Abstract
Breast cancer is the most common invasive cancer in women, affecting morethan 10% of women worldwide. Microscopic analysis of a biopsy remains one ofthe most important methods to diagnose the type of breast cancer. This requiresspecialized analysis by pathologists, in a task that i) is highly time- andcost-consuming and ii) often leads to nonconsensual results. The relevance andpotential of automatic classification algorithms using hematoxylin-eosinstained histopathological images has already been demonstrated, but thereported results are still sub-optimal for clinical use. With the goal ofadvancing the state-of-the-art in automatic classification, the Grand Challengeon BreAst Cancer Histology images (BACH) was organized in conjunction with the15th International Conference on Image Analysis and Recognition (ICIAR 2018). Alarge annotated dataset, composed of both microscopy and whole-slide images,was specifically compiled and made publicly available for the BACH challenge.Following a positive response from the scientific community, a total of 64submissions, out of 677 registrations, effectively entered the competition.From the submitted algorithms it was possible to push forward thestate-of-the-art in terms of accuracy (87%) in automatic classification ofbreast cancer with histopathological images. Convolutional neuronal networkswere the most successful methodology in the BACH challenge. Detailed analysisof the collective results allowed the identification of remaining challenges inthe field and recommendations for future developments. The BACH dataset remainspublically available as to promote further improvements to the field ofautomatic classification in digital pathology.