Photography-based Retinal Vasculometry and Visual Field Progression in Primary Open Angle Glaucoma.

Abstract

To evaluate whether retinal vasculometry features derived from baseline fundus photographs are associated with subsequent longitudinal visual field (VF) change in glaucoma suspects and eyes with primary open-angle glaucoma. Retrospective cohort study. Retinal vasculometry metrics, including length-diameter ratio, tortuosity, branching measures, fractal dimension, and vessel-density measures, were derived from baseline fundus photographs using an automated deep-learning system. Linear mixed-effects models evaluated associations between baseline vasculometry and longitudinal VF mean deviation (MD) change. Clinical-only, full vasculometry, and parsimonious models were compared, with hierarchy-respecting backward selection performed under maximum likelihood using the Akaike information criterion. A total of 1,595 eyes (757 POAG and 838 glaucoma suspect) from 929 participants with at least 2 years of follow-up and 5 VF examinations were included. A mean of 14.5 VFs were obtained over 10.3 years, and the mean VF MD change was -0.15 dB/year. In the parsimonious model, VF MD slope differed by 0.079 dB/year (95% CI, 0.034-0.123) per 1-SD higher venous fractal dimension, 0.056 dB/year (95% CI, 0.025-0.087) per 1-SD higher arterial fractal dimension, and 0.032 dB/year (95% CI, 0.015-0.050) per 1-SD higher arterial tortuosity (all P<0.001). Baseline retinal vasculometry features were associated with subsequent longitudinal VF change after adjustment for demographic and clinical factors. These findings support further investigation of fundus photograph-derived vascular features as potential prognostic markers, although independent validation is needed before their clinical predictive utility can be established.