Evidential Fusion Network for Multimodal Survival Prediction under Missing Modalities

arXiv:2606.20757v1 Announce Type: new Abstract: Recent multimodal survival prediction models have demonstrated strong predictive performance by leveraging complementary information across modalities. However, such models generally assume data completeness and exhibit limited robustness toward missing modalities, which are frequently encountered in real-world clinical settings. We propose the Evidential Missing Modality Survival Fusion (EMMS) model for multimodal survival prediction under missing...