To systematically investigate into the relationships between
luminal enhancement, convolution kernel, plaque density, and
stenosis severity in coronary computed tomography (CT) angiography. A coronary phantom including 63
stenoses (
stenosis severity, 10-90%; plaque densities, -100 to 1,000 HU) was loaded with increasing solutions of
contrast material (
luminal enhancement, 0-700 HU) and scanned in an anthropomorphic chest. CT data was acquired with prospective triggering using 64-section dual-source CT; reconstructions were performed with soft-tissue (B26f) and sharp convolution kernels (B46f). Two blinded and independent readers quantitatively assessed
luminal diameter and CT number of plaque using electronic calipers. Measurement bias between phantom dimensions and CT measurements were calculated. Multivariate linear regression models identified predictors of bias. Inter- and intra-reader agreements of
luminal diameter and CT number measurements were excellent (ICCs > 0.91, p < 0.01, each). Measurement bias of
luminal diameter and plaque density was significantly (p < 0.01, each) lower (-12% and 58 HU, respectively) with B46f as opposed to B26f, especially in plaque densities >200 HU. Measurement bias was significantly (p < 0.01, each) correlated (ρ = 0.37-55 and ρ = -0.70-85) with the differences between
luminal enhancement and plaque density. In multivariate models, bias of
luminal diameter assessment with CT was correlated with plaque density (β = 0.09, p < 0.05). Convolution kernel (β = -0.29 and -0.38),
stenosis severity (β = -0.45 and -0.38), and
luminal enhancement (β = -0.11 and -0.29) represented independent (p < 0.05,each) predictors of measurement bias of
luminal diameter and plaque number, respectively. Significant independent relationships exist between
luminal enhancement, convolution kernel, plaque density, and
luminal diameter, which have to be taken into account when performing, evaluating, and interpreting coronary CT angiography.