Honeycomb-shaped electrical neural interface for retinal prosthesis

Abstract

We present a 3D configuration of electrode arrays for neural cell stimulation, designed to maximize the use of the migration of retinal cells into the voids in the subretinal space. Walls around each pixel align the electric field vertically, matching the orientation of the bipolar cells in the retina, and thereby reduce the stimulation threshold. These walls also separate the field penetration depth from the pixel width, enabling the pixel size to be reduced to the cell size. The inner retinal cells migrate into the electrode cavities, enabling very efficient stimulation. Because of the one-dimensional arrangement of the electric field along the cavities, the stimulation threshold current density does not increase significantly with the reduction in pixel size, unlike the quadratic increase seen in planar arrays.

Description

Technical Field

[0001] This invention relates to the electrical stimulation of nerve cells. Background Technology

[0002] Conventional stimulation arrays for visual prostheses have a planar configuration of active and return electrodes. This planar configuration presents significant challenges in scaling the prosthesis to have sufficiently small and dense pixels to provide useful visual acuity. As pixel size decreases, two main difficulties arise: increased crosstalk between adjacent pixels and an increase in the pixel stimulation threshold (i.e., the current density required to elicit a cellular response). Increasing the stimulation threshold with decreasing pixel size is particularly problematic because the required current density becomes biologically unsafe at the ideal pixel size for retinal implants. Therefore, mitigating these limitations of planar stimulation arrays would be an advancement in the field. Summary of the Invention

[0003] We present a novel 3D configuration of electrode arrays designed to maximize the migration of retinal cells into the subretinal space. The walls surrounding each pixel are vertically aligned with the electric field, matching the orientation of bipolar cells in the retina and thereby lowering the stimulation threshold. These walls also separate the field penetration depth from the pixel width, enabling pixel size reduction to the cell size. Internal retinal cells migrate into the honeycomb wells, allowing these neurons to reside within the electrode cavities, resulting in highly effective stimulation. Due to the one-dimensional alignment of the electric field along the vertical walls of the honeycomb cavities, the stimulation threshold current density does not increase significantly with decreasing pixel size, unlike the quadratic increase in planar arrays. Compared to planar electrode arrays, this 3D electrode configuration enables vision restoration with smaller pixels, thus achieving higher visual acuity. Similar 3D arrays can be used for electrical neural interfaces with the brain in other applications.

[0004] In a preferred embodiment, the pixel cavity has a depth between 10 μm and 100 μm and a width between 5 μm and 100 μm. The current density at the active central electrode is preferably 0.01 A / cm². 2 and 1A / cm 2 The width of the device as a whole is preferably between 0.5 mm and 5 mm to make it suitable for retinal implantation. The charge injection per pulse on the active electrode is preferably 0.1 mC / cm. 2 and 10mC / cm 2 between.

[0005] Applications typically include retinal prostheses for restoring vision in retinal degeneration, and high-resolution electro-neural interfaces. This work advantageously enables the reduction of pixels in electro-neural interfaces to the cell size. Attached Figure Description

[0006] Figure 1 schematically illustrates the operation of a planar array for nerve cell stimulation.

[0007] Figure 2A The first embodiment of the present invention is shown.

[0008] Figure 2B A second embodiment of the present invention is shown.

[0009] Figure 3A -B is another view of the difference between planar and nonplanar stimulus arrays.

[0010] Figure 3C -D illustrates the use of retinal anatomy for coverage Figure 3A The implant described in -B.

[0011] Figure 4A -B shows an image of the fabricated cavity array.

[0012] Figure 4C An image of a honeycomb implant under the retina of an animal is shown.

[0013] Figure 5 Histology of the retina integrated with the implanted cellular array is shown.

[0014] Figure 6A The simulated potentials of the planar (top) and cellular (bottom) arrays are shown.

[0015] Figure 6B The experimental and simulated threshold current density of planar arrays relative to pixel size are shown compared to safety limits and compared to simulated threshold current density of cellular arrays relative to pixel size. Detailed Implementation

[0016] Section A describes the general principles relevant to embodiments of the invention. Section B provides detailed experimental examples.

[0017] A) General Principles

[0018] To better understand embodiments of the invention, it is helpful to first consider the operation of a conventional planar stimulation array 100 as shown in FIG. 1. This is a cross-sectional view of a row of pixels 102, 104, and 106. “ON” pixels (102 and 106, also marked with a checkmark) generate a current and inject it into an electrolyte through a central active electrode (e.g., 108 of pixel 102), which is then collected by a remote return electrode 110 located outside the pixel (or between pixels). The current is indicated by black arrows. The current through the electrolyte creates a potential gradient (shaded). The current diffuses from the active electrode in all directions and also affects cells in adjacent pixels. For example, “OFF” pixel 104, marked with an X, will receive parasitic stimulation from adjacent pixels, undesirably resulting in low contrast.

[0019] Figure 2A A cross-section of a first embodiment of the invention is shown. Here, the stimulation array 200 includes pixels 202, 204, and 206. “Conduction” pixels 202 and 206 (checkmarks) generate current and inject it into an electrolyte via a central active electrode (e.g., 214 of pixel 202). The current flows upward through the electrolyte (tissue) and is collected by a return electrode 216 at the top of wall 218 to complete the circuit. The current through the electrolyte primarily creates a potential gradient (shading) in front of the “conduction” pixels, allowing for localized stimulation of the retina. Note that compared to pixel 104 in FIG1, Figure 2A The shadow above pixel 204 is reduced. The high-capacitance material 216 deposited on top of the conductive honeycomb wall 218 ensures that most of the current through the electrolyte is collected at the top of the wall, and negligible amounts of current are collected on the sides of the conductive wall due to the low capacitance of the interface between the metal and the electrolyte. This configuration makes it possible to fabricate raised return electrodes in the honeycomb structure without the need for sidewall passivation or insulation.

