EfficientNet-Based Deep Learning Model for Accurate Diabetic Foot Ulcer Classification

Diabetic Foot Ulcer is one of the serious complication issues in diabetes which, if not identified and treated early, leads to severe infection with amputation and diminishes the quality of life. DFU staging using automatic techniques is needed since manual diagnosis takes so much time, hence it becomes prone to human error. Deep learning techniques can offer an efficient alternative so that healthcare professionals will be assisted in classifying diabetic foot ulcers automatically. This study proposes a diabetic foot ulcer classification methodology using a very efficient convolutional neural network model, namely EfficientNet, which is trained on data comprising a dataset of two classes of images: abnormal-analysis (ulcer) with 512 images and normal (healthy skin) with 543 images. The model was optimized for training to find the best difference between these two classes; thus, performance evaluations were carried out based on validation loss and accuracy metrics. The model proposed therefore gives a validation loss of 0.0496 and hence has reached an accuracy of 95%, proving to be effective in the detection and classification of diabetic foot ulcers. The current study points toward the feasibility of advanced deep learning models for the classification of medical images-a promising solution toward the early detection and treatment that could reduce the risks of complications from diabetic foot ulcers. These results therefore show that EfficientNet, through its optimized architecture, can be a very promising and highly accurate strategy in the classification of DFUs and could thus be integrated into healthcare systems for better patient outcomes.