Automatic Aorta Detection in Non-contrast 3D Cardiac CT Images Using Bayesian Tracking Method

• Published in: Medical Computer Vision. Large Data in Medical Imaging pp. 130 - 137
• Main Authors: Zheng, Mingna, Carr, J. Jeffery, Ge, Yaorong
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 2014
• Series: Lecture Notes in Computer Science
• Subjects: Aorta Detection; Aortic Structure; Automatic Aorta; Bayesian Tracking; Vessel Tracking Methods
• ISBN: 9783319055299; 3319055291
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-05530-5_13

Automatic aorta detection is important for the diagnosis and treatment planning of aortic diseases, such as acute aortic dissection and aneurysm. Manually labeling and tracking the aorta in a large amount of non-contrast CT images are time-consuming and labor-intensive. In this paper, we describe a fully automated method to tackle this problem. We apply General Hough Transom(GHT) to detect the approximately circular shape of the aorta on 2D slices. The k-means clustering algorithm is used to identify two initial points for subsequent vessel tracking. In order to correctly detect the centerline of aorta, the proposed method based on the Bayesian estimation framework incorporates aorta-related prior knowledge. Our approach can handle the variations in the radius along the tubular vessel and the morphological differences of the aortic arch. Initial results on 24 CT datasets from a longitudinal cardiovascular study are encouraging.