The Agatston-derived density score for coronary artery calcification demonstrates substantially lower reliability across computed tomography scanners and commercial analysis software than the standard Agatston score, according to a comparative investigation evaluating both phantom models and human subjects. The researchers assessed the density score—an Agatston-derived measurement recognized as an independent predictor of coronary events—to determine whether its quantitative values remain consistent across different imaging hardware and analytical software platforms.
To evaluate measurement consistency under controlled laboratory conditions, the investigators imaged a coronary artery calcification phantom using two distinct CT systems, designated scanner A and scanner B. The resulting scan datasets were processed and evaluated using four commercial coronary artery calcification scoring software packages, labeled a’, b’, c’, and d’. The experimental protocol also examined partial volume effects through repeated imaging acquisitions incorporating incremental positional displacements as well as varying reconstruction slice thicknesses.
In addition to the controlled phantom experiments, the study incorporated a retrospective clinical analysis of 98 human subjects, comprising 49 men, with an overall mean age of 66 ± 7 years. Across this patient cohort, recorded Agatston scores spanned a wide baseline range from 1 to 1546. The researchers analyzed this clinical imaging data to assess the comparative reliability of the Agatston score, the volume score, and the Agatston-derived density score, calculating intra-class correlation coefficients to quantify measurement agreement across the software tools.
The phantom evaluations revealed stark contrasts in measurement stability between the scoring methods across different platforms. The traditional Agatston score demonstrated high reliability across both CT scanners, yielding intra-class correlation coefficients between 0.97 and 0.98, and achieved near-perfect agreement across all software applications, with intra-class correlation coefficients ranging from 0.99 to 1.00. In contrast, one commercial application, software a’, generated erroneous density score values and was excluded from subsequent statistical evaluations. Among the remaining applications, the density score showed lower inter-scanner reliability, with coefficients between 0.43 and 0.62, and cross-software reliability spanning 0.81 to 0.87.
Physical adjustments during imaging further degraded the consistency of the density metric in the phantom trials. Introducing incremental positional displacements reduced density score reliability to intra-class correlation coefficients between 0.72 and 0.81, which remained inferior to the Agatston score’s stability of 0.97 to 0.98. In the human cohort analysis, the divergence was even more pronounced: while the Agatston score maintained excellent software reliability with an intra-class correlation coefficient of 0.99, the density score demonstrated poor overall reliability, recording an intra-class correlation coefficient of only 0.29.
Based on these findings, the study authors concluded that substantial variations in density score measurements occur across specific software platforms, resulting in reliability that is markedly inferior to that of the conventional Agatston score. The investigators noted that this low measurement reliability for traditional Agatston-derived density scoring could compromise its practical ability to predict future coronary events. Consequently, the researchers emphasized that software- and scanner-dependent variability must be carefully accounted for during the study design phase and the interpretation of findings in coronary artery calcification cohort studies.
Subject of Research: Measurement reliability of CT coronary artery calcification density versus Agatston scores
Article Title: Low reliability of CT density score versus the Agatston score of coronary artery calcification – a phantom and human study
Article References: Wu, H.-F., Huang, E. P.-Y., Chen, Y.-C., Wang, C.-C., Liu, H.-H., Chen, C.-M., & Wu, M.-T. (2026). Low reliability of CT density score versus the Agatston score of coronary artery calcification – a phantom and human study. BMC Cardiovascular Disorders. https://doi.org/10.1186/s12872-026-06956-6
Image Credits: AI Generated
DOI: 10.1186/s12872-026-06956-6
Keywords: coronary artery calcification, Agatston score, density score, computed tomography, intra-class correlation coefficient, phantom study, coronary events, software reliability, scanner variability, partial volume effects, volume score, reconstruction slice thickness, cardiovascular imaging, cardiac risk prediction, cohort studies, measurement reliability
News Source: Ophelia Keating. (November 13, 2026). Coronary Artery Density Scores Show Low Reliability Compared With Agatston Scoring. Scienmag.



