Age-related macular degeneration (AMD) is a leading cause of irreversible vision loss1, and its prevalence dramatically increases with age: from 1.5% in the US population above 40 years to more than 15% in the subjects older than 802. The atrophic form of AMD (also known as geographic atrophy, GA) results in a gradual loss of photoreceptors in the central macula, which is responsible for high-resolution vision, and severely impairs reading and face recognition. Low-resolution peripheral vision is retained in this condition, enabling orientation and the use of eccentric fixation for visual discrimination at reduced acuity. Therefore, the goal of any treatment strategy should be to restore functional central vision without jeopardizing the surrounding retina and allowing for their simultaneous use.
While PRs gradually disappear in GA, the inner retinal cells survive to a large extent3. To restore sight in the scotoma, we replace the lost photoreceptors with photovoltaic pixels in the subretinal implant, which convert light into electric current to selectively stimulate the secondary neurons in the retina4. These electronic substitutes of photoreceptors replace the two main functions of the natural photoreceptors: (a) the light-to-current conversion, corresponding to the function of the outer segment, and (b) transfer of the visual information to secondary neurons by their polarization in extracellular electric field, substituting the function of the synapse.
To avoid irreversible electrochemical reactions at the electrode-electrolyte interface, stimulation current is pulsed and charge-balanced. On the other hand, to provide steady visual percepts under pulsatile illumination, repetition rate should exceed the frequency of flicker fusion. In preclinical studies, we demonstrated that selective stimulation of bipolar cells without direct activation of the downstream neurons results in preservation of multiple features of the natural retinal signal processing, including flicker fusion, adaptation to static images4, ON and OFF responses with antagonistic center-surround5, and non-linear summation of subunits in RGC receptive fields4. We have also shown that visual acuity matches the pixel pitch with 75 and 55µ m pixels4,6.
The first generation of the human-grade photovoltaic subretinal prosthesis PRIMA (Pixium Vision SA., Paris, France) is 2 mm in width (~ 7° of the visual angle in a human eye), 30 µm in thickness, containing 378 hexagonal pixels of 100 µm in width. Images captured by the camera are processed and projected onto the retina from video glasses using intensified light (Fig. 1). To avoid photophobic and phototoxic effects of bright illumination, we use near-infrared (NIR, 880 nm) wavelength7. Photovoltaic pixels in the implant directly convert the projected pulsed light into local electric current flowing through the retina between the active and return electrodes4,8.
Five patients with GA were implanted in Paris during 2017–2018 (NCT03333954). In four of them, the implant was placed in the subretinal space, but in one it ended up inside the choroid due to patient’s accidental movement during surgery. In one of the four patients, the implant accidentally shifted by about 2 mm from the central position after the fluid-air exchange since the patient did not keep the head in a prone position post implantation. Due to wireless nature of the implant, surgical procedure was relatively short – about 2 hours9. As shown in Table 1, residual natural acuity in the operated eye did not decrease in any of the subjects. Interestingly, in some patients, acuity improved compared to baseline, which could be attributed to either a neurotrophic benefit of subretinal surgery10 or of electrical stimulation11 or just improvement with eccentric fixation after training.
In the first phase of the trial, reported earlier9, prosthetic vision was assessed independently from the remaining natural vision. For this purpose, opaque virtual reality glasses (VR, PRIMA-1) have been used. The projected images covered a horizontal field of 5.1 mm (17.5° on the retina), with approximate resolution of 10.5 µm. Maximum peak retinal irradiance was 3 mW/mm2, well within the thermal safety limits for chronic use of near-infrared light12. Brightness of the percept was controlled by pulse duration, between 0.7 and 9.8 ms, in 0.7-ms increments.
The four patients with subretinal implant placement demonstrated monochromatic (white-yellowish “sun-color”) shaped vision, with flicker fusion above 30 Hz. In three patients with central location of the subretinal implant, acuity closely matched the pixel size: 20/460, 20/500 and 20/550 (1.1, 1.2 and 1.3 pixels). Patient with the off-center implant demonstrated lower acuity: 20/800 (1.9 pixels)9. Patient #1 with the intra-choroidal implant had blurry prosthetic vision, with no discernable acuity.
