Microsoft and Paige Partner to Develop World's Largest AI Model for Cancer Detection

Microsoft and Paige have joined forces to create the world’s largest image-based artificial intelligence (AI) model for cancer detection.

This groundbreaking collaboration aims to revolutionize digital pathology and oncology by developing a computer vision AI model with billions of parameters, surpassing any existing model in terms of scale.

Dr. Thomas Fuchs, founder and chief scientist of Paige, revealed that the amount of data incorporated into the model is significantly larger than anything publicly disclosed by Google or Facebook.

He emphasized the criticality of scale in assisting pathologists and oncologists accurately.

While existing models typically include up to 100,000 slides, Fuchs explained that a truly effective model requires millions of digitized histology slides.

Over the past seven years, Paige has played a pivotal role in digitizing millions of slides to identify cancers.

However, managing such an extensive dataset and growing AI models required more computational power than the company possessed.

This is where Microsoft stepped in, offering its collaboration and resources, including Nuance, Azure, and Microsoft Research, along with their extensive compute capabilities.

The upcoming phase of development will incorporate up to 4 million digitized microscopy slides from a petabyte-scale clinical data archive.

Leveraging Microsoft’s robust supercomputer infrastructure, Paige will train the AI model and deploy it across laboratories and hospitals using Azure.

Fuchs likened the AI model to “ChatGPT for the microscopic world” and emphasized its significance as the foundation model for microscopic imaging.

The vast amount of data used by Paige and Microsoft in building this model exceeds the entirety of Netflix’s content by tenfold.

Fuchs expressed that this model will not only be the largest oncology model but also the largest publicly announced computer vision model.

The benefits of this AI model extend to gaining a meticulous understanding of cancer morphology in normal tissue, serving as a foundation for rare cancer models, mutation prediction, and response prediction.

Significantly, the model is expected to substantially reduce misdiagnoses, a problem that leads to hundreds of thousands of deaths and disabilities each year in the United States alone.

Fuchs envisions tuning the model for all rare cancers and developing detection systems across various cancer types.

A prototype already exists, which could potentially cater to the needs of any cancer patient, not limited to specific types such as breast or lung cancer.

Paige has gained recognition in the healthcare industry for developing the first Large Foundation Model, utilizing over a billion images from half-a-million pathology slides encompassing multiple cancer types. The company has also obtained FDA approval for a clinical AI application in digital pathology, marking a significant milestone.

Razik Yousfi, SVP of Engineering at Paige, expressed confidence in the collaboration, stating that by combining Microsoft’s expertise and computational power with Paige’s AI technology and knowledge in digital pathology, they will advance the state-of-the-art in cancer imaging. Ultimately, the aim is to improve the lives of millions of people affected by cancer every day.

Source: Fox News