[0020] More precisely, a first embodiment of the invention is a device for electrical stimulation of nerve cells, comprising an array of cavities configured to allow nerve cells to migrate within cavities. Each cavity has a base plate (e.g., Figure 2A 212) and conductive walls (e.g., Figure 2A (218 in the text). Each cavity has a first electrode disposed on its base plate (e.g., Figure 2A (214) and has a second electrode disposed on the top of its wall and perpendicularly separated from its corresponding base plate electrode (e.g., Figure 2A (216 in the original text). During device operation, an ion current flows through the contents of the cavity. The capacitance per cavity of the second electrode is greater than the capacitance per cavity of the conductive wall, thereby preferentially collecting the ion current on the side of the second electrode relative to the conductive wall. Preferably, the capacitance per unit area of ​​the second electrode is at least 100 times greater than the capacitance per unit area of ​​the conductive wall.

[0021] Figure 2B A cross-sectional view of a second embodiment of the invention is shown. Here, the stimulation array 250 includes pixels 252, 254, and 256. “Conduction” pixels 252 and 256 (checkmarks) generate current and inject it into an electrolyte via a central active electrode (e.g., 214 of pixel 252). The current flows upward through a honeycomb well surrounded by electrically insulating walls 262 and is then collected by a remote return electrode 264 outside the pixel. The current through the electrolyte primarily creates a potential gradient (shading) in front of the “conduction” pixels, allowing for significantly better local stimulation of the retina compared to the absence of vertical walls. Note the difference between pixel 104 in Figure 1 and... Figure 2B The shadow above pixel 254 is reduced.

[0022] More precisely, a second embodiment of the invention is a device for electrical stimulation of nerve cells, comprising an array of cavities configured to allow nerve cells to migrate within cavities. Each cavity has a base plate (e.g., Figure 2B 212 in the middle) and electrically insulating walls (e.g., Figure 2B (262 in the text). Each cavity has a first electrode disposed on its base plate (e.g., Figure 2B (214 in the text). The device includes a common return electrode disposed outside the cavity array (e.g., Figure 2B (See 264 above). During device operation, ion current flows through the contents of the cavity. Electrically insulating walls improve stimulation efficiency and reduce crosstalk between adjacent cavities of the device by forcing the ion current to travel vertically within the cavity.

[0023] Figure 2A The preferred configuration of the embodiment of -B is as follows. The depth of the cavity is preferably between 10 μm and 100 μm. The width of the cavity is preferably between 5 μm and 100 μm. The depth of the cavity is preferably greater than the width of the cavity. The cavity array is preferably periodic, and in this case, the cavities are preferably hexagonal. In a preferred embodiment, the nerve cells are retinal cells. Preferably, during operation, the injected charge density at the first electrode of the cavity is 0.1 mC / cm³. 2 and 10mC / cm 2 between.

[0024] In the preceding description, it is convenient to refer to cavities as having walls. However, in array structures (e.g., such as...) Figure 4A In most cases, adjacent cavities share a wall. For any shared wall, allocating portions of the wall to two adjacent cavities is arbitrary. The practice of this invention is entirely independent of how this theoretical partitioning is thought to be performed. Similarly, in most cases, local return electrodes (e.g., Figure 2A(216) is similarly shared between adjacent cavities. For any shared return electrode, allocating a portion of the return electrode to two adjacent cavities is arbitrary. The practice of this invention is entirely independent of how this theoretical partitioning is thought to be performed.

[0025] B) Specific experimental examples

[0026] B1) Introduction

[0027] Electronic methods for restoring vision are rapidly developing, with several systems approved for clinical use in patients blinded by hereditary retinal degeneration (retinitis pigmentosa, RP), and others in clinical trials. Patients affected by RP and implanted with either the preretinal prosthesis Argus II (Second Vision Medical Instruments, Sylmar, California) or the subretinal Alpha IMS / AMS (Retina Implants, Reutlingen, Germany) have shown improved performance in walking and visual search, with the best reported visual acuity of 20 / 1260 and 20 / 546, respectively. While encouraging, these benefits are insufficient to help treat the most common retinal degeneration, age-related macular degeneration (AMD), in which patients lose high-resolution central vision but retain peripheral visual acuity typically not less than 20 / 400.

[0028] A visual acuity of 20 / 200 is the legal limit for blindness in the United States, geometrically corresponding to a pixel pitch of approximately 50 μm. Safe charge injection across the electrode-electrolyte interface limits the minimum electrode size. Furthermore, crosstalk between adjacent electrodes increases as the pixel size decreases. This latter problem can be addressed by providing a circumferential return electrode in each pixel, but this approach further reduces the penetration depth of the electric field into the tissue.

[0029] Using cylindrical electrodes can reduce electrode-tissue separation in the subretinal space. Retinal cells in the inner nuclear layer (INL) migrate to fill the gaps in this 3-D implant, bringing the target neuron closer to the stimulating electrode and thus lowering the stimulation threshold. However, this cylindrical electrode only reduces the stimulation threshold by half and cannot significantly reduce the pixel size below 55 μm. The fundamental problem limiting electrode size is the shape of the electric field extending from one small electrode and returning to another electrode beneath the target cell.

[0030] Here, we offer a novel 3D geometry for subretinal prostheses, which we call a honeycomb structure, to overcome these limitations and enable pixel scaling down to the cell size. In this approach, the return electrode rises high along a vertical insulating wall around each pixel, vertically aligned with the electric field, matching the orientation of bipolar cells in the retina, thereby lowering the stimulation threshold. Figure 3B The raised return electrode 306 and Figure 3A The planar structures are compared. Here, 302 is the stimulation electrode and 304 is the substrate on which the stimulation electrode 302 is disposed.

