Comparison of artery-based methods for ordinal grading of coronary artery calcium on low-dose chest computed tomography

Abstract

Objectives To identify the optimal artery-based method for ordinal grading of coronary artery calcium (CAC) on non-electrocardiogram (ECG)-gated low-dose chest computed tomography (LDCT) among three methods. Methods A total of 120 asymptomatic subjects who underwent both LDCT and ECG-gated calcium scoring CT on the same day were retrospectively enrolled. Three cardiothoracic radiologists independently assessed CAC severity on LDCT (1.25-mm and 2.5-mm slice thickness) and classified it into four categories (none, mild, moderate, or severe) using three artery-based ordinal scoring methods (extent-based scoring, Weston scoring, and length-based scoring). Inter- and intra-observer CAC severity agreements of each method were assessed by Fleiss kappa statistics. Agreements between each method and ECG-gated calcium scoring CT were assessed by weighted kappa statistics. Results The inter-observer agreement was highest with length-based method for both 1.25-mm (Fleiss kappa 0.735 for extent-based method, 0.801 for Weston score, and 0.813 for length-based method) and 2.5-mm slice thickness evaluation (Fleiss kappa 0.755 for extent-based method, 0.776 for Weston score, and 0.833 for extent-based method). Agreement across the three grading methods for the same observer was poor to moderate on 1.25-mm (Fleiss kappa 0.379-0.441) and moderate on 2.5-mm thickness evaluation (Fleiss kappa 0.427-0.461). Agreement of CAC severity between each method and ECG-gated calcium scoring CT was highest with the length-based method for all three observers on both 1.25-mm (weighted kappa 0.773-0.786) and 2.5-mm (weighted kappa 0.794-0.825) LDCT images. Conclusion Among the three artery-based ordinal grading methods, the length-based method appears to be the most reliable for evaluating CAC on non-ECG-gated LDCT.