Photovoltaic retinal prostheses utilizing optically configurable electric field confinement

Abstract

The photovoltaic retinal prosthesis provides optically configurable electric field confinement. A video stream is projected onto the retinal implant. An array of photovoltaic pixels is configured to provide retinal stimulation in response to the video stream. The photovoltaic pixels comprise a common return electrode. Each pixel comprises an active electrode coupled to retinal tissue via a capacitive or faradaic interface. Each pixel comprises a photodiode connected in series between the common return electrode and a corresponding active electrode. The projected video stream is configured based on the source video stream such that one or more pixels of the retinal implant that darken in a next frame of the projected video stream are optically preconditioned (precharged) by the projected video stream in a previous frame to be sufficiently conductive to act as a transient local return electrode in the next frame of the projected video stream.

Description

Technical Field

[0001] The present invention relates to a photovoltaic retinal prosthesis.

Background Art

[0002] Retinal degenerative diseases such as age-related macular degeneration (AMD) and retinitis pigmentosa are the main causes of untreatable visual impairment and legal blindness. Even when photoreceptor cells are irreversibly lost, most of the neurons inside the retina often remain. Since electrical stimulation of secondary retinal neurons (mainly bipolar cells) induces visual perceptual representations, electronic vision restoration becomes possible. AMD patients wearing the subretinal implant PRIMA (Pixium Vision, Paris, France) with 100-μm bipolar pixels showed a prosthetic letter visual acuity of 1.17±0.13 pixels corresponding to Snellen visual acuity of 20 / 460 to 20 / 565. This is a very attractive proof of concept, but for this approach to be widely adopted by AMD patients, it is usually necessary for the prosthetic vision to reliably exceed other peripheral fields that are usually above 20 / 400. The sampling limit for visual acuity of 20 / 200 corresponds to 50-μm pixels, and 20 / 100 corresponds to 25-μm pixels. Similar to natural vision, prosthetic vision is basically limited not only by spatial resolution (i.e., pixel size) but also by the contrast of the stimulation pattern affected by crosstalk between adjacent electrodes. The lateral spread of the electric field, like that of the PRIMA implant, can be suppressed by the local return electrodes of each pixel, but since the penetration depth of the electric field in the tissue is also limited to approximately the pixel radius, it is difficult to scale down such bipolar pixels. As a result, the retinal stimulation threshold in such a shape rapidly increases with the decrease in pixel size, and even with one of the best electrode materials (SIROF) with a pixel size of less than 40 μm, it exceeds the safe charge injection limit.

[0003] One approach to overcome this problem is based on using a three-dimensional honeycomb array to raise the return electrodes to the top of the inner nuclear layer, thereby orienting the electric field vertically in the well. This arrangement decouples the depth of penetration of the visual field from the pixel width, and the vertical visual field aligns with the orientation of bipolar cells in the retina, thus significantly reducing the stimulation threshold. Early animal experiments showed promising results for retinal migration to subretinal wells, but it has not yet been confirmed whether the migrated neurons function against the three-dimensional array. Moreover, the manufacturing process of honeycomb structures with localized return electrodes is far from trivial and requires further development. [Overview of the Initiative] [Means for solving the problem]

[0004] The spatial resolution of retinal prostheses is limited by pixel size and crosstalk from adjacent electrodes. While local return electrodes in bipolar pixels help reduce crosstalk, they excessively restrict the penetration of the electric field into the tissue, thus limiting the effectiveness of nerve stimulation. The potential of the active electrode increases due to the charging of the active electrode, which is capacitively or Faraday-bound to an electrolyte, and the electric field generated by adjacent active electrodes. As the potential across the photodiode of a photovoltaic pixel increases, the photodiode becomes more conductive, effectively transforming into a transient return electrode. Therefore, pre-charging the active electrode ensures that if a pixel becomes dark in the next image, the active electrode becomes an effective return electrode for the next pulse. The distance between the active and return electrodes defines the penetration depth of the electric field into the tissue. Thus, by pre-tuning pixels to be transient returns in the next image frame, the spatiotemporal control of the image projected onto the photovoltaic array allows for flexible control of the lateral and axial confinement of the electric field within the tissue. The depth of stimulation and lateral selectivity can be optimized for each patient, depending on the retinal thickness and the proximity of the implant.

