Automatic heart isolation in 3D CT images

Abstract

In this chapter, we present an automatic heart segmentation algorithm for the diagnosis of coronary artery diseases (CAD). The goal is to visualize the heart from a cardiac CT image with irrelevant tissues such as the lungs, rib cage, pulmonary veins, pulmonary arteries and left atrial appendage hidden so that doctors can clearly see the major coronary artery trees, aorta and bypass arteries if they exist. The algorithm combines a model-based detection framework with data-driven post-refinements to create a mask for a given cardiac CT image that contains only the relevant part of the heart. The marginal space learning [1] technique is used to localize mesh model or landmark points of different cardiovascular structures in the CT volume. Guided by such detected models, local data-driven voxel-based refinements are employed to produce precise boundaries of the heart mask. The algorithm is fully automatic and can process a 3D cardiac CT volume within a few seconds. © 2013 Springer-Verlag.