bcTCGA

该数据集包含了来自六种不同癌症类型的H&E染色组织病理学切片，旨在评估模型在不同数据集之间的相似性和可迁移性。此外，该数据集还用于根据最优传输距离定义组织病理学数据集之间的一种分层距离概念。该数据集规模宏大，每张切片包含多个瓦片，其任务是对癌症类型进行预测以及评估模型的可迁移性。This dataset comprises H&E-stained histopathological slides from six distinct cancer types. Its primary objectives include evaluating the similarity and cross-dataset transferability of machine learning models, as well as defining a hierarchical distance metric between histopathological datasets based on optimal transport distance. As a large-scale dataset, each slide contains multiple tiles, and the downstream tasks involve cancer type prediction and the assessment of model transferability.