Using Computer Vision and Artificial Intelligence to Track the Healing of Severe Burns.

The article focuses on the development and testing of the Skin Abnormality Tracking Algorithm (SATA), a machine learning and computer vision-based pipeline designed to assess burn severity and monitor wound healing using standard 2D color images captured by mobile devices. SATA employs a convolutional neural network (CNN) to classify burn depth into superficial partial-thickness (SPT) and deep partial-thickness (DPT) burns, and uses a novel Boundary Attention Mapper (BAM) for high-resolution wound segmentation, combined with fiducial markers for accurate spatial measurement and colorimetric analysis based on a red–yellow–black–white tissue model. Tested on a single patient with a deep partial-thickness burn over a 6-week clinical period and 2 weeks of remote home monitoring, SATA demonstrated strong correlation with clinician assessments in wound size measurement and provided quantitative tracking of wound color changes relevant to healing. The study highlights SATA's potential as a low-cost, accessible tool for standardized, objective burn wound evaluation both in clinical and remote settings, while noting the need for further validation on larger datasets and refinement in color differentiation.