CAS-Colon: A Comprehensive Colonoscopy Anatomical Segmentation Dataset for Artificial Intelligence Development

CAS-Colon is a comprehensive colonoscopy video dataset comprising 78 high-resolution endoscopy videos (30GB in total, 9.08 hours duration) recorded at 60 fps. Each video is meticulously annotated with ten distinct anatomical regions: terminal ileum, cecum, ascending colon, hepatic flexure, transverse colon, splenic flexure, descending colon, sigmoid colon, rectum, and anal canal. All videos were collected using standard Olympus endoscopy equipment (CV-290) between March-July 2024. The dataset underwent rigorous quality control, with exclusion criteria including poor bowel preparation, unclear imaging, incomplete recording, and unclear anatomical boundaries. All videos were completely de-identified, making this dataset a valuable resource for developing AI applications in colonoscopy.

CAS-Colon是一款综合性结肠镜检查视频数据集，包含78段高分辨率内镜视频，总容量30GB，总时长9.08小时，录制帧率为60fps。每段视频均已被精细标注了10个不同的解剖部位：回肠末端（terminal ileum）、盲肠（cecum）、升结肠（ascending colon）、肝曲（hepatic flexure）、横结肠（transverse colon）、脾曲（splenic flexure）、降结肠（descending colon）、乙状结肠（sigmoid colon）、直肠（rectum）以及肛管（anal canal）。所有视频采集于2024年3月至7月，采用标准奥林巴斯（Olympus）CV-290型内镜设备录制。该数据集经过严格的质量管控，排除标准涵盖肠道准备不佳、成像模糊、录制不完整以及解剖边界不清等情况。所有视频均已完成完全去标识化处理，使得该数据集可作为开发结肠镜检查领域人工智能（AI）应用的宝贵资源。

搜集汇总

背景与挑战

背景概述

CAS-Colon是一个全面的结肠镜视频数据集，包含78个高分辨率内窥镜视频（总计30GB，时长9.08小时），每个视频都详细标注了10个不同的解剖区域。该数据集使用标准奥林巴斯内窥镜设备采集，经过严格质量控制，适用于开发结肠镜AI应用。