A Robust DFU_CVENETB3 Deep Learning Model for Improved Accuracy in Early Ulcer Prediction Based on Diabetic Foot Images

Abstract

In many facets of healthcare, including diagnosis, treatment, and even epidemiology, deep learning (DL) plays a significant role. In the past, medical practice was determined exclusively by the training that physicians had received. However, the importance of artificial intelligence applications has increased along with the number of databases, and the advantages of using DL in medicine have gained more recognition. Diabetes has become a major medical concern and is a fairly common disease. This metabolic disorder increases the chance of developing numerous conditions, including foot diseases, ulcers, amputations, etc. when blood glucose levels are not properly managed. Significant morbidity is a result of diabetic foot ulcers and amputations. Diabetic foot can be avoided by identifying at-risk patients and putting preventative measures in place. We presented a novel model in this paper that combines the characteristics of Vgg16 with EfficientNetB3 and DFU_CVENETB3 Convolutional Neural Network (CNN), a very powerful potential game-changer towards improved prevention of diabetic foot ulcers. This study will introduce deep neural network models to automatically classify early diabetic foot images into normal (healthy) and diseased (DFU) categories. A set of images of diabetic foot ulcers will also be used, and the results will show us that the model based on EfficientNetB3 performed better than traditional CNN models such as VGG16. EfficientNetB3 produced the highest accuracy results, compared to previous works mentioned in the study. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.