High-fidelity reproduction of visual signals by electrical stimulation in the central primate retina

Abstract

Electrical stimulation of retinal ganglion cells (RGCs) with electronic implants provides rudimentary artificial vision to people blinded by retinal degeneration, but cannot reproduce the neural code of the retina. Recent work has demonstrated more precise activation of RGCs using focal electrical stimulation in the peripheral macaque retina, but high-resolution vision requires the central retina. This work probes the effectiveness of focal epiretinal stimulation in the central macaque retina, using large-scale recording and stimulation ex vivo. The functional organization of the major RGC types was similar between the peripheral and central retina, and these cell types could be distinguished by their electrical properties. Electrical stimulation revealed similar RGC activation thresholds and reduced axon bundle activation in the central retina, but diminished stimulation selectivity. However, a novel stimulation algorithm revealed higher expected acuity in the central retina. These results support the possibility of reproducing high-acuity vision in the central retina with an epiretinal implant.

Focal electrical stimulation of human retinal ganglion cells for vision restoration

“…Another approach might be to implant the device at other retinal locations such as the raphe region, which has a relatively low density of axon bundles-results in macaque indicate that a larger fraction of RGCs in the raphe can be activated without axon bundles (Grosberg et al 2017). Finally, as ongoing work suggests, increasing the electrode density of the multielectrode array (MEA) increases the number of activation sites for each cell, significantly expanding the opportunities for selective activation (Gogliettino et al 2022).…”

Precise control of neural activity using dynamically optimized electrical stimulation

“…A single-electrode stimulation scan (42 amplitudes at each of the 512 electrodes individually, 0.1-4 μ A, linearly spaced, 15 times each) was then used to compute an electrical stimulus dictionary. Response probabilities for each electrical stimulus were identified using a custom template matching approach (Gogliettino et al 2023) and axon bundle activation thresholds were determined by automated methods as described above (Tandon et al 2021). Briefly, the custom template matching approach performs unsupervised clustering on the voltage traces for each stimulation pattern, then iteratively compares the difference signals between the clusters to cell waveforms estimated from visual stimulus recordings.…”

Modeling responses of macaque and human retinal ganglion cells to natural images using a convolutional neural network

“…The time course of the STA was calculated by summing the three primaries of the pixels within the RF. Clusters in a 2D space formed by RF diameter and the first principal component of the time course were used to identify distinct cell types [14][15][16]. Cells with measurable light responses were grouped into known functional types.…”

“…The spike-triggered average (STA) stimulus was computed for each cell to summarize its light response properties [1]. Clusters of light response properties (receptive field diameter and STA time course) were used to separate and identify specific cell types, as described previously [14][15][16]. In each of the four recordings examined, nearly complete populations of ON parasol, OFF parasol, ON midget and OFF midget cells were recorded.…”