Eye orientation trends match pointing errors in simultaneous eye tracking and target localization with retinal prostheses

Abstract

Purpose : Determine whether simultaneous eye tracking and target localization can support the link between hand-camera coordination shifts and eye orientations (EOs).

Methods : Three Argus II retinal prosthesis users touched single, randomly-located 7°-wide white targets on an otherwise black touchscreen. Argus II glasses were replaced with eye-tracking glasses. Any 18° × 11° area within the scene camera’s 73° × 55° field of view (FOV) could be used for retinal stimulation. This area was either fixed to one point in the camera’s FOV (head-only scanning, HOS), or adjusted based on EO (eye-head scanning, EHS). Each subject performed 5 or 6 runs of 20 trials, with at least 2 runs per condition. An inertial measurement unit monitored head motion during tests.

Linear models were calculated for localization errors and EOs with respect to time within each trial run. Differences in direction between vectors for errors and EOs were calculated for each run. Correlation significance and 95% confidence intervals (CIs) were determined using bootstrap resampling, and Watson’s test was used for circular uniformity.

Results : The time when the target was viewed was not recorded, so EO trends were based on readings only from the first 0.5s of each trial. The correlation between head angular velocity and EO velocity was 20% less negative within the first 0.5s than all other 0.5s periods (p < 0.0006), suggesting less influence of the vestibulo-ocular reflex (VOR). For HOS, error and EO trend directions were significantly correlated (p < 0.04) and direction differences were biased toward a mean of −10° (p < 0.05, CI: −49°–23°). Directions with EHS were not positively correlated and differences were not significantly nonuniform. When considering all EO readings, however, direction differences with EHS were biased toward a mean of 0.3° (p < 0.05, CI: −30°–64°). Errors with EHS scanning changed at 0.02°/s, while errors shifted at 0.04°/s with HOS (p < 5 × 10^-6).

Conclusions : Average EOs, not considering VOR, shifted in the same directions as localization errors within HOS trial runs, and EHS reduced rates of error change. EHS may therefore be useful for reducing long-term hand-camera coordination shifts. Overall EO changes still matched error changes in EHS, likely because pointing bias was still influenced by gazes outside the camera FOV and the side on which the 7° target was viewed.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.