[0005] Alternatively, photovoltaic pixels can be transformed into transient returns by optically controlling the discharge using photosensitive transistors that respond to different wavelength ranges. These are phototransistors or metal-oxide-semiconductor field-effect transistors (MOSFETs) gated by secondary photodiodes. Such separate optical control elements help discharge the active electrodes within the pixel more quickly and can also optimize current steering within the retina to generate a specified electric field.

[0006] In a first embodiment, the present invention is a method for providing illumination to a photovoltaic retinal prosthesis. The method is characterized by projecting a video stream onto a retinal implant. The retinal implant comprises an array of photovoltaic pixels configured to provide retinal stimulation in response to the video stream. The array of photovoltaic pixels comprises a common return electrode. Each pixel comprises an active electrode coupled to retinal tissue via a capacitive interface or a Faraday interface. Each pixel comprises one or more photodiodes connected in series between the common return electrode and the corresponding active electrode. The method further comprises configuring the projected video stream based on a source video stream such that one or more pixels of the retinal implant that darken in the next frame of the projected video stream are optically pre-conditioned (pre-charged) by the projected video stream to become sufficiently conductive and function as transient local return electrodes between the next frames of the projected video stream.

[0007] In a further embodiment of this method, the transient local return electrode is pre-tuned to reach a bias voltage in the range of 0.2V to 0.7V per photodiode by illumination of the projected video stream.

[0008] More optimally, the transient local return electrodes are pre-tuned to achieve a bias voltage in the range of 0.3V to 0.6V per photodiode.

[0009] In yet another embodiment, the method is characterized in that the image processing period for pre-adjustment and stimulation is shorter than the frame duration of the source video stream.

[0010] In yet another embodiment, the method features a pre-adjustment algorithm that defines the polarity and amplitude of the current at each electrode, optimizes it under the least mean squares error criterion, and estimates a target electric field in biological tissue.

[0011] In a second embodiment, the present invention relates to projecting a video stream onto a retinal implant. Near Eye This is a retinal prosthesis system with a display. The retinal implant includes photovoltaic pixels that convert near-infrared light (e.g., 850-915 nm) projected from the display into an electric current that flows through the biological tissue, stimulating retinal neurons. Each pixel of the retinal implant comprises one or more photodiodes connected in series between an active electrode and a return electrode. The active electrode and return electrode are coupled to the electrolytes of the biological tissue via a capacitive interface or a Faraday interface, and the return electrodes of the pixels are connected to each other. Near Eye The series of images in the video stream projected onto the retinal implant by the display are designed to be conductive enough to function as transient return electrodes by optically pre-tuning designated pixels and accumulating a bias voltage at the electrode-electrolyte interface of these pixels.

[0012] In a further embodiment of this system, the photodiodes are fabricated from crystalline silicon, and the bias voltage is configured to be in the range of 0.2 to 0.7 V per photodiode. More optimally, the bias voltage is greater than 0.3 V per photodiode.

[0013] In yet another embodiment of this system, the penetration depth of the electric field into the retina is optimized according to the patient's anatomical structure by controlling the distance between the illuminated pixels of the current video frame and the pixels that were pre-adjusted in the previous frame but are not illuminated in the current video frame.

[0014] In yet another embodiment of this system, the photovoltaic pixel comprises one or more optically controlled transistors to adjust the discharge current.

