Diabetic Foot Monitoring Based on A Lightweight Model

Diabetic foot monitoring has garnered widespread attention in the medical field. With the annual increase in the number of diabetes patients, effectively and timely detecting and managing foot complications has become an urgent issue to address. This paper proposes the use of a lightweight model, MobileNetV3, for diabetic foot monitoring, and introduces Efficient Local Attention (ELA) to weight images from three channels, enhancing the model's feature extraction capabilities. Additionally, we improved the loss function by integrating the focal loss, aiming to enhance the model's performance and evaluate key data values. This also addresses the issue of uneven data distribution, improving the model's ability to judge critical data thresholds. The effectiveness and accuracy of this method were verified through experiments on clinical datasets of diabetic foot.