Ensemble of Vision Transformers and CNNs for Accurate Diabetic Foot Ulcer Classification

Diabetic Foot Ulcers (DFU) are a significant diabetes complication that can lead to lower limb amputation. Currently, 537 million suffer from diabetes worldwide and it is anticipated to rise to 783 million by 2045. Given the speed at which DFU is developing, prompt action is necessary to avoid the serious consequences of amputation and associated various medical conditions. With the evolution of image-based ML algorithms, automated methods for identifying and assessing DFUs are becoming more common. Existing research works on visual computing techniques concerns tissue classification and detects the DFU's visual appearance. This research study combines ResNet50 with Vision Transformers (ViT) and MobileNet with Vision Transformers (ViT) to develop an ensemble model to classify the existence or absence of a Diabetic Foot Ulcer (DFU). Experimental results of the proposed model when tested on challenging datasets achieved a validation accuracy of 98.6%.