Figure 1: Data creation workflow.
From: A reference set of curated biomedical data and metadata from clinical case reports
To assemble the set of Metadata Acquired from Clinical Case Reports (MACCRs) we first assembled a corpus of 3,100 published case reports. Using a document metadata template including document-level identification and acknowledgement features (i.e., citation data such as title; Medical Subject Headings [MeSH terms]) and concept-level medical content features (e.g., descriptions of patient demography, clinical signs and symptoms, or outcomes), a team of medical experts manually identified text from each document corresponding to each feature. More specific terms were identified through automated approaches. To finalize this dataset, we aggregated all document metadata records into a single file. We normalized categorical features, verified, and cleaned these data, which are available as the MACCR set.
