Abstract:

Diabetic foot ulcers (DFUs) are a leading cause of lower-limb amputations, with over 160,000 diabetics in the United States undergoing amputations annually. Current diagnostic methods lack real-time assessment of ulcer severity, infection risk, and treatment urgency. Clinicians often rely on slow, invasive, and costly procedures such as debridement and Doppler ultrasound. This paper presents a smart diabetic shoe integrated with an AI-powered mobile application to address these limitations. Upon detecting an ulcer through embedded temperature, moisture, and pressure sensors, the patient captures a foot image via the app. An AI-based imaging system, trained on over 16,000 images, classifies ulcer severity using the Wagner scale, detects infection and ischemia non-invasively, and generates depth images alongside wound segmentation and 3D mesh reconstruction for accurate measurement. The system also converts images to thermal representations and provides interactive dashboards to support healing monitoring, wound dressing guidance, and recurrence prediction. The smart shoe further features vibration motors as a preventive measure to promote circulation and reduce ulcer risk. Evaluation against clinical data demonstrated high accuracy across the metrics of all six AI models implemented. The results support improved patient self-care and clinical decision-making by enabling low-cost, non-invasive ulcer pre-classification, saving time, and reducing complication risks.