Comparison of prognostic value between CAD-RADS 1.0 and CAD-RADS 2.0 evaluated by convolutional neural networks based CCTA

Abstract

ObjectivesThe aim of the present study was to investigate the prognostic value of the novel coronary artery disease reporting and data system (CAD-RADS) 2.0 compared with CAD-RADS 1.0 in patients with suspectedcoronary artery disease (CAD) evaluated by convolutional neural networks (CNN) based coronary computed tomography angiography (CCTA). MethodsA total of 1796 consecutive inpatients with suspected CAD were evaluated by CCTA for CAD-RADS 1.0 and CAD-RADS 2.0 classifications. Kaplan-Meier and multivariate Cox models were used to estimate major adverse cardiovascular events (MACE) inclusive of all-cause mortality or myocardial infarction (MI). The C-statistic was used to assess the discriminatory ability of the two classifications. ResultsIn total, 94 (5.2%) MACE occurred over the median follow-up of 45.25 months (interquartile range 43.53–46.63 months). The annualized MACE rate was 0.014 (95% CI: 0.011–0.017). Kaplan-Meier survival curves indicated that the CAD-RADS classification, segment involvement score (SIS) grade, and Computed Tomography Fractional Flow Reserve (CT-FFR) classification were all significantly associated with the increase in the cumulative MACE (all P < 0.001). CAD-RADS classification, SIS grade, and CT-FFR classification were significantly associated with endpoint in univariate and multivariate Cox analysis. CAD-RADS 2.0 showed a further incremental increase in the prognostic value in predicting MACE (c-statistic 0.702, 95% CI: 0.641–0.763, P = 0.047), compared with CAD-RADS 1.0. ConclusionsThe novel CAD-RADS 2.0 evaluated by CNN-based CCTA showed higher prognostic value of MACE than CAD-RADS 1.0 in patients with suspected CAD.