Reproduce “ProtoBridge-CLIP: Observation-Masked Finding Prototypes for Medical Image--Report Representation Learning” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Jinrui Wu, Yun Peng

ProtoBridge-CLIP trains medical image–report models to distinguish affirmed findings, explicit negatives, and unmentioned findings. It combines global image–report matching with text-derived finding prototypes, observation masking, and graded image relations based on affirmed report concepts. On ChestXray14, it achieves 80.2% macro-AUROC, versus 76.9% for a matched global-only control. The results establish benchmark discrimination, not clinical utility or cross-hospital generalization.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.