AI-Enabled Opportunistic Detection of Cardiovascular Disease Risk from Chest Radiographs in the PLCO Cohort
Principal Investigator
Name
Morteza Naghavi
Degrees
M.D.
Institution
HeartLung Technologies
Position Title
President
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCOI-2034
Initial CDAS Request Approval
Mar 30, 2026
Title
AI-Enabled Opportunistic Detection of Cardiovascular Disease Risk from Chest Radiographs in the PLCO Cohort
Summary
Cardiovascular disease (CVD) remains the leading cause of mortality worldwide, yet most individuals at risk remain undiagnosed until late stages of disease. Chest radiography (CXR) is among the most widely performed imaging tests globally, but currently provides limited quantitative cardiovascular risk assessment. Advances in artificial intelligence (AI) enable extraction of latent phenotypic signals from medical images that are not apparent to human readers.
The objective of this project is to develop and validate AI models capable of identifying individuals at elevated risk of cardiovascular events using chest radiographs from the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial. Our team has previously demonstrated that automated volumetric measurements of cardiac chambers derived from non-contrast chest CT scans can predict major cardiovascular outcomes including heart failure, atrial fibrillation, stroke, and cardiovascular mortality. These findings suggest that structural cardiac remodeling detectable on routine imaging contains powerful prognostic information.
In this project, we will extend this concept to chest radiography by developing deep learning models that estimate cardiovascular risk and detect imaging biomarkers of cardiomegaly, chamber enlargement, vascular remodeling, and other cardiopulmonary phenotypes associated with future cardiovascular events. The PLCO cohort provides a uniquely valuable dataset because it includes a large number of screening chest radiographs linked to long-term clinical outcomes.
The resulting models could enable opportunistic cardiovascular risk screening using routine chest radiographs obtained for any clinical indication. If successful, this approach could transform the CXR into a scalable population-level screening tool for early detection of individuals at high risk for cardiovascular disease.
Aims
• Aim 1 – Develop deep learning models to quantify cardiovascular imaging biomarkers from chest radiographs.
We will train AI models to detect cardiomegaly, chamber enlargement, vascular remodeling, and other structural cardiopulmonary features associated with cardiovascular disease using PLCO chest radiographs.
• Aim 2 – Develop AI models to predict future cardiovascular outcomes from chest radiographs.
Using linked clinical outcome data available in PLCO, we will train models that estimate risk of major cardiovascular events including heart failure, atrial fibrillation, stroke, and cardiovascular mortality.
• Aim 3 – Translate CT-derived cardiovascular phenotypes to chest radiography.
Our prior work has demonstrated that automated cardiac chamber volumetry from non-contrast chest CT scans predicts major cardiovascular outcomes. We will develop methods to infer analogous structural phenotypes from CXR images and evaluate their predictive performance.
• Aim 4 – Evaluate the potential of opportunistic cardiovascular screening using routine chest radiographs.
We will assess whether AI-derived features from CXR can identify high-risk individuals who might benefit from further evaluation or preventive interventions.
Collaborators
Morteza Naghavi HeartLung Corporation
Kyle Atlas HeartLung Technologies
Hamed Zarei HeartLung Technologies