Explainable CNN Model for Real Time Diabetic Wound Severity Classification in Telemedicine

Diabetic Foot Ulcers (DFUs) represent a major cause of morbidity and potential amputations in diabetic patients, highlighting the need for efficient and accurate diagnostic methods. This study presents an explainable deep learning model utilizing Convolutional Neural Networks (CNN) to automatically classify diabetic wound severity. The model leverages transfer learning and fine-tuning strategies to enhance its performance on varied image datasets. The primary objective is to develop a real-time, scalable solution that can be integrated into telemedicine platforms for remote management of DFUs. The CNN model was trained and evaluated using standard classification metrics, including precision, recall, and F1-score, achieving an impressive classification accuracy of 92.4%. To ensure clinical interpretability and trust, Grad-CAM was incorporated to visualize the regions of interest the model focuses on during diagnosis, providing transparency in decision-making. The system delivers real-time inference in under two seconds per image, outperforming traditional methods like Support Vector Machines (SVM) and Random Forest in both accuracy and speed. These results demonstrate the model’s robustness across varying image conditions and its potential for integration with Electronic Health Records (EHRs) and telemedicine platforms. The combination of high accuracy, fast processing speed, and explainability ensures that this approach can support early intervention, improve patient outcomes, and facilitate the adoption of AI-driven diagnostic tools in clinical settings.