Diagnostic accuracy of a modularized, virtual-reality-based automated pupillometer for detection of relative afferent pupillary defect in unilateral optic neuropathies.

Purpose: To describe the construction and diagnostic accuracy of a modularized, virtual reality (VR)-based, pupillometer for detecting relative afferent pupillary defect (RAPD) in unilateral optic neuropathies, vis-à-vis, clinical grading by experienced neuro-ophthalmologists.

Methods: Protocols for the swinging flashlight test and pupillary light response analysis used in a previous stand-alone pupillometer was integrated into the hardware of a Pico Neo 2 Eye^® VR headset with built-in eye tracker. Each eye of 77 cases (mean ± 1SD age: 39.1 ± 14.9yrs) and 77 age-similar controls were stimulated independently thrice for 1sec at 125lux light intensity, followed by 3sec of darkness. RAPD was quantified as the ratio of the direct reflex of the stronger to the weaker eye. Device performance was evaluated using standard ROC analysis.

Results: The median (25th - 75th quartiles) pupil constriction of the affected eye of cases was 38% (17 - 23%) smaller than their fellow eye (p<0.001), compared to an interocular difference of +/-6% (3 - 15%) in controls. The sensitivity of RAPD detection was 78.5% for the entire dataset and it improved to 85.1% when the physiological asymmetries in the bilateral pupillary miosis were accounted for. Specificity and the area under ROC curve remained between 81 - 96.3% across all analyses.

Conclusions: RAPD may be successfully quantified in unilateral neuro-ophthalmic pathology using a VR-technology-based modularized pupillometer. Such an objective estimation of RAPD provides immunity against biases and variability in the clinical grading, overall enhancing its value for clinical decision making.