Microsoft is collaborating with Paige to construct the most extensive image-based AI model for cancer identification. Paige specializes in AI-fueled services for pathologists, the doctors who determine diseases such as cancer. Scientists hope that the model may eventually provide aid to pathologists in situations of increased case volumes and fewer personnel. The company's partnership with Microsoft is its latest effort to leverage AI and big data to improve pathology.
On Thursday, Microsoft announced it is joining forces with digital pathology provider Paige to create the biggest AI model for detecting cancer based on images. This model will be trained on a remarkable amount of data, numbering in the billions of images. It is designed to identify both frequent and rare cancers, which are usually problematic for diagnosis. Researchers are hoping it will give a helping hand to doctors in present circumstances of limited personnel and increased demand.
Paige develops solutions which are powered by AI and digital technologies for pathologists - the medical professionals responsible for conducting tests on bodily fluids and tissues for diagnosing patients. In spite of its importance, Fuchs, Paige’s co-founder and chief scientist, says this specialty has not changed much during the past 150 years. As an example, pathologists usually inspect a specimen on a glass slide using a microscope for recognizing cancer. This reliable approach may result in unfavorable outcomes for the patients if something gets missed. In a bid to improve accuracy and speed of pathology, Paige has joined hands with Microsoft in order to use AI and large volumes of data.
The Food and Drug Administration has given the thumbs up to the viewing tool FullFocus from the company, permitting pathologists to assess scanned digital slides on a screen rather than using a microscope. Paige also developed an AI system that can assist pathologists to identify breast cancer, colon cancer and prostate cancer when it surfaces on the display.
Paige is the only company approved by the FDA for pathologists to use AI as a secondary tool for detecting prostate cancer. CEO Andy Moye said this is likely due to the costs associated with storage and data collection. Digitizing just one slide can require a gigabyte of storage, making it difficult for small health systems to afford. Paige, which spun out of the Memorial Sloan Kettering Cancer Center in 2017, had a "wealth of data" to start building AI-powered solutions. The company has ten times more data than Netflix. To create an AI model that could recognize more cancer types, Paige partnered with Microsoft using its cloud storage and supercomputing infrastructure. This model is the larger model ever announced publicly and trains on four million slides to identify both common and rare cancers. Moye said this could be a groundbreaking moment for cancer care. Paige and Microsoft will publish a paper about the model on arXiv, and hopes to submit to the scientific journal Nature. The model must go through rigorous testing and get regulators' approval before it is released. Fuchs said the AI would help pathologists work through cases and arrive at a diagnosis more quickly, especially for patients outside of academic medical centers.
Comments