ConvXGDFU - Ensemble Learning Techniques for Diabetic Foot Ulcer Detection

Abstract

Medical practitioners have been studying Diabetic Foot Ulcers (DFU) as a critical subject for treatment purposes. The fundamental objective is to achieve a mechanism for early detection and identification of DFU, ensuring effective treatment before progressing to a critical stage. The traditional clinical techniques have drawbacks, such as a high diagnosis cost, high clinical workload, and an extended treatment time. Moreover, the cost of delayed detection and treatment can lead to significant significance. Although this approach yields outstanding results, a remote, cost-effective, and easy DFU diagnostic method is required. In recent times, Machine Eearning and Deep Learning methods have proven to be very effective and efficient in medical diagnosis and disease detection. The fundamental objective of this study is to build an efficient Artificial Intelligence model for detecting DFUs. We have proposed a novel Deep Learning model using CNNs and XGBoost for DFU detection. Our proposed model is called ConvXGDFU, which can efficiently classify DFU vs Normal Skin patches. Results show that our devised model achieved an accuracy and F1 score of 99.90% and 99.60% for both classes.