Surgical Activity Recognition

Surgical activity recognition aims to automatically identify and predict actions performed during surgical procedures using computer vision and machine learning. Current research focuses on improving the accuracy and generalizability of these systems, addressing challenges like open-set scenarios (handling unseen activities) and multi-centric variations in surgical techniques, often employing deep learning models such as transformers and 3D convolutional networks, sometimes incorporating spatio-temporal reasoning and human gaze data for enhanced performance. This field is crucial for advancing robotic surgery, improving surgical training, and developing intelligent tools for real-time intraoperative assistance and workflow optimization.