Mimicking natural neural encoding through retinal electrostimulation

Retinal neuroprostheses or `bionic eyes', aim to restore patterned vision to those with vision loss by electrically stimulating the remaining neurons in the degenerate retina. Despite considerable progress over the last two decades, such devices generally stimulate indiscriminately both the `ON' and `OFF' visual pathways in the retina, conveying highly non-physiological signals to the brain. In this study, we used computational models to explore the ability of a new approach to retinal neurostimulation. To elicit neural responses more closely resembling those of natural vision, we demonstrate preferential excitation of the ON and OFF neurons by delivering spatiotemporally patterned stimuli with a multi-electrode array. In particular, the strategy relies on: (1) strategic placement of electrodes at key anatomical positions such as the RGC body and optic disc, and (2) using modulated stimulus waveforms, tailored to each stimulus electrode, to either initiate or suppress neural responses travelling to the cortex. By providing preferential activation of ON and OFF cells, and hence better device-to-biology integration, retinal implants would convey more physiologically-realistic spiking patterns to the visual processing centers in the brain, potentially improving the efficacy of next generation retinal implants.