[0031] Figure 3C The diagram illustrates a planar pixel with circular return, generating a locally confined electric field with shallow vertical penetration. Cells within this electric field are polarized according to a potential difference across their length. The bipolar cell body and axonal terminals are located in the INL (inner nuclear layer) and IPL (inner plexiform layer), respectively. Therefore, the potential relative to the middle of the IPL represents the electric potential. Figure 3D The diagram illustrates how a return electrode on top of the insulating wall creates a vertical dipole confined to a local pixel volume, thereby maximizing the vertical potential drop across the target cell layer. Current amplitudes (arrow lengths) are shown on a logarithmic scale. The potential difference relative to the midpoint of the IPL (57 μm) for a current of 68 nA is shown in grayscale.

[0032] These walls also separate field penetration depth from pixel width, enabling pixel size reduction to the cell size. We first investigated the anatomical integration of this 3D structure with the retina using implants with 20, 30, and 40 μm pixels, then quantified the electrical stimulation capability using an experimentally validated network-mediated retinal stimulation model. Our results demonstrate that this technology opens the door to prosthetic vision with visual acuity exceeding 20 / 100, which is highly beneficial not only for patients completely blinded by RP but also for the restoration of central vision in a large number of AMD patients.

[0033] Previously, we described the migration of retinal cells into pores in the membrane implanted in the subretinal space and its effect on subretinal prostheses. However, in that configuration, the base of the implant had openings that allowed current to propagate beneath the implant. In reality, however, retinal prosthesis implants are typically solid, meaning current does not pass through them, and therefore current can only flow upwards, as in this case... Figure 3C -D is shown. Therefore, we describe here the different geometries and materials of the return electrode for optimal shaping of the electric field with such a stimulation array.

[0034] B2) Result

[0035] B2a) Anatomical Integration

[0036] Figure 4A -C shows a subretinal cellular implant. Figure 4A Images of a 1mm wide device with a 25μm deep cell structure featuring pixel pitches of 40(*), 30(**), and 20μm(***) are shown. The fourth quadrant contains a 10μm pitch structure, exceeding the fabrication limits, and therefore was not unfolded. Since there are no walls, we refer to this as a "flat" region. Figure 4BIt is a higher magnification of a honeycomb with a 30μm pitch. Figure 4C This is an OCT image of the subretinal implant in RCS rats 6 weeks post-surgery. Figure 4A The scale is 200 μm. Figure 4B It is 50μm. Figure 4C It is 100μm.

[0037] To evaluate the integration of honeycomb structures with the degenerating retina, 1 mm diameter silicon arrays were implanted into the subretinal space of RCS rats (P180-300, n=6) for up to 6 weeks. Each array was divided into quadrants containing honeycomb and flat regions with spacing of 40, 30, and 20 μm. Figure 4A An array was fabricated in silicon using a Bosch etching process to define 25 μm deep cellular chambers. Figure 4B In vivo monitoring of implant-retinal integration using optical coherence tomography (OCT) Figure 4C Six weeks after implantation, INL was almost undetectable in OCT above the cellular level. Figure 4C (right), but above the flat quadrant ( Figure 4C (Left) and visible outside the implant, indicating that INL has migrated into the cavity.

[0038] Histological examination confirmed the migration of INL cells, without showing any visible signs of fibrosis or trauma, such as Figure 5 As shown. The retinal structure is still preserved, with well-defined INL, inner plexiform layer (IPL), and ganglion cell layer (GCL). Although some INL cells remain above the cell walls, the cavities are completely filled with dense cells up to the base of the array. The black arrows point to the original wall locations, which were removed and refilled with epoxy resin for sectioning after sample embedding. Figure 5 The scale is 40 μm.

[0039] A comprehensive assessment of retinal integration and immune response was performed using 3D confocal imaging of the fully implanted retina. 3D reconstruction revealed dense INL (DAPI) within most of the observed cellular cavities. A side view of a single cellular row showed complete migration down to the base of the array.

[0040] Within the degenerated RCS-controlled retina, the extended microglial process within the IPL indicates a quiescent state of microglia, while the microglial process beneath the INL extends its process through the degenerated outer plexus layer (OPL). Microglial processes in the IPL using subretinal implants appear similar to those in control retinas, with the extended process indicating a quiescent state of microglia. With planar implants, microglial processes are located close to the device surface. The presence of cortical responses in these active implants indicates that microglial processes on subretinal prostheses do not block electrical stimulation. With cellular implants, the microglial process extends primarily along the top of the wall, with minimal extension into the well.

[0041] The degree of retinal integration was assessed by analyzing the variation in cell density with height relative to the basal layer in cavities of each size. Average 50%, 45%, and 54% of INL cells were found in cavities with intercellular spacings of 40, 30, and 20 μm, respectively. This was because the electric field could extend above the cell walls (…). Figure 6A Therefore, it can stimulate more cells.

[0042] B2b) Modeling in vivo retinal responses

[0043] Figure 6A The figure shows the effect at 0.5 A / cm. 2 The active electrode current density is calculated for the potential of the planar (top) and cellular (bottom) arrays relative to the IPL center (z = 57 μm). Figure 6B Experimental (data points) and calculated thresholds for current density on the active electrode are shown for planar (dashed) and cellular (dotted) devices. Planar models (both binary and linear models, described below) reproduce the trends observed in experimental measurements. Cellular arrays significantly reduce the stimulation threshold (dotted), enabling safe operation of devices with pixels smaller than 40 μm. The maximum charge injection of SIROF (3 mC / cm² for a 10 ms pulse) is shown. 2 (Used as a dashed line.) Here, SIROF is an abbreviation for Sputtered Iridium Oxide Film. Any other biocompatible high-capacitance material can also be used for high-capacitance electrodes, such as IrOx, porous Pt, PEDOT, carbon nanotubes, etc., deposited through various methods (electroplation, chemical deposition, etc.).

