Evaluation of Surgeons’ Performance on Three Robotic Surgical Systems

Minimally invasive surgery (MIS) has become a stan- dard technique due to its benefits over traditional open procedures, such as reduced infection risk and faster recovery. However, it presents challenges including lim- ited dexterity, poor depth perception, and steep learning curves . Robotic systems and computer-assisted surgery (CAS) have been introduced to address these limitations, offering enhanced control, precision, and visualization. Accurate 3D tracking of instruments during procedures is key to enabling effective feedback, dynamic constraints, and objective skill assessments. Traditional tracking meth- ods often require external hardware, while recent trends in computer vision enable image-based solutions using only surgical video data . This study proposes a methodology to estimate the 3D pose of surgical instruments using a visual segmentation model, Segment Anything 2 (SAM2) , in combination with feature tracking. The system extracts high-fidelity masks and applies the Perspective-n-Point (PnP) al- gorithm with known CAD geometries to estimate 3D instrument pose. Validation includes performance evalu- ation across three robotic systems: the da Vinci Surgical System, Hugo RAS System , and Versius Surgical Robotic System , during the anastomosis phase of a gastrectomy procedure.