An Explainable AI Framework for Diabetic Foot Ulcer Detection and Personalized Footwear Recommendation

Diabetic Foot Ulcers (DFUs) are a major complication of diabetes, often leading to infections, hospitalizations, and amputations if not detected early. This study proposes an integrated deep learning and IoT-based system for the early detection, classification, and prevention of DFUs. An EfficientNetB0 model is utilized to classify DFU images with high accuracy, while plantar foot type classification is performed using WaveNet, ResNet, and an Attention Mechanism to analyze plantar pressure data. For personalized footwear recommendation, a Decision Tree model is employed, enhanced with Explainable AI techniques (SHAP, LIME, and Grad-CAM) to provide transparency in decision-making. The system captures real-time foot health parameters using multiple IoT sensors (pressure, temperature, heart rate, gyroscope) embedded in smart footwear, enabling continuous monitoring and risk prediction. The developed models show promising results in improving early DFU diagnosis, personalized prevention strategies, and clinical decision support. Future enhancements aim to scale the solution with larger datasets, real-time cloud integration, and clinical validation. This research provides a novel, explainable, and wearable AI-driven framework for improving diabetic foot care outcomes.