Model integrates clinical and quantitative imaging-based variables to predict long-term cardiac risks
FRIDAY, Dec. 27, 2019 (HealthDay News) — A machine learning (ML) model improves the prediction of long-term risks for myocardial infarction (MI) and cardiac death, according to a study published online Dec. 19 in Cardiovascular Research.
Frederic Commandeur, Ph.D., from the Biomedical Imaging Research Institute at the Cedars-Sinai Medical Center in Los Angeles, and colleagues examined the performance of ML for predicting the long-term risk for MI and cardiac death in 1,912 asymptomatic individuals. A fully automated deep learning model was used to quantify epicardial adipose tissue (EAT) volume and density. Clinical covariates, plasma lipid panel measurements, risk factors, coronary artery calcium (CAC), aortic calcium, and automated EAT measures were used to train ML extreme gradient boosting.
The researchers found that 76 events of MI and/or cardiac death occurred during a mean follow-up of 14.5 ± 2 years. A significantly higher area under the receiver operating characteristic curve for predicting events was seen with ML versus atherosclerotic cardiovascular disease (ASCVD) risk and CAC score (ML, 0.82; ASCVD, 0.77; CAC, 0.77). The risk for suffering events was high for individuals with a higher ML score (hazard ratio, 10.38); in a multivariable analysis, including ASCVD-risk and CAC measures, the correlations persisted (hazard ratio, 2.94). For both genders, age, ASCVD-risk, and CAC were prognostically important.
“Machine learning used to integrate clinical and quantitative imaging-based variables improved prediction of MI and cardiac death compared with standard clinical risk assessment and provided explicit explanation of individualized risk prediction,” the authors write.
Copyright © 2019 HealthDay. All rights reserved.
