Elamur Research⌕
☰
home / papers / 2026 / MedCAT v2: a modular, extensible architecture for clinical named entity recognition and linking under real-world privacy and compute constraints

MedCAT v2: a modular, extensible architecture for clinical named entity recognition and linking under real-world privacy and compute constraints

Paper · 2026 · Mart Ratas, Thomas Searle, Adam Sutton et al.

Summary

MedCAT is an open-source toolkit that detects mentions of clinical concepts in text and maps them to ontologies such as SNOMED-CT and UMLS, a task referred to as NER+L. Its v1 design was tightly bound to spaCy data structures, so trying new tokenizers, NER modules or linkers usually meant forking the code. MedCAT v2 rebuilds the toolkit around a registry. In this design, tokenization, NER and entity linking are separate, swappable components. Add-ons such as MetaCAT (contextual meta-annotation) and RelCAT (relation extraction) plug into the same pipeline, and external packages can register their own components at runtime. Component provenance is stored in model metadata, which supports reproducibility. A compatibility layer loads v1 models without user changes.

The authors tested v2 on MDACE, COMETA, Distemist, the SNOMED Linking Challenge and MedMentions. Accuracy matched v1. Different component choices exposed clear trade-offs between speed and accuracy. This matters for hospitals that must run NLP on-premises, often without GPUs.

Key points

  • New example components:
    • a regex tokenizer for minimal environments;
    • a heuristic "faster linker" that skips disambiguation for unambiguous surface forms;
    • an embedding-based linker, distributed as a separate PyPI extension;
    • a GLiNER-based neural NER component.
  • With the same trained model, v2 reproduces v1's precision, recall and F1 within expected variance. This was checked using a private NHS-trained model that cannot be released.
  • On COMETA, the default spaCy tokenizer with the vector-context linker reached F1 0.6072 in 68.16 s. Combining the regex tokenizer with the faster linker gave F1 0.5681 in only 6.21 s.
  • The embedding linker (abhinand/MedEmbed-small-v0.1) gave the best F1 on the SNOMED Linking Challenge data (0.4215 vs. 0.4112). However, it took 511.19 s versus 75.40 s, so it is impractical without a GPU.
  • The authors conclude that no single configuration is best everywhere and that teams should evaluate components on their own corpora. Openly available models still lag behind institution-specific ones.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓King's College LondonLondon · United Kingdom42 works · 1,617 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksBiochemistry, Genetics and Molecular Biology · 234 worksMolecular Biology · 231 works

🧭 Topics 1Research subjects and languages the work is about.Named entity recognition · 200 works

Cite

Cite as Mart Ratas, Thomas Searle, Adam Sutton and Richard Dobson. 2026. MedCAT v2: a modular, extensible architecture for clinical named entity recognition and linking under real-world privacy and compute constraints. BioNLP 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.bionlp-1.17

@inproceedings{ratas2026medcat,
  title = {{MedCAT v2: a modular, extensible architecture for clinical named entity recognition and linking under real-world privacy and compute constraints}},
  author = {Mart Ratas, Thomas Searle, Adam Sutton and Richard Dobson},
  booktitle = {BioNLP 2026},
  year = {2026},
  doi = {10.18653/v1/2026.bionlp-1.17},
  url = {https://doi.org/10.18653/v1/2026.bionlp-1.17}
}

