Agatston-2.0

Agatston-2.0: A next-generation AI-based coronary calcium quantification approach to improve risk stratification among individuals with zero Agatston scores – Part I : An AI-CVD Study within the Multi-Ethnic Study of Atherosclerosis and Framingham Heart Study

Background: A coronary artery calcium (CAC) score of zero using the conventional Agatston scoring method (Agatston-1.0) is associated with a very low risk of cardiovascular events, a concept commonly referred to as the "Power of Zero." However, a small proportion of individuals with CAC=0 still develop coronary heart disease (CHD). Agatston-1.0 relies on thick CT slices (2.5–3.0 mm) and a fixed attenuation threshold (≥130 Hounsfield Units), which may fail to detect early, small, low-density, or partially calcified coronary plaques. Agatston-2.0 is a novel artificial intelligence (AI) framework that performs automated coronary artery segmentation and continuous voxel-wise calcium quantification without using fixed density thresholds. The framework is compatible with CT scans acquired with slice thicknesses of ≥0.2 mm and generates an AI-derived coronary artery calcium score (AI-CAC).

Objectives: To evaluate the prognostic value of Agatston-2.0 for cardiovascular risk stratification among individuals with a baseline conventional CAC score of zero.

Methods: A total of 3,965 participants with baseline CAC=0 were pooled from the Multi-Ethnic Study of Atherosclerosis (MESA; n=2,816) and the Framingham Heart Study (FHS; n=1,149). Associations between AI-CAC findings and incident coronary heart disease were assessed using Cox proportional hazards models with up to 20 years of follow-up.

Results: Agatston-2.0 detected an AI-CAC score greater than zero in 862 participants (21.7%) who had been classified as CAC=0 by the conventional Agatston method. Participants with AI-CAC>0 experienced a significantly higher 20-year incidence of coronary heart disease compared with those with AI-CAC=0 (7.7% vs. 3.8%, p<0.0001). After adjustment for traditional cardiovascular risk factors, AI-CAC>0 remained independently associated with incident CHD (hazard ratio [HR] 1.71; 95% confidence interval [CI], 1.18–2.47). In addition, AI-CAC>0 was a strong predictor of progression to a positive conventional CAC score during follow-up (adjusted HR 1.95; 95% CI, 1.70–2.24).

Conclusions: The Agatston-2.0 AI framework identifies clinically meaningful coronary calcification in individuals who are classified as CAC=0 using the conventional Agatston scoring method. If these findings are validated in additional prospective cohorts, Agatston-2.0 has the potential to redefine coronary calcium scoring and become the new clinical standard for cardiovascular risk assessment.