Evaluating the influence of stimulation parameters on phosphene detectability for epiretinal prostheses

Abstract

Purpose : Traditional, non-adaptive methods for estimating detection threshold of electrically-evoked phosphenes have not permitted a systematic characterization of stimulation parameters. Instead, the Bayesian adaptive method can accurately estimate thresholds with fewer trials, making it ideal for studies with human subjects. In this study, we implemented the Bayesian method to gain a better understanding of how stimulation parameters influence the thresholds of phosphenes evoked by epiretinal electrodes. Our results will be used to develop advanced stimulation paradigms that increase resolution and minimize perceptual adaptation of visual feedback provided by epiretinal prostheses.

Methods : Three people with Argus II devices participated in this study. 12 electrodes for subject S1 and 6 electrodes for S2 and S3 were selected. Each electrode's threshold was estimated using a Bayesian adaptive method in a yes/no experiment, in which subjects reported if they observed a phosphene after hearing an auditory cue. Using the collected responses, the next stimuli were presented at the amplitude level that maximized the entropy for the threshold and the slope of the Weibull psychometric function. The process continued until the thresholds of all selected electrodes stabilized (the differences between 3 consecutive estimations were smaller than 3 µA) or exceeded 30 trials. Thresholds were estimated for pulse train frequencies of 6, 20, and 60 Hz, each tested with 2 and 4 pulses within a train. We also estimated the thresholds of pulse trains with a fixed train duration of 250 ms at 6, 20, and 60 Hz.

Results : A paired Wilcoxon sum rank test for all pairs of estimated thresholds with different pulse parameters within-subjects showed no significant differences (p>0.05) for two out of three subjects. For S2, the estimated threshold at 60 Hz was unexpectedly greater than the threshold at 20 Hz with 250 ms pulse trains (p=0.02), with root mean square differences of 128 µA.

Conclusions : We investigated the effect of pulse train duration and frequency on estimated detection thresholds using Argus II electrodes. We interpret the results as showing effects of sensory adaptation at high frequencies, causing the thresholds to not decrease as expected. We, therefore, conclude that higher pulse frequencies do not necessarily enable a broader dynamic range of stimulation parameters for Argus II users.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.