Elham Rastegari1, Hesham Ali, PhD1, Carl Nelson, PhD2, Dmitry Oleynikov, MD3, Ka-Chun Siu, PhD3. 1University of Nebraska at Omaha, 2University of Nebraska – Lincoln, 3University of Nebraska Medical Center
Introduction: With the increased sophistication of medical instruments and the growing importance of their functionality, the problem of assessing the proficiency of operating such instruments has emerged as a critical problem in biomedical applications. In this work, we propose a new model to assess how well users operate surgical instruments and track their learning progress in operating these instruments. The proposed approach is based on using population analysis and graph modeling by assessing each individual’s proficiency as it relates to other users and tracking the progress of the proficiency level of each user over time. The study focuses on using the population analysis model with objectively-measured laparoscopic skills measures to better understand the learning patterns of skills acquisition using a virtual training system.
Methods: Participants were recruited from the local Medical Center in Omaha, Nebraska (4 professional surgeons and 6 medical students). They were asked to perform a set of training tasks (Peg Transfer, Needle Passing, Eye-Hand Coordination, and Matching) using the Portable Camera Aided Surgical Simulator, PortCASTM, once a week for four weeks. Once those tasks were performed, several kinematics based parameters were extracted, including time to task completion, average speed and smoothness of performance path. Then, a population analysis model that uses correlation networks was created for each training session. In these networks, every subject is represented by a node and the edge connecting two nodes represents the level of similarity in operating the instrument. The weight on each edge is obtained using the Pearson correlation coefficient calculated from the extracted parameter vector. In this study, a threshold of 95% correlation with the significance value of 0.05 was set to create the population analysis model.
Results: We performed clustering analysis on the correlation networks obtained from laparoscopic skills scores on the networks created for all training sessions. In the network obtained for the first week’s data, we can identify two separate clusters that represent two groups of subjects with different levels of laparoscopic surgical expertise, with the first group containing the expert surgeons and the second group containing the medical students. In the networks obtained from data collected in weeks 2-4, higher levels of correlations between members of the two groups can be observed, reflecting the fact that medical students are gaining higher levels of proficiency in operating the instruments. In the final week’s network, four students are connected to the expert group with only two students remaining in a separate cluster indicating that their proficiency level remains lower than the expert level in performing several tasks using PortCASTM.
Conclusions: This study provides promising evidence in favor of the innovative population analysis model to study and visualize the progress of learning fundamental laparoscopic skills (the changes of cluster formation). The model creates multiple correlation networks that clearly captures the change of learning progress in trainees across a period of training, and possibly identifies individuals in need of additional training. Additional analyses are warranted by using training tasks with different levels of complexity.
Presented at the SAGES 2017 Annual Meeting in Houston, TX.
Abstract ID: 98671
Program Number: ETP720
Presentation Session: Emerging Technology Poster Session (Non CME)
Presentation Type: Poster