胸腔镜术中意外事件（出血）识别模型数据集

胸腔镜手术出血事件识别是评估决策类医疗行为。数据均来自北京大学人民医院胸腔镜手术数据库，所有手术均由同一团队进行，从数据库中逐个筛选出包含出血的手术视频，对每个手术视频逐帧进行分析，进而筛选出出血发生的具体事件点，后利用计算机视觉算法对出血的特征进行识别，从而构建以术中出血为代表的识别模型。该数据集采集于研究开展期间，数据离线导出至项目专用硬盘。Video-assisted thoracoscopic surgery (VATS) bleeding event recognition is a medical decision-making behavior evaluation task. The dataset is sourced from the Video-assisted Thoracoscopic Surgery Database of Peking University People's Hospital. All surgeries included in this dataset were performed by the same surgical team. We first screened out surgical videos containing bleeding events one by one from the database, performed frame-by-frame analysis on each video, and identified the specific timestamps where bleeding occurred. Subsequently, computer vision algorithms were employed to extract and recognize the characteristics of bleeding, thereby constructing a recognition model targeting intraoperative bleeding as the representative endpoint. This dataset was collected during the study period, and the raw data was exported offline to a project-specific hard drive.