PuSH - Publikationsserver des Helmholtz Zentrums München

Ding, T.* ; Wagner, S. ; Song, A.H.* ; Chen, R.J.* ; Lu, M.Y.* ; Zhang, A.* ; Vaidya, A.J.* ; Jaume, G.* ; Shaban, M.* ; Kim, A.* ; Williamson, D.F.K.* ; Robertson, H.* ; Chen, B.* ; Almagro-Pérez, C.* ; Doucet, P.* ; Sahai, S.* ; Chen, C.* ; Chen, C.S.* ; Komura, D.* ; Kawabe, A.* ; Ochi, M.* ; Sato, S.* ; Yokose, T.* ; Miyagi, Y.* ; Ishikawa, S.* ; Gerber, G.* ; Peng, T. ; Le, L.P.* ; Mahmood, F.*

A multimodal whole-slide foundation model for pathology.

Nat. Med., DOI: 10.1038/s41591-025-03982-3 (2025)
Verlagsversion Forschungsdaten DOI PMC
Open Access Hybrid
Creative Commons Lizenzvertrag
The field of computational pathology has been transformed with recent advances in foundation models that encode histopathology region-of-interests (ROIs) into versatile and transferable feature representations via self-supervised learning. However, translating these advancements to address complex clinical challenges at the patient and slide level remains constrained by limited clinical data in disease-specific cohorts, especially for rare clinical conditions. We propose Transformer-based pathology Image and Text Alignment Network (TITAN), a multimodal whole-slide foundation model pretrained using 335,645 whole-slide images via visual self-supervised learning and vision-language alignment with corresponding pathology reports and 423,122 synthetic captions generated from a multimodal generative AI copilot for pathology. Without any fine-tuning or requiring clinical labels, TITAN can extract general-purpose slide representations and generate pathology reports that generalize to resource-limited clinical scenarios such as rare disease retrieval and cancer prognosis. We evaluate TITAN on diverse clinical tasks and find that it outperforms both ROI and slide foundation models across machine learning settings, including linear probing, few-shot and zero-shot classification, rare cancer retrieval, cross-modal retrieval and pathology report generation.
Altmetric
Weitere Metriken?
[➜Einloggen]
Zusatzinfos bearbeiten [➜Einloggen]
Publikationstyp Artikel: Journalartikel
Dokumenttyp Wissenschaftlicher Artikel
Schlagwörter Artificial-intelligence; Cancer; Heterogeneity; Prediction; Resource; Bag
ISSN (print) / ISBN 1078-8956
e-ISSN 1546-170X
Zeitschrift Nature medicine
Verlag Nature Publishing Group
Verlagsort New York, NY
Begutachtungsstatus Peer reviewed
Institut(e) Helmholtz Artifical Intelligence Cooperation Unit (HAICU)
Förderungen DoD
AMED Practical Research for Innovative Cancer Control
Siebel Foundation
Tau Beta Pi Fellowship
German Academic Exchange Service (DAAD)
Add-on Fellowship of the Joachim Herz Foundation
Helmholtz Association
National Institute of Health (NIH) National Institute of General Medical Sciences (NIGMS)
BWH President's Fund, MGH Pathology