Early Detection of Diabetic Foot Ulcers through CoAtnet Model: A Scalable Solution for Healthcare Integration

Diabetic foot ulcers are one of the serious complications of diabetes mellitus affecting millions of patients across the globe and could lead to infections and amputations if diagnosed late. The recent trends in machine learning, with high promise, have been embodied in deep learning to solve the detection task for diabetic foot ulcers accurately. However, existing methods, such as the traditional CNN, faced the challenge of focusing on the critical ulcerous regions and thus were not productive in the real-time performances. Thus, this present research focuses on improving a deep learning-based model by employing CoAtNet, which is a customized attention mechanism to develop convolutional mechanisms for enhancing the outcomes related to DFU detection from foot images. The model was trained on the DFUC dataset, along with data augmentation techniques such as rotation, flipping, and contrast adjustment to enhance robustness. Therefore, accuracy was around 98% and F1-score stood at 90%. The suggested solution has outstanding potential for the early detection of DFUs, especially in resource-poor health environments, thus improving the outcome of patients and reducing the occurrence of severe complications like amputation.