Accurate Surface Area Measurement of Biological Tissues Based on an Improved EfficientSAM Model

The development of a non-invasive and robust three-dimensional optical imaging method enabled rapid and precise surface area measurements of flexible biological tissues through computer vision techniques. This paper intends to establish accurate and non-invasive measurement methodologies, which have become an urgent requirement for various clinical applications, including skin grafting. In this study, we explore the performance of the EfficientSAM model in wound segmentation tasks, employing targeted fine-tuning strategies to enhance its effectiveness. Key metrics like computational cost, accuracy of segmentation, generalization of the model, and robustness are taken into consideration. We also propose a novel loss function, achieving approximately a 24% improvement in Intersection over Union (IoU) scores, which can support precise surface area calculation and improve biomedical application outcomes.