Advances in Multimodal Imaging and Artificial Intelligence for the Diagnosis and Management of Diabetic Retinopathy: From Biomarkers to Personalized Care

Background and Objectives: Diabetic retinopathy (DR) remains a leading cause of preventable vision loss. Staging based on colour fundus photography captures only part of a disease that involves the whole neurovascular unit and the retinal periphery. This narrative review summarizes the structural (optical coherence tomography, OCT), vascular (OCT angiography and ultra-widefield imaging) and functional (microperimetry, electrophysiology) biomarkers of DR, grades the evidence supporting each of them, and examines how artificial intelligence (AI) is changing screening, prognostication and treatment selection. Materials and Methods: PubMed/MEDLINE, Scopus and Web of Science were searched for publications from January 2000 to May 2026, giving priority to randomized trials, prospective multicentre cohorts, systematic reviews and externally validated algorithms. Results: Among imaging biomarkers, disorganization of the retinal inner layers, outer retinal integrity and ultra-widefield angiographic non-perfusion have the most consistent prognostic evidence, whereas OCT angiography metrics remain device-dependent. AI has progressed from research classifiers to autonomous systems with regulatory authorization in the United States and the European Union, to prospectively evaluated deployments in national programmes and low-resource settings, and to time-to-event progression models and foundation models. Image-based prediction improves discrimination over clinical risk factors by modest margins, and biomarker-guided treatment of diabetic macular oedema still rests largely on retrospective series. Cross-device standardization of quantitative metrics, representative datasets, prospective outcome trials and health-economic evidence remain the main gaps. Conclusions: Individualized screening intervals are the application closest to routine practice, whereas phenotype-driven therapy and precision monitoring remain promising but unproven.