2D Wound Image Segmentation Using Deep Learning Algorithms

The purpose of wound segmentation is to accurately identify and delineate wound areas in images, enabling precise analysis of metrics like size, shape, and spread. This aids healthcare professionals in assessing wound severity and monitoring treatment progress. This research focuses on developing an automated wound image segmentation tool using the Deep learning Algorithms. The objective is to enhance wound assessment accuracy by precisely identifying and segmenting wound areas in digital images. Three models are employed to find the effectiveness in segmenting the wound which includes U- Net, DeepLabV3+ and Mask-CNN. The deep learning models are trained on a massive dataset of wound pictures and masks; they then provide segmented outputs, which are examined for wound characteristics and severity. The U-Net model has obtained better performance in segmenting the wound with an accuracy of $\mathbf{9 2. 3} \%$. This work seeks to incorporate advanced capabilities into a user-friendly interface for practical use in clinical settings. Initial results demonstrate promising segmentation accuracy, with plans for further enhancements in future.