MIMIC-CXR Reports

该数据集是从MIMIC-CXR数据库中提取的614份报告的一个子集，其中包含307份无错误的报告以及307份由GPT-4生成的带有错误的相关合成报告。这些合成报告基于真实报告，包含了多种错误类型。总的来说，该数据集共有614份报告，旨在用于放射科报告中的错误检测任务。This dataset is a subset of 614 reports extracted from the MIMIC-CXR database, which includes 307 error-free authentic reports and 307 error-containing related synthetic reports generated by GPT-4. These synthetic reports are developed based on real clinical reports and cover multiple types of errors. Overall, this dataset has a total of 614 reports and is designed for error detection tasks in radiology reports.