基于Snake改进模型的心脏MR图像左心室分割方法

提出一种基于Snake改进模型的心脏MR图像左心室分割方法。首先对梯度矢量流GVF模型进行改进,提出基于扩展邻域的S型函数梯度矢量流ENSGVF模型,该模型可获得更大的捕获域,并能解决深度凹陷及弱边界泄露的问题。然后将ENSGVF作为新的外力条件,构造ENSGVF Snake模型,用于内外膜分割。对于内膜分割,引入圆形约束项,消除由于图像灰度不均匀造成的局部极小问题。进而利用内膜分割结果构造新的外力场和约束,实现外膜的精确自动分割。实验结果表明,该算法能有效解决分割中存在的弱边界、图像灰度不均匀、乳突肌干扰等问题,提高了精确度。

A Segmentation Method of Left Ventricle in Cardiac Magnetic Resonance Images Based on Improved Snake Model

A novel method for segmenting cardiac magnetic resonance images based on snake model is proposed. An external force called extended neighborhood sigmoid gradient vector flow ENSGVF is presented as the improvement of gradient vector flow (GVF) for snake which has a good performance on deep and narrow concavity convergence, capture range and weak edge preserving. In terms of the segmentation of endocardium, considering that the left ventricle is roughly a circle, a circle shape constraint is adopted on the basis of ENSGVF snake models, which can eliminate the unexpected local minimum caused by image inhomogeneity and papillary muscle. For the segmentation of epicardium, making full use of the segmentation result of endocardium, a new external force field and a new shape constraint are constructed to achieve automatic precise segmentation. The experimental results show that the proposed method can address the following challenges: (1) lake of edge inhomogeneity; (2) image inhomogeneity; (3) effect of papillary muscle. And it improves the rate of accuracy.