Quantitative measurements of coronary arterial stenosis, plaque geometry, and composition are highly reproducible with a standardized coronary arterial computed tomographic approach in high-quality CT datasets

Abstract

Computed tomographic (CT) coronary angiography provides a noninvasive method for coronary plaque detection and quantification, but data are limited on reproducibility of a quantitative evaluation. Intrarater and interrater reliability of a semiquantitative and highly standardized, fully quantitative approach was evaluated in 480 coronary segments in 30 patients. Quantitative vessel-wall and plaque geometrical parameters (minimal lumen diameter [MLD], minimal lumen area [MLA], percentage of atheroma volume [PAV], and remodeling index [RI]) and compositional parameters (calcified plaque volume [CAP] and % of CAP [%CAP], noncalcified plaque [NCP] and % of NCP [%NCP], high-density NCP volume [HD-NCP] and % of HD-NCP [%HD-NCP] and low-density NCP volume [LD-NCP] and % of LD-NCP [%LD-NCP]) were measured. Semiquantitative agreement was evaluated by weighted κ; quantitative agreement was evaluated by concordance correlation coefficient (CCC) and Bland-Altman analysis. Intraobserver agreement for MLD, MLA, and RI was excellent (CCC: 0.96, 0.96, and 0.84, respectively). Intraobserver agreement for %CAP, %HD-NCP, and %LD-NCP was also excellent (CCC: 0.99, 0.98,and 0.96, respectively). Interobserver agreement for MLD, MLA, PAV and RI was excellent (CCC: 0.98, 0.99, 0.96,and 0.86, respectively). Interobserver agreement for %CAP, % NCP, %HD-NCP, and %LD-NCP was also excellent (CCC: 0.99, 0.99, 0.98,and 0.90, respectively), and mean differences were small. Quantitative analysis showed statistically significant differences in both geometrical and compositional parameters between normal segments and those with plaque. Standardized, quantitative analysis of coronary CTA datasets is reproducible for the measurement of plaque geometrical and compositional parameters and can quantify differences between normal and abnormal segments in high-quality datasets.