EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Medical Concept Normalization of German Clinical Expressions to SNOMED CT

Helena Adam, Akhila Abdulnazar, Roland Roller, Stefan Schulz, Markus Kreuzthaler

Studies in Health Technology and Informatics · 2026

DOI: 10.3233/shti260995 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Studies in Health Technology and Informatics, 2026. Verfasst von Helena Adam und 4 weiteren Autor:innen. Beteiligte Einrichtung: Institute of Medical Informatics, Statistics and Documentation, Medical University of Graz, Graz, Austria. Hauptthemen laut MEDLINE: Systematized Nomenclature of Medicine, Electronic Health Records, Natural Language Processing. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Introduction

Clinical narratives in electronic health records frequently contain clinical expressions describing medical conditions. Their free-text format limits interoperability and automated processing. Medical concept normalization (MCN) addresses this challenge by mapping textual expressions to standardized terminology concepts, such as those from SNOMED CT. However, clinical language is characterized by abbreviations, spelling variants, and short forms, which complicate automatic normalization.

Methods

In this work, we investigate the MCN of short German clinical expressions to SNOMED CT by comparing a direct large language model (LLM)-based normalization approach using GPT-5.4 with a hybrid retrieval approach that combines a medBERT.de bi-encoder for embedding-based retrieval and retrieval-augmented generation (RAG) reranking using GPT-5 variants.

Results

The LLM-only baseline achieves a Recall@1 of 0.235, Recall@3 of 0.297, and Recall@5 of 0.303. In contrast, the embedding-based bi-encoder retrieval approach achieves a Recall@1 of 0.681, a Recall@3 of 0.783, and a Recall@5 of 0.812. Incorporating RAG-based LLM reranking further improves Recall@1 to 0.771, while Recall@3 and Recall@5 reach 0.809 and 0.812, respectively.

Conclusion

The results demonstrate that combining domain-specific embedding-based retrieval with LLM-based reranking substantially improves MCN performance compared to a standalone LLM approach. This hybrid strategy supports robust semantic matching between German clinical expressions and standardized terminology concepts, facilitating structured representation and enabling more effective secondary use of clinical text.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Systematized Nomenclature of MedicineElectronic Health RecordsNatural Language ProcessingLarge Language ModelsGermanyHumansInformation Storage and Retrieval

Stichwörter der Autor:innen

German clinical textSNOMED CTbi-encoder retrievallarge language modelsmedical concept normalizationrerankerretrieval-augmented generation

Autor:innen und Institutionen

  • Institute of Medical Informatics, Statistics and Documentation, Medical University of Graz, Graz, Austria.
  • CBmed GmbH - Center of Biomarker Research in Medicine, Graz, Austria.
  • German Research Center of Artificial Intelligence (DFKI), Berlin, Germany.

Zitieren

APA

Adam, H., Abdulnazar, A., Roller, R., Schulz, S., & Kreuzthaler, M. (2026). Medical Concept Normalization of German Clinical Expressions to SNOMED CT. Studies in Health Technology and Informatics. https://doi.org/10.3233/shti260995

Vancouver

Adam H, Abdulnazar A, Roller R, Schulz S, Kreuzthaler M. Medical Concept Normalization of German Clinical Expressions to SNOMED CT. Studies in Health Technology and Informatics. 2026. doi:10.3233/shti260995

BibTeX
@article{Adam2026,
  author = {Adam, Helena and Abdulnazar, Akhila and Roller, Roland and Schulz, Stefan and Kreuzthaler, Markus},
  title = {Medical Concept Normalization of German Clinical Expressions to SNOMED CT},
  journal = {Studies in Health Technology and Informatics},
  year = {2026},
  doi = {10.3233/shti260995}
}
Kostenlos starten →