Brim
An interpretable bridged multimodal fusion model that links histology and molecular data for pan-cancer prognosis.
Overview
Brim is an interpretable multimodal fusion system built for the clinical reality that many cancer patients do not arrive with complete molecular profiling. The model learns a bridge between pathology images, genomics, and transcriptomics so that prognosis modeling remains useful even when one or more modalities are missing.
Rather than treating multimodal AI as a best-case laboratory exercise, Brim is explicitly designed for incomplete data and translational practicality. The result is a pan-cancer risk modeling framework that preserves predictive strength while making its cross-modal reasoning easier to inspect.