[0044] To evaluate the benefits of the cellular array, we used a network-mediated retinal stimulation model. To validate the model, we first compared the modeling results with in vivo stimulation thresholds measured in rats with planar subretinal photovoltaic implants of various pixel sizes, and then calculated the stimulation thresholds for cellular arrays of various sizes.

[0045] A complete modeling of this system requires simulating the electric field generated by the device. Figure 6A The process of converting retinal responses into responses of internal retinal neurons, applying subsequent network-mediated processing to RGC activity, and finally converting retinal outputs into cortical visual evoked potentials (VEPs) is an extremely complex modeling task with multiple unknowns. This task can be simplified by the following assumptions: (1) network-mediated stimulation elicits RGC activity following an sigmoid curve, and (2) cortical responses are driven by the sum of retinal signals and also follow an sigmoid curve. We approximate the sigmoid dependence of network-mediated retinal responses on electric fields from previous experiments through two extremes: (1) simulating a step function of a binary transition across the stimulation threshold and (2) simulating a linear function of the response to a gradually increasing neural output with increasing stimulation. Thus, the total retinal response is calculated by (1) integrating the cellular responses over the INL volume with binary coefficients, i.e., only calculating the portion of the INL volume above the stimulation threshold, or (2) integrating the cellular responses over the INL volume with cellular responses proportional to their polarization. Thus, we refer to these two models as “binary” and “linear”, respectively.

[0046] Assuming a steady-state current, the electric field in the retina was calculated using a finite element model of the entire array in COMSOL Multiphysics 5.0 by solving Maxwell's equations for the potential using the electrostatics module. The calculated field was then converted to a retinal response using both a binary and a linear model. Each model had only one fitting parameter used to correlate its retinal output with the magnitude of the cortical response. For the binary model, a subset of INL cells should be activated to elicit a cortical response, while for the linear model, the slope of the linear fit should be activated. Irradiance was converted to current density based on the pixel geometry and photocurrent conversion efficiency of our 2-diode pixels. To fit the binary model, we therefore calculated the electric field versus irradiance and the percentage of INL cells above the stimulation threshold for all pixel sizes (4.8 mV for a 10 ms anodic pulse, see Methods). Using our previously recorded experimental thresholds for 140, 70, 55, and 40 μm pixels, we observed that 8.27 ± 1.42% of the INL volume should be above the stimulation threshold to elicit a VEP response. Both the binary and linear models produced very similar ratios of stimulus threshold to pixel size, indicating that the crucial factor is the shape of the electric field, rather than the details of the S-shaped response curve.

[0047] B2c) Stimulation threshold of cellular array

[0048] Using binary model parameters validated by comparison with in vivo stimulation thresholds obtained using planar implants, we calculated the stimulation thresholds for the cellular array (see Methods). These thresholds are significantly lower than those for planar pixels of the same size in terms of current density on the active electrodes for a 10 ms pulse, and do not increase much with decreasing pixel size. Figure 6B Due to (a) the radial expansion of the electric field from the active electrode and (b) the coplanar return electrode, the planar array with circumferential return suffers a rapid decrease in potential along the vertical axis. Figure 3C Therefore, for a 10ms pulse (>30mA / cm) 2 Or charge density >3mc / cm 2 For planar pixels smaller than 40 μm, a current density greater than the SIROF safe charge injection limit is required. Placing the return electrode on top of the insulating honeycomb wall surrounding the pixel forces the current to flow primarily upwards from the active electrode. Figure 3D This significantly increases the depth to which the potential exceeds the stimulation threshold. Furthermore, using a cellular array, the penetration depth of the electric field into the tissue is set by the height of the wall, thus separating it from the pixel width. Therefore, the stimulation threshold in terms of current density is independent of the pixel width. Figure 6B This allows pixels to be reduced to a size limited only by retinal migration (i.e., by cell size). Table 1 summarizes the improvements in stimulation threshold.

[0049] Table 1. Calculated stimulus threshold current densities for different pixel sizes. An asterisk indicates that the current density of a 10ms pulse exceeds the SIROF charge injection limit.

[0050]

[0051] In addition to the stimulation threshold, the cellular electrodes significantly improve the spatial selectivity of electrical stimulation (i.e., the contrast between adjacent pixels), which is crucial for high visual acuity. To replicate the grating pattern used for in vivo visual acuity assessment, we simulated the electric field distribution from our array, which has alternating rows of on and off pixels. For both electrode configurations, increasing the current density resulted in an increase in the positive potential above the on pixels and an increase in the negative potential above the off pixels. The insulating walls of the cellular structure prevent lateral diffusion of the electric field, thereby widening the dynamic range of selective activation of the on pixels. Furthermore, the electric field extends above the walls of the cellular structure to stimulate cells up to 40 μm away from the cavity substrate within safe charge injection limits, allowing up to 99% activation of the total inner retina without crosstalk.

[0052] B3) Discussion

[0053] Our study provides a pathway to significantly improve the spatial resolution of retinal prostheses compared to the limitations of current systems. Until recently, the best prosthetic visual acuity achieved in clinical trials was 20 / 546, observed in two patients using Alpha IMS / AMS subretinal implants [2,20]. With 70 μm pixels, this prosthesis performed twice as poorly relative to its sampling limit, leading some to conclude that continued reduction in pixel size does not improve visual acuity. However, the bottleneck for these devices is likely electrical rather than biological: the unipolar configuration of such implants, where active electrodes in each pixel share a common remote return electrode, results in strong crosstalk between adjacent electrodes, significantly reducing spatial contrast. Alternative designs that provide both active and return electrodes in each pixel improve the localization of the electric field and the associated spatial resolution. In fact, a recent clinical trial of a photovoltaic subretinal implant with 100 μm pixels utilizing local return electrodes showed visual acuity as high as 20 / 460 (only 15% lower than the sampling limit for that pixel size (20 / 400)). Furthermore, in rats, similar implants with 70 μm and 55 μm pixels also provided grating sharpness that matched the sampling limit. This suggests that visual sensitivity can reach the sampling density limit of the stimulus array if the electric field in each pixel is appropriately limited. However, further reduction in pixel size is limited by a rapid increase in the stimulation threshold, which rises to a safe limit close to that of a flat pixel with a width of 40 μm. Figure 6B ).

