Join

BACH's dataset is publicly available under the CC BY-NC-ND license. The citation should refer to 10.1016/j.media.2019.05.010. Teams must notify the organisers of the challenge about any publication that is (partly) based on the results or data published on this site, in order to the challenge organization being able to maintain a list of publications associated with the challenge.

Motivation¶

Breast cancer is one of the leading cancer-related death causes worldwide, specially on women. However, early diagnosis significantly increases treatment success. For the purpose of early diagnosis, proper analysis of histology images is essential. Specifically, during the diagnosis procedure, specialists evaluate both overall and local tissue organization via whole-slide and microscopy images. However, the large amount of data and complexity of the images makes this task time consuming and non-trivial. Because of this, the development of automatic detection and diagnosis tools is challenging but also essential for the field.

There are two goals in this challenge. The part A of the challenge consists in automatically classifying H&E stained breast histology microscopy images in four classes: normal, benign, in situ carcinoma and invasive carcinoma. The part B consists in performing pixel-wise labelling of whole-slide images in the same four classes. For these purposes, challenge participants will be provided with approx. 400+ labeled microscopy images, and 10 pixel-wise labeled and 20 non-labeled whole-slide images. Participation in only one of the parts of the challenge is allowed.

This challenge is a follow-up of the one of Bioimaging2015. Challenge participants will be able to (partially) benchmark their method by comparing with this article.

Challenge organization¶

The challenge organization results from a cooperation between Universidade do Porto, Instituto de Engenharia de Sistemas e Computadores, Tecnologia e Ciência (INESC TEC) and Instituto de Investigação and Inovação em Saúde (i3S), Portugal. The team as a whole has experience in machine learning and computer vision, has medical expertise, and has experience in organizing previous challenges. Specifically, the present grand challenge is organized by:

Araújo, Teresa (PhD student) - INESC TEC and Faculty of Engineering of Universidade do Porto
Aresta, Guilherme (PhD student) - INESC TEC and Faculty of Engineering of Universidade do Porto
Eloy, Catarina (MD, PhD) - i3S/IPATIMUP and Faculty of Medicine of Universidade do Porto
Polónia, António (MD) - i3S/IPATIMUP and Faculty of Medicine of Universidade do Porto
Aguiar, Paulo (PhD) - i3S/INEB, CMUP and Faculty of Medicine of Universidade do Porto

For questions related to the challenge, please use the Forum page. Alternatively, send us an e-mail to iciar2018.histochallenge@gmail.com.

Partners¶

The following text is mostly related to the participation on the challenge as part of the ICIAR 2018 conference, but remain here for reference purposes:

The performance of the developed methods will be evaluated using an independent test set. The best performing methods will be eligible for prizes.

Important dates¶

Challenge start and training set release: November 1, 2017
Paper and code submission deadline: January 22, 2018 February 1, 2018
Test set release: January 23, 2018 February 5, 2018
Test set prediction submission deadline: February 1, 2018 February 12, 2018
Results announcement: February 28, 2018 March 12, 2018

Participation guidelines¶

This challenge is held as part of the ICIAR 2018 conference. Participation on the challenge requires a submission of a paper describing the approach and achieved results to the conference proceedings. Please read the Rules section.