A novel approach for localization and extraction of left ventricle in MSCT data

The accurate extraction of the left ventricle (LV) from 3D high resolution cardiac Multi Slice CT (MSCT) datasets plays a crucial role in diagnostic findings and relevant parameters computation. However, quantization noise, neighboring structure noise and misalignment of slices lead the segmentation to leakage and false boundary. In this paper,a novel approach for the 3D segmentation has been developed to extract the epi-cardium and end-ocardium boundaries of the left ventricle of the heart. Our hybrid segmentation algorithm is divided into three parts. The first part, identify the LV from slices utilizing the structure continuity; The second part, combines intensity with position-priors to construct the myocardium fuzzy affinity map in the polar coordinator system; The last part performs the improved live-wire algorithm incorporating fuzzy connectedness to delineate the myocardium boundaries. Experimental results and the 3D surface reconstruction show the accuracy and advantage of our method for the segmentation of the left ventricle from real CT data.