Identification of coronary arteries in CT images by Bayesian analysis of geometric relations among anatomical landmarks. (December 2019)

• Record Type: Journal Article
• Authors: Jeon, Byunghwan; Jang, Yeonggul; Shim, Hackjoon; Chang, Hyuk-Jae
• Is Part Of: Pattern recognition. Volume 96(2019:Dec.)
• Publication Date: 2019-12
• Subjects: Computed tomography angiography -- Bayesian -- Localization -- Coronary artery -- Multiple target -- Curve analysis -- Curvature and torsion; AA ascending aorta -- CCA connected component analysis -- CTA computed tomography angiography -- CTO chronic total occlusion -- CVD cardiovascular disease -- DA descending aorta -- FN false negative -- FOV field of view -- FP false positive -- FPR false positive rate -- GT ground truth -- LA left atrium -- LAA left atrial appendage -- LAD left anterior descending artery -- LCA left coronary artery -- LCX left circumflex artery -- LGM local geometric method -- MSL marginal space learning -- OP outermost point -- RCA right coronary artery -- ROC receiver operating characteristic -- ROI region of interest -- TN true negative -- TP true positive -- FP false positive -- FN false negative -- TPR true positive rate; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2019.07.003
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: Our fully automated method localizes multiple objects in CT angiography images. Bayesian approach models hearts geometric relations as prior distributions. Method is applied to identify three coronary arteries. Experiments demonstrate robustness and accuracy compared with existing methods. Method is robust on challenging cases such as vessels with chronic total occlusion. Abstract: We propose a robust method for the identification of coronary arteries in computed tomography angiography (CTA) images. Utilizing geometric relations among the target and reference objects, which are assumed to follow a Gaussian distribution, an anatomic and geometric model is designed by Bayesian inference, which provides robust geometric priors for the target object localization. As a prerequisite process for the identification of coronary arteries, partially broken coronary artery segments found in CTA images are grouped and reconnected by geometric analysis of higher order curves connecting the broken segments. The geometric properties such as curvature and torsion represent naturalness and consistency between the vessel segments. As a problem to identify coronary arteries from CTA images, we demonstrate the robustness and accuracy of the proposed method in comparison with existing methods including commercial workstations on a variety of CTA cases.