Machine learning models of vessel features from coronary computed tomography (CT) angiography better discriminate patients with versus without subsequent death and cardiovascular events, according to a study published online June 25 in Radiology.

Kevin M. Johnson, M.D., from the Yale University School of Medicine in New Haven, Connecticut, and colleagues analyzed coronary CT angiography and developed four model types using machine learning. For comparison, five conventional vessel scores were computed. A total of 6,892 patients underwent coronary CT angiography between February 2004 and November 2009.

The researchers identified 380 deaths of all causes; 70 patients died of coronary artery disease and 43 had a nonfatal myocardial infarction. The area under the receiver operating characteristic curve (AUC) was 0.77 for all-cause mortality for machine learning versus 0.72 for the conventional Coronary Artery Disease Reporting and Data System (CAD-RADS) score. For coronary artery heart disease deaths, the AUC was 0.85 and 0.79 for machine learning and CAD-RADS, respectively. In deciding to start statins, if the choice is made to tolerate treatment of 45 patients to prevent one death from coronary disease, using the machine learning score ensures that 93 percent of patients with events will be administered the drug compared with 69 percent using CAD-RADS.

"Machine learning can improve the use of vessel features to discriminate between patients who will have an event and those who will not," the authors write.

-----------------------------------------------------------------------
Editor's Note:

If you have any suggestion to the content, 

please email: Julia.Zhang@ubmsinoexpo.com