RadGraph: CheXpert Results
Dataset Description
RadGraph is a dataset of entities and relations in full-text chest X-ray radiology reports based on a novel information extraction schema designed to structure radiology reports. Our paper (accepted to the NeurIPS 2021 Track on Datasets and Benchmarks): https://arxiv.org/abs/2106.14463 Here, we release the subset of the RadGraph dataset corresponding to CheXpert reports, which include RadGraph annotations for the CheXpert test set (test.json) and the CheXpert inference set (CheXpert_graphs.json). - test.json contains two independent sets of board-certified radiologist annotations for 50 deidentified CheXpert reports (1,473 entities and 1,106 relations) - CheXpert_graphs.json contains annotations automatically generated by a deep learning model called the RadGraph Benchmark for 500 deidentified CheXpert reports (13,783 entities and 9,908 relations) with mappings to the associated chest radiographs Documentation for the CheXpert_test.json (same as test.json) and CheXpert_graphs.json files, along with the full RadGraph dataset, can be accessed at the following link: https://doi.org/10.13026/hm87-5p47.