Quantification of Coronary Artery Calcium on the Basis of Dual-Energy Coronary CT Angiography

Abstract

To evaluate the feasibility of using virtual noncontrast material-enhanced (VNC) computed tomographic (CT) series derived from dual-energy CT imaging studies for coronary artery calcium quantification. This HIPAA-compliant study was institutional review board approved; all patients provided written informed consent. Thirty-six patients prospectively underwent noncontrast-enhanced CT calcium scoring followed by coronary CT angiography performed in dual-energy mode. By using different reconstruction algorithms, three VNC series were generated and evaluated for noise and efficiency of virtual iodine removal. Two readers independently quantified calcium on VNC images and true noncontrast-enhanced conventional calcium scoring series. A leave-one-out cross validation was used to assess the accuracy of calcium score prediction from VNC series by means of linear regression. CT value histograms of the VNC series closely resembled the profile in the true noncontrast-enhanced series. There was excellent correlation between calcium volumes on the VNC series and true noncontrast-enhanced series on a per-patient (r = 0.94, P < .001, n = 36) and per-vessel (r = 0.94, 0.91, and 0.92 for the three coronary arteries, all P < .001, n = 36 each) level. The ability of a linear regression model to predict actual calcium scores from calcium volumes on VNC series was excellent (r = 0.82). Multiethnic Study of Atherosclerosis rankings that were derived from the predicted calcium scores also showed excellent agreement (intraclass correlation coefficient = 0.909). Coronary artery calcium identification and quantification based on dual-energy coronary CT angiographic studies may obviate the need for dedicated CT calcium scoring studies.