Lucy Starling¹, Juju Fars¹, Po-Chun Tseng¹, Junyun Zhang², Dennis M. Levi³, Laura Breach⁴, Vijay K. Tailor-Hamblin^5,6, Pedro Quijada Leyton⁷, William Latham⁸, Stephen Taylor¹, I. Betina Ip¹, Holly Bridge¹; ¹University of Oxford, Oxford, UK, ²Peking University, Beijing, China, ³UC Berkeley, USA, ⁴Oxford Eye Hospital, John Radcliffe Hospital, Oxford, UK, ⁵NIHR Moorfields Biomedical Research Centre, London, UK, ⁶Moorfields Eye Hospital NHS Foundation Trust, London, UK, ⁷University College London, UK, ⁸Goldsmiths, University of London, UK

Virtual reality (VR) training offers a promising platform for vision science research conducted outside of the laboratory and for visual rehabilitation. Here we aimed to determine whether an established computer-based training for amblyopia could be implemented into VR. We tested whether a home-based VR training protocol, based on the dichoptic de-masking task described in Liu and Zhang (2018), could induce perceptual learning in participants with amblyopia (N=18, 12 female, aged 18-50,15 training days). Normally sighted controls (N=16, 13 female, aged 21-30, 5 training days) provided normative performance measures. The training task was to indicate which of two sinusoidal gratings, presented to the weak eye, had a higher contrast. The stronger eye was presented with visual noise. The delta (contrast difference between the two gratings) was used to quantify changes during training. In addition, both groups performed tests of visual acuity, contrast sensitivity, interocular balance and stereopsis before and after training. VR training improved performance on the dichoptic de-masking task for participants with amblyopia (first day mean delta = 20.42 (SD=4.09), final day mean delta = 12.55 (SD=5.79), t(17)=5.89, p<.001) and controls (first day mean delta = 18.13 (SD=2.02); final day mean delta = 13.49 (SD=3.62); t(14)=4.19, p<.001). Participants with amblyopia significantly improved their interocular balance (t(17)=2.386, p=.011), and five gained stereopsis. There was no difference in visual acuity or contrast sensitivity in the amblyopic participants, and those with healthy vision showed no change in any outcome measures. While all except one participant completed the required training amount, feedback suggested training was repetitive, tedious, and lacked variety. These findings suggest that VR training outside of the lab is feasible and can be completed with high compliance. Training resulted in plasticity in controls and visual rehabilitation in amblyopes, opening avenues for larger scale studies in a variety of visual conditions.

Acknowledgements: This work was funded by the Medical Research Council (MR/V034723/1 to HB and IBI), and The Royal Society (University Research Fellowship to HB, Dorothy Hodgkin Research Fellowship to IBI). LS is funded by a Linacre EPA Cephalosporin scholarship.