A team led by researchers at Harvard-affiliated Massachusetts General Hospital and the Broad Institute of MIT and Harvard has created an artificial intelligence-based technique for detecting people at risk for atrial fibrillation.
Atrial fibrillation, or an erratic and frequently fast heart beat, is a common disorder that can cause clots in the heart to travel to the brain and cause a stroke. The research was published in the journal Circulation.
Based on information from electrocardiograms (noninvasive tests that capture the electrical impulses of the heart) in 45,770 individuals getting primary care at MGH, the researchers created an artificial intelligence-based strategy to forecast the likelihood of atrial fibrillation within the following five years.
The researchers next tested their strategy on three big data sets from studies involving a total of 83,162 people. When paired with established clinical risk variables for predicting atrial fibrillation, the AI-based technique proved synergistic in predicting atrial fibrillation risk. The technique was also highly predictive in subgroups of people, such as those who had previously suffered from heart failure or stroke.
Senior author Steven A. Lubitz, a cardiac electrophysiologist at MGH, an associate member of the Broad Institute, and associate professor of medicine at Harvard Medical School, says, “We see a role for electrocardiogram-based artificial intelligence algorithms to assist with the identification of individuals at greatest risk for atrial fibrillation.”
According to Lubitz, the algorithm might be used as a kind of pre-screening tool for individuals who are now experiencing undiagnosed atrial fibrillation, leading physicians to look for it using longer-term cardiac rhythm monitoring, which could lead to stroke preventive measures.
The outcomes of the study also show the potential of artificial intelligence (AI) — specifically, machine learning — to revolutionise medicine. “Machine learning is set to assist clinicians and researchers make enormous progress in strengthening cardiac care,” says co-author Anthony Philippakis, chief data officer at the Broad and co-director of the institute’s Eric and Wendy Schmidt Center. “As a data scientist and former cardiologist, I’m looking forward to seeing how machine learning–based methods may be combined with the testing and clinical treatments we use on a daily basis to enhance risk prediction and care for patients with atrial fibrillation.”
“The adoption of such algorithms might encourage doctors to adjust significant risk factors for atrial fibrillation, perhaps reducing the likelihood of getting the condition completely,” says co–lead author Shaan Khurshid, an electrophysiology clinical and research fellow at MGH.
Samuel Friedman, Christopher Reeder, Paolo Di Achille, Nathaniel Diamant, Pulkit Singh, Lia X. Harrington, Xin Wang, Mostafa A. Al-Alusi, Gopal Sarma, Andrea S. Foulkes, Patrick T. Ellinor, Christopher D. Anderson, Jennifer E. Ho, and Puneet Batra are co-lead authors; co-authors include Samuel Friedman, Christopher Reeder, Paolo Di Achill.
The National Institutes of Health, the American Heart Association, the Doris Duke Foundation, and the Leducq Foundation all contributed to this research.