In the second phase of the study, starting at 18–24 months post-op, we introduced augmented reality glasses (AR, PRIMA-2), which allow unobstructed natural vision by the fellow eye and by the peripheral field of the operated eye, simultaneously with prosthetic central vision in the treated eye (Fig. 2a). The projected images covered a horizontal field of 5.3 mm (18.5°) on the retina, with a resolution of 6.7 µm, as illustrated in Fig. 2b. This design provided improved beam homogeneity and easier alignment, compared to VR glasses (PRIMA-1). The maximum retinal irradiance was increased to 3.5 mW/mm², with the same range of pulse durations as in the VR glasses. This system allows the use of electronic magnification (x1, x2, x4 and x8) between the camera and the image projection onto the implant. As shown in Table 1, perceptual thresholds 18–24 months after the implantation, measured with PRIMA-2 glasses, were slightly lower than the thresholds measured during the first 6 months - in the first phase of the trial9. Patient #3 passed away due to unrelated cause before the second phase of the trial.
Prosthetic visual acuity with PRIMA-2 glasses was measured using Landolt C optotypes. To mimic the crowding effect of the letter charts, the Landolt rings were surrounded by a square frame (Fig. 3a). At each trial, subjects reported the font orientation (up, down, left or right), and its size was then adjusted, depending on the response. The visual acuity was determined using the Freiburg Visual Acuity Test (FrACT) software13,14. For a stable perception under pulsatile illumination, 30 Hz repetition rate was applied. In the first set of the tests, computer-generated Landolt optotypes were projected into the eye directly from the AR glasses without using a camera. As shown in Table 1, patients #2 and #5 demonstrated prosthetic acuity at the level similar to that observed with VR glasses in the first phase of the trial (20/500, 20/460), but patient #4 significantly improved compared to the earlier result - from 20/800 to 20/438. This is potentially due to easier alignment of the display to the off-center location of the implant with improved glasses. The average acuity in the four patients with the subretinal implant placement was 1.17 ± 0.13 pixels at the latest measurement, corresponding to logMAR 1.39, or 20/500 on a Snellen scale.
In the second set of the acuity tests, letters were displayed at 40 cm distance from the subject, so patients used camera and were allowed to apply their preferred electronic magnification (1, 2, 4 or 8). To ensure that prosthetic acuity is measured rather than the residual natural vision, in these tests the fellow eye was covered. In addition, contrast of the electronic image was inverted from the original black letter on white background to white letters on black background (white patterns stimulate the retina), and patients were asked about the color of the percept. With magnification, all three participants of the second trial demonstrated significant improvement in prosthetic acuity: to the level of 20/98, 20/71 and 20/63, respectively. As shown in Table 1, these values significantly exceeded their residual natural acuity in the treated eye, and for patients 4 and 5, even in the (better) fellow eye.
Video S1 illustrates a test of prosthetic vision using an ETDRS chart, with 4x magnification and a contrast reversal. Video S2 illustrates a reading test with 4x magnification and a control experiment (Video S3) where patient is attempting to read the same word without the PRIMA glasses.
To evaluate the effect of background light on prosthetic vision when the transparent AR glasses are used, Landolt C optotypes have been presented on the glasses display directly, without using the camera, while intensity of the background visible light was varied. In this experiment, subjects with both eyes open were placed 40 cm in front of a wide LCD screen, where a homogeneous white illumination at 16 levels (ranging from 1.4 to 256 cd/m2) was presented. Prosthetic patterns were presented at maximum brightness: 3.5 mW/mm2 of NIR irradiance with 9.8 ms pulse duration. As shown in Table 1, subjects 2 and 4 did not have a problem seeing the Landolt C in front of the screen even with the highest background luminance (256 cd/m2). Subject 5 had difficulties with luminance above 64 cd/m2, and therefore was provided later with a shaded lens (65% attenuation of white light) to allow using the device in a bright office environment.
It is important to note that patients could simultaneously use prosthetic and residual natural vision from both, the study eye, and the fellow eye. For example, in a setup shown in Fig. 3b, green bars of various orientations were presented on a large screen for natural vision and another set of bars was simultaneously presented just on the NIR display inside the glasses. The patient was asked about both orientations and colors, as illustrated in the video S4 for binocular vision and in S5 for monocular vision. In both cases, bars were perceived simultaneously and orientations detected correctly.
In summary, this trial confirmed the safety and stability of the PRIMA implant over 24–30 months follow-up in four patients with geographic atrophy. Prosthetic central vision in the former scotoma represents shaped monochromatic perception matching the presented patterns and, most importantly, is perceived in conjunction with the residual peripheral vision, thus enabling natural orientation and central discrimination. Spatial resolution was, on average, 1.2 pixels of the implant, corresponding to letter acuity of about 20/500, and using electronic magnification, all patients with subretinal implant demonstrated acuity exceeding 20/100. Further advancements in the photovoltaic pixel design15, video glasses and image processing promise even more functional restoration of sight for numerous patients suffering from atrophic macular degeneration.