Chest x-ray flags patients who could use targeted MACE prevention

Monday, November 30 | 1:30 p.m.- 2:30 p.m. | M6-SSCA03-1 | Room E352 

Chest x-ray-based screening could help identify individuals who need targeted prevention for major adverse cardiovascular events (MACE).

This could be especially useful for patients who don't have the lab results available to calculate the Predicting Risk of Cardiovascular Disease EVENTs (PREVENT) risk score, according to session presenter Jay Chandra, MD, of Harvard University in Boston.

"Guidelines for cardiovascular prevention hinge on the PREVENT 10-year MACE risk score, which requires serum cholesterol and other lab data generally only available if the patient has consistent longitudinal care," Chandra and colleagues noted. "Because chest radiographs are widely available, deep-learning models applied to routine chest x-rays can enable opportunistic cardiovascular risk assessment."

In this afternoon session, Chandra will share results from a study that used DL models to predict coronary artery calcium (CAC) scores, cardiac radiomics, and MACE from chest x-ray images. For the research, Chandra's team reconstructed posteroanterior and lateral chest x-rays from low-dose lung cancer screening chest CTs from 16,464 National Lung Screening Trial (NLST) participants; it extracted CAC and cardiac radiomics features from each CT using CT-based segmentation models, then trained RESNET-18 convolutional neural networks (CNN) to predict CAC and cardiac radiomics features from the reconstructed chest x-rays.

In the study's second stage, the group further honed the CNNs to assess real chest x-rays from 3,280 patients in the Mass General Brigham Biobank who had both a chest x-ray and a chest CT within three years. It combined radiomics features into a single composite risk score and tested those scores in 19,693 additional Mass General Brigham Biobank patients with chest x-rays and long-term follow-up for MACE.

What did the team discover? PREVENT, chest x-ray-predicted CAC, and chest x-ray-derived composite radiomics risk scores had concordances of 0.66, 0.64, and 0.65 respectively in predicting incident MACE.

Get the rest of the details by attending this Monday afternoon session.

Page 1 of 2
Next Page