- The largest image-based AI model for cancer detection in the world is being created by Microsoft and Paige.
- The AI model will be employed in digital pathology and cancer.
- Paige made headlines in the medical field when she created the first Large Foundation Model.
The largest image-based artificial intelligence (AI) model for cancer detection in the world is being created by Microsoft in collaboration with the digital pathology startup Paige.
The AI model, which has been set up with billions of parameters to give a computer vision AI that is orders of magnitude larger than any comparable model now in use, will be employed in digital pathology and cancer.
Microsoft and Paige
According to Dr. Thomas Fuchs, the model’s developer and lead scientist, the amount of data provided is “orders of magnitude” bigger than anything made available by Google or Facebook.
For seven years, Paige has worked to digitize millions of slides to find tumors, but with such a large computing cast, Paige needed additional capacity to complete the task.
Microsoft steps in and collaborates with Nuance, Azure, and Microsoft Research to employ resources, compute, and hundreds of GPUs at an unprecedented scale to construct models.
Up to 4 million digitalized microscope slides from a petabyte-scale clinical data library will be used in the following development phase. The advantages of the model will aid in a more thorough understanding of the morphology of cancer in healthy tissue, which can then be applied as the basis for models of rare cancers, mutation prediction, and therapeutic response prediction.
Since misdiagnoses in the United States result in hundreds of thousands of fatalities and severe impairments each year, it is hoped that the approach will greatly reduce misdiagnosis in medicine.
Fuchs eventually aims to create detection methods on top of the model or tune it for all rare tumors. There is already a prototype that could soon help all cancer patients, not just those with specific types like breast or lung cancer.
Paige made headlines in the medical field when she created the first Large Foundation Model, which makes use of more than a billion images from 500,000 pathology slides including various cancer types.