GastroVision
收藏资源简介:
GastroVision是由西北大学放射学系和印度理工学院古瓦哈提分校共同创建的一个多中心开放访问的胃肠道内窥镜图像数据集,包含来自挪威和瑞典两家医院的8000张图像,涵盖27个不同的类别,包括解剖标志、病理异常、息肉移除案例和正常发现。该数据集旨在通过提供大规模、精确标记和多样化的数据,解决临床实践中人工智能系统集成所面临的挑战,如数据可用性、偏见结果、数据质量和在不同分布下未见数据集上的表现不佳。GastroVision的应用领域包括开发用于胃肠道疾病检测和分类的人工智能算法,以及为计算机视觉和机器学习研究人员提供公平比较的基础,以推动胃肠道内窥镜研究的发展。
GastroVision is a multi-center open-access gastrointestinal endoscopy image dataset co-created by the Department of Radiology of Northwestern University and the Indian Institute of Technology Guwahati. It contains 8,000 images sourced from two hospitals in Norway and Sweden, covering 27 distinct categories including anatomical landmarks, pathological abnormalities, polypectomy cases, and normal findings. This dataset aims to address the challenges encountered in integrating artificial intelligence systems into clinical practice, such as limited data availability, biased outcomes, substandard data quality, and poor performance on out-of-distribution unseen datasets, by providing large-scale, accurately labeled and diverse data. Applications of GastroVision include developing artificial intelligence algorithms for gastrointestinal disease detection and classification, as well as providing a benchmark for fair comparison among computer vision and machine learning researchers to advance gastrointestinal endoscopy research.

- 1GastroVision: A Multi-class Endoscopy Image Dataset for Computer Aided Gastrointestinal Disease Detection西北大学放射学系 · 2023年



