274. AUTOMATED SURGICAL STEP RECOGNITION FOR ROBOT-ASSISTED MINIMALLY INVASIVE ESOPHAGECTOMY USING ARTIFICIAL INTELLIGENCE. (24th September 2022)

• Record Type: Journal Article
• Authors: Takeuchi, Masashi; Kawakubo, Hirofumi; Matsuda, Satoru; Kitagawa, Yuko
• Is Part Of: Diseases of the esophagus. Volume 35(2022)Supplement 2
• Publication Date: 2022-09-24
• Subjects: Esophagus -- Diseases -- Periodicals; 616.32
• DOI: 10.1093/dote/doac051.274
• Languages: English
• ISSNs: 1120-8694

Abstract

Abstract: More robot-assisted minimally invasive esophagectomy (RAMIE) procedures have been performed owing to three-dimensional field of view, image stabilization, and flexible joint function, however, both the surgeon, and surgical team require proficiency. Therefore, assessing the surgical learning curve and determining the surgical difficulty are essential for reducing complications. This study aimed to establish an artificial intelligence (AI)-based automated surgical phase recognition system for RAMIE by analyzing robotic surgical videos and evaluate the learning curve. This study enrolled 31 patients who underwent RAMIE. We annotated the videos into the following nine surgical phases, namely preparation, lower mediastinal dissection, upper mediastinal dissection, azygos vein division, subcarinal lymph node dissection (LND), right recurrent laryngeal nerve (RLN) LND, left RLN LND, esophageal transection, and post-dissection to completion of surgery to train the AI for automated phase recognition. We divided all patients into two groups, namely early period (20 patients) and late period (11 patients), after which the relationship between the surgical phase duration and learning curve was assessed. Frames were extracted from each video at a rate of one frame per second (fps), with an average of 14517 ± 4770 frames per video, which were used to train and test the AI model. Four-fold cross validation was applied to evaluate the performance of the present model. The AI had an accuracy of 84%. The preparation (p = 0.012), post-dissection to completion of surgery (p = 0.003), and 'no step' (p < 0.001) phases predicted by the AI were significant shorter in the late period than in the early period. We established an automated surgical phase recognition system for RAMIE using deep learning. Specific phase durations were significantly associated with surgeons' learning curve, and similar tendencies were observed on phase durations predicated by the AI. Our AI-based system can be useful for not only education or OR efficiency but also evaluating surgical skill.