Untitled Item

The MIDOG++ dataset represents an extension of the data set used in the MIDOG 2021 and 2022 challenges. We provide region of interest images from 503 histological specimens of seven different tumor types with variable morphology: breast carcinoma, lung carcinoma, lymphosarcoma, neuroendocrine tumor, cutaneous mast cell tumor, cutaneous melanoma, and (sub)cutaneous soft tissue sarcoma. The human and canine samples were processed and stained at different human and veterinary pathology laboratories with standard H&E dye and digitized with different digital whole slide image scanners. We provide labels for 11,937 mitotic figures that have been differentiated against 14,351 imposter cells in a blinded consensus by two pathologists and a final decision by a third pathologist for disagreed labels.

MIDOG++数据集是MIDOG 2021与MIDOG 2022挑战赛所用数据集的扩展版本。本数据集提供了来自7种形态学特征各异的肿瘤类型的503份组织学标本的感兴趣区域（Region of Interest, ROI）图像，涵盖乳腺癌、肺癌、淋巴肉瘤、神经内分泌肿瘤、皮肤肥大细胞瘤、皮肤黑素瘤以及皮下软组织肉瘤。上述人类与犬类样本由多家人类病理与兽医病理实验室采用标准苏木精-伊红（Hematoxylin and Eosin, H&E）染色剂完成处理与染色，并通过多款数字全切片病理扫描设备完成数字化。我们为11937个有丝分裂象（mitotic figure）提供了标注，该标注的生成流程为：两名病理学家通过盲法共识将这些有丝分裂象与14351个冒充细胞（即疑似有丝分裂的非有丝分裂细胞）进行区分；对于标注意见存在分歧的样本，则由第三名病理学家作出最终判定。