Fig. 4
Concordance statistics for AI determination of CACV. a Correlation of expert and AI-calculated CACV volumes by Spearman’s method. R2 = 0.792. Two-tailed ICC = 0.904 (95% CI 0.857–0.936), p = < 0.0001. α = 0.05. The AI and expert CACV volumes have an excellent agreement that reliably follows a generalized linear trend with few outliers. b Bland-Altman plot of the quantitative differences between AI and expert measurements of CACV volume. AI and expert have a mean volume difference of 65.96 for the agreement of CACV volume. The CACV-AI software accurately predicts the CACV volume within acceptable limits compared to the expert. c Concordance of expert and AI reads of CACV binned into CACV > 0 and CACV = 0 groups, representing patients with and without CACV, respectively. Expert and AI excellently agree (Cohen’s kappa = 0.846, 95% CI 0.726–0.965) about CACV status with excellent screening parameters (sensitivity = 0.929, specificity = 0.960). CAC-AI volume, coronary artery calcium-artificial intelligence volume; CAC-Expert volume, coronary artery calcium-expert volume; SD, standard deviation; PPV, positive predictive value; NPV, negative predictive value; CI, confidence interval