patients. Cyanoacrylate and thrombin glues do not allow easy replacement of the implant and can be associated with toxicity. ²⁸  

ERG test results are not significantly different in the operated eyes compared with the normal fellow eyes in all dogs. However, a full-field ERG is a mass response of the retina and does not specifically represent the area of retina under the retinal implant. We used two measures to make a better assessment of the functionality of the retinal area to which the array was fixed (area centralis). VEPs and 31-Hz ERG are potentials dominated by the area centralis. ²⁹ The ganglion cells of the area centralis project to the visual cortex nearest to the cortical surface, which means activity from the area centralis comprises the bulk of the VEP. Similarly, the 31-Hz ERG is a cone response, and cones are in greater number in the area centralis. VEPs and 31-Hz ERG responses add further support to the full-field ERG and show that the retina underlying the implant likely remains functional to near normal levels. Focal ERG with the incident rays focused on the retina under the electrode array would be yet another method to provide information regarding retinal function from the implanted area. ³⁰ However, in lieu of the above electrophysiological results and the histology showing near normal retinal anatomy, we believe that focal ERG testing would not add significantly. Focal ERG may also be limited in our case, because the metal electrodes would block the incident light and may also scatter it beyond what is covered by the background illumination of a typical focal ERG machine. 

In conclusion, we designed and fabricated an electrode array and then successfully implanted it epiretinally over a prolonged period. Although, in the future improvements in the material, fixation, and surgical technique will undoubtedly be made, we have a prototype design and method of implantation. These techniques will allow long-term retinal electrical stimulation, the next step toward the development of a retinal prosthesis. 

 Reprint requests: Mark S. Humayun, Wilmer Ophthalmological Institute, the Johns Hopkins University, School of Medicine, Maumenee 738, 600 N. Wolfe Street, Baltimore, MD 21287.

 Supported by grants from the National Eye Institute (1R 29 EY 11888-01), Research to Prevent Blindness, The Whitaker Foundation, and the Ross Clinician Scientist Award.

 Submitted for publication July 6, 1998; revised February 18, 1999; accepted April 6, 1999.

 Proprietary interest category: N.

 

Figure 1.

 

Multielectrode array held next to human eye. The array was approximately 5 × 3 mm and contained 25 platinum disk electrodes embedded in a silicone matrix.

Figure 2.

 

(A) Preoperative color fundus photograph in dog 3 showing the tapetal region superiorly and junction of tapetal and nontapetal region inferiorly (magnification, ×10). (B) Color fundus photograph in the same animal at 24-week postoperative period. Electrode array (open arrow) and retinal tacks (white arrow) remain in good apposition to the retina. Hyperpigmentation around the tack area was observed (magnification,× 10).

Figure 3.

 

(A) Arteriovenous phase of a preoperative fluorescein angiogram in dog 2 showing normal retinal vascular perfusion (magnification, ×10). (B) Postoperative fluorescein angiogram in the same animal at 40 weeks. The arteriovenous phase shows good retinal vascular perfusion. Some staining is evident. Hyperpigmented areas, retinal tacks, and pixels of electrode array appear as blocked fluorescence (magnification, ×10). (C) Late frame (3 minutes after injection) of a fluorescein angiogram at 40 weeks in the same animal. The hyperpigmented areas exhibit staining with fluorescein. (D) Late frame (10 minutes after injection) of a fluorescein angiogram at 40 weeks in the same animal. The hyperpigmented areas evident at 3 minutes have not enlarged, indicating that the leakage is contained within the affected area of RPE. There is no evidence of choroidal neovascularization.

Figure 4.

 

(A) Electroretinograms (raw waveform) of dog 3 in the operated and fellow eye recorded at 24 weeks postoperative. (B) Average of normalized data of ERG in all dogs recorded at each follow-up. The data were represented as percentage of the unoperated eye in the same animal. (C) 31 Hz ERG data. Raw waveform shows sinsusoidal nature of response. (D) Fourier transform of the photopic data shows magnitude peaks at 31 Hz and higher harmonics. The control and operated eyes have almost identical magnitudes.

Figure 5.

 

(A) VEPs (raw waveform) of dog 3 in the operated and fellow eyes recorded at 24 weeks postoperative. (B) Averages of normalized data of VEPs in all dogs recorded at each follow-up. Data are represented as percentage of the unoperated eye in the same animal.

Figure 6.

 

(A) Light microscopic photomicrograph of retina in the normal eye in dog 4 showing normal retinal histology (magnification, ×400). (B) Light microscopic photomicrograph of the retina under electrode array (3 months after implantation) in dog 4 showing normal appearance of outer nuclear layer (ONL), inner nuclear layer (INL), and ganglion cell layer (GCL) of the retina, there is an artifactual retinal separation due to tack removal during tissue processing. RPE, tapetum, and choriocapillaris appear normal (magnification, ×400). (C) Transmission electron microscopic photomicrograph of the area under the electrode array showing intact RPE, tapetum (T), and outer segments of photoreceptors (arrowhead). Scale bar, 1 μg.