0028Artificial Intelligence Prediction Models for Diabetic Foot Ulcer Progression and Limb Loss: A Systematic Review

Objective To systematically evaluate artificial intelligence (AI) prediction models developed for diabetic foot ulcer progression, healing failure, infection severity, and lower-extremity amputation risk. Methods A systematic review was conducted according to PRISMA guidelines using PubMed, Embase, Scopus, and Web of Science through April 2026. Studies evaluating machine learning, deep learning, computer vision, or predictive analytics for diabetic foot ulcer assessment or amputation prediction were included. Data regarding model design, clinical variables, predictive performance, and implementation potential were extracted. Results Fifty-one studies met inclusion criteria. AI models commonly incorporated wound imaging, laboratory markers, vascular parameters, and electronic health record data to predict healing outcomes and limb loss risk. Deep learning image-analysis systems demonstrated strong diagnostic accuracy for ulcer classification and tissue assessment. Predictive performance was highest in models integrating multimodal clinical and imaging data. Major limitations included retrospective study designs, small datasets, inconsistent external validation, and limited evaluation in diverse populations. Few studies assessed prospective clinical implementation. Conclusions AI-based prediction systems demonstrate promising utility for early identification of diabetic foot ulcer complications and amputation risk. However, larger prospective studies and standardized validation frameworks remain necessary before widespread clinical integration into diabetic limb-preservation care pathways. Is this an encore abstract? No