Document resource
Background Retinal microvasculature reflects systemic vascular health and is associated with cardiovascular disease risk. Advances in artificial intelligence have enabled extraction of complex retinal features that function as non-invasive cardiovascular risk biomarkers. Dr Noon CVD is an AI-based retinal model designed to estimate future cardiovascular event risk; however, its relationship with underlying coronary artery disease burden remains incompletely characterized.Objective To evaluate the alignment of Dr Noon CVD, a non-invasive AI-based retinal cardiovascular risk biomarker, with angiographically defined coronary artery disease severity, and to compare its risk classification patterns with Framingham and Pooled Cohort Equation (PCE) risk scores.Methods We conducted a retrospective case series of 20 patients undergoing coronary angiography at Hanoi Medical University Hospital, Vietnam. Clinically significant disease was defined as ≥70% stenosis in any major epicardial vessel, chronic total occlusion, or left main stenosis ≥50%. Multivessel severe disease was defined as involvement of two or more major epicardial vessels or left main disease. Patients were stratified using Dr Noon CVD, Framingham, and PCE risk categories. Risk classification rates were evaluated across angiographic severity strata to assess concordance between risk stratification and disease burden.Results Fourteen patients (70%) had multivessel severe coronary artery disease, and 19 (95%) had at least one severe angiographic lesion. Dr Noon CVD classified 71% of patients with multivessel severe disease as high risk, compared with 57% for PCE and 21% for Framingham. Among patients without multivessel severe disease, Framingham most frequently classified patients as non–high risk. When moderate and high Dr Noon CVD risk categories were combined, 86% of patients with multivessel severe disease were classified as positive, without a corresponding increase in misclassification among non-multivessel cases.Conclusions In this angiography-enriched case series, Dr Noon CVD demonstrated meaningful alignment with coronary disease burden despite being designed as a prognostic model. These findings support the potential role of AI-based retinal cardiovascular biomarkers in risk reclassification among patients with established coronary artery disease and highlight the need for prospective evaluation in broader at-risk populations.Abstract 327 Figure 1