Wang, V.

Background Retinal vascular imaging offers a non‐invasive window into the central nervous system health and has shown promise in the study of neurodegenerative diseases. Emerging evidence highlights that Parkinson's disease (PD) exhibits structural and functional changes in the retina, including vascular alterations. This pilot study evaluated whether retinal vascular morphological features, quantified using the deep learning tool Automorph, discriminate PD from non‐PD subjects in the UK Biobank cohort. Methods We analyzed color fundus photographs (CFPs) from 303 PD participants (mean age 62.6 ± 5.4 years, hypertension (HTN) = 67%, diabetes mellitus (DM) = 13%) and 516 non‐PD controls (age = 55.9±8.1, HTN = 22.8%, DM = 3%) enrolled in the UK Biobank. Six retinal vascular morphology features were quantified: fractal dimension, vessel density, average width, distance tortuosity, squared curvature tortuosity, and tortuosity density. General linear models were used to assess group differences while adjusting for age, hypertension, and diabetes. We also evaluated the association between retinal features and age, and the left‐right eye retinal fractal measures correlations using Pearson correlation and Fisher Z‐test. Results Compared to non‐PD individuals, PD patients showed significantly altered retinal vessel average width, both without (p = 0.001) and with (p = 0.002) vascular covariate adjustment. Group‐by‐age interaction effects were also significant for fractal dimension (p = 0.010) and vessel density (p = 0.023). Furthermore, inter‐eye correlation of vessel density differed significantly between groups (p = 0.034), suggesting altered bilateral vascular symmetry in PD. Conclusion Measurable alterations in retinal vascular morphology in PD support the potential of retina vascular imaging for non‐invasive neurodegenerative disease research. These results warrant further validation in larger cohorts, with improved analytic robustness and automated retinal vascular grading.