Automatic segmentation of the left-ventricular cavity and atrium in 3D ultrasound using graph cuts and the radial symmetry transform

Abstract

In this paper, we propose a graph-based method for fully-automatic segmentation of the left ventricle and atrium in 3D ultrasound (3DUS) volumes. Our method requires no user input and can segment volumes with open and closed mitral valves. We utilize the radial symmetry transform to determine a central axis along which the 3D volume is warped into a cylindrical coordinate space. A graph is constructed for the volume in this space and a min-cut algorithm is applied to segment the left ventricle and atrium from the background. Since segmentation in the cylindrical coordinate space is defined as finding a boundary between the left (interior) and right (exterior) sides, we obviate the need for user specified seeds. The segmented results are transformed back to the Cartesian coordinate space. Experiments using intraoperative 3D ultrasound data show promising results.