Kolbinger Fiona R, Kirchner Max, Pfeiffer Kevin, Bodenstedt Sebastian, Jenke Alexander C, Barthel Julia, Carstens Matthias, Dehlke Karolin, Dietz Sophia, Emmanouilidis Sotirios, Fitze Guido, Freitag Martin, Holderried Fabian, Jacobi Thorsten, Kanjo Weam, Leitermann Linda, Mees Sören Torge, Pistorius Steffen, Prudlo Conrad, Seiberth Astrid, Schultz Jurek, Thiel Karolin, Ziehn Daniel, Speidel Stefanie, Weitz Jürgen, Kather Jakob Nikolas, Distler Marius, Saldanha Oliver Lester
Department of Visceral, Thoracic and Vascular Surgery, University Hospital and Faculty of Medicine Carl Gustav Carus, TUD Dresden University of Technology, Dresden, Germany.
Else Kroener Fresenius Center for Digital Health, Faculty of Medicine and University Hospital Carl Gustav Carus, TUD Dresden University of Technology, Dresden, Germany.
medRxiv. 2025 Sep 7:2025.09.05.25335174. doi: 10.1101/2025.09.05.25335174.
The limited availability of diverse and representative training data poses a critical barrier to the development of clinically relevant computational tools for intraoperative surgical decision support. As surgical interventions are not routinely recorded and data annotation often requires domain knowledge, existing open-access surgical video datasets with high-quality annotations are available only for a restricted number of surgical procedures, such as cholecystectomy, and typically originate from a single institution. Appendix300 is a comprehensive dataset of video footage from 330 surgical procedures, including 312 full-length recordings of laparoscopic appendectomies in pediatric and adult patients treated at five German centers, along with 18 control recordings of non-inflamed appendices collected during non-appendectomy laparoscopic surgeries. Clinical metadata for each recording includes patient demographics, medical history, clinical symptoms, preoperative laboratory parameters, and postoperative histopathological findings. Additionally, we provide annotations of the grade of appendicitis based on the intraoperative presentation. This dataset enables new and clinically relevant validation tasks for computer vision in laparoscopic surgery, thereby enhancing the breadth and translational relevance of AI-based surgical video analysis.
多样且具代表性的训练数据有限，这对开发用于术中手术决策支持的临床相关计算工具构成了关键障碍。由于手术干预并非常规记录，且数据标注通常需要领域知识，现有的带有高质量标注的开放获取手术视频数据集仅适用于有限数量的手术程序，如胆囊切除术，并且通常来自单一机构。附录300是一个包含330个手术程序视频片段的综合数据集，其中包括在德国五个中心治疗的儿科和成年患者的312例腹腔镜阑尾切除术的全长记录，以及在非阑尾切除腹腔镜手术期间收集的18例非炎症阑尾的对照记录。每个记录的临床元数据包括患者人口统计学、病史、临床症状、术前实验室参数和术后组织病理学结果。此外，我们还根据术中表现提供阑尾炎分级的标注。该数据集为腹腔镜手术中的计算机视觉带来了新的且与临床相关的验证任务，从而增强了基于人工智能的手术视频分析的广度和转化相关性。
