Machine learning assessment of retinal blood flow links metabolic dysfunction and accelerated microvascular aging.

ABSTRACT: Microvascular aging impairs tissue perfusion and contributes to age-related organ dysfunction; however, direct approaches to assess microvascular aging remain limited. Here, we developed machine learning models to estimate chronological age from retinal blood flow data acquired by laser speckle flowgraphy (LSFG) in 1,008 adults undergoing health checkups. Using 18 predefined parameters and 3,253 automatically extracted time-series features, the best model for all participants achieved a mean absolute percentage error (MAPE) of 10.3% between predicted and chronological age. In addition, sex-specific models for women and men outperformed the model for all participants (MAPE 8.6% in women and 9.3% in men), allowing us to explore sex differences in microvascular aging patterns. Based on bias-c

SUBMITTER: Magi S 

PROVIDER: S-EPMC12827333 | biostudies-literature | 2026 Jan

REPOSITORIES: biostudies-literature