Retinal Fundus Imaging as a Biomarker for ADHD: Machine Learning for Screening and Executive Function Stratification

This study analyzed retinal fundus photographs from 648 children and adolescents using machine learning to develop ADHD screening models, achieving high accuracy (AUROC 0.955–0.969) with vessel density as the most predictive feature. The research also explored whether retinal features could stratify executive function deficits in ADHD, representing a novel noninvasive approach to complement existing diagnostic methods.