[0015] In a third embodiment, the present invention is a retinal prosthesis comprising an array of photovoltaic pixels configured to provide retinal stimulation in response to a received video stream. The array of photovoltaic pixels comprises a common return electrode. Each pixel comprises an active electrode that capacitively or Faraday-couples to retinal tissue. Each pixel comprises one or more photodiodes connected in series with the return electrode and the corresponding active electrode. Each pixel further comprises an optically controllable conductance element, and illumination of the array of photovoltaic pixels by a secondary illumination pattern selects one or more pixels of the retinal implant that function as local return electrodes by activating the corresponding conductance element. The conductance element is not sensitive to NIR light but is sensitive to a different wavelength range (e.g., the visible wavelength range). This secondary illumination pattern is emitted in the wavelength range that affects the conductance element.

[0016] The embodiments described later do not require pre-adjustment (pre-charging) of pixels and enable direct control of pixel conductivity using the secondary wavelength of light. This embodiment simplifies the control of pixel conductivity, thereby simplifying the control of electric field confinement within the retina. [Brief explanation of the drawing]

[0017] [Figure 1]This figure shows a near-eye projection system for a retinal prosthesis, according to an exemplary embodiment of the present invention, comprising a video camera on augmented reality glasses, an image processing device, a projector that provides images captured by the camera using near-infrared (e.g., 850-915 nm) light to a retinal implant, and a subretinal photovoltaic array. [Figure 2A] Figure 2A shows a photovoltaic array having 40 μm pixels according to an exemplary embodiment of the present invention, which is connected to a common return electrode at the periphery (1). A high-magnification view of the array (Figure 2B) shows individual pixels of a hexagonal array having an active electrode (2) in the center and a light-receiving region (3) around the periphery. [Figure 2B] Figure 2A shows a photovoltaic array having 40 μm pixels according to an exemplary embodiment of the present invention, which is connected to a common return electrode at the periphery (1). A high-magnification view of the array (Figure 2B) shows individual pixels of a hexagonal array with an active electrode (2) in the center and a light-receiving area (3) around the periphery. [Figure 3] This figure shows how the conductivity of a diode increases exponentially with forward bias, according to an exemplary embodiment of the present invention. [Figure 4A] Figure 4A shows the calculated potential above the electrode array in the retina when the central pixel is injected with an anode current of 1 μA, according to an exemplary embodiment of the present invention. Top figure: Top view of the electrode array surface. Dashed circles indicate the electrodes of six adjacent pixels. Contour lines show the potential in mV relative to a reference electrode placed at a distance. Bottom figure: Side view taken at y=0. Arrows indicate the direction of local current. The circuit diagram schematically shows the photodiode and capacitive electrode-electrolyte interface of the photovoltaic pixel, and their return electrodes are connected to each other. Similarly, Figure 4B shows the photodiodes of six adjacent pixels optically pre-tuned (pre-charged) to 0.54 V before activation of the central pixel. [Figure 4B]Figure 4A shows the calculated potential above the electrode array in the retina when the central pixel is injected with an anode current of 1 μA, according to an exemplary embodiment of the present invention. Top figure: Top view of the electrode array surface. Dashed circles indicate the electrodes of six adjacent pixels. Contour lines show the potential in mV relative to a reference electrode placed at a distance. Bottom figure: Side view taken at y=0. Arrows indicate the direction of local current. The circuit diagram schematically shows the photodiode and capacitive electrode-electrolyte interface of the photovoltaic pixel, and their return electrodes are connected to each other. Similarly, Figure 4B shows the photodiodes of six adjacent pixels optically pre-tuned (pre-charged) to 0.54 V before activation of the central pixel. [Figure 5] This figure shows a side-by-side comparison of the potentials of adjacent pixels with and without pre-adjustment (upper panel) according to an exemplary embodiment