Abstract
Purpose :
To simplify the imagery presented to Argus II retinal implant wearers by filtering out information outside the desired depth range using the disparity map from two head-mounted cameras. We compared the functionality of distance filtered (DF) imagery with unfiltered (NF) imagery in depth discrimination tasks.
Methods :
The data were acquired from Argus II users in two separate experiments (T1 and T2). In T1, a person with a white shirt stood in front of the subject randomly at 1, 3 or 6 m. The subject was asked to determine at which distance the person was standing. Responses ("1", "3" or "6") and timing of the response were recorded. In T2, two identical white cups were placed randomly at distances of 53±5cm and 78±5cm on the left and right side of a black table. The subject was asked to determine which cup was closer. The response ("right" or "left") and response time were recorded. The Wilcoxon signed-rank test was used for pairwise comparisons between subjects and between conditions.
Results :
Four Argus II users aged 62-87 performed T1 with 120 trials. Three of the subjects performed T2 with 80 trials. There were no significant differences between subjects in either test, so data were pooled across subjects for analysis between conditions. Results illustrated that subject performance was significantly better with DF compared to NF (in T1 – mean accuracy: 86% with DF and 35% with NF, p<0.005; in T2 – mean accuracy: 81% with DF and 65% with NF, p<0.005). Significantly longer time was spent to determine the response with DF compared to NF (mean ± standard deviation response times: in T1: 13.75±9.90 s and 11.78±7.64 s respectively, p=0.002; in T2: 78.98±67.95 s and 32.08±26.96 s respectively, p = 0.001).
Conclusions :
Subjects detected the distance of the target more accurately with DF than using NF. Even though the unfiltered camera image would have allowed subjects to make the discrimination based on size of the target, performance with this camera was close to chance, presumably due to more (non-target) phosphenes with the NF. The differences in response times may be attributed to unfamiliarity of the DF system and/or guessing with NF system. More data is required to study the learning effect.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.