Okumura Shintaro, Tsunoda Shigeru, Hisamori Shigeo, Kitano Shoichi, Ueno Kohei, Sakaguchi Masazumi, Yoshida Yu, Sakamoto Takashi, Yamamoto Takehito, Okamura Ryosuke, Kasahara Keiko, Maeda Masahiro, Hoshino Nobuaki, Itatani Yoshiro, Hida Koya, Obama Kazutaka
Department of Gastrointestinal Surgery, Kyoto University Hospital, 54 Shogoin Kawahara-Cho, Sakyo-Ku, Kyoto, 606-8507, Japan.
Surg Endosc. 2025 Sep 13. doi: 10.1007/s00464-025-12205-2.
Although the high accuracy of artificial intelligence (AI) for recognizing surgical anatomy has been reported, its effective usage remains unclear. In this study, we investigated the utility of AI in surgical education for medical students.
Fifth-grade medical students were recruited to investigate the educational utility of EUREKA™. After an introductory lecture, they watched a video of distal gastrectomy with or without the suggestion of the connective tissue and the pancreas by EUREKA™ and then drew dissection lines in still images captured from the video. The distance between the lines drawn by students and the optimal dissection line determined by an expert surgeon was integrated to evaluate how well the students appropriately recognized the dissection line. Students filled out questionnaires after the study. A total of 45 operative video frames from radical gastrectomies performed with three different robotic systems were analyzed. The accuracy of the EUREKA™ recognition of the connective tissue and the pancreas was assessed using Dice and Intersection over Union (IoU) as a measurement tool.
Twelve students participated in the study, and nine students drew dissection lines. All students completed questionnaires. The students could recognize dissection lines more appropriately with the EUREKA™ suggestion, and the deviations between the dissection lines drawn by the students and the optimal dissection lines were significantly reduced. From the questionnaires completed by the students, eight students agreed with the possibility of AI to facilitate their understanding of the operation, and two students agreed with the potential of AI to increase the number of medical students who choose gastrointestinal surgery as their career. There were no differences in the DICE and IoU scores of the connective tissue and the pancreas between the three robotic systems, suggesting the versatility of the EUREKA™ system.
AI may facilitate students' understanding of surgery.
尽管已有报道称人工智能（AI）在识别手术解剖结构方面具有很高的准确性，但其有效应用仍不明确。在本研究中，我们调查了AI在医学生外科教育中的效用。
招募五年级医学生以研究EUREKA™的教育效用。在一次入门讲座后，他们观看了有或没有EUREKA™提示结缔组织和胰腺的远端胃切除术视频，然后在从视频中捕获的静态图像中绘制解剖线。将学生绘制的线与专家外科医生确定的最佳解剖线之间的距离进行整合，以评估学生对解剖线的识别程度。学生在研究结束后填写问卷。对使用三种不同机器人系统进行的根治性胃切除术的45个手术视频帧进行了分析。使用Dice和交并比（IoU）作为测量工具评估EUREKA™对结缔组织和胰腺的识别准确性。
12名学生参与了研究，9名学生绘制了解剖线。所有学生都完成了问卷。在EUREKA™的提示下，学生能够更准确地识别解剖线，学生绘制的解剖线与最佳解剖线之间的偏差显著减小。从学生填写的问卷中，8名学生同意AI有助于他们理解手术，2名学生同意AI有潜力增加选择胃肠外科作为职业的医学生数量。三种机器人系统之间结缔组织和胰腺的DICE和IoU分数没有差异，表明EUREKA™系统具有通用性。
AI可能有助于学生理解手术。
