Reproduce “MedAdaptAgent: An Automatic Harness for Adapting Medical Imaging Models to New Sites” 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
Hanxue Gu, Zhonghao Liu, Xingrui Gu, Kang Wang, Yang Yang
MedAdaptAgent automates adaptation of pretrained medical-imaging models for new sites. It inspects the model and site, selects feasible adjustments under resource constraints, compares candidates with the source, and packages a choice or stops. Across ten U-Net segmentation transfers, 20 annotations raised mean Dice from 0.483 to 0.774, with no harmful replacements. This image-level result is not a clinical safety guarantee.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.