Large-scale pathology dataset for mitotic figure detection

We released a large-scale data set for mitotic figure detection as open data. Unlike previous data sets, it is the only data set that provides a complete annotation map for the histology whole slide image. It includes 44880 mitotic figures, verified in blind review by two expert pathologists. We were able to show that the data set has a high overall consistency.

The paper describing the data set was released in Scientific Data (SpringerNature) and is, as the data itself, open access and creative commons:

Another version of the data set is available as DICOM WSI image on kaggle here, including a jupyter notebook that shows how to do mitotic figure detection as an object detection task here