Literature · OpenMed
OpenMed PathologyDetect
A DistilBERT model for extracting disease entities from clinical text and biomedical literature.
What OpenMed PathologyDetect is
OpenMed PathologyDetect is a token-classification model developed by OpenMed for named entity recognition (NER) across clinical records, medical texts, and scientific papers. It focuses on identifying disease mentions, histopathological descriptions, and tissue abnormalities within unstructured text.
The model assists researchers and clinicians in processing biomedical literature and patient documentation by automatically surfacing disease concepts. It supports downstream tasks including biomedical knowledge graph construction, adverse event monitoring, medication extraction, and literature mining for drug discovery.
How it works
The model is built on the distilbert-base-multilingual-cased architecture and fine-tuned for token classification. It takes tokenised biomedical text as input and predicts BIO tags (specifically B-Disease and I-Disease) to delineate entity boundaries.
Training and fine-tuning are conducted on the NCBI Disease corpus using the AdamW optimizer alongside a learning rate schedule. The NCBI Disease corpus provides gold-standard annotations across 793 PubMed abstracts, containing 6,892 disease mentions mapped to 790 concepts in Medical Subject Headings (MeSH) and Online Mendelian Inheritance in Man (OMIM).
Key facts
- Base architecture
- distilbert-base-multilingual-cased
- Model size
- 135M parameters
- Task
- Token classification (Named Entity Recognition)
- Training dataset
- NCBI Disease corpus
- Supported entity labels
- B-Disease, I-Disease
- F1 score
- 0.86
- Precision
- 0.85
- Recall
- 0.87
- Accuracy
- 0.96
What you can use it for
- Clinical Text Mining
- Extracting mentions of diseases, lesion descriptions, and tissue-level abnormalities from electronic health records and clinical notes.
- Biomedical Literature Mining
- Scanning PubMed abstracts and research articles to catalog disease names for systematic reviews and drug discovery workflows.
- Knowledge Graph Construction
- Populating biomedical ontologies and knowledge graphs by extracting structured disease entities and their contextual relationships.
- Adverse Event and Patient Monitoring
- Surfacing diagnostic mentions and disease occurrences from patient data to support safety monitoring and clinical analytics.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Runs on On-demand GPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
OpenMed PathologyDetect
Strengths and limits
Good at
- Trained on the gold-standard NCBI Disease corpus mapped to standard MeSH and OMIM concepts.
- Achieves an accuracy of 0.96 and an F1 score of 0.86 on the NCBI Disease benchmark.
- Employs a lightweight 135M-parameter DistilBERT base, enabling faster inference than larger transformer architectures.
Watch out for
- Only tags disease entities using B-Disease and I-Disease labels, offering no native tags for chemical, gene, or anatomical mentions.
- Displays lower F1 performance (0.86) compared to larger variants in the OpenMed series, such as PubMed-109M (0.9110) and PubMed-335M (0.9086).
Inputs and outputs
You provide
Text
Entities are labelled token by token.
You get back
Labels with confidence scores, highest first.
Licence
Check the licence before commercial use · See model card
The See model card licence carries terms worth reading before you depend on it.
Questions
Can I use OpenMed PathologyDetect commercially?
Check the licence before commercial use. The See model card licence carries terms worth reading before you depend on it.
How do I run OpenMed PathologyDetect without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers on CPU only, typically in 12s, and a run costs about 28 credits. Visitors without an account get 1,000 credits a month.
What entity types can this model detect?
The model identifies disease entities, tagging individual tokens as either beginning a disease mention (B-Disease) or continuing inside one (I-Disease).
What base architecture does this model use?
The model is based on distilbert-base-multilingual-cased, adapted for token-level classification.
What dataset was used to fine-tune the model?
It was fine-tuned on the NCBI Disease corpus, which contains 793 PubMed abstracts annotated with 6,892 disease mentions.
How does this model handle token aggregation?
When integrated with standard transformers pipelines, token predictions can be aggregated using strategies such as simple, first, average, max, or left unaggregated (none).
How to cite it
Maziyar Panahi. OpenMed NER: Open-Source, Domain-Adapted State-of-the-Art Transformers for Biomedical NER Across 12 Public Datasets. arXiv:2508.01630 [cs.CL], 2025.
Written from OpenMed PathologyDetect's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.