The MAMA-MIA Challenge: Advancing Generalizability and Fairness in Breast MRI Tumor Segmentation and Treatment Response Prediction
The MAMA-MIA Challenge was established to improve the generalizability and fairness of AI models used in breast MRI tumor segmentation and treatment response prediction. This challenge provided standardized datasets and evaluation protocols to address issues arising from heterogeneous data and varied assessment methods in existing AI models. Twenty-six international teams participated, with results indicating significant performance variability and trade-offs between overall accuracy and subgroup fairness across different patient demographics. AI
IMPACT Promotes the development of more robust and equitable AI systems for breast cancer imaging by providing standardized benchmarks and resources.