Association Between Video-Derived Digital Operative Metrics and GEARS Scores in Robotic Surgery and Development of an Exploratory Multivariable Model

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Background: Robotic-assisted surgery has been increasingly adopted across surgical specialties, but objective and efficient assessment of robotic surgical skills remains challenging. The Global Evaluative Assessment of Robotic Skills (GEARS) is commonly used for expert-based evaluation; however, it requires manual review and may not fully capture detailed operative processes. This study aimed to investigate the associations between video-derived digital operative metrics and expert-derived GEARS scores and to conduct an exploratory multivariable analysis of metrics associated with GEARS scores. Methods: Data were collected from 42 surgeons who completed a standardized robotic surgery training program. On the final training day, anonymized animal-operation videos were evaluated by blinded experts using the GEARS scale. Surgical videos were processed to extract 43 digital operative metrics related to instrument visibility, spatial positioning, camera control, clutch use, switching use, and motion characteristics. GEARS scores and digital operative metrics were compared according to annual laparoscopic surgical volume, including dichotomized volume groups, specialty-based subgroups, and volume tertiles. Stepwise multiple linear regression was used as an exploratory variable-selection approach to identify video-derived metrics associated with expert-derived GEARS scores. Results: After false discovery rate correction, no significant differences in GEARS scores or digital operative metrics were observed across annual laparoscopic surgical volume groups, specialty subgroups, or volume tertiles. The final regression model retained three predictors: mean camera-use count per minute, minimum distance from the left-side center point, and maximum distance from the right-side center point. The model had an apparent in-sample R2 of 0.509 (F = 9.588; p < 0.001). This value describes model fit within the present sample and has not been corrected for optimism. Conclusions: Video-derived digital operative metrics, particularly those reflecting camera control and instrument spatial distribution, were associated with expert-derived GEARS scores and may provide complementary quantitative information for expert-based robotic surgical skill assessment.