AI algorithm predicts aneurysm risk from EHR, genetic data

Stanford researchers are able to predict the risk of abdominal aortic aneurysm by using an artificial intelligence algorithm that combines genome-sequence information and electronic health record data.

A new study from the Stanford University School of Medicine, published September 6 in Cell, is good news for the millions of people who are afflicted each year with the often fatal cardiovascular disease—abdominal aortic aneurysm (AAA)—that rarely shows symptoms.

“No one has ever set up a predictive test for it and, just from a genome sequence, we found that we could actually predict with about 70 percent accuracy who is at high risk for AAA,” says Michael Snyder, professor and chair of genetics at Stanford, who noted that when other details from EHRs were added—such as whether a patient smoked and his or her cholesterol levels—the accuracy of the machine-learning algorithm increased to 80 percent.

Stanford Med-CROP.jpg

Also See: Virtual reality enables 3D view of anatomy before surgery

The research, funded by the National Institutes of Health, showed that the algorithm—called Hierarchical Estimate From Agnostic Learning (HEAL)—not only identified disease-associated components in AAA by aggregated learning from population genomes but also, when combined with personal EHR data, reached a level similar to or better than many existing clinical screening tests.

As a result, the authors say HEAL could be potentially developed into a clinically actionable test for early screening of AAA.

“By modeling personal genomes with EHRs, this framework quantitatively assessed the effectiveness of adjusting personal lifestyles given personal genome baselines, demonstrating its utility as a personal health management tool,” conclude the study’s authors. “Our study presents a new framework for disease genome analysis, which can be used for both health management and understanding the biological architecture of complex diseases.”

NIH Director Francis Collins, MD, agreed with their assessment and wrote in a September 7 tweet that “this machine-learning method could be a model for other diseases.”

Going forward, Snyder and his research team are looking to use HEAL to help detect the genetic underpinnings of preterm birth and autism—infants born prematurely have a higher risk of autism than infants delivered at term.

“I see a future in which everyone will be born with their genome sequenced, or shortly thereafter,” adds Snyder. “Both your single-gene and your complex disease risk will be used to predict your overall disease risk, and then you can take action based on that information.”

For reprint and licensing requests for this article, click here.