Reproduce “Meta-CLIP: Metadata-Conditioned Contrastive Learning for Zero-Shot Disease Inference” 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
Haozhe Luo, Nirosh Sivanesan, Shelley Zixin Shu, Francesco Nitti, Mauricio Reyes
Meta-CLIP uses DICOM metadata during training to weight unmatched image–report comparisons between studies with similar acquisition conditions. Its image and text encoders require no DICOM input at inference. On CT-RATE’s validation split, macro-AUROC rises from 73.08% with CT-CLIP to 74.49%, a 1.41-point gain (95% paired case-bootstrap CI: 0.96–1.85). This zero-shot result does not establish acquisition invariance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.