PRIME: Prototype-Driven Multimodal Pretraining for Cancer Prognosis with Missing Modalities

Create an OSS framework for multimodal medical prognosis that handles missing modalities natively using prototype-driven learning. This fills a critical gap in clinical machine learning research.

Suggested repo: prime-med

"Cancer prognosis with missing data: A robust multimodal framework for fragmented clinical cohorts."

Estimated effort: 70h