CT comparison of visual and computerised quantification of coronary stenosis according to plaque composition

Abstract

To compare the diagnostic performance of computerised quantification with visual assessment for the detection of significant coronary stenosis using MDCT, and to determine the impact of plaque composition on diagnostic procedure. We retrospectively evaluated 1564 coronary segments of 127 patients who underwent 64-slice MDCT and quantitative coronary angiography (QCA). The lesions were analysed with both methods of visual assessment and computerised quantification using an automatic vessel contour detection tool, and the results were compared with the QCA results. Plaques detected with MDCT were classified as calcified, mixed, and non-calcified according to plaque composition. The sensitivity and PPV of visual assessment (computerised quantification) were 95% (86%) and 76% (81%), respectively. Bland-Altman analysis demonstrated a mean difference of -5.2 ± 21.6% for all lesions, 2.2 ± 23.7 for calcified plaques, and -12.0 ± 17.2% for non-calcified plaques. The correlation coefficients and limits of agreement between CTA and QCA were 0.48 and ± 46.5% (0.74 and ± 33.7%) in the lesions with calcified plaques (non-calcified plaques). The computerised quantification decreases the sensitivity due to underestimation of non-calcified plaques compared with visual assessment, and had a poorer correlation and a larger limit of agreement in the lesions with calcified plaque compared with non-calcified plaques.