graph-tcga-brca
收藏资源简介:
Graph-TCGA-BRCA 是一个基于 TCGA-BRCA 组织病理学数据集的图级分类数据集,专为乳腺癌细粒度亚型分类设计。该数据集将 224x224 的病理图像块转换为细胞图(cell-graph),其中节点代表检测到的细胞核,边编码空间邻近关系。节点特征包含细胞形态、纹理和颜色强度信息,边特征则为微米级的欧几里得距离。数据集包含 11,149,500 个图和 2 个临床相关类别(浸润性导管癌和浸润性小叶癌),每个图平均具有 96 维节点特征和 1 维边特征。该数据集适用于图神经网络在乳腺癌病理分析中的自监督学习和分类任务研究,采用 PyTorch Geometric 数据格式存储,并遵循 CC BY-NC-SA 4.0 许可协议。
Graph-TCGA-BRCA is a graph-level classification dataset based on the TCGA-BRCA histopathology dataset, purpose-built for fine-grained subtyping of breast cancer. This dataset converts 224×224 histopathological image patches into cell-graphs, where nodes represent detected cell nuclei and edges encode spatial proximity relationships. Node features include information on cell morphology, texture, and color intensity, while edge features are micron-scale Euclidean distances between corresponding nodes. The dataset comprises 11,149,500 graphs and 2 clinically relevant categories, namely invasive ductal carcinoma and invasive lobular carcinoma. On average, each graph has 96-dimensional node features and 1-dimensional edge features. This dataset is suitable for research on self-supervised learning and classification tasks of graph neural networks in breast cancer histopathological analysis, is stored in the PyTorch Geometric data format, and complies with the CC BY-NC-SA 4.0 license.