[0054] The fundamentally different geometry of the stimulation array described in this work—a honeycomb structure—solves this basic limitation. Due to the confinement of the pixels by the insulating walls, current flows upward from the active electrodes at the bottom of the cavity. Figure 3B and Figure 3D This change in the shape of the electric field significantly reduces the stimulation threshold. Furthermore, the unidirectional flow of current fundamentally alters the scaling of the stimulation threshold relative to pixel size. In spherical geometry, an electrode smaller than the distance to the target cell can be approximated as a point source. Since the required current does not change with electrode size in this case, the current density increases inversely with the electrode area, i.e., quadratically with the electrode radius. If the ratio of electrode size to pixel width is maintained, smaller pixels will require higher current densities, limited by material properties. However, with one-dimensional current, the situation is different: the current density required for a given potential drop does not change with pixel width, and therefore pixels can be much smaller while maintaining the same current density on the same electrode. For photovoltaic pixels, this means that as long as the relative dimensions of the electrode and pixel width remain constant, the threshold irradiance should remain nearly identical for all pixel sizes.

[0055] Because intraretinal neurons can migrate into spaces within the subretinal space, the cellular design is particularly well-suited for subretinal placement. Our study demonstrates that intraretinal neurons readily migrate into wells as small as 18 μm wide (20 μm pixel pitch). Tissue survival after 6 weeks indicates that diffusion of oxygen and nutrients from retinal vessels located above the implant is sufficient to sustain cells within a 25 μm high wall. Since no lower bound on the integral was observed in our study, the pixel width may continue to decrease, but certainly not below a cell size of approximately 10 μm. Determining the precise minimum within this range will require further experimentation. Even without any further reduction in pixel size, an array with 20 μm pixels should achieve spatial resolution matching the natural visual acuity of rats and better than 20 / 100 in humans.

[0056] Most of the INL migration into the cavity did not appear to affect the rest of the retinal structures, where the INL, IPL, and GCL are clearly depicted. Since all connections between INL cells and RGCs are located above the cell walls in the IPL, we expected that retinal signal processing would not be affected by this migration. Lateral connections between bipolar cells via horizontal cells connecting to photoreceptor ends in the outer plexiform layer (OPL) are likely absent in degenerated retinas due to the lack of photoreceptors. The amount of microglia found at the cell walls was similar to that of the planar device, and since the latter elicits VEP throughout the animal's lifespan, the immune response to both implants appears acceptable. Interestingly, for the cell, the microglia are located above the walls, allowing neurons to migrate closer to the stimulating electrodes.

[0057] Existing subretinal prostheses with long-range return electrodes can also incorporate vertical walls around pixels. In such a unipolar configuration, the polarization of bipolar cells migrating into the well should become more efficient due to the vertical alignment of the electric fields, thus reducing the stimulation threshold and crosstalk between adjacent pixels, as... Figure 2B As shown. However, without local return, the potential inside a dark pixel is still affected by its illuminated neighboring pixels, therefore the dynamic range of the pattern presented on such an array will be lower than that of a bipolar electrode configuration, such as... Figure 2A As shown.

[0058] In a steady state, the current density on an electrode surface is proportional to its capacitance per unit area. Therefore, if the electrode is composed of two materials with very different capacitances, the current will primarily flow to the material with the higher capacitance. This phenomenon allows for the fabrication of vertical walls for 3D electrodes using conductive materials, provided their capacitance is much lower than the capacitance of the material deposited on top of the wall for return electrodes. Although the walls are made of conductive materials, because of their very low capacitance per unit area, they absorb very little current in the electrolyte, thus behaving as if they were made of insulators in the liquid. For example, the walls could be electroplated with gold, with a capacitance in salt water approximately 0.01 mF / cm². 2 The return electrode at the top of the wall can be made of iridium oxide, with a capacitance of 1-10 mF / cm. 2 Due to the conductive properties of walls made of metal, IrOx can be electroplated onto the top of such walls. Alternatively, IrOx can be deposited by sputtering, but the conductive wall will connect this coating to the circuitry on the device surface. The walls can be made of other metals, including platinum, aluminum, molybdenum, etc.

[0059] The sidewalls can also be coated with a non-conductive material to prevent any current from flowing between the electrolyte and the sidewalls. For example, oxidation of aluminum or molybdenum makes its surface non-conductive. The walls can also be insulated by additional processes, including but not limited to atomic layer deposition or photolithography of non-conductive materials. Alternatively, they can be made of an insulator. In this case, the return electrode deposited on top of such a wall should be connected to the circuit via a conductive trace deposited on top of the wall for this purpose.

[0060] The pixels in the implant can be photovoltaic, meaning that light falling on the pixel is converted into an electric current using a photodiode connected between the active and return electrodes. Alternatively, the current can be transmitted to the electrodes via a wired connection.

