Global Versus Local Kinematic Skills Assessment on Robotic-Assisted Hysterectomies

Different methods have been proposed to evaluate surgical skills from observer-based scoring to recent data-driven approaches. However, most of these methods assess the surgical performance considering the procedure as a whole, avoiding detailed performance insights. In this study, we focused on the most challenging phases of robotic-assisted hysterectomies to compare the performance of expert and intermediate surgeons using the surgical process model methodology. We recorded surgical video and kinematic data of fifty-two robotic-assisted laparoscopic hysterectomies performed by five experts and three intermediate surgeons. We annotated the video in eight phases. We computed twenty-five automated performance metrics (APMs); seven for each of the right, left, and endoscope robotic arms, and four global ones. For the global analysis, only four APMs differed significantly between experts and intermediates. However, interpreting these APMs was difficult. For local analysis, we observed that 23 APMs were significantly different for at least one phase. We found that the two most challenging phases had APMs that highlighted difficulty due to the presence of the uterus, lack of confidence in anatomical knowledge, and difficulty in moving the endoscope. Such results of the local analysis allow us to propose appropriate training for surgeons.