EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Benchmarking machine learning for bowel sound pattern classification – From tabular features to pretrained models

Zahra Mansour, Verena Nicole Uslar, Dirk Weyhe, Danilo Hollosi, Nils Strodthoff iD

PLOS One · 02.01.2026 · 21 (1) · e0338911

DOI: 10.1371/journal.pone.0338911 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS One, online erschienen am 02.01.2026. Verfasst von Zahra Mansour und 4 weiteren Autor:innen. Beteiligte Einrichtung: Division AI4Health, Department for Health Services Research, Faculty of Medicine and Health Sciences, Carl von Ossietzky Universität Oldenburg, Olden…. Hauptthemen laut MEDLINE: Machine Learning. Die Arbeit wurde bisher 9-mal zitiert (Crossref) und nennt 56 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: MWK Niedersachsen (via Fraunhofer IDMT, institute part HSA, project Connected Health)., Faculty VI, Oldenburg University.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The development of electronic stethoscopes and wearable recording sensors opened the door to the automated analysis of bowel sound (BS) signals. This enables a data-driven analysis of bowel sound patterns, their interrelations, and their correlation with different pathologies. This work leverages a BS dataset collected from 16 healthy subjects that was annotated according to four established BS patterns. This dataset is used to evaluate the performance of machine learning models to detect and/or classify BS patterns. The selection of considered models covers models using tabular features, convolutional neural networks based on spectrograms and models pre-trained on large audio datasets. The results highlight the clear superiority of pre-trained models, particularly in detecting classes with few samples, achieving an AUC of 0.89 in distinguishing BS from non-BS using a HuBERT model and an AUC of 0.89 in differentiating bowel sound patterns using a Wav2Vec 2.0 model. These results pave the way for more comprehensive understanding of bowel sounds in general and future machine-learning-driven diagnostic applications for gastrointestinal examinations.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Machine LearningHumansBenchmarkingMaleNeural Networks, ComputerSoundAdultFemale

Autor:innen und Institutionen

  • Division AI4Health, Department for Health Services Research, Faculty of Medicine and Health Sciences, Carl von Ossietzky Universität Oldenburg, Oldenburg, Germany.
  • Fraunhofer IDMT, Institute Part HSA, Oldenburg, Germany.
  • University Clinic for Visceral Surgery, Faculty of Medicine and Health Sciences, Carl von Ossietzky Universität Oldenburg, Oldenburg, Germany.

ORCID-Profile: Nils Strodthoff

Förderung

  • MWK Niedersachsen (via Fraunhofer IDMT, institute part HSA, project Connected Health).
  • Faculty VI, Oldenburg University

Zitieren

APA

Mansour, Z., Uslar, V. N., Weyhe, D., Hollosi, D., & Strodthoff, N. (2026). Benchmarking machine learning for bowel sound pattern classification – From tabular features to pretrained models. PLOS One, 21(1), e0338911. https://doi.org/10.1371/journal.pone.0338911

Vancouver

Mansour Z, Uslar VN, Weyhe D, Hollosi D, Strodthoff N. Benchmarking machine learning for bowel sound pattern classification – From tabular features to pretrained models. PLOS One. 2026;21(1):e0338911. doi:10.1371/journal.pone.0338911

BibTeX
@article{Mansour2026,
  author = {Mansour, Zahra and Uslar, Verena Nicole and Weyhe, Dirk and Hollosi, Danilo and Strodthoff, Nils},
  title = {Benchmarking machine learning for bowel sound pattern classification – From tabular features to pretrained models},
  journal = {PLOS One},
  year = {2026},
  volume = {21},
  number = {1},
  pages = {e0338911},
  doi = {10.1371/journal.pone.0338911}
}
Kostenlos starten →