Reproduce “HARE: an entity and relation centric evaluation framework for histopathology reports” 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
Yunsoo Kim, Michal W. S. Ong, Alex Shavick, Honghan Wu, Adam P. Levine
HARE evaluates generated histopathology reports by extracting pathology-specific entities and relationships, embedding entity terms to match semantic counterparts, and combining entity and relation F1 scores. Against expert ratings on 231 nonzero-scored reports from a 600-report set, it outperformed GREEN in correlation and regression measures, including Pearson r = 0.606 and RMSE = 0.134. This evidence is not clinical validation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.