Scholarlink

<p>Subepithelial lesions (SELs) of the gastrointestinal tract encompass a heterogeneous spectrum of histology, ranging from benign to malignant. Their detection during colonoscopy is critical yet challenging, as endoscopic interpretation is highly operator-dependent with variable diagnostic performance. Artificial intelligence (AI) offers great promise for assisting endoscopists in SEL detection and assessing, yet progress has been limited by the lack of large, diverse, and well-annotated datasets. To address this, we present CAD-SEL dataset, the first publicly available dual-modality colonoscopy dataset for SEL detection. The dataset comprises 4,912 images from 641 patients, collected across six medical centers in China, with paired white light endoscopy and endoscopic ultrasound images. Expert endoscopists provided histology-based multi-class annotations and clinically relevant binary classifications (potentially malignant versus benign) ensuring both research flexibility and AI-driven applicability. Initial technical validation with several state-of-the-art object detection models demonstrates the dataset’s potential to support AI-based computer-assisted detection systems. CAD-SEL dataset provides a high-quality open-access benchmark to advance AI-assisted SEL analysis, support reproducibility, and facilitate broader clinical translation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CAD-SEL: A dual-modal colonoscopy dataset of subepithelial lesion

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月4日

查看全文
  • Louzhe Xu,
  • Shurong Chen,
  • Chunxiao Li,
  • Jingjing Zhou,
  • Xiaohan Shen,
  • Jinghua Wang,
  • Xinjue He,
  • Tianlian Yan,
  • Han Ma,
  • Yizhen Zhang,
  • Lan Li,
  • Shu Bian,
  • Yue Huang,
  • Chao Lu,
  • Yangyang Xiong,
  • Jiang Liu,
  • Qiang Chen,
  • Lina Ye,
  • Qiaona He,
  • Yini Ke,
  • Youming Li,
  • Xinli Mao,
  • Chaohui Yu,
  • Ting Li,
  • Yi Chen

摘要

Subepithelial lesions (SELs) of the gastrointestinal tract encompass a heterogeneous spectrum of histology, ranging from benign to malignant. Their detection during colonoscopy is critical yet challenging, as endoscopic interpretation is highly operator-dependent with variable diagnostic performance. Artificial intelligence (AI) offers great promise for assisting endoscopists in SEL detection and assessing, yet progress has been limited by the lack of large, diverse, and well-annotated datasets. To address this, we present CAD-SEL dataset, the first publicly available dual-modality colonoscopy dataset for SEL detection. The dataset comprises 4,912 images from 641 patients, collected across six medical centers in China, with paired white light endoscopy and endoscopic ultrasound images. Expert endoscopists provided histology-based multi-class annotations and clinically relevant binary classifications (potentially malignant versus benign) ensuring both research flexibility and AI-driven applicability. Initial technical validation with several state-of-the-art object detection models demonstrates the dataset’s potential to support AI-based computer-assisted detection systems. CAD-SEL dataset provides a high-quality open-access benchmark to advance AI-assisted SEL analysis, support reproducibility, and facilitate broader clinical translation.