Instrument Segmentation

Instrument segmentation in surgical videos aims to automatically identify and delineate surgical instruments, improving surgical workflow and patient safety. Current research focuses on enhancing accuracy and robustness using deep learning models, including transformers and adaptations of the Segment Anything Model (SAM), often incorporating multimodal data (e.g., audio commands, kinematic data) to improve segmentation precision and address challenges like occlusions and variations in surgical scenes. These advancements have significant implications for computer-assisted surgery, enabling improved surgical guidance, automated analysis of surgical procedures, and enhanced surgical training.