cancer-subtype

This paper proposes UMMT, a token-level cross-modal transformer with contrastive multi-task learning for breast cancer subtype classification and survival prediction, achieving state-of-the-art results on METABRIC and TCGA-BRCA datasets.