Advancing Generalizability and Fairness in Breast MRI Tumour Segmentation and Treatment Response Prediction (MAMA-MIA)

Breast cancer remains the most common cancer among women and a leading cause of female mortality worldwide. Dynamic contrast-enhanced MRI (DCE-MRI) is a highly sensitive imaging modality for evaluating breast tumours, guiding pre-surgical treatment planning, and assessing therapeutic responses. Despite its growing adoption, a standardized benchmark for using DCE-MRI in breast cancer research—particularly for primary tumour segmentation and treatment response prediction—remains absent, hindering advancements in personalized care.

The MAMA-MIA Challenge addresses this gap by providing a comprehensive framework for two critical and complementary tasks: (1) segmentation of primary tumours in DCE-MRI and (2) prediction of pathological complete response (pCR) to neoadjuvant chemotherapy (NAC). This dual focus integrates imaging data with clinical outcomes to enhance tumour characterization and individualized treatment planning. The challenge also prioritizes generalizability and fairness, evaluating AI models not only for accuracy but also for equitable performance across medical centres and demographic subgroups, such as age, menopausal status, and breast density.

Participants will gain access to the largest homogenized DCE-MRI dataset to date, featuring 1506 annotated scans from over 20 centres, enriched with clinical variables like tumour subtype [1, 2]. External validation includes 574 cases from three distinct European centres (Spain, Lithuania, and Poland), showcasing diverse imaging protocols and clinical settings. By advancing tumour segmentation, treatment response prediction, and fairness, the challenge seeks to optimize NAC regimens, refine surgical planning, and set new standards for ethical AI in medical imaging—driving innovation in breast cancer care.