Angelou0516/MoNuSeg

MoNuSeg（多器官核分割）数据集包含H&E染色的组织病理学图像（来自TCGA WSIs，放大倍数为40x），并附有每个核的二进制分割掩码。该数据集用于MICCAI 2018挑战赛。数据集包含37个训练样本和14个测试样本，共51个样本。图像大小为1000x1000 RGB，包含8种器官类别（未知、乳腺、肾脏、肝脏、前列腺、膀胱、结肠、胃）。测试样本还包括肺和脑（标记为未知）。数据集的地面真实数据是由单个注释者提供的语义二进制掩码（0=组织，1=核），通过OR组合所有核实例多边形得到。数据集的列包括患者ID、器官标签、图像、掩码和核数量。数据集的来源是RationAI/MoNuSeg parquet mirror，许可证为CC BY-NC-SA 4.0。

The MoNuSeg (Multi-Organ Nucleus Segmentation) dataset contains H&E-stained histopathology images (from TCGA WSIs at 40x magnification) with per-nucleus binary segmentation masks. It was used for the MICCAI 2018 challenge. The dataset includes 37 training samples and 14 test samples, totaling 51 samples. The image size is 1000x1000 RGB, with 8 organ classes (Unknown, Breast, Kidney, Liver, Prostate, Bladder, Colon, Stomach). The test set also includes lung and brain (labeled as Unknown). The ground truth is a single-annotator semantic binary mask (0=tissue, 1=nucleus), derived by OR-combining all per-nucleus instance polygons. The dataset columns include patient ID, organ label, image, mask, and nucleus count. The source of the dataset is RationAI/MoNuSeg parquet mirror, and the license is CC BY-NC-SA 4.0.