Machine learning score using stress CMR may predict death in patients with CAD
Click Here to Manage Email Alerts
A score derived from machine learning that included information from stress cardiac magnetic resonance effectively predicted 10-year all-cause death in patients with known or suspected CAD, researchers reported at EuroEcho 2021.
“This is the first study to show that machine learning with clinical parameters plus stress cardiac magnetic resonance (CMR) can very accurately predict the risk of death,” Théo Pezel, MD, researcher in cardiovascular imaging at Johns Hopkins Hospital, said in a press release. “The findings indicate that patients with chest pain, dyspnea or risk factors for cardiovascular disease should undergo a stress CMR exam and have their score calculated. This would enable us to provide more intense follow-up and advice on exercise, diet and so on to those in greatest need.”
Pezel and colleagues analyzed 31,762 patients with known or suspected CAD (mean age, 64 years; 66% men) who were referred to a center in Paris for stress CMR to evaluate myocardial perfusion and wall motion from 2008 to 2018.
During 206,453 patient-years of follow-up (median, 6 years per patient), 8.4% of patients died. Compared with those who survived, those who died were older and more likely to be men, have diabetes, have hypertension, have obesity, be current or former smokers at baseline, have known CAD, have known MI, have had prior PCI, have had prior CABG and have peripheral atheroma, Pezel said during a presentation, noting that those who died were also less likely to have a family history of CAD compared with those who survived.
According to the release, machine learning was used to select the clinical and CMR parameters that could predict death and to build an algorithm based on those parameters to predict 10-year risk for death, resulting in a patient score of 0 (low risk) to 10 (high risk).
Among the parameters that independently predicted death were diabetes (P for interaction = .047), smoking history (P for interaction = .014) and known CAD (P for interaction = .047), Pezel said during the presentation.
The machine learning score had an area under the curve of 0.76 (99.5% CI, 0.74-0.78) for prediction of 10-year mortality in the cohort, which was superior to the C-CMR-10 score (0.68; 99.5% CI, 0.66-0.7), the ESC score (0.66; 99.5% CI, 0.64-0.68), the QRISK-3 score (0.64; 99.5% CI, 0.62-0.66) and the Framingham Risk Score (0.63; 99.5% CI, 0.61-0.65), Pezel said during the presentation.
“Stress CMR is a safe technique that does not use radiation,” Pezel said in the release. “Our findings suggest that combining this imaging information with clinical data in an algorithm produced by artificial intelligence might be a useful tool to help prevent cardiovascular disease and sudden cardiac death in patients with cardiovascular symptoms or risk factors.”
Reference:
- Machine learning predicts risk of death in patients with suspected or known heart disease. www.escardio.org/The-ESC/Press-Office/Press-releases/Machine-learning-predicts-risk-of-death-in-patients-with-suspected-or-known-heart-disease. Published Dec. 11, 2021. Accessed Jan. 4, 2022.
Collapse
Pezel T, et al. Young Investigator Award - Clinical Science. Presented at: EuroEcho 2021; Dec. 9-11, 2021 (virtual meeting).
Read more about
Click Here to Manage Email Alerts