Abstract DP331: Automated Identification of Retinal Vascular Biomarkers in the Field using a Mobile Laboratory and Portable Fundus Cameras

Objective: To assess the link between retinal vascular health and cognitive performance in a diverse population, using a novel mobile laboratory platform and portable color fundus photography. Background: The retina is an accessible window into brain health. Studies that had highlighted the link between retinal vasculature and cognition included clinical populations or narrowly defined research cohorts, with limited generalizability. To address the need to expand this research to demographic and geographic diverse communities, we launched a mobile laboratory platform across rural and urban Arizona zip codes. Methods: 240 subjects were recruited in our prospective community-based study. We assessed cognition using the MindCrowd web-based testing platform and analyzed verbal paired associates (VPA) task, a validated measure of verbal associative memory sensitive to age-related cognitive decline. Non-mydriatic color fundus photographs were obtained using the portable hand-held Aurora camera (Optomed), two images per eye (disc- and macula-centered). Image processing and vascular feature quantification were automatically conducted using Automorph software. Linear regression models examined the associations between retinal vascular metrics and VPA, adjusting for age, sex, education attainment, and risk factors. Results: Our cohort’s mean age was 45 +/-16.7 years, 32.9% were male and 39.5% had a college degree. In disc-centered fundus images, fractal dimension strongly associated with chronological age (p = 1.04E-10). Greater artery tortuosity in the retinal zones B and C was associated with lower VPA (p = 0.02 and p = 0.004, respectively). In macula-centered images, fractal dimension showed the strongest association with age (p = 4.31E-11), reinforcing its utility as a robust biomarker of vascular aging. Males had wider average artery caliber in zone B (p=0.02). Zone B vein fractal dimension (p=0.002) and vein vessel density (p=0.006) were the strongest predictors of VPA. Conclusions: This study demonstrates the safety, feasibility, and utility of a mobile laboratory to collect high-quality non-mydriatic color photos and cognitive data across a geographically diverse cohort. Arterial tortuosity and venous fractal dimension and density, particularly in retinal zone B, were predictive of cognitive performance, after adjusting for vascular health covariates. Retina may become a scalable and accessible platform for studying brain aging in the general population.