Song Yiming, Du Sijia, Wang Ruilan, Liu Fei, Lin Xiaolu, Chen Jinnan, Li Zeyu, Li Zhao, Yang Liuyi, Zhang Zhengjie, Yan Hao, Zhang Qingwei, Qian Dahong, Li Xiaobo
Division of Gastroenterology and Hepatology, Shanghai Institute of Digestive Disease, NHC Key Laboratory of Digestive Diseases, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China.
School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China.
Sci Data. 2025 May 31;12(1):918. doi: 10.1038/s41597-025-05251-x.
Colorectal cancer often arises from precancerous polyps, where accurate size assessment is vital for clinical decisions but challenged by subjective methods. While artificial intelligence (AI) has shown promise in improving the accuracy of polyp size estimation, its development depends on large, meticulously annotated datasets. We present Polyp-Size, a dataset of 42 high-resolution white-light colonoscopy videos with polyp sizes precisely measured post-resection using vernier calipers to submillimeter precision. Unlike existing datasets primarily focused on polyp detection or segmentation, Polyp-Size offers validated size annotations, diverse polyp features (Paris classification, anatomical location and histological type), and standardized video formats, enabling robust AI models for size estimation. By making this resource publicly available, we aim to foster research collaboration and innovation in automated polyp measurement to ultimately improve clinical practice.
结直肠癌通常起源于癌前息肉，准确的大小评估对于临床决策至关重要，但主观方法对此提出了挑战。虽然人工智能（AI）在提高息肉大小估计的准确性方面已显示出前景，但其发展依赖于大量精心注释的数据集。我们展示了Polyp-Size，这是一个包含42个高分辨率白光结肠镜检查视频的数据集，息肉大小在切除后使用游标卡尺精确测量至亚毫米精度。与现有主要专注于息肉检测或分割的数据集不同，Polyp-Size提供了经过验证的大小注释、多样的息肉特征（巴黎分类、解剖位置和组织学类型）以及标准化的视频格式，从而能够构建用于大小估计的强大AI模型。通过公开提供此资源，我们旨在促进自动息肉测量方面的研究合作与创新，以最终改善临床实践。