[0061] For cells to migrate into the honeycomb cavities, the width of the cavities should exceed the cell size, i.e., be greater than 5 micrometers. When the width of the cavity significantly exceeds its depth, the benefits of the cavity become negligible. Figure 6B As shown, retinal stimulation becomes negligible for pixels larger than 100 micrometers. For subretinal implantation, i.e., to facilitate the tiling of a rigid array to follow the curvature of the human eye, the array size should be in the range of 0.5 to 5 mm, more typically 1–3 mm. At these sizes, the array does not require the use of the flexible materials described earlier; however, larger arrays can be combined with a flexible substrate to conform to the eye. The cavity depth should allow cells to migrate from the nuclear layer, approximately the thickness of the nuclear layer plus the subretinal debris layer, i.e., in the range of 20–70 micrometers. Tissue viability 6 weeks after implantation indicates that additional nutrient flow and tissue survival do not require perforation at the bottom of the well.

[0062] The sidewalls can be designed to allow for implant removal while improving mechanical stability within the subretinal space. As this study demonstrates, a completely smooth, vertical wall should not exert excessive mechanical force on the tissue during device removal. However, tissue migrating into the cavity provides a method for laterally anchoring the device relative to the retina. Additional overhangs at the top of the wall can be introduced by electroplating over a groove that guides the electroplating process. In this case, the cavity opening d o <d c By making it smaller than its width, implantation stability along the z-axis can be further improved. However, this may make device removal more invasive, so the exact configuration can be determined based on the patient's age and the likelihood of removal.

[0063] As shown in Table 1, the stimulation threshold is between 0.01 and 1 A / cm. 2 Within this range, the pulse duration is 10 ms. Therefore, electrode materials should be selected that allow for the injection of 0.1-10 mC / cm. 2 The charge density within a certain range.

[0064] If the implant is tilted relative to the incident light, the reflective sidewalls of the honeycomb can help direct radiation to the photosensitive area of ​​the implant at the bottom of the well. For this purpose, it is advantageous to use electroplated metal walls due to the high reflectivity of metals to visible and infrared light. Other materials and coatings with high reflectivity can also be used for this purpose, especially for light incident on the walls at near-grazing angles.

[0065] B4) Method

[0066] B4a) Passive cell implants

[0067] Passive cellular implants are fabricated from crystalline silicon wafers by using two mask layers to create patterns for deep silicon etching. A 2μm negative photoresist (AZ5214-IR) is spin-coated onto the wafer, primarily hexamethyldisilazane (HMDS), and treated to define the cellular walls. This photoresist is further treated with UV light for 15 minutes to improve selectivity in subsequent etching processes. A 25μm deep cavity is formed in the exposed silicon area using a Bosch etching process. After removing the cellular defining photoresist, a photoresist (7.5% SPR220-7, 68% MEK, and 24.5% PGMEA) is sprayed onto the wafer to a thickness of 14μm and treated to define release trenches around a 1mm wide array. A second Bosch process is applied to create these release trenches, and then the photoresist is removed. The wafer was coated with a protective 60 μm thick photoresist, and then underwent back-side grinding at a thickness ranging from 50 μm to 50 μm from the cellular substrate (San Jose Polishing & Cutting Services, California, USA). Subsequent etching of the remaining excess silicon in XeF2 gas completed the implant release. The resulting structure was... Figure 4AAs shown in section -B, the cavities are arranged in hexagonal honeycomb patterns with wall thicknesses of 4, 3, and 2 μm, respectively, on a 10 μm thick substrate, with a pitch of 40, 30, and 20 μm. The fourth quadrant was designed for 10 μm pitch honeycombs, but these exceeded the process limits of our lithography system and were not developed. We refer to this region as the "flat" quadrant in this text. A 50 nm thick oxide layer is grown on the surface of the silicon implant to prevent its in-bulk dissolution.

[0068] B4b) Animals and Surgery

[0069] All experimental procedures were performed in accordance with institutional guidelines and the ARVO statement regarding the use of ophthalmic and vision research animals. Animal care and subsequent implantation were performed using rats with retinal degeneration from the Royal College of Surgeons (RCS) habitat maintained by the Stanford Animal Facility. N=6 animals were implanted with a cellular array, with implantation occurring between P180 and P300 to ensure complete photoreceptor degeneration. Animals were anesthetized with a mixture of ketamine (75 mg / kg) and toluidine (5 mg / kg) administered intramuscularly. A 1.5 mm incision was made in the sclera and choroid 1.5 mm posterior to the limbus. The retina was lifted by injecting saline, and the implant was inserted into the subretinal space. The conjunctiva was sutured with nylon 10-0, and postoperative topical antibiotics (bacitracin / polymyxin B) were administered to the eye. Successful surgery and retinal reattachment were verified using optical coherence tomography (OCT) (Heidelberg Engineering HRA2-Spectralis, Heidelberg, Germany). Animals were euthanized 6 weeks post-implantation. Other animals in the control group had flat, active implants and were euthanized after an in vivo study lasting approximately 6 months.

[0070] B4c) Integrated retinal imaging

[0071] Animals were euthanized by intracardiac injection of Beuthanesia, and the eyes were enucleated and rinsed with phosphate-buffered saline (Thermo Fisher Scientific, Sunnyvale, CA; PBS, Gibco). The anterior segment and lens were removed, and after positioning the implant under a stereomicroscope, the eyecup was cut into a 3mm x 3mm square centered on the implant and fixed in 4% paraformaldehyde (PFA, USA; EMS, PA) at 4°C for 12 hours. The implant was left in situ to prevent tissue damage or remodeling due to its removal. The sample was permeated for 3 hours at room temperature with 1% Triton X-100 (Sigma-Aldrich, CA, USA) PBS solution. Samples were placed in 10% bovine serum albumin (BSA) blocking buffer at room temperature for 1 hour, and then incubated with two primary antibodies at room temperature for 12 hours. Rabbit anti-IBA1 (1:200; Wako Chemicals, Virginia, USA) and mouse anti-glutamine synthetase (GS, 1:100; Novus Biologicals, Colorado, USA) were added to 0.5% Triton X-100, 5% BSA in PBS. Samples were washed in 0.1% Triton X-100 PBS (PBS-T) at room temperature for 6 hours and incubated with two secondary antibodies at room temperature for 12 hours: donkey anti-rabbit Alexa Fluor 488 (1:200; Thermo Fisher Scientific, Sunnyvale, California, USA) and donkey anti-mouse CY3 (1:200; Jackson Immuno Research Inc., Pennsylvania, USA). Counterstaining was performed in PBS with 4',6-diamidinyl-2-phenylindole (DAPI). After washing in PBST for 6 hours, the samples were sealed with Vectashield (H-1000; Vector Laboratories, Burlingame, California, USA).

