Fully automated derivation of coronary artery calcium scores and cardiovascular risk assessment from contrast medium-enhanced coronary CT angiography studies.

Objectives: Performance evaluation of a fully automated system for calculating computed tomography (CT) coronary artery calcium scores from contrast medium-enhanced coronary CT angiography (cCTA) studies. Methods: One hundred and twenty-seven patients (58 ± 11 years, 71 men) who had undergone cCTA as well as an unenhanced CT calcium scoring study where included. Calcium scores were computed from cCTA by an automated image processing algorithm and compared with calcium scores obtained by standard manual assessment of unenhanced CT calcium scoring studies. Results were compared vis-a-vis (1) absolute calcium score values, (2) age-, gender- and race-dependent percentiles, and (3) commonly used calcium score risk classification categories. Results: One hundred and nineteen out of 127 (93.7%) studies were successfully processed. Mean Agatston calcium score values obtained by traditional non-contrast CT calcium scoring studies and derived from contrast medium-enhanced cCTA did not significantly differ (235.6 ± 430.5 vs 262.0 ± 499.5; P > 0.05). Calcium score risk categories and Multi-Ethnic Study of Atherosclerosis (MESA) percentiles showed very high correlation (Spearman rank correlation coefficient = 0.97, P < 0.0001/0.95, P < 0.0001) between the two approaches. Conclusions: Calcium score values automatically computed from cCTA are highly correlated with standard unenhanced CT calcium scoring studies. These results suggest a radiation dose- and time-saving potential when deriving calcium scores from cCTA studies without a preceding unenhanced CT calcium scoring study. Key Points: • CT coronary calcium scoring is now widely used for cardiac risk stratification • Derivation of calcium scores from coronary CT angiography saves time and radiation • Automatically derived scores are comparable to conventional coronary artery calcium scores • Patient risk stratification is similar, whether using automatically derived or conventional scores [ABSTRACT FROM AUTHOR]