Autonomous Non-Technical Surgical Skill Assessment and Workload Analysis in Laparoscopic Cholecystectomy Training

Despite the fact that Minimally Invasive Surgery (MIS) offers undeniable clinical benefits, such as less tissue damage, smaller scars and faster recovery, it requires extensive training from the surgeons, including technical and non-technical skills. Coping with stress and distractions, maintinaing situation awareness, prompt decision making, advanced communication, leadership and teamwork are all essential in MIS. Workload-which represents the human effort to perform a task-shows a strong correlation with non-technical skills. In this paper, a MIS training experiment is introduced, developed to autonomously assess non-technical surgical skills based on sensory data (im-age and force). For this, a surgical phantom and adequate workflow were designed to simulate a stressful laparoscopic cholecystectomy tasks, such as peritoneum dissection and cystic artery clipping. The experiment included the simulation of an abrupt situation (cystic artery bleeding). 20 training session were recorded from 7 subjects (3 non-medicals, 2 residents, 1 expert surgeon and 1 expert MIS surgeon). Analysis of the surgical workload and autonomous skill classification based on surgical tool tracking and force measurements were presented. Workload was tested for the two groups (medical and control) with the Surgical Task Load Index (SURG-TLX) workload assessment tool. Unpaired t-tests showed significant differences between the two groups in the case of mental demands, physical demands and situational stress (p <0.0001, 95 % confidence interval (CI)), and also in the case of task complexity (p <0.05). There were no significant differences in temporal demands and distraction levels. Learning curve in workload was studied with paired t-tests; only task complexity resulted significant difference between the first and the second trials. Autonomous non-technical skill classification was done based on image data with tracked instruments based on Channel and Spatial Reliability Tracker (CSRT) and force data. Time series classification was done by a Fully Convolutional Neural Network (FCN), which resulted high accuracy on temporal demands classification based on the $z$ component of the used forces (85 %) and 75 % accuracy for classifying mental demands/situational stress with the $x$ component of the used forces validated with Leave One Out Cross-Validation (LOOCV). It suggests there are non-technical skills and workload components which can be classified autonomously with objectively measured data.