Fig. 4

Bland–Altman plots of Agatston scores of CAC. Dashed lines show 95% limits of agreement. The average score from ECG-gated CAC scoring and AI-based automated quantification on non-gated CT is plotted against the difference between the two methods, the difference is Agatston score on chest CT minus the measurement on cardiac CT. The plots reveal underestimated calcium scores with the automated method on non-gated CT and an increasing difference with a high average score. A regression of absolute difference (X) is performed, and 95% limits of agreement (Y) are calculated as follows: Y = (− 8.25 + 5.26 *X^0.5)*1.96 *(π/2)^0.5 for Total Agatston score (A), Y = (− 17.53 + 10.96*X^0.5)*1.96 *(π/2)^0.5 for Agatston scores of Left main trunk (LM) (B), Y = (− 19.95 + 7.38*X^0.5)*1.96 *(π/2)^0.5 for Agatston scores of left anterior descending artery (LAD) (C), Y = (− 50.70 + 13.76*X^0.5)*1.96 *(π/2)^0.5 for Agatston scores of circumflex (CX) (D) and Y = (− 19.22 + 10.07*X^0.5)*1.96 *(π/2)^0.5 for Agatston scores of Right coronary artery (RCA)(E) respectively