(A) segmentation method of left ventricle using hierarchical edge maps in MR cardiac long-axis images

Abstract

Researchers who analyze cardiac images have focused on short-axis view images. However, the segmentation of left ventricle in magnetic resonance cardiac long-axis view is required to analyze cardiac function accurately. Many segmentation algorithms that are fundamentally based on an active contour model use a semi-automatic method by user intervention of drawing initial contours and contrast enhancement process that depends on the interested image property is used for robust edge map. Also in order to efficiently segment the MR cardiac long-axis view images, we use hierarchical edge maps obtained with multiple σ values of the Gaussian function for blurring. Experiments are performed for multiphase MR cardiac images from Medinus and GE MRI systems. The experimental results verify that the proposed method can acceptably segment endocardial and epicardial walls of left ventricle in MR cardiac long-axis view images.