Anand Malpani, BTech, Balazs Vagvolgyi, David D Yuh, MD, Chi Chiung Grace Chen, MD, Hiep T Nguyen, MD, Rajesh Kumar, PhD. Johns Hopkins University, Yale University School of Medicine, Children’s Hospital, Boston
Introduction: Robotic surgery training methods are currently being developed using both real models and simulators. Training tasks, models, and metrics are all being developed and reported. However, published research quantifies skill as a monolithic quantity and do not associate metrics with specific aspects of robotic surgery (e.g. surgical skills, or man-machine interaction aspects) or specific aspects of skill (e.g. a proficient technique or safe approach).
Methods: We describe analysis of data collected from a novel automated, objective framework in robotic surgery. Using a new portable recording system compatible with the da Vinci surgical system, we have acquired longitudinally (approximately monthly) stereo instrument video, time-stamped and synchronized with instrument and hand motion data from several benchmarking training tasks (e.g. suturing) from 28 subjects (trainee and expert surgeons, with none to more than 20 years of medical experience) over an approximately 2 year period. Here we report on new analysis of the collected data aimed at detecting a subject’s ability to perform safe manipulation, as well style aspects of safe manipulation across trainee and expert subjects.
Using the calibration parameters for the endoscopic cameras, the motion of the instruments was projected into the camera workspace with accuracy greater than 40 pixels, sufficient to overlay the motion vector visually on the instrument shaft. The instrument motion outside the field of view (unseen instrument motion, or placement of instruments outside the field of view) in the integrated motion and video data was then automatically segmented. This unseen motion then was analyzed with respect to its distance from the field of view, cumulative motion statistics, and for comparison across the two classes. In addition, the image zoom and camera field of view were also assessed for each skill class.
Results: Results show that experts possess a much greater situational awareness. While the cumulative distance moved by the instruments outside of field of view – which is commonly used to assess safe motion – was greater for the experts (e.g. experts 0.7m versus trainees 0.2m in one test), the envelope of such motion around the field of view was smaller (as was the standard deviation), compared to the trainees. Therefore, unseen instrument motion alone was not representative of skill in our data. The location of such unseen with respect to the field of view was also important. This finding was confirmed by measuring the average working distance maintained by the subjects (experts << trainees), resulting in experts often moving or placing instrument outside of the smaller field of view, but in “safe” or “known” areas.
Conclusions: We report on assessment of unseen instrument motion as a measure of safety during robotic surgery training. We are now incorporating the location of the motion into the unseen/unsafe motion measurements to create better indicators of manipulation skill. Such measures will weigh the unsafe motion according to time (greater being worse), and distance (greater being worse) from the field of view of the camera.
Session Number: Poster – Poster Presentations
Program Number: P583
View Poster