Comparison between different kernel reformatting filters in 3D quantitative analysis of MDCT coronary angiography

Abstract

Coronary angiography with multidetector-row computed tomography (MDCT-CA) allows quantification of coronary artery stenosis with a high level of accuracy; however, a better estimation of stenosis can be achieved by using appropriate reformatting filters, especially in stents and calcified segments. Quantitative computed tomography angiography (QCTA) is intended to overcome the limitations of the visual score. The aim of this study was to evaluate the accuracy of QCTA with different filters in comparison with quantitative coronary angiography (QCA) and visual score. Two blinded operators visually scored 17 consecutive patients referred for MDCT-CA with a per-segment analysis. The degree of stenosis was classified as 0-20%, 20-50% (wall irregularities), 50-70% (significant disease) and 70-100% (vessel occlusion). Each segment was then analysed using the electronic callipers of the QCTA system with 15 different filters. No contour editing was performed. Data were compared with QCA and conventional coronary angiography (CCA). Comparison between QCTA, visual score and QCA were performed using Spearman's rank correlation. Of 25 segments analysed (mean 1.4 diseased segment per patient), 375 measurements were considered. Good correlation was found between the visual score and QCA [Pearson correlation coefficient (rho=0.852; p<0.0001)] and between QCA and CCA (rho=0.804; p<0.0001). Moderate correlation was found between QCA and QCTA only using two filters (rho=0.444; p<0.0001 for YA filter and rho=0.450; p<0.0001 for YB filter). Overall QCTA accuracy is low if contour editing is not applied, especially in calcified vessels. Certain filters can help to better estimate the exact percentage of stenosis.