New deep learning models excel in organ donation image segmentation

Deep learning models, detectron2 and yolov8, significantly improved automated segmentation in organ donation photography, achieving internal validation Intersection over Union (IoU) scores of 0.93 and 0.94 for kidneys, and 0.97 for livers. Compared to traditional methods, which peaked at IoU 0.54, these new models demonstrated superiority in accuracy and speed, completing tasks in just 0.13-1.5 seconds per image. This advancement could transform surgical photography practices.