Uncertainty-aware ensemble of foundation models differentiates glioblastoma from its mimics
Accurate pathological diagnosis is crucial in guiding personalized treatments for patients with central nervous system cancers. Distinguishing glioblastoma and primary central nervous system lymphoma is particularly challenging due to their overlapping pathology features, despite the distinct treatments required. To address this challenge, we establish the Pathology Image Characterization Tool with Uncertainty-aware Rapid Evaluations (PICTURE) system using 2141 pathology slides collected worldwide. PICTURE employs Bayesian inference, deep ensemble, and normalizing flow to account for the uncertainties in its predictions and training set labels. PICTURE accurately diagnoses glioblastoma and primary central nervous system lymphoma with an area under the receiver operating characteristic curve (AUROC) of 0.989, with the results validated in five independent cohorts (AUROC = 0.924-0.996). In addition, PICTURE identifies samples belonging to 67 types of rare central nervous system cancers that are neither gliomas nor lymphomas. Our approaches provide a generalizable framework for differentiating pathological mimics and enable rapid diagnoses for central nervous system cancer patients.
Authors
- G Larghero
- Ting‐Wan Kao (ORCID: https://orcid.org/0000-0003-2673-3168)
- Shih‐Yen Lin (ORCID: https://orcid.org/0000-0001-7028-481X)
- Adelheid Wöehrer (ORCID: https://orcid.org/0000-0001-8435-8189)
- Jeffrey A. Golden (ORCID: https://orcid.org/0000-0001-9933-5502)
- Keith Lloyd Ligon (ORCID: https://orcid.org/0000-0002-7733-600X)
- Tsung-Hua Lee
- Cheng Che Tsai
- David Cook (ORCID: https://orcid.org/0009-0007-6196-4179)
- Eliana Marostica (ORCID: https://orcid.org/0000-0002-2061-243X)
- Ying‐Chun Lo (ORCID: https://orcid.org/0000-0002-8893-0645)
- Junhan Zhao (ORCID: https://orcid.org/0000-0002-0316-8365)
- Thomas Roetzer-Pejrimovsky (ORCID: https://orcid.org/0000-0001-8875-7304)
- Omar Arnaout (ORCID: https://orcid.org/0000-0003-3890-8615)
- Liza Mathews (ORCID: https://orcid.org/0000-0002-9977-7631)
- David M. Meredith (ORCID: https://orcid.org/0000-0002-7054-8624)
- MacLean P. Nasrallah (ORCID: https://orcid.org/0000-0003-4861-0898)
- Natalie N. C. Shih (ORCID: https://orcid.org/0009-0007-5025-1026)
- Christian Engel (ORCID: https://orcid.org/0000-0003-2130-7514)
- Yutian Wang (ORCID: https://orcid.org/0009-0005-0984-9777)
- Kun‐Hsing Yu (ORCID: https://orcid.org/0000-0001-9892-8218)
- Jung-Hsien Chiang (ORCID: https://orcid.org/0000-0002-4657-6705)
- Nipon Chaisuriya (ORCID: https://orcid.org/0000-0003-1241-8503)
- Raphaël Attias
- Chia‐Jen Liu (ORCID: https://orcid.org/0000-0001-7945-9292)
- Shih-Chieh Lin (ORCID: https://orcid.org/0000-0002-6671-9902)
- Dmytro Vremenko
Institutions
- Cedars-Sinai Medical Center (US)
- Brigham and Women's Hospital (US)
- Harvard University (US)
- Innsbruck Medical University (AT)
- National Yang Ming Chiao Tung University (TW)
- Khon Kaen University (TH)
- Harvard University Press (US)
- Taipei Veterans General Hospital (TW)
- National Yang Ming University Hospital (TW)
- Dana-Farber Cancer Institute (US)
- Mayo Clinic in Arizona (US)
- Dana-Farber/Harvard Cancer Center (US)
- Artificial Intelligence in Medicine (Canada) (CA)
- Medical University of Vienna (AT)
- University of Pennsylvania (US)
- National Cheng Kung University (TW)
Publication Details
- Journal
- Nature Communications
- Published
- 2025-09-29
- DOI
- https://doi.org/10.1038/s41467-025-64249-6
- Citations
- 3
- Primary Topic
- Cell Image Analysis Techniques
- Type
- article
- Field-Weighted Citation Impact
- 4.90
Funders
- U.S. Department of Defense
- Brigham and Women's Hospital
- National Institute of General Medical Sciences
- DOD Peer Reviewed Cancer Research Program