Biosemantics Group

Biomedical text mining and clinical information extraction

Much biomedical information is available only as unstructured data, i.e. free text, for example in scientific publications and electronic health records. The Biosemantics group studies, develops, and applies natural language processing methods, including advanced machine-learning techniques and large language models, to turn such texts into structured information, across multiple European languages and healthcare settings.

Biomedical data trajectories

Electronic health records contain longitudinal patient data recorded over time. We develop methods to extract meaningful patterns and interpretable clinical insights from these trajectories, representing how patients’ conditions, treatments, and outcomes evolve. This supports a better understanding of disease progression, prognostic prediction, patient-history summarisation, clinical-code validation, and the generation of synthetic data.

Biomedical terminologies

Biomedical terminologies can be used to identify concepts in unstructured text and connect information across data sources. We use these terminologies for concept recognition and investigate methods for mapping concepts between different terminology systems.

Biosignal analysis

The group also has a long history in the analysis of biosignals, with a particular focus on electrocardiograms (ECGs). We developed the Modular ECG Analysis System (MEANS) to automatically measure and interpret ECGs. MEANS has been used in numerous research projects, including population-based epidemiological studies in the Netherlands, Germany, and the United States, and has been licensed to several biomedical companies.

We are participating in the following projects:

4D PICTURE. The central aim of the 4D PICTURE project is to transform health care delivery decision-making processes in oncology by redesigning patients’ care paths and integrating evidence-based decision-support tools. Its mission is that individualized medicine becomes truly person-centered by empowering clinicians and cancer patients to engage in treatment decision-making.

SYNTHIA. SYNTHIA is a multidisciplinary collaboration providing a 360° perspective on advancing healthcare applications through synthetic data. Launched in September 2024, SYNTHIA is the first Innovative Health Initiative (IHI) project focused on addressing the need for privacy-preserving healthcare data solutions. It aims to accelerate personalised-medicine research and innovation while protecting patient privacy. Our team contributes expertise in synthetic-data generation, the OMOP Common Data Model, and integrating other data modalities, including genomics, wearables, and imaging.

TeamPerMed. TeamPerMed is a Centre for Data-Enriched Personalised Medicine, founded in 2023 by the University of Tartu and Tartu University Hospital. Together with the University of Helsinki, Erasmus MC and Erasmus University Rotterdam, it develops scalable AI- and real-world-data–based tools for personalised prevention and treatment of chronic and hereditary diseases. The project combines genomic, health, lifestyle and environmental data to identify people at increased risk and support clinical decision-making through tools such as polygenic risk scores, pharmacogenomics and clinical decision-support systems.

NAKO. The German National Cohort (NAKO Gesundheitsstudie) is a nation-wide, prospective population-based study. Its overall aim is to investigate the causes underlying major chronic diseases, such as cardiovascular diseases, cancer, diabetes, respiratory and infectious diseases, and dementia. We are responsible for the analysis of all standard 12-lead ECGs that are collected in the project.

Past projects:

European Health Data and Evidence Network, EHDEN. The EHDEN project aspires to be the trusted observational research ecosystem to enable better health decisions, outcomes and care. Its mission is to provide a new paradigm for the discovery and analysis of health data in Europe, by building a large-scale, federated network of data sources standardized to a common data model.

eTRANSAFE. The eTRANSAFE project is a large collaborative effort of academia, SMEs and pharmaceutical companies aiming at the development of a data platform and new methods and tools to improve translational safety assessment and research.

1. Seinen TM, Kors JA, van Mulligen EM, Fridgeirsson E, Rijnbeek PR. The added value of text from Dutch general practitioner notes in predictive modeling. J Am Med Inform Assoc. 2023. doi: 10.1093/jamia/ocad160.
2. Sanz F, Pognan F, Steger-Hartmann T, et al. eTRANSAFE: data science to empower translational safety assessment. Nat Rev Drug Discov. 2023;22:605. doi: 10.1038/d41573-023-00099-5.
3. Eikelboom WS, Singleton EH, van den Berg E, et al. The reporting of neuropsychiatric symptoms in electronic health records of individuals with Alzheimer's disease: a natural language processing study. Alzheimers Res Ther. 2023;15:94. doi: 10.1186/s13195-023-01240-7.
4. Vlietstra WJ, Vos R, van Mulligen EM, Jenster GW, Kors JA. Identifying genes targeted by disease-associated non-coding SNPs with a protein knowledge graph. PLoS One. 2022;17:e0271395. doi: 10.1371/journal.pone.0271395.
5. Seinen TM, Fridgeirsson EA, Ioannou S, et al. Use of unstructured text in prognostic clinical prediction models: a systematic review. J Am Med Inform Assoc. 2022;29:1292. doi: 10.1093/jamia/ocac058.
6. Visser JJ, de Vries M, Kors JA. Automatic detection of actionable findings and communication mentions in radiology reports using natural language processing. Eur Radiol. 2022;32:3996. doi: 10.1007/s00330-021-08467-8.
7. Vlietstra WJ, Vos R, van den Akker M, van Mulligen EM, Kors JA. Identifying disease trajectories with predicate information from a knowledge graph. J Biomed Semant. 2020;11:9. doi: 10.1186/s13326-020-00228-8.
8. van den Berg ME, Rijnbeek PR, Niemeijer MN, et al. Normal values of corrected heart-rate variability in 10-second electrocardiograms for all ages. Front Physiol. 2018;9:424. doi: 10.3389/fphys.2018.00424.
9. Rijnbeek PR, van den Berg ME, van Herpen G, Ritsema van Eck HJ, Kors JA. Validation of automatic measurement of QT interval variability. PLoS One. 2017;12:e0175087. doi: 10.1371/journal.pone.0175087.
10. Kors JA, Clematide S, Akhondi SA, van Mulligen EM, Rebholz-Schuhmann R. A multilingual gold-standard corpus for biomedical concept recognition: the Mantra GSC. J Am Med Inform Assoc. 2015;22:948. doi: 10.1093/jamia/ocv037.
11. Afzal Z, Pons E, Kang N, Sturkenboom MC, Schuemie MJ, Kors JA. ContextD: an algorithm to identify contextual properties of medical terms in a Dutch clinical corpus. BMC Bioinformatics. 2014;15:373. doi: 10.1186/s12859-014-0373-3.

Collaboration within Erasmus MC

Department of Public Health (Dr. J. Rietjens, Dr. I. Korfage)

Department of Epidemiology (Dr. M. Kavousi)

Collaboration outside Erasmus MC

Leiden Institute of Advanced Computer Science (Dr. S. Verberne)

Amsterdam University Medical Center (Dept. Cardiology, Dr. P. Postema)

German Cancer Research Center (NAKO Gesundheitsstudie, Dr. H. Greiser)

Euretos (Dr. O. Beckerhof)

Dr. Erik van Mulligen
Assistant Professor
e.vanmulligen@erasmusmc.nl

Dr. Tom Seinen
Assistant Professor
t.seinen@erasmusmc.nl

Stergios Ntanavaras,
PhD Candidate
s.ntanavaras@erasmusmc.nl

Paula Araujo Rabinovich,
PhD Candidate
p.araujorabinovich@erasmusmc.nl

Dr. Reinder Vos
Emeritus Professor
r.vos@erasmusmc.nl

Dr. Jan Kors
Associate Professor