Multiclass Classification of Diabetic Foot Ulcers Using EfficientNet

A diabetic foot ulcer (DFU) is among the most serious complications of diabetes. When DFUs are not detected early and managed properly, they cause serious complications like infections, hospital admissions, and possibly lower-extremity amputation. This paper presents a deep learning framework powered by artificial intelligence to classify diabetic foot ulcers into four clinically relevant categories, specifically: infection, ischemia, combined infection and ischemia, and cases where there is no ulcer present. A convolutional neural network (CNN) based EfficientNet is constructed to classify DFU photos across the four classification categories. The model is developed using the Roboflow DFU multiclass dataset and includes preprocessing and data augmentation techniques that improve generalizability. Results demonstrate that the AI-based deep learning classification of DFUs reached an accuracy of 91.8% overall, with very high values of precision, recall, and F1-score for each category.