The Application of a Deep Learning Image Recognition Model Based on Multi-Scale Convolutional Attention Mechanism for Clinical Open Wound Detection

Wound detection is essential for preoperative evaluations and postoperative monitoring in clinical settings. This study presents a deep learning-based image recognition model that uses a multi-scale convolutional attention mecha-nism for automatic wound detection and classification, along with a depth estimation model to assess wound healing. Featuring 1,449 clinically annotated wound images, the model combines YOLOv8, for instance, segmentation with SegNext for en-hanced spatial attention, achieving a detection mAP50 of 0.885 and a classification mAP50 of 0.801. The ArUco Marker-based method allows for precise wound area estimation with an error margin of less than 0.6%.