I\u2019m working on a system that processes large medical record packets and generates a chronological timeline with evidence citations (think: turning hundreds or thousands of pages of medical records into a structured chronology).<\/p>\n
Right now I\u2019m trying to find datasets that resemble real world medical record packets so I can test robustness. Most of the datasets I\u2019ve found so far are either:<\/p>\n
\u2022 purely structured EHR tables (diagnoses, labs, etc.)
\u2022 small sets of individual clinical notes
\u2022 synthetic datasets<\/p>\n
What I\u2019m ideally looking for:<\/p>\n
\u2022 Long clinical documents (discharge summaries, physician notes, operative reports)
\u2022 Multi-document patient records
\u2022 Collections of clinical PDFs or reports
\u2022 Narrative-heavy hospital documentation
\u2022 Anything resembling actual chart records rather than isolated notes<\/p>\n
Datasets I already know about:<\/p>\n
\u2022 MIMIC-IV \/ MIMIC-IV-Note (waiting for credentials, anyone have a mirror?)
\u2022 i2b2 \/ n2c2 clinical NLP datasets (registration to download it is closed?)
\u2022 MTSamples medical transcription dataset<\/p>\n<\/div>\n
submitted by \/u\/deputy1389 <\/a>
[link]<\/a><\/span> [comments]<\/a><\/span><\/p>
