A First Attempt to Translate Images in Optic Nerve Stimuli for a Visual Prosthesis

Abstract

Abstract: : Purpose: Electrical stimulation of the optic nerve has been shown to induce the perception of small sized phosphenes at positions in the visual field that are related to the stimulus parameters (Delbeke et al., ARVO 2002). This study represents a first attempt to use that information in order to translate camera images in a stimulation stream that would convey useful visual information in terms of a topologically organized collection of phosphenes. Methods: A model of the phosphene position according to the stimulus parameters has been fitted to previously collected experimental data. These equations are then exploited to derive a list of all the phosphene locations that can be accessed. Accordingly, a table is obtained that gives the stimulus parameters for each position of the visual field that can be activated. Images from a camera mounted on glasses worn by an implanted blind volunteer are reduced to 32 by 64 pixels with 1° resolution. After edge detection, processed image pixels in coincidence with one of the listed positions are randomly selected, excluding the last 10 occurrences if other choice among the coincidences is available. This process is sequentially repeated at a selected speed, up to 50 Hz. Results: No phosphene can be generated in the lateral neither in the upper visual fields. Initial trials have indicated that, in untrained conditions and with our volunteer, the preferred phosphene transmission rate was 12.5 per second. Despite these limitations, satisfactory results have been obtained in daily life-like tasks (Lambert et al., this meeting). Conclusions: The method described is still open to many improvements, including simultaneous and interleaved stimuli, adaptation to higher frame rates and further image pre-processing. Results obtained with the rather simple paradigm already implemented are much promising for the development of a full fledged visual prosthesis.

Keywords: electrophysiology: non-clinical • perceptual organization • computational modeling