of the present invention. The calculated potential in the medium is calculated when a current of 1 μA is injected into the central pixel. Note that the electric field is more tightly confined in the lower panel. [Figure 6A] Figure 6A shows a photovoltaic array having 40 μm pixels illuminated by a diffraction grating pattern, according to an exemplary embodiment of the present invention. Figure 6A is a side view of the potential above the photovoltaic array having 40 μm pixels illuminated by a diffraction grating pattern with stripes equal in width to 1 pixel, showing both with and without pre-adjustment (solid line). With pre-adjustment, the contrast between the bright and dark stripes is 100%, but without pre-adjustment, the potential above the illuminated stripes is only about 25% higher than above the dark stripes. Such low contrast makes it impossible to detect the visual response of rats to the alternating grating. [Figure 6B]Figure 6A shows a photovoltaic array having 40 μm pixels illuminated by a diffraction grating pattern, according to an exemplary embodiment of the present invention. Figure 6A is a side view of the potential above the photovoltaic array having 40 μm pixels illuminated by a diffraction grating pattern with stripes equal in width to 1 pixel, showing both with and without pre-adjustment (solid line). With pre-adjustment, the contrast between the bright and dark stripes is 100%, but without pre-adjustment, the potential above the illuminated stripes is only about 25% higher than above the dark stripes. Such low contrast makes it impossible to detect the visual response of rats to the alternating grating. [Figure 7A] The amplitudes of visually evoked potentials for alternating gratings in rats using photovoltaic arrays with 20 μm and 40 μm pixels are shown according to exemplary embodiments of the present invention. The alternating grating response decreases as the fringe width decreases, reaching a noise level below the visual acuity limit (horizontal line). For 40 μm pixels (Figure 7A), this limit coincides with a pixel size of 40 μm (indicated by the arrow). For 20 μm pixels (Figure 7B), the limit is set by the rat's innate spatial resolution, which is approximately 27 μm (also indicated by the arrow). [Figure 7B] The amplitudes of visually evoked potentials for an alternating grid in rats using photovoltaic arrays with 20 μm and 40 μm pixels are shown according to exemplary embodiments of the present invention. The alternating grid response decreases as the fringe width decreases, reaching a noise level below the visual acuity limit (horizontal line). For 40 μm pixels (Figure 7A), this limit coincides with a pixel size of 40 μm (indicated by the arrow). For 20 μm pixels (Figure 7B), the limit is set by the rat's innate spatial resolution, which is approximately 27 μm (also indicated by the arrow). [Figure 8A] Figure 8A shows a circuit diagram of a photovoltaic pixel comprising a phototransistor capable of independently controlling the discharge current with light of different wavelengths, according to an exemplary embodiment of the present invention. Replace the phototransistor in Figure 8A with a MOSFET controlled by a secondary photodiode. [Figure 8B] Figure 8A shows a circuit diagram of a photovoltaic pixel including a phototransistor that can independently control a discharge current by light of different wavelengths according to an exemplary embodiment of the present invention. Replace the phototransistor in Figure 8A with a MOSFET controlled by a secondary photodiode. [Figure 9] According to an exemplary embodiment of the present invention, it is a diagram showing that each frame of a video stream is divided into a pre-adjustment stage and a stimulation stage. During the pre-processing stage, pixels that darken during the stimulation stage are illuminated with an intensity below the threshold to accumulate charge at the electrode-electrolyte interface, thereby making these diodes conductive (Image 1). In a shorter stimulation stage, bright pixels are illuminated above the stimulation threshold (Image 2), and dark pixels, being conductive, serve as local return electrodes.

Mode for Carrying Out the Invention

[0018] The photovoltaic retinal prosthesis is activated by near-infrared (NIR) light projected from augmented reality glasses (Figure 1). Images are acquired by a camera attached to the glasses, processed by a pocket computer, and projected onto the eye using pulsed NIR light. Each pixel of the implant (Figures 2A - 2B) converts this light into a pulsed current flowing through the retina, thereby stimulating neighboring intraretinal neurons.

