Literature · OpenMed
OpenMed AnatomyDetect
A 109M-parameter transformer for detecting anatomical structures, body parts, and tissues in biomedical text.
What OpenMed AnatomyDetect is
OpenMed AnatomyDetect is a biomedical named-entity recognition (NER) model developed by OpenMed and authored by Maziyar Panahi. Built using an ElectraMed 109M backbone (with an e5-base-v2 base architecture), it is tailored to identify mentions of anatomical structures, body parts, organs, and physiological systems across medical literature and clinical records.
The model addresses the challenge of automatically extracting fine-grained anatomical terminology from unstructured text. This aids downstream tasks in biomedical data processing, such as clinical decision support, medical coding, surgical planning documentation, and biomedical knowledge graph construction.
How it works
The model performs token classification using the Hugging Face Transformers framework. Built on the e5-base-v2 architecture and trained with the AdamW optimizer using learning rate scheduling, it processes tokenised English biomedical text.
Fine-tuning was conducted on the curated ANATOMY dataset using cross-validation on a held-out test set. The model tags text using the standard BIO (Beginning, Inside, Outside) format, specifically outputting B-Anatomy and I-Anatomy labels to mark continuous spans of anatomical terms.
Key facts
- Base architecture
- e5-base-v2
- Backbone model
- ElectraMed 109M
- Parameter count
- 109 million
- Task
- Token classification (named entity recognition)
- Target entities
- B-Anatomy, I-Anatomy
- Training dataset
- ANATOMY corpus
- Precision
- 0.88
- Recall
- 0.89
- F1 score
- 0.88
- Accuracy
- 0.97
- Language
- English
What you can use it for
- Clinical documentation and text mining
- Extracting body parts, organs, and tissue references from electronic health records, clinical notes, and surgical documentations.
- Biomedical literature mining
- Parsing academic papers and biomedical publications to automatically index anatomical structures for research databases and knowledge graphs.
- Automated medical coding support
- Assisting clinical decision-support pipelines and healthcare informatics systems that require entity extraction for anatomical coding.
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 AnatomyDetect
Strengths and limits
Good at
- Achieves an accuracy of 0.97 and an F1 score of 0.88 on anatomical entity recognition benchmarks.
- Features a compact 109M-parameter footprint suitable for fast batch inference on standard hardware.
- Fine-tuned directly on a corpus dedicated to anatomical and physiological nomenclature.
Watch out for
- Entity detection is limited strictly to anatomical classes (B-Anatomy, I-Anatomy) and does not label other clinical concepts such as diseases, chemicals, or dosages.
- Input text is restricted to English biomedical documentation.
- Outperformed on F1 score on the ANATOMY benchmark by larger models within the same family, such as ElectraMed-560M.
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 AnatomyDetect 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 AnatomyDetect 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 does the model extract?
The model outputs BIO tags for anatomical entities, specifically B-Anatomy for the start of an anatomical span and I-Anatomy for continuation tokens.
What dataset was used to train the model?
It was trained and evaluated on the curated ANATOMY dataset, which contains annotated mentions of organs, body parts, and physiological systems.
How does the model handle entity aggregation?
When run within a standard token classification pipeline, token predictions can be merged into complete entity spans using aggregation strategies such as simple, first, average, or max.
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 AnatomyDetect'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.