EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

ConMIL: interactive and contrastive text-guided multiple instance learning for whole slide image classification

Anxuan Han, Alexandra Jolley, Lisa M Butler, Weitong Chen

Bioinformatics · 2026

DOI: 10.1093/bioinformatics/btag679 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Bioinformatics, 2026. Verfasst von Anxuan Han und 3 weiteren Autor:innen. Beteiligte Einrichtung: South Australian Immunogenomics Cancer Institute (SAiGENCI), Adelaide University, Adelaide, South Australia 5005, Australia. Hauptthemen laut MEDLINE: Multiple-Instance Learning Algorithms, Image Interpretation, Computer-Assisted, Image Processing, Computer-Assisted. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: South Australian Immunogenomics Cancer Institute, Australian Government, Australian Research Council Early Career Industry Fellowship.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Motivation Whole-slide image (WSI) classification in computational pathology typically relies on Multiple Instance Learning (MIL) for weakly supervised analysis. Recent pathology vision-language models have inspired text-guided approaches, but these methods typically use text for representation alignment or region localization, rather than directly incorporating semantic signals into MIL attention weighting. Furthermore, these approaches often rely on static prompts and provide limited insight into the learned nonlinear transformations performed by the classifier. Results We propose ConMIL, an interactive contrastive text-guided MIL framework for WSI classification. At its core, ConMIL introduces a contrastive semantic-guided attention mechanism that uses paired positive and negative pathology-specific text embeddings to directly modulate MIL attention weighting. This mechanism is complemented by human-in-the-loop prompt refinement to improve semantic specificity and a Kolmogorov–Arnold Network (KAN) classifier that enables visualization and quantitative inspection of learned nonlinear transformations. Experiments on CAMELYON16, TCGA-BRCA, and BRACS demonstrate that ConMIL consistently outperforms representative MIL baselines while producing pathology-consistent attention heatmaps and inspectable nonlinear transformations. Availability The source code for ConMIL is available at https://github.com/anxuanhan/ConMIL

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Multiple-Instance Learning AlgorithmsImage Interpretation, Computer-AssistedImage Processing, Computer-AssistedHumansAlgorithms

Autor:innen und Institutionen

  • South Australian Immunogenomics Cancer Institute (SAiGENCI), Adelaide University, Adelaide, South Australia 5005, Australia.
  • SA Pathology, Anatomical Pathology and Genetics and Molecular Pathology Directorates, Adelaide, South Australia 5000, Australia.
  • Precision Cancer Medicine Theme, South Australian Health and Medical Research Institute (SAHMRI), Adelaide, South Australia 5000, Australia.
  • School of Computer Science and Information Technology, Adelaide University, Adelaide, South Australia 5005, Australia.

Förderung

  • South Australian Immunogenomics Cancer Institute
  • Australian Government
  • Australian Research Council Early Career Industry Fellowship

Zitieren

APA

Han, A., Jolley, A., Butler, L. M., & Chen, W. (2026). ConMIL: interactive and contrastive text-guided multiple instance learning for whole slide image classification. Bioinformatics. https://doi.org/10.1093/bioinformatics/btag679

Vancouver

Han A, Jolley A, Butler LM, Chen W. ConMIL: interactive and contrastive text-guided multiple instance learning for whole slide image classification. Bioinformatics. 2026. doi:10.1093/bioinformatics/btag679

BibTeX
@article{Han2026,
  author = {Han, Anxuan and Jolley, Alexandra and Butler, Lisa M and Chen, Weitong},
  title = {ConMIL: interactive and contrastive text-guided multiple instance learning for whole slide image classification},
  journal = {Bioinformatics},
  year = {2026},
  doi = {10.1093/bioinformatics/btag679}
}
Kostenlos starten →