metadata

pretty_name: MultiCaRe Articles
license: cc-by-4.0
task_categories:
  - text-classification
  - text-retrieval
language:
  - en
size_categories:
  - 100K<n<1M

MultiCaRe: Open-Source Clinical Case Dataset

MultiCaRe aggregates open-access, de-identified clinical case reports from PubMed Central’s OA corpus, pairing article-level metadata and abstracts with case narratives and figure images/captions. The normalization makes it easy to map from images → cases → articles.

Source and process: parse PMC OA case reports; extract metadata/abstracts; download/process figures; align captions; build a hierarchical taxonomy for image labels.
Scale: 85k+ OA articles, 160k+ images (v2.0).
Tasks: article-level retrieval/classification, summarization; joins to cases/images for multimodal tasks.
Citation: MDPI DATA paper — https://www.mdpi.com/2306-5729/10/8/123; Zenodo — https://zenodo.org/records/13936721.

This repository: per-article dataset Per-article dataset with bibliographic metadata and abstracts (one file: articles.parquet).

Schema

article_id: PMCID (primary key)
title, journal, year
doi, pmid, pmcid
mesh_terms, major_mesh_terms, keywords
link, license, case_amount
abstract: article abstract

Quick start

from datasets import load_dataset
art = load_dataset("openmed-community/multicare-articles", split="train")
row = art[0]
print(row["title"])
print(row["abstract"][:600])

Join examples

from datasets import load_dataset
art = load_dataset("openmed-community/multicare-articles", split="train")
cas = load_dataset("openmed-community/multicare-cases", split="train")

aid = cas[0]["article_id"]
article = art.filter(lambda e: e["article_id"] == aid)[0]
print(article["title"])  # matching article

Notes

Use article-level splits to avoid leakage when combining with images/cases.