Automated Severity Classification Using Convolutional Neural Networks: A Deep Learning Approach to Diabetic Foot Ulcer Assessment

: This paper elaborates on the development of designing and training a CNN to distinguish between mild, moderate, and severe DFUs. There were four distinct parts to the research project: The main activities involved are data acquisition and data pre-processing, model architectural design and model building, model training and model assessment, and the final phase which is model analysis and result interpretation. Pictures of DFUs were used in the dataset and performances such as normalization and augmentation were used to enhance the dataset to ensure that all was in order. The CNN was designed to learn and extract data from images; the two sets and convolutional layers, max, and fully connected layers. Achieving 96 % good precision, 85 % good recall, and F1 between 0.97 and 0.98 for all severity levels, the model used a confusion matrix to distinguish between training and testing. Class A had a real positive rate of 0% to 20%, Class B of 20% to 40%, Class C of 40% to 60%, Class D of 60% to 80%, and Class E of 80% to 100%. Providing a useful tool for the treatment of diabetic foot ulcers, these results show that the model is resilient and can improve diagnosis accuracy in clinical settings.