Evaluation of a Virtual Reality implementation of a binocular imbalance test

Abstract

The purpose of this study was (1) to implement a test for binocular imbalance in a Virtual Reality headset, (2) to assess its testability, reliability and outcomes in a population of clinical patients and (3) to evaluate the relationships of interocular acuity difference, stereoacuity and binocular imbalance to amblyogenic risk factors. 100 volunteers (6 to 70 years old, mean 21.2 ± 16.2), 21 with no amblyogenic risk factors and 79 with amblyopia or a history of amblyopia participated. Participants were classified by amblyogenic risk factor (24 anisometropic, 25 strabismic and 30 mixed) and, for those with strabismus, also by refractive response (16 accommodative and 39 non-accommodative). We characterized our sample using three variables, called the 'triplet' henceforth: interocular acuity difference, stereoacuity and imbalance factor. Binocular imbalance showed high test-retest reliability (no significant difference between test and retest in a subgroup, n = 20, p = 0.831); was correlated with Worth 4 dots test (r = 0.538, p<0.0001); and correlated with both interocular acuity difference (r = 0.575, p<0.0001) and stereoacuity (r = 0.675, p<0.0001). The mean values of each variable of the triplet differed depending on group classification. Mixed and non-accommodative groups showed the worst mean values compared with the other groups. Among participants with strabismus, strabismic vs mixed subgroups did not show significant differences in any variable of the triplet, whereas the accommodative vs non-accommodative subgroups showed significant differences in all of them. According to a univariate logistic model, any variable of the triplet provides a good metric for differentiating patients from controls, except for binocular imbalance for anisometropic subgroup. The proposed binocular imbalance test is feasible and reliable. We recommend monitoring amblyopia clinically not only considering visual acuity, but also stereoacuity and interocular imbalance. Stereoacuity on its own fails because of the high percentage of patients with no measurable stereoacuity. Binocular imbalance may help to fill that gap.

Conflict of interest statement

I have read the journal's policy and the authors of this manuscript have the following competing interests: SM and JP promoted, with the support of the University of Oviedo, the creation of the startup VisionaryTool. Both have assisted VisionaryTool, S.L. (www.visionarytool.com) to create a commercial version of the VR imbalance test described in this manuscript (University of Oviedo contract FUO-EM-19-099). VisionaryTool has not had any role (writing, analysis, or control over publication) in the production of the paper. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Figures

Fig 1. Binocular imbalance test.

Prior to the test itself, a dichoptic nonius alignment screen is presented (bottom). The test consists of a series of trials, each one divided into three steps (from bottom to top): firstly, a fusion frame of a dichoptic nine square grid surrounded by a high contrast frame is presented to facilitate fixation. After the experimenter presses the spacebar, the computer displays two dichoptic letters, randomly selected and filtered to the target interocular contrast, for 200 msec. Finally, both letters are presented binocularly one next to the other with enhanced contrast and size. The observer’s task is to indicate which letter he/she has perceived (two-alternative forced choice). The computer adjusts the imbalance ratio following a 1 up/ 1 down staircase for the next trial, until a valid threshold is obtained.

Fig 2. Binocular imbalance repeatability Bland-Altman plot.

Binocular imbalance results in first and second measurements. N = 20 volunteers.

Fig 3. Triplet variables box plots.

Graphical display of the continuous variables (LogMAR acuity difference, log10 stereoacuity and imbalance factor) using boxplots for each classification groups: amblyogenic factor (control, anisometropic, strabismus and mixed) and deviation nature (accommodative and non-accommodative). The box plot shows the mean and distribution in quartiles of the data. Circles represents each single participant.