BiomarkersHumans3,965 participantsCohort study

AI detects coronary calcium in adults with zero scores

In individuals with conventional zero scores, AI-detected calcium tracked a 7.7% 20-year coronary heart disease incidence compared with 3.8% in those without.

Figure 1 from American Journal of Preventive Cardiology — Figure 1Naghavi et al.

American Journal of Preventive Cardiology4 Oct 2026, 05:45 BST

In a pooled analysis of 3,965 human participants from the Multi-Ethnic Study of Atherosclerosis and the Framingham Heart Study with conventional coronary artery calcium scores of zero, an artificial intelligence tool identified previously undetected calcification. Conventional Agatston scoring uses fixed thresholds and thick slices that can miss early, small, or low-density plaques. The new Agatston-2.0 method applies continuous voxel-wise quantification without fixed attenuation thresholds to detect early arterial changes.

The AI framework detected calcium in 862 participants, representing 21.7% of the cohort. Over up to 20 years of follow-up, participants with positive AI-derived calcium scores experienced higher coronary heart disease incidence (7.7% versus 3.8%) and higher adjusted risk of incident disease (hazard ratio 1.71). A positive AI score also predicted progression to a positive conventional calcium score (adjusted hazard ratio 1.95).

Why it matters

Coronary calcification is a hallmark of vascular aging and cardiovascular risk. Identifying subclinical calcification in people previously thought to have zero burden could improve early detection of age-related arterial disease.

Caveats

The study is observational and pooled data from two cohorts, meaning the AI framework requires validation in additional populations before it can be used as a standard clinical measure.