Sensor-free computer-vision hand-motion entropy and video-analysis of technical performance during open surgery on fresh cadavers: Report of methodology and analysis

Computer vision (CV) hand motion analysis, video-recorded instrument organization and use, and handactivity summarized by Shannon Joint Entropy (variations in seemingly random movements), have potential for unbiased evaluation of surgical technical performance. We hypothesized that Individual Procedure Performance Scores (IPS), as determined by co-located, trained observers, and hand-motion analysis show congruence among operators by levels and types of surgical training and experience. Surgical residents, expert trauma surgeons and anatomists were video-recorded performing axillary artery exposure and control (AA) on fresh cadavers. Color-coded gloves permitted motion-tracking and CV hand motion analysis. Timing and instrument-use metrics were obtained through video reviews. IPS, idle and active times, elapsed time to division of pectoralis minor, type and number of instruments used, and calculated Shannon Joint Entropy values discriminated among the three groups of participants. Experts had lowest entropy values, idle and active times, and shortest time to divide pectoralis minor, using fewer instruments. Residents improved with training, not reaching expert levels, but showed deterioration 12-18 months later. Anatomists' results differed substantially from expert and resident surgeons. Surgeon's IPS scores mirrored the results of hand motion analysis. Conclusions Hand motion analysis techniques, which were congruent with IPS evaluations, can discriminate levels of surgical skill and training.