News
Paige Releases 2nd-Gen Virchow AI Models for Cancer Diagnostics
Microsoft-backed Paige, a provider of generative AI pathology solutions, recently launched the latest versions of its Virchow pathology foundation model.
Virchow2 and Virchow2G are now generally available, the company announced earlier this month.
The new Virchow models were trained on very large datasets. According to Paige, their training data comprised:
- Over 3 million pathology slides from over 800 labs in 45 countries.
- Anonymized data from nearly a quarter of a million patients, covering a broad swath of races, ethnicities, genders and regions.
- Over 40 different tissue types.
The company is also touting Virchow2G as being "the largest pathology model ever created" at 1.8 billion parameters.
Paige worked with Microsoft on developing these new Virchow models. Last September, the two companies announced a partnership enabling Paige to tap Microsoft's substantial Azure cloud compute power and supercomputing infrastructure to train its massive new foundation models.
"Our collaboration with Microsoft has been pivotal in the development of Virchow," said Paige founder and chief scientist Thomas Fuchs. "Virchow's immense scale unlocks key information that can be used to drive groundbreaking innovations, enabling precise diagnostics, targeted treatments, and personalized patient care."
Virchow2G is particularly performant, noted Paige senior vice president Razik Yousfi, who said the model "outperforms anything in the industry and continues to grow in knowledge and capability."
Paige built its Virchow models to help pathologists identify and diagnose cancers more efficiently, potentially improving patient treatments and outcomes. It's using its second-generation Virchow models to power a medical application that can be used to detect cancer, as well as to help organizations create prebuilt AI modules for medical, life sciences and pharmaceutical research.