Reproduce “Revisiting Performance Claims for Chest X-Ray Models Using Clinical Context” 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

Andrew Wang, Jiashuo Zhang, Michael Oberst

The authors evaluate chest X-ray models against prior history in MIMIC-CXR and MIMIC-IV. They estimate pre-X-ray disease risk from earlier discharge summaries, then stratify and match DenseNet121 evaluations across 13 labels. Edema AUROC is 0.900 in the lowest-risk quarter and 0.812 in the highest. Matching lowers AUROC for all 13 labels, but findings apply only to patients with prior records.

Summary by alphaXiv.

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