A novel vision processing method for detection and avoidance of well and poorly color-contrasted obstacles for prosthetic vision: preliminary results for the 44 channel suprachoroidal retinal prosthesis

Abstract

Purpose : A novel vision processing method, Local Background Enclosure Vision Processing (LBE VP), was evaluated with three recipients unilaterally implanted with the second-generation suprachoroidal retinal prosthesis with end stage retinitis pigmentosa (RP), NCT03406416. LBE VP aims to highlight objects closer in proximity to the recipient, irrespective of color. We evaluated the effectiveness of LBE VP compared to the state-of-the-art Intensity-based vision processing (I VP) and System Off (SO) for an orientation and mobility task.

Methods : The three recipients (2 male, 1 female; aged 48 – 67 years) aimed to walk along an indoor corridor-like environment (white walls) and detect and avoid six common (pool of 10 obstacles, e.g., mannequin, chair, overhanging object; black or white) obstacles. The effectiveness of LBE VP was compared to the current state-of-the-art, I VP and SO in a RCT design.

Results : For the mean percentage of all obstacles detected, LBE (N = 39; M 59.83%±24.71) performed better than I (N = 39; M 45.73%±9.91) VP (P < .0001); however, both performed better than SO (N = 38; M .00%±.00; P < .0001). For poorly color-contrasted obstacles, LBE (M 58.97%±31.96) was better (P < .0001) than I VP (M 5.98%±12.96) and SO (M .00%±.00) for obstacle detection, while I VP performed no better than SO. Collision tallies with all obstacles were similar for LBE (M 2.92±2.08) and I VP (M 3.62±2.47), however both had fewer collisions than SO (M 4.95±2.57; P < .05). For poorly color-contrasted obstacles, LBE (M 1.51±1.39) had fewer collisions than SO (M 2.53±2.15; P < .05) and was trending (P = .069; Cohen’s Effect Size = .535 – moderate) with fewer collisions than I VP (M 2.46±2.09), while I VP performed no differently than SO.

Conclusions : LBE VP exhibited superior performance in detecting objects irrespective of color contrast in comparison to state-of-the-art, I VP and SO for these retinal prosthesis recipients. In turn, LBE VP showed promise of improving the performance of prosthetic vision in avoidance of poorly color-contrasted obstacles. Novel VP algorithms such as LBE may improve orientation and mobility functioning for environments with objects present irrespective of their color contrast.

This is a 2020 ARVO Annual Meeting abstract.