Opportunistic Cardiovascular Risk Assessment Using Routine Head CT in the Emergency Department.

Area of Science:

• Radiology
• Cardiology
• Artificial Intelligence

Background:

• Noncardiac computed tomography (CT) scans, including head CT (CTH), are common and may offer cardiovascular risk insights.
• The utility of CTH for cardiovascular risk assessment has not been previously studied.

Purpose of the Study:

• To develop and validate deep learning models for predicting incident cardiovascular disease (CVD) and estimating coronary artery calcium (CAC) scores from CTH.
• To assess the performance of these models against traditional clinical risk factors.

Main Methods:

• A retrospective cohort study of 27,990 patients undergoing CTH for CVD prediction and 2,313 for CAC scoring.
• Deep learning models extracted imaging features from CTH; risk factors were from electronic health records.
• Performance was compared against the American Heart Association PREVENT risk model using concordance index (C-index) and area under the receiver operating characteristic curve.

Main Results:

• The CTH model achieved a C-index of 0.82 for CVD prediction, outperforming the PREVENT model (C-index 0.75).
• For CAC estimation, the CTH+PREVENT model yielded a C-index of 0.76 and AUC of 0.80 for CAC >100.
• 15.7% of patients were reclassified, with younger, higher-risk individuals showing more vascular calcifications and brain infarcts.

Conclusions:

• Routine CTH scans enhance traditional risk factors for cardiovascular risk stratification.
• Deep learning analysis of CTH identifies subclinical disease, particularly in younger patients.
• Clinical integration of CTH analysis can improve CVD detection and prevention cost-effectively.