Reproduce “Findings-aware Language-Image Pretraining for CT Volume Interpretation” 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

Takuro Shimaya, Junya Inoh, Masahiro Saiko

FiLIP trains a chest-CT image encoder with report-derived finding phrases as targets, comparing cases by findings rather than patient identity. The method uses intra-image and inter-image losses to guide visual features. On 16 finding classes, mean AUC was 0.799 on CT-RATE versus 0.778 for fVLM; FiLIP significantly outperformed fVLM in-domain and externally, supporting the approach only on evaluated chest-CT tests.

Summary by alphaXiv.

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