Using artificial intelligence to better predict heart attacks, strokes, and death
Clinicians currently assess individual risk for future heart attacks, strokes, or death by looking at clinical measures such as cholesterol levels, blood pressure, and body weight. However, for reasons that are not well-understood, these approaches have relatively poor accuracy for predicting such health outcomes.
NIH researchers led by IRP senior investigator Ronald Summers, M.D., Ph.D., in collaboration with researchers at the University of Wisconsin at Madison, leveraged NIH’s Biowulf Supercomputer to analyze over 9,000 computed tomography (CT) scans using five automated artificial intelligence tools. The suite of tools used the CT scans to assess patients’ muscle size, bone mineral density, hardening of arteries, and the volume and accumulation of fat in their livers and around their internal organs. The five CT biomarkers more accurately predicted the patients’ future risk of heart attack, stroke, and death than the patients’ body mass index (BMI) values and Framingham risk scores (FRS), two commonly used means of predicting those sorts of health outcomes.
This new, AI-based method could substantially improve the ability of physicians to predict their patients’ future risk of heart attack, stroke, or death. These more accurate predictions, in turn, could lead to improved treatments and lifestyle adjustments that protect patients’ health. More generally, the study demonstrates the potential of using artificial intelligence to mine the data produced by widely used medical procedures like CT scans in order to enhance patient care.
Pickhardt PJ, Graffy PM, Zea R, Lee SJ, Liu J, Sandfort V, Summers RM. (2020). Automated CT biomarkers for opportunistic prediction of future cardiovascular events and mortality in an asymptomatic screening population. Lancet Digital Health. April;2(4):e192–200. doi: 10.1016/S2589-7500(20)30025-X.
This page was last updated on Friday, January 14, 2022