
Mayo Clinic says its Mayo Clinic Digital Pathology has already had notable early achievements. [Image courtesy of Mayo Clinic]
Rochester, Minnesota–based Mayo Clinic said the collaboration has already resulted in promising early achievements.
“Mayo Clinic is reimagining what is possible in disease detection and prediction, both within its own system and globally. We are doing this by using large, diverse datasets to build powerful artificial intelligence models in pathology. This will make diagnoses faster, more accurate, and more efficient, improving treatment approaches and speeding new cures to patients,” Jim Rogers, CEO of Mayo Clinic Digital Pathology, said in a news release.
According to some estimates, only about a tenth of pathology labs in the U.S. have digitized at this point. In many cases, cancer patients looking for a second opinion or alternative treatment option at a different health system need to have physical slides mailed to the new location.
However, where digitization has already taken place, there are signs that digitization combined with artificial intelligence could take oncology to the next level. For example, at our DeviceTalks West show in October 2024, UCSF’s Dr. Zoltan G. Laszik and Philips’ digital pathology business leader Kevin Coady demonstrated how AI tools could enable pathologists to assess prostate cancer more accurately. (Discover more medtech insights at DeviceTalks Boston, April 30–May 1, and DeviceTalks Minnesota, June 11.)
Mayo Clinic, with its 20 million digital slides linked to 10 million patient records, seeks to show even more what is possible when artificial intelligence gets combined with digital pathology.
“AI-driven insights can accelerate diagnostics, enhance precision medicine and revolutionize patient care,” said Nvidia VP of Healthcare and Life Sciences Kimberly Powell. “By digitizing and harnessing the power of vast datasets through its Digital Pathology platform, powered by Nvidia’s accelerated computing, Mayo Clinic is helping pave the way for a future with faster medical breakthroughs, better treatments and improved outcomes for patients across the globe.”
“Merging Mayo Clinic’s data and expertise with our advanced machine learning capabilities will produce breakthrough foundation models and AI products that advance the field of precision medicine and meaningfully improve patient care,” said Viktor Matayas, CEO of Aignostics.
Over less than two months, Mayo Clinic and Berlin-based Aignostics created a novel foundation model called Atlas. They trained it on 1.2 million unidentified histopathology whole slide images from Mayo Clinic and Charité – Universitätsmedizin Berlin. They published their findings on Jan. 9, reporting that Atlas achieved state-of-the-art performance across 21 public benchmark datasets, even though it was neither the largest model by parameter count nor training dataset size. They’re presently developing and deploying new features enabled by the model and plan to build new models, including one being trained on 5 million slides.
The Nvidia collaboration is focused on creating infrastructure to build and deploy foundation models to accelerate generative AI advances in pathology and beyond. Nvidia has a healthcare-specific full-stack computing architecture for artificial intelligence called Nvidia Clara. Mayo Clinic is using it to build models that its officials think will open new frontiers in medicine.
“These new capabilities using digital pathology data will unlock this critically important clinical information for building AI solutions for advanced diagnosis and care of patients, and that will improve the lives of patients globally,” said Dr. Matthew Callstrom, chair of Mayo Clinic Radiology in the Midwest and medical director for Generative AI and Strategy.