Heart HealthAI Detects Hidden Coronary Calcium in People Told They Have a Zero Score
The standard Agatston coronary calcium score (CAC) has a well-known blind spot: its fixed 130 HU threshold and thick 2.5–3 mm slices miss small, low-density plaques. Researchers developed Agatston-2.0, an AI framework using automated 3D coronary segmentation and continuous voxel-wise calcium quantification with no fixed threshold. Applied to 3,965 participants with a conventional CAC=0 from MESA and the Framingham Heart Study, the AI detected meaningful calcification in 21.7%. Over 20 years, those flagged by the AI had nearly double the coronary heart disease incidence compared to those with a truly clean scan. The AI score also strongly predicted conversion to a positive conventional CAC score, suggesting it captures real early-stage atherosclerosis, not noise.
