Fusion is not one-size-fits-all: Cross-Modal Representation Alignment for Time-to-Event Modeling

In three linesMultimodal fusion framework for time-to-event prediction (PE mortality, CVD outcomes) aligning CT and longitudinal EHR representations using foundation models. Four strategies tested (late fusion, contrastive alignment, cross-attention, co-attention) on 3,099–2,951 patients. Contrastive fusion improves concordance index by 1.5–5.4% vs unimodal baselines.

Summary generated by Claude — human-verified