[0019] This paper introduces an approach to high-resolution prosthetic vision using planar subretinal implants that utilizes the conductivity of diodes under forward bias by spatiotemporal modulation of photovoltaic pixels, enabling high penetration depth and high contrast of electric fields within tissue. The potential of the active electrode increases as charge accumulates at the electrode-electrolyte interface and the potential from adjacent pixels in the electrolyte couples. Since the forward conductance of the photodiode increases exponentially with the electrode potential (Figure 3), some pixels become sufficiently conductive and thus converted into transient local returns, which help confine the electric field generated by other pixels in the array. Therefore, by pre-charging some of the active electrodes, if a pixel becomes dark in the next image projected onto the array, it becomes a valid return electrode for the next image. The distance between the active and return electrodes defines the penetration depth of the electric field into the tissue. By pre-tuning pixels in this way to become transient returns in the next image frame, spatiotemporal modulation of the image projected onto the photovoltaic array allows for flexible control of the lateral and axial confinement of electric fields within the tissue. Depending on the thickness of the retina and the proximity of the implant, it is possible to optimize the depth of stimulation and lateral selectivity for each patient. Such a unipolar photovoltaic array is shown in Figures 2A-2B. Each pixel comprises an active electrode (2) and a photosensitive area (3) connected to a common return electrode located at the end of the implant (1).

[0020] Modeling such behavior involves (a) characterizing the spatial coupling between pixels using static finite element modeling, and then (b) calculating the dynamics of the photovoltaic circuit in a multidimensional form describing the potential in the electrolyte as a function of space and time. As shown in Figures 4A–4B, 5, and 6A–B, the computational model demonstrated that with pre-adjustment, the confinement of the field around the illuminated pixels is far better than without pre-adjustment. These results were validated by comparison with potential measurements in the electrolyte. Most importantly, in vivo measurements of diffraction grating visual acuity in animals using 40 μm pixels demonstrated that this approach allows for a resolution limited by pixel size (Figure 7A), as predicted by the computational modeling (Figures 6A–6B), which was not possible with pixels smaller than 55 μm in other configurations. Furthermore, with 20 μm pixels, as shown in Figure 7B, rat visual acuity is limited by a natural spatial resolution of approximately 27 μm (1.2 cpd).

[0021] The conductance of the pn junction increases exponentially with the forward bias voltage, and becomes significant only when the sum of (a) the voltage rise due to charge accumulation at the electrode-electrolyte interface and (b) the potential rise due to the current generated in adjacent pixels in the electrolyte exceeds the diode's turn-on voltage, which is approximately 0.5V for a silicon-based photodiode. Therefore, when the forward bias falls below the turn-on voltage, further electrode discharge becomes inefficient. A shunt resistor in parallel with the photodiode can accelerate the discharge of the active electrode, but the discharge current changes over time and depends on the coupling between pixels.

[0022] To provide more reliable and independent control of pixel discharge and transient return, optical control can be integrated using photosensitive transistors within each pixel that respond to a different wavelength range than the NIR light used in the primary photodiode. For example, the photovoltaic pixels in Figures 4A-4B can be replaced with those in Figure 8A. Each pixel has a primary photodiode and a phototransistor in parallel to control the discharge. The phototransistor is protected by a dichroic coating that blocks the NIR wavelength used to drive the primary photodiode and allows secondary wavelengths (such as visible light) to pass through. To prevent unintended discharge due to ambient light, this wavelength must be blocked with glass. To convert the pixel into a transient return, the phototransistor is turned on by light of the secondary wavelength, causing a current to flow in the opposite direction to the photocurrent flowing through the primary photodiode. The intensity of the secondary light may be significantly lower than that of the primary NIR beam, and the amplitude of the negative current is controlled by the intensity profile of the secondary light. To obtain higher current gain and greater input resistance, a MOSFET controlled by a secondary photodiode may be used instead of a phototransistor (Figure 8B).

