一种基于广义梯度矢量流Snake模型的心脏MR图像分割方法  

A Method for Segmentation of the Cardiac MR Images Based on GGVF Snake

作　　者：武玉伟[1] 梁佳[1] 王元全[1,2] 

机构地区：[1]北京理工大学计算机科学技术学院,智能信息技术北京市重点实验室,北京100081 [2]天津理工大学计算机科学与技术学院,智能计算与软件新技术天津市重点实验室,天津300191

出　　处：《中国图象图形学报》2010年第4期598-606,共9页Journal of Image and Graphics

基　　金：国家自然科学基金项目(60602050,60805004)

摘　　要：提出了一种基于广义梯度矢量流Snake模型的心脏核磁共振图像左心室内、外膜分割方法。首先构造了一种基于目标边缘的方向广义梯度矢量流(edge-based directional generalized gradient vector flow,EDGGVF)Snake模型,该模型在传统GGVF的基础上,结合目标边缘图梯度方向信息,将左心室内、外膜区分为正边缘和负边缘,从而实现左心室内外膜的全自动分割。其次,根据左心室近似为圆形的形状特点,引入了圆形能量约束,有利于克服由于图像灰度不均、乳突肌等引起的局部极小。实验结果表明,该方法可以高效准确地自动分割出左心室内、外膜。In this paper, a novel method based on generalized gradient vector flow （GGVF） Snake model is proposed for segmentation of the left ventricle cardiac magnetic resonance （MR） images. Firstly, an edge-based directional generalized gradient vector flow （EDGGVF） Snake model is proposed as an improvement to GGVF, which differentiates cardiac endocardium and epicardium into positive and negative boundaries by incorporating the gradient orientation information of the images edge map. In addition, a circle-shape based energy for the Snake model is adopted considering the shape of the left ventricle. With this energy, the Snake contour can overcome the unexpected local minimum stemming from image inhomogeneity and papillary muscle. Experimental results show the method is able to segment LV endocardium and epicardium accurately and effectively.

关 键 词：梯度矢量流 SNAKE模型 图像分割 左心室 形状约束 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]