Vision Transformer Based Diabetic Foot-Ulcer Detection: A Study
Ramya Mohan, N. Sri Madhava Raja, Robertas Damaševičius, David Taniar, S. Prabha, V. Rajinikanth
Abstract
The global rise in diabetes is driven by various factors. Poor management can lead to severe complications, emphasizing the urgent need for heightened awareness and enhanced prevention and management strategies. Diabetes can lead to severe foot ulcers (FU), which, if left untreated, may result in incurable wounds or even amputation. The proposed research aims to develop a tool for automatically detect the FU using the Vision Transformer (ViT). The stages of this tool include the following sections; (i) image collection, resizing, and contrast enhancement, (ii) implementation of ViT to extract the image features, and (iii) binary classification and performance confirmation. The planned work is executed with various patch sizes and the reached results are discussed. The performance of the ViT is also verified against the deep-learning methods like VGG16, VGG19, ResNet50 and ResNet101 and the experimental outcome confirms that the ViT approach is efficient in providing better detection accuracy (98.58%).