Deep learning Applications in Cardiovascular Risk Prediction Using Retinal Fundus Imaging: A literature Review

Abstract

This literature review examines the use of deep learning (DL) models in analyzing retinal fundus images for predicting cardiovascular risks. It compiles current findings from peer-reviewed research conducted between 2020 and 2025, which have developed and validated DL algorithms for forecasting cardiovascular events and risk factors using retinal photographs. A thematic narrative method was utilized to identify methodological trends, performance measurements, and clinical implications across diverse study designs. Models such as Reti-CVD, Reti-CAC, and networks for predicting retinal age show strong correlations with traditional risk indicators, like coronary artery calcium (CAC) and carotid intima-media thickness (CIMT), highlighting the retina as a non-invasive marker of systemic vascular health. The review concludes that while DL-based retinal biomarkers offer significant predictive potential, additional real-world validation and regulatory alignment are necessary before they can be implemented clinically