Automated ascending aorta delineation from ECG-gated computed tomography images.

Abstract

The instrumental role of comprehensive geometrical quantification in contemporary, effective descriptions of aortic growth and disease is well established. General or specific purpose algorithms are being developed to provide automatic landmark detection and high accuracy measurements. In the present study, an objective method for automated delineation of the ascending aorta is introduced, based on geometrical properties of the aortic wall. In the proximal ascending aorta, the method identifies the sinotubular junction by tracing the mean surface curvature transition region from the origins of the coronary arteries to the location where the aorta acquires its tubular shape. In the distal ascending aorta, the brachiocephalic artery origin is defined by a split centreline cross section within the brachiocephalic artery-aortic arch bifurcation region. The method's accuracy of detection was quantified against the manual border identification performed by two experienced observers on 3D aortic reconstructions of 44 computed tomography examinations. Median (method, observer) distance and inclination measurements ranged from 0.89 [1.02] mm and 4.66 [5.07]°, respectively, in the proximal border, to 2.18 [2.39] mm and 7.13 [4.77]° in the distal. Accuracy of border detection was found to be high compared to interobserver variability and relevant automatic and manual methodology results previously reported in literature. Delineation of the ascending aorta on a three-dimensional aortic reconstruction with automated identification of the sinotubular junction (proximal border) and of the origin of the brachiocephalic artery (distal border).