AI-driven mobile solution for early detection and management of diabetic foot ulcers

Abstract

Diabetic Foot Ulcers (DFUs) are one of the most serious and common complications of diabetes mellitus, with an estimated 15% to 25% of people with diabetes developing a DFUs during their lifetime. To combat misinformation and promote treatment adherence, it is proposed to develop an integrated follow-up framework, capable of intelligent treatment monitoring. The application integrates a Deep Learning (DL) approach to analyse images submitted by patients and provide personalized feedback to help adapt and optimize treatment. The proposed tool aims not only to provide educational information but also facilitate remote communication between the patient and the healthcare professional, contributing to the improvement of existing ulcers and the early detection of new lesions. The architecture of a mobile application for this purpose is outlined, and the routing of information in the application via APIs is also explained so that data can be recorded and captured efficiently. The joint implementation of the Deep Learning, YOLO/RetinaNet, and Segment Anything Model (SAM) models to classify and segment, respectively, the images submitted to the application is likewise described. Preliminary results indicate that the YOLOv11n and RetinaNet with resnet50+FPN backbone models achieved mAP@50 of 0.844 and 0.811, respectively.