Oral Spotlight and Best Paper Award Shortlist at MICCAI 2025 in Daejeon

Our paper Bias and Generalizability of Foundation Models across Datasets in Breast Mammography, by Elodie Germani, İlayda Selin Türk, Fatima Zeineddine, Charbel Mourad and Shadi Albarqouni, was selected for an oral spotlight at MICCAI 2025 in Daejeon, South Korea, and was one of the five papers shortlisted for the MICCAI 2025 Best Paper Award.

The paper shows that foundation-model features for mammography carry dataset-specific signals: models that perform well on pooled benchmarks can still be biased across datasets and under-represented subgroups, such as extreme breast densities and age groups, and domain adaptation alone does not remove these disparities. The work is part of our FAIR-Mammo project and our collaboration through EEDA.

Congratulations, Elodie, and thanks to all co-authors!