[0072] 3D imaging was performed using a Zeiss LSM 880 confocal inverted microscope and Zeiss ZEN Black software. Z-stacking was used to acquire the image plane through the total thickness of the retina, with upper and lower limits defined 10 μm below the internal limiting membrane (ILM) and the cellular cavity substrate, respectively. Stacks were acquired at the center of each cellular quadrant using 40x oil immersion objectives, with an acquisition area >225 μm x 225 μm, a 360 nm z-step, and 0.55 μm pinholes.

[0073] B4d) Image Analysis

[0074] The confocal dataset was analyzed using the Fiji distribution in ImageJ. To analyze cell density within the pores and above the implant, we first maximized the contrast of the individual XY planes to ensure 0.3% channel saturation to correct for brightness variations at different Z positions within the stack. Noise was then removed from the XY planes, and the background was subtracted. A Gaussian blur filter (σ = 3 pixels, 0.42 μm) was applied to smooth brightness variations within individual cells. The XY planes were then passed through an edge detection filter, and the final image was composed of an OR combination of the processed and edge-detected images minus the background. For cell density analysis, the channel thresholds (default method) were adjusted to provide a binary representation of the cells. Cell density in the XY planes was then calculated as a percentage of the area occupied by cells, taking into account the area occupied by the cell walls. To account for local variations in retinal histology, each cell was analyzed independently, where cell density was normalized to the maximum value within the cell stack. The percentage of INL contained within the cavities was calculated as follows:

[0075]

[0076] Where D(z) is the relative density of each cell in the XY plane, which is a function of the height (z) at which it is separated from the base (z=0), z' is the wall height (25μm), and z” is the end of the INL, defined as the point where D(z)<0.02.

[0077] Six implanted devices were imaged and analyzed, each containing cellular cells of 40, 30, and 20 μm in size. In three devices, the 20 μm cellular cells were damaged, and these devices were therefore excluded from analysis.

[0078] B4e) Histological preparations

[0079] Following confocal imaging, samples were rinsed in buffer and fixed in 1.25% glutaraldehyde solution for 24 hours at room temperature. They were then post-fixed in osmium tetroxide for 2 hours at room temperature and dehydrated in fractionated ethanol and propylene oxide. After overnight infiltration in epoxy resin (DMP-30-free) at room temperature (Electron Microscopy Science - Araldite - EMbed, RT13940, Mollenhauer kit), samples were placed in an oven at 70°C for 36 hours. The epoxy blocks were then trimmed until the silicon implants were exposed. To prevent damage to the slicing blade and the formation of silicon debris from the honeycomb structure, the silicon implants were removed using XeF2 etching (Xactixe-1, 23°C, 3 Torr). The blocks were then refilled with epoxy resin and placed in a vacuum desiccator for two hours, followed by overnight baking at 70°C. This refilling of the voids left after etching the implants provided structural support during sectioning. 700 nm thick sections (cut by Reichart Ultracut E) were stained with toluidine blue for optical microscopy examination.

[0080] B4f) Modeling of electric fields and retinal stimulation

[0081] Assuming a steady-state current, the electric field in the retina is calculated using a fully arrayed 3D finite element model in COMSOL Multiphysics 5.0 by solving Maxwell's equations for the potential using the electrostatics module. The modeled array has a diameter of 1 mm and a thickness of 30 μm, and consists of hexagonal pixels of various sizes, as listed in Table 2. Return electrodes are connected to a single mesh.

[0082] Pixel size [μm] Number of pixels <![CDATA[Active / Pixel [μm 2 > <![CDATA[Return / Pixel [μm 2 > 140 37 1018 3823 70 157 254 1173 55 250 154 853 40 502 79 407 30 930 44 255 20 2172 20 112

[0083] Table 2. Number of pixels and their geometry in the modeled photovoltaic array

[0084] The electric field is calculated within a volume (cube, side = 10 mm), where ground (0 potential) is defined at the edges. A modeled spur is used as a closed system, where all injected current from the active electrode is collected on the return electrode. Boundary conditions on the electrode surfaces are defined with a uniform current density, corresponding to steady state.

[0085] Under steady-state conditions, the current density at the electrode-electrolyte interface is proportional to the electrode capacitance per unit area. Therefore, even without sidewall insulation, the current primarily flows through the top coated with SIROF. For example, if a SIROF coating is used as the return electrode material, its capacitance is approximately 1000 times higher than that of gold (10 mF / cm²). 2 For 0.01mF / cm 2 The ratio of the area of ​​the return electrode placed on top of the wall to the area of ​​the sidewall of each pixel is equal to the ratio of half the wall's width to its height. For a wall with an aspect ratio of 5:1, the ratio of the area of ​​the top of the wall to its sidewall is approximately 10:1. Therefore, the capacitance of each pixel's sidewall will be 100 times smaller than the capacitance of the SIROF return electrode deposited on top of the cell. Thus, even if the walls are made of metal, they will accept very low current in the electrolyte, and therefore they function almost as if they had insulated sidewalls. This greatly simplifies the fabrication of the cell by allowing for simple electroplating without the need for insulation of the sidewalls.

