一种双水平集模型分割左心室膜的方法

Double level set model for segmentation of left ventricular membrane

针对目前左心室内外膜分割方法存在分割出的轮廓正则性差、分割心室膜模糊边缘不完全和分割效率低等问题,提出新双阱势函数和各向异性梯度矢量流(AGVF)改进水平集模型能量函数,用融合0水平集和k水平集的双水平集同时分割左心室内外膜的方法。首先,用改进Hough变化圆检测算法定位心室内外膜初始位置;然后利用双水平集模型同时分割内外膜。观察和对比分析实验结果,该方法能够分割出平滑的左心室内外膜轮廓;能够分割出符合临床定义的左心室内外膜轮廓,且分割MRI图像左心室内外膜轮廓重叠率平均提高到0.9569。

In view of the existing endocardium and epicardium contour of left ventricle segmentation methods,there are some problems such as the poor regularity of segmented contour,segmenting the fuzzy edge of ventricular membrane incompletely and low segmentation efficiency.This paper proposed a new double well potential function and anisotropy gradient vector flow(AGVF)energy function improved level set model,with the fusion of zero level set and k level set of the double level set,the division of left ventricular inner and outer membrane method at the same time.First of all,with the modified Hough change detection algorithm,positioned ventricular endocardium and epicardium initial position;then,it used the double level set model to segment the endocardium and epicardium simultaneously.The experimental results show that the proposed method can segment smooth endocardium and epicardium contour of left ventricle.The analysis of experimental datum demonstrates that the proposed method can segment endocardium and epicardium contour of left ventricle confirms to the clinical definition,and increases averagely the overlap rate of the endocardium and epicardium contour on MRI images to 0.9569.