aceA foundation model for clinical-grade computational pathology and rare cancers detection
The analysis of histopathology images with artificial intelligence aims to enable clinical decision support systems and precision medicine. The success of such applications depends on the ability to model the diverse patterns observed in pathology images. To this end, we present Virchow, the largest foundation model for computational pathology to date. In addition to the evaluation of biomarker prediction and cell identification, we demonstrate that a large foundation model enables pan-cancer detection, achievin…
"Sarcoidosis - Asteroid body" by Pulmonary Pathology is licensed under CC BY-SA 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-sa/2.0/.
In a July 2024 Nature Medicine paper, researchers introduced Virchow, a foundation model for computational pathology, and tested its ability to support clinical decision-making from histopathology images.
The reported 0.95 specimen-level AUC across common and rare cancers suggests foundation models could reduce dependence on large labeled datasets for rare cancer detection, though the abstract does not detail patient populations, deployment settings, or prospective clinical validation.
- Virchow is described as the largest foundation model for computational pathology to date.
- Evaluated on biomarker prediction and cell identification in addition to cancer detection.
- Pan-cancer detector built on Virchow achieved 0.95 specimen-level AUC across nine common and seven rare cancers.
- With less training data, the Virchow-based detector matched tissue-specific clinical-grade models in production and outperformed them on some rare variants.
Virchow, the largest pathology foundation model to date, enables pan-cancer detection with 0.95 specimen-level AUC across nine common and seven rare cancers and matches production tissue-specific models using less training data.
The rundown
The authors present Virchow as a foundation model trained to model diverse patterns in histopathology images, evaluated for biomarker prediction, cell identification, and cancer detection.
As of the July 2024 publication, the Virchow-based pan-cancer detector reported 0.95 specimen-level AUC across 16 cancer types and demonstrated comparable performance to tissue-specific clinical-grade models in production while using less training data, with gains on some rare variants.
Sources
- Peer-reviewedNature Medicine2024-07-22
How should this claim be treated?
The debate