Abstract

Diabetes is a chronic health condition that affects the way the body converts food into energy. One of the most common complications for patients with diabetes is diabetic foot ulcers (DFU). DFUs are usually located in the lower part of the leg. In general, DFU, if not treated early, will be costly and can lead to foot amputations, and death rates can increase. The study offers classification of ischemia, infection, and normality and prediction of diabetic foot ulcers with a deep learning approach. The proposed methodology uses a deep convolutional neural network model that includes four stages: preprocess augmentation, model training, validation, and classification of ischemia, infection, and normal. The model used is EfficientNet. The model that has been trained on the computer is then converted into a lite classification model. Convert and optimize models using TF Lite. The resulting model will be smaller and more optimal, so it can be used in mobile applications with the help of the TF Lite Interpreter. Android apps on mobile devices that have made the accuracy level reach 0.88, the precision value is 0.96, the recall is 1.00, and the F1 score is 0.97.