EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Zero-Shot TNM Staging from German Pathology Reports Using Pre-Trained Transformer Models (BB-TEN)

Hasan Taha, Werner O Hackl, Sabrina B Neururer, Patricia Gscheidlinger, Christoph Hickmann, Helmut Mühlböck, Bernhard E Pfeifer

Studies in Health Technology and Informatics · 07.05.2026

DOI: 10.3233/shti260089 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Studies in Health Technology and Informatics, erschienen am 07.05.2026. Verfasst von Hasan Taha und 6 weiteren Autor:innen. Beteiligte Einrichtung: Health Data Competence Center, Tirol Kliniken GmbH, Innsbruck, Austria. Hauptthemen laut MEDLINE: Neoplasm Staging, Natural Language Processing. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

TNM staging is essential for cancer registries but is often embedded in unstructured pathology reports, requiring manual extraction. Transformer-based models have recently been proposed for automated TNM classification from English pathology texts.

Objectives

To assess the robustness and cross-lingual transferability of pre-trained English-language TNM classifiers applied without fine-tuning to German pathology reports.

Methods

Three publicly available transformer-based TNM classifiers (T, N, M) were applied to a synthetic German pathology dataset of 109 breast, lung, and prostate cancer reports, using expert-assigned TNM labels as the gold standard.

Results

The models achieved high precision and specificity but moderate sensitivity for tumor and nodal staging and low sensitivity for metastasis, frequently abstaining by predicting “Unknown.” Conclusion: English-trained TNM classifiers can extract staging information from German pathology reports in a zero-shot setting with high reliability when predictions are made, but reduced recall. This conservative behavior supports their use as high-precision screening tools in registry workflows and could be improved through limited domain adaptation.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Neoplasm StagingNatural Language ProcessingHumansGermanySensitivity and SpecificityMachine LearningRegistriesProstatic NeoplasmsLung NeoplasmsReproducibility of ResultsMaleBreast NeoplasmsClassification AlgorithmsFemale

Stichwörter der Autor:innen

Artificial IntelligenceMachine LearningNatural Language ProcessingNeoplasm StagingPathologyRegistries

Autor:innen und Institutionen

  • Health Data Competence Center, Tirol Kliniken GmbH, Innsbruck, Austria.
  • Division for Digital Health and Telemedicine, UMIT TIROL - Private University For Health Sciences and Health Technology, Hall i.T., Austria.
  • Tyrolean Federal Institute for Integrated Care, Tirol Kliniken GmbH, Innsbruck, Austria.

Zitieren

APA

Taha, H., Hackl, W. O., Neururer, S. B., Gscheidlinger, P., Hickmann, C., Mühlböck, H., & Pfeifer, B. E. (2026). Zero-Shot TNM Staging from German Pathology Reports Using Pre-Trained Transformer Models (BB-TEN). Studies in Health Technology and Informatics. https://doi.org/10.3233/shti260089

Vancouver

Taha H, Hackl WO, Neururer SB, Gscheidlinger P, Hickmann C, Mühlböck H, et al. Zero-Shot TNM Staging from German Pathology Reports Using Pre-Trained Transformer Models (BB-TEN). Studies in Health Technology and Informatics. 2026. doi:10.3233/shti260089

BibTeX
@article{Taha2026,
  author = {Taha, Hasan and Hackl, Werner O and Neururer, Sabrina B and Gscheidlinger, Patricia and Hickmann, Christoph and Mühlböck, Helmut and Pfeifer, Bernhard E},
  title = {Zero-Shot TNM Staging from German Pathology Reports Using Pre-Trained Transformer Models (BB-TEN)},
  journal = {Studies in Health Technology and Informatics},
  year = {2026},
  doi = {10.3233/shti260089}
}
Kostenlos starten →