Shifting gazes with visual prostheses: Long-term hand-camera coordination

Embargo until
Date
2018-03-26
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
Purpose: Prosthetic vision is young, and many aspects of its use remain unexplored. Hand-camera coordination, the prosthetic correlate of hand-eye coordination, relies heavily on how the camera is aligned with the eye. It is unknown whether users of prostheses can adapt to using misaligned cameras, or whether requirements for proper alignment remain constant over time. Methods: Four blind subjects implanted with Argus II retinal prostheses participated in this study. Each subject attempted to touch a single 4°–7° white target that was randomly located on an otherwise black touchscreen in a target localization task. Touch response accuracy was used to determine the necessary adjustment to eye-camera alignment, the optimal camera alignment position (OCAP). Subjects attended over 100 sessions across up to 5.3 years. S1–S3 were given misaligned cameras for over 1 year. Adaptation was measured through changes in localization errors. Outside that period of intentional misalignment, cameras were aligned to maximize localization accuracy. During the final year, localization tasks were performed in alternation with eye tracking. S2–S4 also participated in 1-day experiments with simultaneous eye tracking and target localization. Results: Subjects were not able to significantly reduce localization error when cameras were misaligned. When trying to maximize localization accuracy, necessary OCAPs changed significantly over time. OCAP trend directions within days and trial runs matched changes between the beginnings of days and runs. Changes between the end of a day or run and the beginning of the next tended to point in the opposite direction of the previous trend, indicating a reset of OCAP changes. Changes in eye orientations correlated significantly with changes in OCAPs. Eye-orientation trends displayed the same reset behavior between days and runs as OCAPs. Simultaneous eye tracking and localization showed agreement between eye-orientation and localization-error trend directions. Adjusting camera alignment with eye-tracking data slowed changes in localization errors. Conclusions: Users of current visual prostheses cannot passively adapt to camera misalignments. OCAPs are not constant with time. Prosthesis users who desire maximum pointing accuracy will require regular camera realignments. Camera alignments based on eye tracking can reduce both transient and long-term changes in localization that are related to eye movements.
Description
Keywords
neuroprosthesis, neuroprostheses, visual prosthesis, visual prostheses, retinal prosthesis, retinal prostheses, retinal implant, prosthetic vision, bionic eye, localization, coordination, eye tracking, pointing, reaching, blindness, low vision, adaptation, stability, alignment, camera alignment, retinitis pigmentosa, retinal degeneration, photoreceptor degeneration, drift, phosphene, Argus, Second Sight, Dagnelie, CAP, OCAP
Citation