Automatically Seeded Region Growing Approach for Automatic Segmentation of Ascending Aorta

Region Growing is frequently used for segmentation purposes especially for objects with homogenous intensity. The problem is how the region growing approach begins, how it proceeds and how it finally ends. In this paper we present a modified region growing approach for the purpose of automatic segmentation of the ascending aorta from Computed Tomography Angiography (CTA) images. The ascending aorta in CTA images has homogenous intensity which makes region growing a good approach for its segmentation. Moreover the approach presented in this paper is fully automatic to suit clinical non-invasive diagnosis purposes. For developing our region growing approach; the geometric, spatial and intensity features of the ascending aorta has been exploited. These features provide the region growing methodology with the automatic seed from which it begins and then continues to segment the whole ascending aorta from the aortic arch down to ostia points. The proposed algorithm is tested and validated on the Computed Tomography Angiography database provided by the Rotterdam Coronary Artery Algorithm Evaluation Framework. The proposed algorithm has three main advantages: 1) it's fully automatic, i.e. no user interaction needed, 2) real-time performance even with large datasets, 3) robust since parameters values are the same for all the tested datasets.