Paige demonstrates first clinical-grade performance for cancer detection using a foundation model ‘Virchow’

image credit- shutterstock

Paige, the pioneer in AI-driven pathology solutions designed to improve the fight against cancer, in collaboration with Microsoft, announced publication of a novel pioneering study in Nature Medicine. The study highlights the capabilities of Virchow, the first million-slide foundation model for cancer. Virchow demonstrated clinical grade cancer detection showcasing state-of-the-art performance across 16 types of cancer, including seven rare cancers.

The extensive dataset used to train Virchow empowers it to deliver advanced capabilities and accuracy that allow for the ability to create novel applications in cancer diagnosis that previously weren’t possible due to limited data. Virchow works by breaking down each gigapixel H&E slide into thousands of image tiles and characterizing intricate cancer patterns that may not be visible to the human eye. This is made possible by self-supervised learning, which does not require task-specific data and can leverage the information gained from billions of image tiles extracted from 1.5 million whole slides, containing diverse cancer subtypes.

In addition to cancer detection, the study also highlights Virchow's state-of-the-art performance across a variety of biomarker prediction tasks. "Biomarker prediction is crucial for tailoring treatments to individual patients and sets the path for personalized precision medicine.” "Virchow's superior performance in biomarker prediction tasks across multiple tissue types demonstrates its transformative potential in clinical and pharmaceutical settings”, said explained Dr. Siqi Liu, PhD, Director of AI Science at Paige.

Empowered by this technology, Paige has made significant strides in developing transformative products such as Paige Pan Cancer Suite, a single application capable of detecting cancers across the broad array of tissue types, and PRISM, a multi-modal, slide-level Foundation Model providing comprehensive written diagnostic assessments.