EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A comparative study of explainability methods for whole slide classification of lymph node metastases using vision transformers

Jens Rahnfeld, Mehdi Naouar iD, Gabriel Kalweit, Joschka Boedecker, Estelle Dubruc, Maria Kalweit iD

PLOS Digital Health · 15.04.2025 · 4 (4) · e0000792

DOI: 10.1371/journal.pdig.0000792 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS Digital Health, online erschienen am 15.04.2025. Verfasst von Jens Rahnfeld und 5 weiteren Autor:innen. Beteiligte Einrichtung: Collaborative Research Institute Intelligent Oncology (CRIION), Freiburg, Germany. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Recent advancements in deep learning have shown promise in enhancing the performance of medical image analysis. In pathology, automated whole slide imaging has transformed clinical workflows by streamlining routine tasks and diagnostic and prognostic support. However, the lack of transparency of deep learning models, often described as black boxes , poses a significant barrier to their clinical adoption. This study evaluates various explainability methods for Vision Transformers, assessing their effectiveness in explaining the rationale behind their classification predictions on histopathological images. Using a Vision Transformer trained on the publicly available CAMELYON16 dataset comprising of 399 whole slide images of lymph node metastases of patients with breast cancer, we conducted a comparative analysis of a diverse range of state-of-the-art techniques for generating explanations through heatmaps, including Attention Rollout, Integrated Gradients, RISE, and ViT-Shapley. Our findings reveal that Attention Rollout and Integrated Gradients are prone to artifacts, while RISE and particularly ViT-Shapley generate more reliable and interpretable heatmaps. ViT-Shapley also demonstrated faster runtime and superior performance in insertion and deletion metrics. These results suggest that integrating ViT-Shapley-based heatmaps into pathology reports could enhance trust and scalability in clinical workflows, facilitating the adoption of explainable artificial intelligence in pathology.

Autor:innen und Institutionen

  • Collaborative Research Institute Intelligent Oncology (CRIION), Freiburg, Germany.
  • University of Freiburg, Freiburg, Germany.
  • BrainLinks-BrainTools, Freiburg, Germany.
  • Department of Pathology, University Hospital and University of Lausanne, Lausanne, Switzerland.

ORCID-Profile: Mehdi Naouar, Maria Kalweit

Zitieren

APA

Rahnfeld, J., Naouar, M., Kalweit, G., Boedecker, J., Dubruc, E., & Kalweit, M. (2025). A comparative study of explainability methods for whole slide classification of lymph node metastases using vision transformers. PLOS Digital Health, 4(4), e0000792. https://doi.org/10.1371/journal.pdig.0000792

Vancouver

Rahnfeld J, Naouar M, Kalweit G, Boedecker J, Dubruc E, Kalweit M. A comparative study of explainability methods for whole slide classification of lymph node metastases using vision transformers. PLOS Digital Health. 2025;4(4):e0000792. doi:10.1371/journal.pdig.0000792

BibTeX
@article{Rahnfeld2025,
  author = {Rahnfeld, Jens and Naouar, Mehdi and Kalweit, Gabriel and Boedecker, Joschka and Dubruc, Estelle and Kalweit, Maria},
  title = {A comparative study of explainability methods for whole slide classification of lymph node metastases using vision transformers},
  journal = {PLOS Digital Health},
  year = {2025},
  volume = {4},
  number = {4},
  pages = {e0000792},
  doi = {10.1371/journal.pdig.0000792}
}
Kostenlos starten →