Measuring visual information gathering in individuals with ultra low vision using virtual reality

Abstract

People with ULV (visual acuity ≤ 20/1600 or 1.9 logMAR) lack form vision but have rudimentary levels of vision that can be used for a range of activities in daily life. However, current clinical tests are designed to assess form vision and do not provide information about the range of visually guided activities that can be performed in daily life using ULV. This is important to know given the growing number of clinical trials that recruit individuals with ULV (e.g., gene therapy, stem cell therapy) or restore vision to the ULV range in the blind (visual prosthesis). In this study, we develop a set of 19 activities (items) in virtual reality involving spatial localization/detection, motion detection, and direction of motion that can be used to assess visual performance in people with ULV. We estimated measures of item difficulty and person ability on a relative d prime (d') axis using a signal detection theory based analysis for latent variables. The items represented a range of difficulty levels (- 1.09 to 0.39 in relative d') in a heterogeneous group of individuals with ULV (- 0.74 to 2.2 in relative d') showing the instrument's utility as an outcome measure in clinical trials.

Figure 1

Examples of 2, 3 and 4-AFC spatial localization/detection tasks at 3 difficulty levels. (a) A 2-AFC task to detect whether a person is wearing a tie or not: black (high contrast), blue (medium contrast), and yellow (low contrast). (b) A 3-AFC task to detect whether a hanging towel is on the left, on the right, or missing. (c) A 4-AFC task to determine the location of the missing plate: 3, 6, 9, or 12 o’clock. For a complete set of tasks see the supplement.

Figure 2

Causes of the ULV among participants in the study. *Other conditions included Achromatopsia, Albinism, Leber’s Congenital Optic Neuropathy, Hydrocephalus, Bardet-Biedl Syndrome CMV Retinitis, Cone-Rod Dystrophy and Microphthalmia.

Figure 3

Estimated item measures across different tasks. More positive d′ values indicate more difficult items while more negative d′ values indicate easier items. Error bars represent the 95% CI of the estimates (see supplement for individual item measures). The number in parenthesis indicates the visibility level: (1) for high, (2) for medium, and (3) for low visibility.

Figure 4

Item measures for spatial localization (a), direction of motion (b) and motion detection (c) tasks. (1) indicates the level with highest visibility, (2) indicates level with medium visibility and (3) indicates the level with least visibility.

Figure 5

Shows the comparison of item measures across the three categories: spatial localization, motion detection and direction of motion.

Figure 6

Distribution of person measures across participants. By definition, the sample d’ is zero. Error bars represent the 95% CI of the estimates.

Figure 7

The relationship between visual acuity measured using BRVT and person measures estimated using SDT analysis for individual participants.

Figure 8

Comparison of item measures estimated using SDT and Rasch analysis. Item measures are stratified according to the number of response alternatives $m$ = 2, 3, and 4 because SDT equates chance performance for all $m$-AFC tasks at $d^{{}^{\prime }}=0$ while Rasch analysis does not. The diamond shape symbols (blue) show $m$-AFC = 4, square shape symbols (pink) show $m$-AFC = 3 and triangle shape symbols (green) show $m$-AFC = 2.