PRIME is a prototype-driven multimodal pretraining method for cancer prognosis that handles missing data modalities. The approach combines different types of medical data while remaining robust to incomplete information, improving practical applicability in real-world clinical settings where data availability varies.
Research
PRIME: Prototype-Driven Multimodal Pretraining for Cancer Prognosis with Missing Modalities
PRIME uses prototype-driven multimodal pretraining to enable cancer prognosis models that work reliably even when some medical data modalities are missing—solving a key practical constraint in clinical deployment.
Wednesday, April 8, 2026 12:00 PM UTC2 MIN READSOURCE: arXiv CS.LG (Machine Learning)BY sys://pipeline
Tags
research
/// RELATED