Time-to-Collision Estimation With Age-Related Macular Degeneration Using Visual and Auditory Cues: Which Cues are Most Important?

Abstract

We measured time-to-collision (TTC) judgments from participants with age-related macular degeneration (AMD), and normal vision (NV) controls, with an audiovisual virtual reality system that simulated vehicles approaching in a 3D traffic environment. The vehicle was presented visually only, aurally only, or both simultaneously, allowing us to determine the relative importance of visual and auditory cues with psychophysical reverse correlation. Results indicated that TTC judgments were based on both auditory and visual cues in the AMD and NV groups; the AMD group relied, at least in part, on their residual vision. A multimodal advantage was not observed in either group. TTC estimation in the AMD group was surprisingly similar to that in the NV group. However, the AMD group showed a higher relative importance of “heuristic” cues compared to more reliably accurate cues favored by the NV group, suggesting that similar performance may be achieved through different cue-weighting strategies.

Worldwide, there are more than 250 million people with low vision or blindness (Flaxman et al., 2021). Yet, little is known about how people with different degrees of visual impairment rely on hearing instead of vision in judgments of collision, which is important for safe mobility such as crossing a street. One strategy to help individuals with vision loss avoid collisions is sensory substitution, in which sound is used instead of, or to supplement, visual information. However, sensory substitution studies commonly examined people with complete or nearly complete vision loss (e.g., Maidenbaum et al., 2014) although most people with visual impairment have some residual vision. Further, there are relatively few studies on judgments of collision by people with visual impairment.

We measured time-to-collision (TTC) judgments in 25 adults with age-related macular degeneration (AMD) in both eyes and 25 adults with normal vision (NV). An interactive virtual reality system depicted an approaching vehicle from the viewpoint of a pedestrian and provided relevant visual and auditory distance and motion cues (see Oberfeld et al., 2022). Participants underwent hearing exams to ensure they could hear the vehicle sounds, and ophthalmological exams to confirm AMD or normal vision status. The vehicle was presented visually only, aurally only, or both visually and aurally. Similar to our prior studies (DeLucia et al., 2016; Keshavarz et al., 2017), we used a psychophysical reverse-correlation approach (Ahumada & Lovell, 1971; Azen & Budescu, 2003) to determine the relative importance of visual and auditory cues. The vehicle approached for 3 s and then was removed (no longer seen or heard). Participants subsequently pressed a button when they thought the vehicle would reach them. We measured TTC estimates, and the relative weights of the vehicle’s velocity, as well as actual TTC, distance, optical size, and sound intensity when the vehicle was removed.

Results indicated that TTC judgments were based on both visual and auditory cues in the NV and AMD groups. The latter relied, at least in part, on their residual vision despite having AMD in both eyes. In the vision-only condition, the relative importance of heuristic cues of distance and optical size compared to TTC was greater in the AMD group compared to the NV group. To assess for a multimodal advantage, the mean absolute error (difference between the actual and estimated TTC at time of vehicle removal) was compared across the three modality conditions with a repeated measures analysis of variance. The difference between the AMD and NV groups was not significant, p > .05, and a multimodal advantage was not observed in either group; that is, performance did not improve when both visual and auditory information were presented compared to either one alone. These results are consistent with those of a prior study showing that street-crossing decisions were relatively accurate and comparable between NV and AMD groups (Hassan & Snyder, 2012), although TTC estimates and the relative importance of different types of cues were not measured.

In the current study, TTC estimates in the AMD group were surprisingly similar to those in the NV group. However, the AMD group showed a higher relative importance of “heuristic” cues compared to the more reliably accurate cues favored by the NV group. These findings suggest that similar performance may be achieved through different cue-weighting strategies. However, our scenes showed one vehicle approaching in one direction on a one-lane road with a limited range in vehicle parameters (e.g., size, speed, TTC). This may have allowed the AMD group to use the less reliable heuristic cues to achieve similar performance to the NVG group. Greater differences may be found with a wider range of parameters and more difficult traffic scenarios.