Motivation

Breast cancer (BC) is the most frequently diagnosed cancer in women worldwide and the leading cause of cancer death in most countries. According to current guidelines, human epidermal growth factor receptor 2 (HER2) quantification must be routinely tested, along with estrogen and progesterone receptors, in all patients with invasive BC, recurrences, and metastases. HER2 is a transmembrane protein receptor with tyrosine kinase activity and is amplified and/or overexpressed in approximately 15–20% of BC. The overexpression and/or amplification of HER2 has been associated with aggressive clinical behavior but with a high probability of response to HER2 targeted therapy. Many clinical trials have demonstrated that HER2-targeted therapy administrated during and/or after chemotherapy results in a significant improvement in disease-free and overall survival only in patients with BCs showing HER2 amplification or overexpression. Consequently, the correct identification of HER2 positive BC selects patients expected to benefit from targeted therapy, making HER2 a helpful marker for therapy decision making in patients with BC. In most laboratories, HER2 evaluation begins with the analysis of protein expression by immunohistochemistry (IHC) resulting in the following scenarios: negative (score 0 or 1+), equivocal (score 2+), positive (score 3+), and indeterminate. If the IHC result is equivocal or indeterminate, reflex testing should be performed with in situ hybridization (ISH) assays for the assessment of HER2 amplification. HER2 BC are associated with certain morphological features such as high histological grade, both in invasive and in situ lesions.


Aim

Unlike previous Challenges that evaluated the staining patterns present in IHC, this Grand Challenge new edition proposes to find an image analysis algorithm to identify with high sensitivity and specificity HER2 positive BC from HER2 negative BC specimens evaluating only the morphological features present on the hematoxylin and eosin (HE) slide.





Challenge organization

The challenge organization results from a cooperation between Instituto de Investigação and Inovação em Saúde (i3SIPATIMUP, INEBand Instituto de Engenharia de Sistemas e Computadores, Tecnologia e Ciência (INESC TEC), from Porto University, 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:

For questions related to the Grand Challenge, please contact Eduardo Conde-Sousa.


Important dates

  • Challenge start and training set release: October 1st, 2019
  • Short method description and code submission deadline: until December 31st, 2019
  • Test set release: between January 1st, 2020, and January 6th, 2020
  • Test set prediction submission deadline: until January 15th, 2020
  • Results announcement: until January 19th, 2020


Participation guidelines

This challenge is held as part of the European Congress on Digital Pathology. Please read the Rules section.