AI-enabled Cardiac Chambers Volumetry and Calcified Plaque Characterization in Coronary Artery Calcium (CAC) Scans...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Coronary artery calcium (CAC) scans traditionally use the Agatston Score for coronary heart disease (CHD) prediction.
• CAC scans contain underutilized data for predicting a broader range of cardiovascular disease (CVD) events.

Purpose of the Study:

• To evaluate if artificial intelligence (AI) algorithms applied to CAC scans can enhance the prediction of all CVD events, not just CHD.
• To compare the predictive performance of AI-analyzed CAC scans (AI-CAC) against the standard Agatston Score.

Main Methods:

• AI-enabled automated plaque characterization and cardiac chamber volumetry were applied to CAC scans from 5830 individuals in the Multi-Ethnic Study of Atherosclerosis (MESA).
• The study analyzed 15-year CVD event outcomes, comparing the predictive discrimination of AI-CAC versus the Agatston Score using time-dependent area under the curve (AUC) and Net Reclassification Index (NRI).

Main Results:

• AI-CAC demonstrated significantly higher AUC values across 1-, 5-, 10-, and 15-year follow-ups compared to the Agatston Score (e.g., 0.816 vs. 0.729 at 15 years, p<0.0001).
• The category-free Net Reclassification Index (NRI) for AI-CAC was consistently superior to the Agatston Score (0.29-0.31, p<.0001).
• Specific AI-CAC plaque characteristics (number, location, density) further improved risk prediction, particularly in the CAC 1-100 cohort (NRI 0.342).

Conclusions:

• AI-CAC significantly and consistently improves the prediction of all cardiovascular events over 15 years compared to the traditional Agatston Score.
• This AI-driven approach enhances risk stratification for cardiovascular disease, offering potential for earlier and more accurate interventions.