To develop and evaluate a new method based on a 3D adaptive stochastic method for detection and quantification of coronary artery calcifications (CAC) with multi-slice CT.
METHOD AND MATERIALS
A new 3D method was developed to detect CAC without a predefined threshold, which is required with the Agatston based method (ABM). The new method is based on the measurement of the proportion of calcium in each voxel using a modified Expectation-Maximisation Algorithm (mEMA). A stationary cardiac phantom containing 6 calcium elements of predetermined volume and mass was used to evaluate reproducibility and accuracy of the method regarding partial volume effect and noise. Fifteen different protocols (16-row CT scan) were used to scan the phantom 90 times. Volumes and masses obtained from the mEMA were compared with those from the ABM with the standard 130HU threshold for the different protocols. Coefficients of variation were calculated for each protocol and each phantom element regarding the volume and the mass. Reproducibility was then evaluated in 35 patients who were scanned twice.
A significant difference was found between volumes obtained by the mEMA and the ABM (p=0.01). The mean error for the ABM measurements was 87.5%, compared with 6.1% for the mEMA. The mean coefficients of variation in volume and mass measurements were 9.5±9% and 2.5±1%, respectively with the mEMA and 79±50% and 8.7±10% with the ABM. These results were confirmed in patients with 287mm3 difference in volume of plaques measured with the mEMA compared with a 739mm3 difference measured with the ABM.
Our results clearly demonstrate that reproducibility and accuracy of the ABM are poor. The new adaptive technique, by design is less sensitive to partial volume effect and noise and improves reproducibility in detecting and quantifying CAC.
J.D.,X.Y.,X.L.,M.A.: J Dehmeshki, Xujiong Ye, Xin Yu Lin and M Abaei work for Medicsight PLC company