Pd36-08 Artificial Intelligence-Based Surgical Phase Recognition in Robot-Assisted Radical Prostatectomy and Cross-Surgeon External Validity Verification

You have accessJournal of UrologySurgical Technology & Simulation: Artificial Intelligence III (PD36)1 May 2024PD36-08 ARTIFICIAL INTELLIGENCE-BASED SURGICAL PHASE RECOGNITION IN ROBOT-ASSISTED RADICAL PROSTATECTOMY AND CROSS-SURGEON EXTERNAL VALIDITY VERIFICATION Yuichiro Konnai, Keishiro Fukumoto, Masashi Takeuchi, Mio Tanigawa, Rei Takeuchi, Yota Yasumizu, Nobuyuki Tanaka, Toshikazu Takeda, Kazuhiro Matsumoto, Shinya Morita, Takeo Kosaka, Ryuichi Mizuno, Hiroshi Asanuma, and Mototsugu Oya Yuichiro KonnaiYuichiro Konnai , Keishiro FukumotoKeishiro Fukumoto , Masashi TakeuchiMasashi Takeuchi , Mio TanigawaMio Tanigawa , Rei TakeuchiRei Takeuchi , Yota YasumizuYota Yasumizu , Nobuyuki TanakaNobuyuki Tanaka , Toshikazu TakedaToshikazu Takeda , Kazuhiro MatsumotoKazuhiro Matsumoto , Shinya MoritaShinya Morita , Takeo KosakaTakeo Kosaka , Ryuichi MizunoRyuichi Mizuno , Hiroshi AsanumaHiroshi Asanuma , and Mototsugu OyaMototsugu Oya View All Author Informationhttps://doi.org/10.1097/01.JU.0001008916.72488.6a.08AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: The application of Artificial Intelligence (AI) in various fields has been advancing, though its practical utilization in surgery is still in nascent stages. We developed an AI designed to automatically recognize surgical phases in robot-assisted radical prostatectomy (RARP) and verified its external validity using surgical videos from different surgeons. METHODS: We analyzed 102 RARP cases, comprising 81 consecutive cases from one surgeon and 21 from five other surgeons. Sixty-five out of 81 cases were used for AI development, while the remaining 16 and 21 cases from the first and other surgeons, respectively, validated the AI's accuracy and external validity. We classified surgical operations into nine phases (Table 1). Each video was annotated with information about which time corresponds to which phase independently by 2 well-trained surgeons. We used TeCNO as the AI development model and evaluated the AI using precision true positive/(true positive+false positive) and recall true positive/(true positive+false negative). RESULTS: For AI development, 919,231 frames were utilized. Testing involved 216,357 frames from the same surgeon and 249,553 frames from five different surgeons. Using the developed AI to analyze surgical videos from the same surgeon, out of nine phases, precision exceeded 90% in five phases and recall exceeded 90% in eight phases, with an overall precision of 94% and recall of 93% (Figure 1). In contrast, when applying the AI to videos from different surgeons, at least 80% precision was achieved in five phases and at least 80% recall in five phases, with both overall precision and recall reaching 83%. CONCLUSIONS: The AI we developed not only showed high accuracy, but was also able to maintain this accuracy when adapted to different surgeons. By providing deeper evaluations of surgical videos, our AI can significantly contribute to the quality assessment of surgeries, offering valuable feedback to surgeons and enhancing the effectiveness of surgical education. Download PPT Source of Funding: None © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e795 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Yuichiro Konnai More articles by this author Keishiro Fukumoto More articles by this author Masashi Takeuchi More articles by this author Mio Tanigawa More articles by this author Rei Takeuchi More articles by this author Yota Yasumizu More articles by this author Nobuyuki Tanaka More articles by this author Toshikazu Takeda More articles by this author Kazuhiro Matsumoto More articles by this author Shinya Morita More articles by this author Takeo Kosaka More articles by this author Ryuichi Mizuno More articles by this author Hiroshi Asanuma More articles by this author Mototsugu Oya More articles by this author Expand All Advertisement PDF downloadLoading ...