[0023] By optically configuring negative currents for various pixels, it becomes possible to optimize current steering for field confinement in the retina. Each electrode x = [x1, x2, ..., x] generates the target electric field v. N ] T To find the optimal current, we utilize the linearity of the electric field and formulate the problem as minimizing the difference between the actual electric field and the target electric field under the criterion of least mean squares error (MMSE):

[0024]

number

[0025] Here, U is the transformation matrix from the current at each electrode to the electric field at the retina. The solution to equation (1) can be efficiently calculated in real time using a one-step matrix-vector multiplication. However, this solution may involve currents with large amplitudes that exceed the safe limits of electrode charge injection. We introduce a method to prevent the current amplitude from becoming excessively large by using L-2 regularization in the optimization:

[0026]

number

[0027] Here, .'' 2 is the ridge parameter that determines the trade-off between the strength of regularization and the similarity with the specified electric field. The real-time solution of (2) can be computed with a single matrix-vector multiplication:

[0028]

number

[0029] Signal processing, pixel pre-adjustment, and stimulus sequencing can be implemented as follows: At a typical frame rate of 30 Hz, each frame lasts approximately 32 ms. The typical illumination time for a stimulated pixel is 1–10 ms. When an image frame is acquired by the camera, sufficiently dark pixels are designated as transient returns (e.g., by thresholding). (1) In the pre-adjustment phase of the frame, these pixels are exposed to light at an intensity below the stimulation threshold for a sufficient time to accumulate the necessary charge and become conductive during the stimulation phase of the frame. For a 40 μm wide photovoltaic pixel, the typical range of the stimulation current is 0.01–1 μA. As shown in Figure 3, this corresponds to a voltage range of approximately 0.4–0.6 V. Since part of the potential of the return pixels during the stimulation phase is supplied from adjacent active pixels, the cumulative voltage at the electrode-electrolyte interface is as low as 0.2 or 0.3 V. If the photovoltaic pixel contains multiple diodes in series, this voltage drop is measured per diode. Since the stimulation current density near the electrode is the same for pixels of any size, this voltage range should not depend much on the pixel size. (2) During the stimulation phase, the designated bright pixels are exposed to bright light of an intensity and duration (typically 1-10 ms) corresponding to the desired charge injection, while the dark pixels return current to the return electrode. In this configuration, the stimulation is never delayed by more than one frame relative to image acquisition.

[0030] Avoiding stimulation during the pre-adjustment phase limits the maximum current for transient return, and consequently the maximum charge accumulation, which may not be sufficient to enable conduction at high frame rates. An alternative sequence for signal processing for pixel pre-adjustment and stimulation is as follows: All pixels remain low-light by default, ready to be converted into transient returns. When an image frame is acquired by the camera, sufficiently bright pixels are identified (e.g., by thresholding). (1) During the frame pre-adjustment phase, these bright pixels are kept dark to discharge electrodes, while other pixels remain illuminated at a default level. (2) During the stimulation phase, designated bright pixels are exposed to bright light as described above, while dark pixels are kept dark to sink current as transient returns. With such an arrangement, the stimulation threshold should be approximately five times higher during the pre-adjustment phase than during the anode simulation phase with photocurrent, as the current to the pixels designated to be bright is cathode, thus reducing the likelihood of accidental stimulation. Furthermore, even if such cathode stimulation were to occur during the pre-adjustment phase, these pixels should be activated by the photocurrent a few milliseconds later, so it shouldn't cause much disruption.

[0031] By combining the two strategies described above, it becomes possible to balance minimizing energy consumption with avoiding unintended stimuli.

[0032] Other approaches may be based on image multiplexing. With a typical image refresh rate of 30 Hz, each frame lasts approximately 32 milliseconds. The photovoltaic stimulation pulses typically range from 0.8 to 8 ms, with at least four pulses active within each frame. Thus, pixels in the image are divided into four groups and activated sequentially. In this way, pixels illuminated in the previous group are charged, acting as local returns for pixels activated in the next group. Another advantage of this method is that only a quarter of the pixels are activated simultaneously, and even fewer if the image is sparse. This significantly reduces potential accumulation, improving the localization of the stimulus. By utilizing the varying pulse durations across different pixels, it is possible to divide pixels into more groups or to activate them asynchronously.

