Reproduce “AR-RRG: Anatomy-Guided Retrieval-Driven Radiology Report Generation” 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
Xiang Zhang, Shijie Song, Wanqing Zhao, Wei Zhou, Rui Cao, Dekui Wang, Qiyao Hu, Xianlin Peng and 3 more
AR-RRG generates chest-X-ray reports by combining anatomically constrained lesion detection, whole-report templates, finding-specific phrases, similar-case consensus, and visual tokens from a frozen encoder. A LoRA-tuned language model uses these cues to draft reports. On MIMIC-CXR, adding neighborhood consensus raises 14-finding CheXbert Micro-F1 from 0.526 to 0.582, while adding 625 false positives; the gain concerns label recovery, not image-level correctness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.