The assessment of coronary arteries is an essential step when diagnosing coronary heart diseases. There exists a wide range of specialized algorithms for the segmentation of the coronary arteries in Computed Tomography Angiography datasets. In general, these algorithms have to be initialized by manually placing a seed point at the origins of the coronary arteries or within the ascending aorta. In this paper we present a fast and robust algorithm for the automatic detection of the ascending aorta in Computed Tomography Angiography datasets using a two-level threshold ray propagation approach. We further combine this method with an aorta segmentation and coronary artery tree detection algorithm to achieve a fully automatic coronary artery segmentation.