Shape Discrimination using Video to Stimulation Software for Intracortical Visual Prosthesis (ICVP)

Abstract

Purpose : The ICVP allows the creation of visual percepts for people with neural substrate damage preceding the visual cortex. We developed dedicated camera input to implant-stimulation software that allows mapping a video stream to visual percepts generated by the implant to an estimated spatial configuration. We tested the hypothesis that a subject implanted with the ICVP will be able to discriminate between similar shapes using the current low-resolution phosphene map.

Methods : The ICVP consists of transcranially controlled wireless floating microelectrode arrays (WFMAs) with 16 electrodes per WFMA. The ICVP camera program takes a video input of arbitrary resolution and allocates a set of pixels for each phosphene in the map. The camera sensor size and focal length are used to project each pixel into visual space for the correct reproduction of an image. The program operates on a binary (on/off) stimulation scheme across the average intensity of the pixels within a phosphene region, relative to a threshold. Stimulation was limited to 4 electrodes per WFMA on 6 WFMAs, with parameters/electrode: 200 Hz, 200 μs, and a scaled maximum cathodic current < 60μA. The inaugural ICVP implantee was tested on the ability to discriminate geometric shapes and between (similar) letters – e.g., ‘N’ versus ‘Z’ – subtending ~ 25° on a 4’ x 8’ screen at 6’ distance from the head-worn camera.

Results : During exploratory testing, the subject could identify a horizontal, a vertical, and both diagonal bars; the subject also correctly identified an unknown shape as a circle. Given a randomized forced choice between Z and N, the subject correctly identified the letter 5/5 times (p = 2^-5 < 0.05, Binomial Test). The subject could locate contrasting objects, including a person standing at a 6’ distance, but could not identify object shapes. Due to the small number of phosphenes (6) included in the map, shape discrimination required the subject to scan the outline of each shape. The subject could not identify the images presented without scanning.

Conclusions : The subject demonstrated some ability to differentiate between structurally similar shapes using a low-resolution map. We anticipate that shape discrimination may improve following ongoing tests to create a more detailed map, and reduce the need for visual scanning. This could increase the probability of functional spatial vision with up to several hundred electrodes.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.