EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Weakly Supervised Deep Learning Can Analyze Focal Liver Lesions in Contrast-Enhanced Ultrasound

Adil Oezsoy, James Alexander Brooks, Marko van Treeck, Yvonne Doerffel, Ulrike Morgera, Jens Berger, Marco Gustav, Oliver Lester Saldanha, Tom Luedde, Jakob Nikolas Kather, Tobias Paul Seraphin, Michael Kallenbach

Digestion · 2025

DOI: 10.1159/000545098 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Digestion, 2025. Verfasst von Adil Oezsoy und 11 weiteren Autor:innen. Beteiligte Einrichtung: Department of Gastroenterology, Hepatology and Infectious Diseases, University Hospital Düsseldorf, Medical Faculty at Heinrich-Heine-University, Düs…. Hauptthemen laut MEDLINE: Deep Learning, Liver Neoplasms, Image Interpretation, Computer-Assisted. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Introduction

Assessing the malignancy of focal liver lesions (FLLs) is an important yet challenging aspect of routine patient care. Contrast-enhanced ultrasound (CEUS) has proved to be a highly reliable tool but is very dependent on the examiner’s expertise. The emergence of artificial intelligence has opened doors to algorithms that could potentially aid in the diagnostic process. In this study, we evaluate the performance of a weakly supervised deep learning model in classifying FLLs as malignant or benign.

Methods

Our retrospective feasibility study was based on a cohort of patients from a tertiary care hospital in Germany undergoing routine CEUS examination to evaluate malignancy of FLL. We trained a weakly supervised attention-based multiple instance learning algorithm during 5-fold cross-validation to distinguish malignant from benign liver tumors, without using any manual annotations, only case labels. We aggregated the on-average best performing cross-validation cycle and tested this combined model on a held-out test set. We evaluated its performance using standard performance metrics and developed explainability methods to gain insight into the model’s decisions.

Results

We enrolled 370 patients, comprising a total of 955,938 images extracted from CEUS videos or manually captured during the examination. Our combined model was able to identify malignant lesions with a mean area under the receiver operating curve of 0.844 in the cross-validation experiment and 0.94 (95% CI: 0.89–0.99) in the held-out test set. The accuracy, sensitivity, specificity, and F1-Score of the combined model in finding malignant lesions in the held-out test, yielded 80.0%, 81.8%, 84.6%, and 0.81, respectively. Our exploratory analysis using visual explainability methods revealed that the model appears to prioritize information that is also highly relevant to expert clinicians in this task.

Conclusion

Weakly supervised deep learning can classify malignancy in CEUS examinations of FLLs and thus might one day be able to assist doctors’ decision-making in clinical routine.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Deep LearningLiver NeoplasmsImage Interpretation, Computer-AssistedHumansRetrospective StudiesUltrasonographyContrast MediaFeasibility StudiesMaleFemaleMiddle AgedAgedLiverAdultROC CurveGermanyDiagnosis, Differential

Stichwörter der Autor:innen

Artificial intelligenceContrast-enhanced ultrasoundHepatologyLiver tumorSonography

Substanzen

Contrast Media

Autor:innen und Institutionen

  • Department of Gastroenterology, Hepatology and Infectious Diseases, University Hospital Düsseldorf, Medical Faculty at Heinrich-Heine-University, Düsseldorf, Germany.
  • Else Kroener Fresenius Center for Digital Health, Medical Faculty Carl Gustav Carus, Technical University Dresden, Dresden, Germany.
  • Medical Outpatient Department, Charité Universitätsmedizin Berlin, Berlin, Germany.
  • Department of Gastroenterology, Hepatology, Infectious Diseases and Rheumatology, Ernst von Bergmann Klinikum, Potsdam, Germany.
  • Department of Medicine I, University Hospital Dresden, Dresden, Germany.

Zitieren

APA

Oezsoy, A., Brooks, J. A., van Treeck, M., Doerffel, Y., Morgera, U., Berger, J., Gustav, M., Saldanha, O. L., Luedde, T., Kather, J. N., Seraphin, T. P., & Kallenbach, M. (2025). Weakly Supervised Deep Learning Can Analyze Focal Liver Lesions in Contrast-Enhanced Ultrasound. Digestion. https://doi.org/10.1159/000545098

Vancouver

Oezsoy A, Brooks JA, van Treeck M, Doerffel Y, Morgera U, Berger J, et al. Weakly Supervised Deep Learning Can Analyze Focal Liver Lesions in Contrast-Enhanced Ultrasound. Digestion. 2025. doi:10.1159/000545098

BibTeX
@article{Oezsoy2025,
  author = {Oezsoy, Adil and Brooks, James Alexander and van Treeck, Marko and Doerffel, Yvonne and Morgera, Ulrike and Berger, Jens and Gustav, Marco and Saldanha, Oliver Lester and Luedde, Tom and Kather, Jakob Nikolas and Seraphin, Tobias Paul and Kallenbach, Michael},
  title = {Weakly Supervised Deep Learning Can Analyze Focal Liver Lesions in Contrast-Enhanced Ultrasound},
  journal = {Digestion},
  year = {2025},
  doi = {10.1159/000545098}
}
Kostenlos starten →