[0086] Two electrode configurations were studied: 1) a plane with local return ( Figure 3A ) and 2) have local return cellular ( Figure 3BIn both configurations, a common return electrode collects the current generated by all active pixels, such that the total collected current is the sum of the injected currents on each individual active electrode. The sidewalls of the cell are non-conductive. In our previous work using active devices, we evaluated spatial resolution in volume by projecting grating patterns of various spatial frequencies at 100% contrast. The maximum resolution corresponds to alternating rows (on or off rows) of active pixels. To replicate this configuration in simulations, we calculated the electric field distribution using the same activation scheme and analyzed the field at the center of the array, where crosstalk between adjacent pixels is highest. To correlate the electrical simulations with the light intensity of our photovoltaic spur, the total current per pixel was calculated based on the diode area of ​​the 2-diode configuration and the measured photocurrent conversion efficiency: 0.40, 0.31, 0.26, and 0.24 A / W for 140, 70, 55, and 40 μm pixels, respectively.

[0087] The retinal stimulation threshold was assessed using a network-mediated activation model. In this approach, we assume that the network-mediated stimulation threshold is defined by the voltage drop across the bipolar cell. In an external electric field, the intracellular mediator becomes isoelectric within one microsecond, resulting in hyperpolarization and depolarization of the cell membranes near and far from the anode, respectively. Using the retinal network-mediated stimulation threshold current density of a large electrode with a 10 ms pulse from the literature, the average resistance of the retina (1000 Ωcm), and the average length of the bipolar cell (estimated from the middle of the INL to the middle of the IPL (37 μm)), we calculated a potential difference threshold of 4.8 mV from the somatic cell to the axonal terminal for anodic stimulation. The -21 mV cathodic threshold was calculated based on network-mediated activation curves measured in the rat retinas with both anodic and cathodic stimulation, which were scaled to match the calculated anodic threshold.

[0088] It is worth noting that our model is based on stimulation current densities from the literature, while the calculated transcellular voltage is linearly proportional to retinal resistivity, a point on which there is no consensus in the literature. We use transcellular voltage only as a means of assessing the boundaries of activated areas in tissue relative to a stimulation threshold. We assume that any change in retinal resistivity will linearly affect the potential of the same current, including the threshold potential. Therefore, these changes will not affect the boundaries of activated areas, as they are calculated relative to a stimulation threshold.

[0089] We approximate the sigmoid dependence of network-mediated retinal responses to increasing electric fields from previous experiments using two extremes: (1) a step function simulating the binary transition across the stimulus threshold and (2) a linear function simulating the gradually increasing neural output in response to increasing stimulus. Therefore, the total retinal response is calculated by (1) integrating the cellular response over the INL volume with binary coefficients, i.e., only calculating the portion of the INL volume above the stimulus threshold, or (2) integrating the cellular response over the INL volume with the cellular response proportional to its polarization.

Claims

1. An apparatus for electrical stimulation of neural cells, the apparatus comprising: an array of cavities configured to allow neural cells to migrate within the cavities; wherein each cavity has a floor and a conductive wall; wherein each cavity has a first electrode disposed on its floor; wherein each cavity has a second electrode disposed on the top of its wall and vertically separated from its corresponding first electrode; wherein during operation of the apparatus, ionic current flows through the contents of the cavities; wherein in each cavity, the second electrode has a greater capacitance than the conductive wall, whereby the second electrode preferentially collects the ionic current relative to the sides of the conductive wall.

2. The apparatus of claim 1, wherein the cavities have a depth between 10 μm and 100 μm.

3. The apparatus of claim 1, wherein the cavities have a width between 5 μm and 100 μm.

4. The apparatus of claim 1, wherein the cavities have a depth greater than their width.

5. The apparatus of claim 1, wherein the array of cavities is periodic.

6. The apparatus of claim 5, wherein the cavities are hexagonal.

7. The apparatus of claim 1, wherein the neural cells are retinal cells.

8. The device of claim 1, wherein in operation, the injected charge density at the first electrode of the cavity is between 0.1 mC / cm 2 and 10 mC / cm 2 .

9. The apparatus of claim 1, wherein the second electrode has a capacitance per unit area at least 100 times greater than the conductive wall.

10. The apparatus of claim 1, wherein the apparatus is configured to convert light falling on the apparatus into current using a photodiode connected between the first electrode and the second electrode.

11. An apparatus for electrical stimulation of neural cells, the apparatus comprising: an array of cavities configured to allow neural cells to migrate within the cavities; wherein each cavity has a floor and an electrically insulating vertical wall; wherein each cavity has a first electrode disposed on its floor; wherein the apparatus has a common return electrode disposed outside the array of cavities; wherein during operation of the apparatus, ionic current flows through the contents of the cavities; whereby the electrically insulating vertical wall improves stimulation efficiency and reduces cross-talk between adjacent cavities of the apparatus by forcing the ionic current to travel vertically within the cavities.

12. The apparatus of claim 11, wherein the cavities have a depth between 10 μm and 100 μm.

13. The apparatus of claim 11, wherein the cavities have a width between 5 μm and 100 μm.

14. The apparatus of claim 11, wherein the cavities have a depth greater than their width.

15. The apparatus of claim 11, wherein the array of cavities is periodic.

16. The apparatus of claim 15, wherein the cavities are hexagonal.

17. The apparatus of claim 11, wherein the neural cells are retinal cells.

18. The device of claim 11, wherein in operation, the injected charge density at the first electrode of the cavity is between 0.1 mC / cm 2 and 10 mC / cm 2 .

19. The apparatus of claim 11, wherein the apparatus is configured to convert light falling on the apparatus into current using a photodiode connected between the first electrode and the common return electrode.

Images