[0033] To speed up image processing for determining dark pixels for pre-adjustment and reduce latency, the continuity of natural visual input can be used to predict the next frame based on the previous image in the video sequence. For this purpose, predictive tracking algorithms such as Kalman filters and exponential smoothing can be applied.

Claims

1. A method for providing illumination to a photovoltaic prosthetic retina, which is carried out by a retinal prosthesis device equipped with a near-eye display, (a) The step of the near-eye display projecting a video stream, which includes a sequence of frames, onto a retinal implant, wherein each frame of the video stream is divided into a first image in a pre-adjustment phase and a second image in a stimulation phase, (b) The retinal prosthesis comprises the step of composing the first image of each frame projected on a source video stream so that one or more pixels of the retinal implant that are darkened in the second image in each frame of the projected video stream are optically pre-tuned by the projected video stream so that they are sufficiently conductive to act as transient local return electrodes during the projection of the second image in each frame of the projected video stream, The retinal implant comprises an array of photovoltaic pixels configured to provide retinal stimulation in response to the video stream, The array of photovoltaic pixels is provided with a common return electrode. Each pixel is equipped with an active electrode that connects to retinal tissue via a capacitive interface or a Faraday interface. A method wherein each pixel comprises one or more photodiodes connected in series between a common return electrode and a corresponding active electrode.

2. The method according to claim 1, wherein the transient local return electrode is pre-tuned to reach a bias voltage in the range of 0.2V to 0.7V per photodiode by illumination from the projected video stream.

3. The method according to claim 2, wherein the transient local return electrode is pre-tuned to reach the bias voltage in the range of 0.3 to 0.6 V per photodiode.

4. The method according to claim 1, wherein the image processing units for pre-adjustment and stimulation are configured to be shorter than the frame units of the source video stream.

5. The method according to claim 1, further comprising a pre-adjustment algorithm for defining the polarity and amplitude of the current at each electrode, optimizing under the least mean squares error criterion, and estimating a target electric field in biological tissue.

6. A retinal prosthetic system, The retinal implant features a near-eye display for projecting a video stream containing a sequence of frames. Each frame of the video stream is divided into a first image in the pre-adjustment phase and a second image in the stimulation phase. The retinal implant includes a photovoltaic pixel that converts the light from the display into an electric current flowing through biological tissue to stimulate retinal neurons. Each pixel of the retinal implant comprises one or more photodiodes connected in series between the active electrode and the return electrode. The active electrode and the return electrode are bound to the electrolyte of the biological tissue via a capacitive interface or a Faraday interface. The return electrodes of the aforementioned pixels are connected to each other, A system in which, in each frame of the sequence of images of the video stream projected onto the retinal implant by the near-eye display, the first image is designed to be optically pre-tuned so that, during the projection of the second image, the designated pixel is sufficiently conductive to function as the transient return electrode by accumulating a bias voltage at the electrode-electrolyte interface of the pixel.

7. The aforementioned photodiode is made of crystalline silicon, The system according to claim 6, wherein the bias voltage is configured to be in the range of 0.2 to 0.7 V per photodiode.

8. The system according to claim 6, wherein the bias voltage exceeds 0.3V per photodiode.

9. The system according to claim 6, wherein the penetration depth of the electric field into the retina is optimized for the patient's anatomical structure by controlling the distance between the illuminated pixels in the second image of each frame and the pixels that are pre-adjusted in the first image but not illuminated in the second image.

10. The system according to claim 6, wherein the photovoltaic pixel comprises one or more optically controlled transistors for adjusting the discharge current.

Images