Machine learning is more accurate at predicting the long-term risk of potentially life-threatening cardiac events compared to standard clinical assessments, according to a new study published in Cardiovascular Research.
A team from Cedars-Sinai Medical Center and Friedrich-Alexander University Erlangen-Nuremberg in Germany employed AI to interpret clinical and cardiovascular imaging-based metrics to greatly improve the prediction of heart attack and cardiac death. With further research, this approach could help personalize cardiovascular risk assessments.
“With constant growth of artificial intelligence across various disciplines, especially in cardiology, robust cardiac risk assessment will benefit from quantification of automated imaging biomarkers, increasing the relevant information available for clinical decision making,” Frederic Commandeur with Cedars-Sinai’s Biochemical Research Institute in Los Angeles, and colleagues wrote.
“These promising results suggest that ML has a potential for clinical implementation to improve risk assessment,” the team added.
Commandeur et al. sought to determine if a machine learning “boosting” approach—combining clinical information, visually determined CT angiography measures and epicardial adipose tissue parameters—could better predict severe cardio events. They noted that AI-based approaches have been used in the past to predict all-cause mortality in those with suspected coronary artery disease, but never for global cardio risk stratification in asymptomatic patients.
The researchers prospectively analyzed data from more than 1,900 patients involved in the imaging arm of the EISNER (Early Identification of Subclinical Atherosclerosis by Noninvasive Imaging Research) trial. Each individual underwent coronary artery calcium scoring via cardiac CT scanning, and included 15-year follow-up information for review.
Machine learning-produced risk scores were closely aligned with real-life observed outcomes, the team noted. And, those predictions were more accurate than atherosclerotic cardiovascular disease scores—the clinical standard used by cardiologists—as well as coronary artery calcium scoring.
The research was limited by its low number of subjects and small rate of observed events (only 76 patients experienced heart attack or cardiac death during follow-up), the researchers acknowledged. Going forward, however, they see big things in store for machine learning and cardiovascular care.
“In the future, we foresee machine learning working in the background of standard coronary calcium scoring and electronic reporting software, gathering the variables automatically and allowing ‘on-the-fly’ risk score computation by integrating all relevant measures of clinical risk and imaging biomarkers from coronary calcium scoring scans,” the researchers concluded. “We expect that such personalized prevention tools would help physicians find the right answers for their patients, whose ‘lives and medical histories shape the algorithms.’”