Mp07-17 Is AI Seeing What I’m Seeing? Validation of Touch Surgery’s Enterprise System in Calculating Warm Ischemia Time During Robotic-Assisted Partial Nephrectomy

You have accessJournal of UrologySurgical Technology & Simulation: Artificial Intelligence I (MP07)1 May 2024MP07-17 IS AI SEEING WHAT I'M SEEING? VALIDATION OF TOUCH SURGERY'S ENTERPRISE SYSTEM IN CALCULATING WARM ISCHEMIA TIME DURING ROBOTIC-ASSISTED PARTIAL NEPHRECTOMY Adam Bobrowski, Chelsea Angeles, Simon Czajkowski, and Jason Y. Lee Adam BobrowskiAdam Bobrowski , Chelsea AngelesChelsea Angeles , Simon CzajkowskiSimon Czajkowski , and Jason Y. LeeJason Y. Lee View All Author Informationhttps://doi.org/10.1097/01.JU.0001008728.41882.d7.17AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: The adoption of robotic-assisted surgery (RAS) has underscored the need for comprehensive training in RAS for new and practicing surgeons alike. Despite the critical importance of timely feedback, implementation of coaching programs is hindered by a lack of standardization, scalability and human resources. Artificial intelligence (AI) poses a possible solution. Digital Surgery (Medtronic) has developed the Touch SurgeryTM Enterprise (TSE), a novel product that seeks to provide video-based coaching and analysis of RAS videos utilizing AI. As a first step in validating accuracy of this AI algorithm, we sought to assess its ability of calculating warm ischemia time (WIT) in robotic-assisted partial nephrectomy (RAPNx). METHODS: We conducted a retrospective review of all patients who underwent a RAPNx for a solitary tumor at our institution between November 2019 – November 2023, with complete video recordings of their surgeries uploaded on the TSE platform. Once uploaded to the TSE platform, videos are automatically anonymized by the RedactORTM algorithm and broken down into procedural phases by recognition AI models for RAPNx. This provides detailed information on surgical workflow, operative times as well as the total WIT. The TSE computed WIT was compared to both the dictated WIT in the operative notes, as well as the WIT calculated on independent review of each surgical recording (reference standard). RESULTS: A total of 32 patients had complete data for analysis. All RAPNx cases were performed by a single surgeon – 25 daVinci Si/Xi, 7 HUGOTM RAS. Mean age was 52.6 years, 11 (34.4%) were female, and average WIT was 20.4mins. The majority (28/32) had RCC with a clear cell variant being most common (78.6%) and were pT1a (78.6%). The mean difference between TSE computed WIT and surgeon dictated WIT was 155.8 sec (SD=235.3 sec). Compared to the reference standard, the mean difference in WIT computed by TSE and as dictated in operative note was 68.6 sec (SD=159.3 sec) and 120.0 sec (SD=213.3 sec), respectively. For both TSE WIT (87.5%) and dictated WIT (59.4%), the majority of cases were within 1min of the reference standard WIT. There were 4 cases where TSE computed WIT and 3 cases where dictated WIT was>5 mins off the reference standard. CONCLUSIONS: Our study demonstrated that the AI algorithm embedded within the TSE system can provide WIT more accurately than surgeon documented times during RAPNx, with the exception of a few variant cases. This project acts as a first step in validating this algorithm in providing objective performance data from RAPNx recordings. Our group aims to validate whether this algorithm can accurately assess other aspects of this and other urological surgeries. Source of Funding: None © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e111 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Adam Bobrowski More articles by this author Chelsea Angeles More articles by this author Simon Czajkowski More articles by this author Jason Y. Lee More articles by this author Expand All Advertisement PDF downloadLoading ...