References

22 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Shubham Agarwal, Vlad Dinu, Thomas Searle, Mart Ratas, Anthony Shek, Dan F. Stein, James Teo, and Richard Dobson. 2025a. RelCAT: Advancing extraction of clinical inter-entity relationships from unstructured electronic health records. Preprint, arXiv:2501.16077. — [arXiv:2501.16077](https://arxiv.org/abs/2501.16077) find
  2. Shubham Agarwal, Thomas Searle, Mart Ratas, Anthony Shek, James Teo, and Richard Dobson. 2025b. A framework for flexible extraction of clinical event contextual properties from electronic health records. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volu… find
  3. Marco Basaldella, Fangyu Liu, Ehsan Shareghi, and Nigel Collier. 2020. COMETA: A corpus for medical entity linking in the social media. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 3122–3137, Online. Association for Computational Linguistic… find
  4. Daniel Capurro, Meliha Yetisgen, Erik Eaton, Robert Black, and Peter Tarczy-Hornoch. 2014. Availability of structured and unstructured clinical data for comparative effectiveness research and quality improvement: A multi-site assessment. eGEMs (Generating Evidence & Methods to improve patient outcom… find
  5. Hua Cheng, Rana Jafari, April Russell, Russell Klopfer, Edmond Lu, Benjamin Striner, and Matthew Gormley. 2023. MDACE: MIMIC documents annotated with code evidence. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7534– 7550, T… find
  6. Rory Davidson, Will Hardman, Guy Amit, Yonatan Bilu, Vincenzo Della Mea, Aleksandr Galaida, Irena Girshovitz, Mikhail Kulyabin, Mihai Horia Popescu, Kevin Roitero, Gleb Sokolov, and Chen Yanover. 2025. Snomed ct entity linking challenge. Journal of the American Medical Informatics Association, 32(9)… find
  7. Luis B. Elvas, Ana Almeida, and João C. Ferreira. 2025. Natural language processing in medical text processing: A scoping literature review. International Journal of Medical Informatics, 204:106049. find
  8. Hadi Kharrazi, Laura J Anzaldi, Leilani Hernandez, Ashwini Davison, Cynthia M Boyd, Bruce Leff, Joe Kimura, and Jonathan P Weiner. 2018. The value of unstructured electronic health record data in geriatric syndrome case identification. J. Am. Geriatr. Soc., 66(8):1499–1507. find
  9. Zeljko Kraljevic, Thomas Searle, Anthony Shek, Lukasz Roguski, Kawsar Noor, Daniel Bean, Aurelie Mascio, Leilei Zhu, Amos A. Folarin, Angus Roberts, Rebecca Bendayan, Mark P. Richardson, Robert Stewart, Anoop D. Shah, Wai Keong Wong, Zina Ibrahim, James T. Teo, and Richard J.B. Dobson. 2021. Multi-d… find
  10. Leibo Liu, Victoria Blake, Matthew Barman, Blanca Gallego, Timothy Churches, Georgina Kennedy, Sze-Yuan Ooi, Geoffrey P Delaney, and Louisa Jorm. 2025. Using natural language processing to extract information from clinical text in electronic medical records for populating clinical registries: a syst… find
  11. Aurelie Mascio, Zeljko Kraljevic, Daniel Bean, Richard Dobson, Robert Stewart, Rebecca Bendayan, and Angus Roberts. 2020a. Comparative analysis of text classification approaches in electronic health records. In Proceedings of the 19th SIGBioMed Workshop on Biomedical Language Processing, pages 86–94… find
  12. Aurelie Mascio, Željko Kraljević, Daniel Bean, Richard Dobson, Robert Stewart, Rebecca Bendayan, and Angus Roberts. 2020b. Comparative analysis of text classification approaches in electronic health records. In Proceedings of the 19th sigbiomed workshop on biomedical language processing, pages 86–9… find
  13. Bram Van Es, Leon C Reteig, Sander C Tan, Marijn Schraagen, Myrthe M Hemker, Sebastiaan RS Arends, Miguel AR Rios, and Saskia Haitjema. 2023. Negation detection in dutch clinical texts: an evaluation of rule-based and machine learning methods. BMC bioinformatics, 24(1):10. find
  14. Honghan Wu, Minhong Wang, Jinge Wu, Farah Francis, Yun-Hsuan Chang, Alex Shavick, Hang Dong, Michael TC Poon, Natalie Fitzpatrick, Adam P Levine, and 1 others. 2022. A survey on clinical natural language processing in the united kingdom from 2007 to 2022. NPJ digital medicine, 5(1):186. find
  15. Urchade Zaratiana, Nadi Tomeh, Pierre Holat, and Thierry Charnois. 2024. GLiNER: Generalist model for named entity recognition using bidirectional transformer. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technol…
  16. Antonio Miranda-Escalada, Luis Gascó, Salvador Lima-López, Eulàlia Farré-Maduell, Darryl Estrada, Anastasios Nentidis, Anastasia Krithara, Georgios Katsimpras, Georgios Paliouras, and Martin Krallinger. 2022. Overview of distemist at bioasq: Automatic detection and normalization of diseases from cli… find
  17. Sunil Mohan and Donghui Li. 2019. Medmentions: A large biomedical corpus annotated with umls concepts. arXiv preprint arXiv:1902.09476. — [arXiv:1902.09476](https://arxiv.org/abs/1902.09476) find
  18. Kawsar Noor, Lukasz Roguski, Xi Bai, Alex Handy, Roman Klapaukh, Amos Folarin, Luis Romao, Joshua Matteson, Nathan Lea, Leilei Zhu, and 1 others. 2022. Deployment of a free-text analytics platform at a uk national health service research hospital: Cogstack at university college london hospitals. JMI… find
  19. Mart Ratas. 2026. MedCAT v2 migration guide. https://github.com/CogStack/cogstack-nlp/blob/ main/medcat-v2/docs/migration_guide_v2.md. Accessed: 2026-04-24. ↗
  20. Mart Ratas, Zeljko Kraljevic, Anthony Shek, Thomas Searle, and Xi Bai. 2026. Medcat v2. find
  21. Tom M Seinen, Jan A Kors, Erik M van Mulligen, and Peter R Rijnbeek. 2025. Using structured codes and free-text notes to measure information complementarity in electronic health records: Feasibility and validation study. J Med Internet Res, 27:e66910. find
  22. Liesbet Van Bulck, Meghan Reading Turchioe, Maxim Topaz, and Jiyoun Song. 2024. Exploring the full potential of the electronic health record: the application of natural language processing for clinical practice. European Journal of Cardiovascular Nursing, 24(2):332–337. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ratas-2026-medcat-v2-modular-extensible-architecture/