Article Abstract

Fully automatic volume quantification of aortic valve calcium in coronary computed tomography angiography

Authors: Xinpei Gao, Pieter H. Kitslaar, Michiel A. de Graaf, Alexander Broersen, Shengxian Tu, Boudewijn P. F. Lelieveldt, Arthur J. H. A. Scholte, Jouke Dijkstra, Johan H. C. Reiber


Background: Aortic valve calcium (AVC) quantification has significant meaning for the prognosis of coronary and cardiovascular disease. The extent of calcium is also correlated to the occurrence of paravalvular regurgitation after the procedure of trans-catheter aortic valve replacement (TAVR). In this study, we evaluated a new fully automatic approach to detect and quantify the presence and degree of AVC.
Methods: This retrospective study consisted of 68 patients who went through computed tomography (CT) angiography because of potential coronary artery disease. The first step is the automatic detection of the aortic root. Next, double-oblique images were automatically reconstructed to properly display the aortic valve. Finally, the calcium on the valve was segmented by thresholding, and the volume score was calculated. For the validation, a reference standard was set by an experienced cardiologist using a semi-automatic method.
Results: In this study, the median of the reference standard calcium volume was 0.54 mm3 (range: 0 to 1131.23 mm3, 25th–75th percentile: 0 to 12.00 mm3). The median of the automatic calcium volume score was 0.46 mm3 (range: 0 to 1136.17 mm3, 25th–75th percentile: 0 to 11.22 mm3). The median difference was 1.82 mm3 (25th–75th percentile: 0 to 5.08 mm3), with the Spearman rank correlation coefficient of 0.81 (P<0.001). The Bland-Altman analysis illustrated that the bias between the automatic result and the reference measurement was −1.1 mm3 with limits of agreement between −16.2 and +14.1 mm3. The specificity of our automatic approach was 82.4%; the sensitivity was 85.3%. The mean processing time was 90 seconds.
Conclusions: This study demonstrated that a new fully automatic approach for AVC quantification is accurate.

Article Options

Download Citation