This retrospective study of 417 patients evaluated whether quantitative retinal vascular features from fundus images can non-invasively detect coronary artery disease (CAD). Using AI-based analysis, several retinal parameters—especially reduced fractal dimension (FD) and vessel density (VD), along with changes in vessel diameter and optic disc morphology—were significantly associated with CAD. A combined diagnostic model integrating retinal features with clinical risk factors achieved good performance (AUROC ≈0.80), suggesting retinal imaging is a promising, interpretable tool for early CAD screening.