Endocardium wall extraction from noisy cardiac MR images using modified gradient vector flow algorithm

Magnetic resonance imaging is a non invasive method to detect any type of deformation of heart during a cardiac cycle. The importance of segmentation holds more true in case of medical images where decisions are solely dependent on it. To make thorough use of this functional information, it is necessary to segment endocardium i.e inner wall of left ventricle of human heart for proper diagnosis. To achieve this, an improved external force for active contour called as modified gradient vector flow is used to overcome the drawbacks of original framework. MGVF is computed by diffusion of gradient of a gray scale image. For pre-processing, anisotropic diffusion filtering is used for noise removal and preserving edges. This new external force pushes the contour or snake more closer to the inner boundary of left ventricle. This paper has extracted the left ventricle wall with better accuracy. The pre-processing has sufficiently reduced the noise and preserves LV edges.