The MIDOG challenge allows the use of all publicly available datasets for training or model selection. However, we provide some notable datasets that participants should be aware of:
The MIDOG++ dataset contains mitosis annotations for 11,937 mitotic figures across 503 individual tumor cases, spanning 7 different tumor types. It is the largest dataset of its kind to date.
The MITOS_WSI_CMC dataset contains whole slide image (WSI) annotations for mitosis in canine breast cancer. In total, 13,907 mitotic figure annotations are provided across 21 WSIs.
The MITOS_WSI_CCMCT dataset contains whole slide image annotations for mitosis in canine mast cell tumor. In total, 44,880 mitotic figure objects were annotated across 32 whole slide images.
The AMi-Br dataset containts atypical mitotic figure subclassification for the MIDOG 2021 and the TUPAC16 challenge datasets. The dataset provides the original coordinates but also patches extracted around the objects. In total, 3,720 annotations are provided, split into 832 atypical and 2,888 normal mitotic figures.
The MIDOG 2025 Atypical Training Set contains atypical mitotic figure subclassification for the entire MIDOG++ dataset, encompassing 11,939 mitotic figures from all 7 domains of MIDOG++ for which we provide a three expert blinded majority vote result.
Downloads are available from multiple sources:
– deepmicroscopy.org [images] [labels]
– google drive [images+labels]
– zenodo: [images+labels]
Please note that is a partial overlap between this dataset an the Ami-Br dataset, in that the first 150 tumor cases were both sourced from the MIDOG 2021 training set.
Zhuoyan Shen et al., who are participants in MIDOG 2025, have made available 3024 pathologist-annotated atypical and normal mitoses.
Their dataset can be found on zenodo: [images+labels]