Shopping with glaucoma. Quantifying the impact of mild glaucomatous visual field loss using a virtual reality supermarket.

Technical Summary: Shopping with Glaucoma – Quantifying the Impact of Mild Glaucomatous Visual Field Loss Using a Virtual Reality Supermarket

Problem Statement
Current methods for evaluating the real-world impact of new glaucoma treatments rely heavily on Patient Reported Outcome Measures (PROMs), acuity tests, and standard visual field assays. The authors argue that these conventional assays may be insufficient due to their subjective nature, inherent noise, and practical limitations. Specifically, there is a need for direct, objective measures of functional vision that can detect subtle deficits in everyday tasks. This gap is particularly critical for patients with mild binocular visual field (VF) loss, a group where conventional PROMs are considered least sensitive, yet where preventative interventions are likely to be most beneficial.

Methodology
The study employed a novel virtual reality (VR) shopping task designed to simulate a real-world grocery environment.

• Experimental Setup: Eight glaucoma patients (confirmed by ophthalmologist diagnosis) navigated a VR supermarket using a Meta Quest Pro headset (722g). Participants physically walked through the environment without controllers or external trackers, relying on optical hand tracking for reach-and-grasp interactions.
• Task Protocol: Participants were given a virtual shopping list of 30 items distributed across six aisles. They were required to locate items (which varied in visibility due to lighting, label legibility, and item density) and place them into a virtual trolley. The task continued until all items were successfully deposited. Feedback was provided immediately upon item placement.
• Participant Selection: The cohort consisted of eight individuals with glaucoma. While some exhibited moderate to advanced monocular VF damage, the study specifically selected for mild binocular VF loss (Integrated Visual Field [IVF] Mean Deviation < -6dB) to test sensitivity in this specific population.
• Data Analysis: Beyond standard metrics like completion time and error counts, the study utilized VR headset sensors to compute movement metrics, including head, eye, and body translations. Movement vectors were normalized by time to avoid confounding total movement with duration. Statistical analysis primarily used Pearson's r correlations, with results confirmed via non-parametric Kendall tau tests due to violations of normality.

Key Contributions
The primary contribution of this work is the development and validation of a VR-based functional assessment tool capable of detecting performance variations associated with mild binocular visual field loss. The study introduces a composite performance metric derived from multiple behavioral variables (time, head/body movement, grasping efficiency) and demonstrates its correlation with clinical visual field data.

Results

• Overall Performance: There was a significant correlation between the magnitude of binocular VF loss (measured by IVF-MD) and the combined task performance metric (r = 0.35, P < 0.001).
• Specific Behavioral Correlations: Patients with greater VF loss exhibited distinct behavioral changes:
  • Reduced head movement (r = 0.29, P < 0.001).
  • Reduced body translation (r = 0.29, P < 0.001).
  • Increased number of grabbing movements for target items (indicating hesitation or dropped items) (r = -0.17, P = 0.008).
  • Increased time spent moving items from shelf to trolley (r = -0.20, P = 0.002).
• Non-Correlations: Performance was not correlated with the worse eye's Mean Deviation (P = 0.865). Furthermore, no significant associations were found between binocular VF loss and specific sub-tasks such as time to find items, time spent reading the list, or the number of incorrect items placed in the trolley.

Significance and Claims
The authors position this study as a "proof-of-principle" demonstration that virtual reality can objectively measure performance variations in meaningful real-world tasks for patients with mild glaucoma. The paper claims that the observed correlations suggest VR tasks have promise for evaluating the benefits of new treatments, particularly in the mild VF loss stage where other measures fail.

The authors remain modest regarding the effect size, acknowledging it is relatively small (r = 0.35) and that considerable individual variability exists, which they attribute to the complex, unrestricted nature of the "real-world" simulation. They do not claim the current metric is a definitive clinical standard but suggest it could be refined (e.g., by introducing low-contrast or low-light elements) and potentially used for longitudinal tracking of disease progression or treatment efficacy in future clinical trials. The study concludes that the next steps involve assessing patient acceptability and standardizing the procedure.