Year: 2016
Type: article
Abstract: You have accessJournal of UrologySurgical Technology & Simulation: Training & Skills Assessment II1 Apr 2016MP20-08 COMPUTER-GENERATED ASSESSMENT OF TECHNICAL SURGICAL SKILLS (CATS) Thomas Osinsk, Devansh Arpit, Safwan Wshah, and Ahmed Ghazi Thomas OsinskThomas Osinsk More articles by this author , Devansh ArpitDevansh Arpit More articles by this author , Safwan WshahSafwan Wshah More articles by this author , and Ahmed GhaziAhmed Ghazi More articles by this author View All Author Informationhttps://doi.org/10.1016/j.juro.2016.02.2777AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookTwitterLinked InEmail INTRODUCTION AND OBJECTIVES Minimal invasive surgery (MIS) is playing an increasing role in surgery, inherent to their design; the procedure is easily captured by the cameras and can be stored for review offline. Videos captured in-vivo during a surgical procedure are often post-analyzed in order to evaluate the quality, identify errors, assess surgeons skill level, and to provide feedback. This analysis is currently done manually, requiring many hours of laborious inspection by expert surgeons or trained operators. We introduce an automated computer vision system that is able to automatically augment human review. METHODS A simulated task was chosen for initial software validation. 13 participants completed a vesicourethral anastomosis (VUA) using the daVinci robot on a dry-lab physical model (3-Dmed Franklin, Ohio). 8 critical segments that convey the quality of the procedure were chosen for analysis (3, 6, 9 & 12 O’clock sutures at both bladder and urethral components). 3 skilled urologists graded films in a blinded fashion using needle positioning, needle entry and needle driving components of a modified Objective Structured Assessment of Technical Skill scale. Each component was scored. Using similar scoring scale the CATS software graded these actions using deep learning algorithms. The key idea behind these algorithms is to automatically discover the underlying patterns for any given data, or in other words, automatic representation learning. RESULTS A total of 109 surgical action videos with expert rating were analyzed. The software was trained on a randomly chosen 80% split of this data and test on the rest. CATS software was able to assess the simulated performance at an accuracy of 86.37% vs expert assessment of the ground truth data. CONCLUSIONS The developed software demonstrated high rates of accuracy compared to the expert ratings. Further validation using live surgery is required to further enhance the benefits of reduction in the time, cost, and labor involved in evaluating surgical videos. © 2016FiguresReferencesRelatedDetails Volume 195Issue 4SApril 2016Page: e215 Advertisement Copyright & Permissions© 2016MetricsAuthor Information Thomas Osinsk More articles by this author Devansh Arpit More articles by this author Safwan Wshah More articles by this author Ahmed Ghazi More articles by this author Expand All Advertisement Advertisement PDF downloadLoading ...
Source: The Journal of Urology
Language: English
Subfield: Surgery
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2334793474