Surgical Phase Recognition

Surgical phase recognition (SPR) aims to automatically identify the sequential stages of a surgical procedure from video data, improving surgical workflow analysis and computer-assisted interventions. Current research heavily utilizes deep learning, focusing on transformer-based architectures and state-space models like Mamba, often incorporating multi-scale temporal modeling and attention mechanisms to capture both short- and long-term dependencies within surgical videos. This field is significant for enhancing surgical training, enabling real-time intraoperative planning and risk assessment, and ultimately improving patient outcomes through more efficient and safer procedures.