A Machine Learning Analysis of Pressure and Temperature Sensor Placement Optimization for Diabetic Foot Ulcer Prediction

Diabetic foot ulcers (DFUs) are a serious complication of diabetes, often leading to infection and lower-limb amputation. Early detection of biomechanical and thermal risk factors is essential for prevention. This study introduces a wearable smart compression sock integrating plantar temperature, pressure, and blood oxygen sensors, coupled with machine learning models for real-time DFU risk assessment. Data were collected from twenty participants (16-53 years) across four plantar locations (M1, M5, hallux, calcaneus) during structured activities (sitting, standing, walking, jogging). Temperature-only, pressure-only, and combined temperature-pressure models were evaluated. The combined model achieved the highest predictive performance, with mean F1 scores of 0.996\u20130.997 across locations and 0.983\u20130.996 across activities. M5 and calcaneus consistently provided the most informative signals, and dynamic activities enhanced sensitivity to acute mechanical stress. Predicted distributions closely matched ground truth measurements, demonstrating model reliability. These findings support the feasibility of multimodal wearable systems for continuous activity-aware monitoring of DFU risk, offering practical insights for personalized prevention.