Computer-Aided Diagnosis of Diabetic Foot Ulcers: A Feature-Based Transfer Learning Investigation

Diabetic foot ulcers (DFUs) can lead to severe infections and amputations if not detected early. This study aims to identify and extract features from DFU datasets using a class of deep learning approach known as feature-based transfer learning for improved diagnosis. The dataset contains normal and abnormal feet images with existing ulcers. A pre-trained convolutional neural network model, viz. VGG19 is utilized to extract discriminative features from the images. Two supervised learning methods, namely k-Nearest Neighbours (kNN) and Logistic Regression (LR), are then trained on these images to classify them as normal or abnormal. The data is split into 70:15:15 train, validation and test sets. It was illustrated from the study that the VGG19 + LR pipeline is able to distinguish the classes of DFUs well, suggesting its promising potential for computer-aided diagnosis.