A curated catalog of datasets, corpora, benchmarks, and public data resources for medical AI, clinical NLP, biomedical NLP, imaging, knowledge graphs, retrieval, and evaluation.
Important: this repository is an index, not a redistribution point. Always verify the source license, data-use agreement (DUA), patient-data restrictions, and commercial-use terms before using a dataset.
| I want data for… | Start here |
|---|---|
| Medical image classification, detection, segmentation, registration | Medical imaging |
| Clinical records, ICU/EHR modeling, concept embeddings | Clinical & EHR |
| Clinical or biomedical named entity recognition | Clinical & Biomedical NER |
| Entity linking / terminology normalization | Entity Linking & Normalization |
| Relation / triple extraction and knowledge graph construction | Knowledge Extraction & KGs |
| Biomedical QA, retrieval, semantic indexing, clinical summarization | BioASQ |
| Clinical NLP shared tasks using real clinical notes | n2c2 / i2b2 |
| Precision-medicine / clinical information retrieval | TREC |
| National/public healthcare data | National Healthcare Data |
| Biomedical literature and document corpora | Biomedical Literature |
| Medical speech | Medical Speech |
| Classic tabular medical ML datasets | UCI Medical Datasets |
| Medical-AI challenges and competitions | Challenges & Benchmarks |
| Patient acuity, ML metrics, agentic-AI evaluation | Evaluation & Agentic Benchmarks |
Imaging · Clinical text · Biomedical literature · Structured EHR · Claims · Terminology/KG · Speech · Physiological signals · Population/public-health data
flowchart TD
A[Clinical / Biomedical Text] --> B[NER]
B --> C[Entity Normalization]
C --> D[Relation Extraction]
D --> E[Knowledge / Triple Extraction]
E --> F[Knowledge Graph / Retrieval]
F --> G[QA · Reasoning · Validation]
Clinical NLP and biomedical-literature NLP overlap, but they are not identical. Clinical datasets commonly contain EHR notes, discharge summaries, medications, tests, problems, PHI, temporal expressions, and procedures. Biomedical corpora commonly use PubMed/PMC and annotate diseases, chemicals, genes/proteins, species, variants, cell lines, and relations.
| Label | Meaning |
|---|---|
OPEN |
Public download or openly accessible resource; still verify license/terms |
REGISTRATION |
Account or challenge registration required |
DUA |
Data Use Agreement / credentialed access required |
REQUEST |
Access by request to authors/organizers |
RESTRICTED |
Usage restrictions or controlled access apply |
MIXED |
Access varies by collection/subtask |
Never copy DUA-controlled clinical data into this repository. For example, n2c2 datasets are available to approved users under a DUA and may not be redistributed through GitHub.
GitHub repository search works well once the catalog is split into topic files. Useful search terms include:
NER · SNOMED · RxNorm · MIMIC · MRI · relation extraction · knowledge graph · BioASQ · access:open
Dataset entries also use lightweight metadata such as:
task:ner modality:clinical-text access:dua terminology:snomed-ct
- All dataset categories
- Benchmarks and shared tasks
- Biomedical NLP
- Contributing
- Migration / preservation note
#Dis This repository is provided for informational and research-discovery purposes only. Dataset descriptions can become stale; the authoritative source is always the dataset owner or challenge organizer.