基于预测-校正改进Snake的心脏MR图像左心室外轮廓分割

A NEW TWO-STEP-DEFORMATION SNAKE WITH APPLICATION TO LV EPICARDIUM SEGMENTATION OF CARDIAC MR IMAGES

在心脏MR图像中,由于左右心室结合处的灰度非常靠近,在左心室外轮廓上形成弱边界,基于Snake模型分割左心室外轮廓时就会有边界泄漏的问题,本文先定义了两种局部信息用于边缘增强,并构造了合理的外力场,然后将传统Snake模型的形变结果作为对轮廓新位置的预测,基于左心室外轮廓形状的先验知识对预测的结果进行校正,使得Snake的形变由预测、校正两步来完成。实验结果表明,这种预测-校正两步形变Snake模型对心脏MR图像在心室外轮廓分割有较好的效果。

The traditional snake model will encounter the boundary leaking problem when used for the left ventricle (LV) epicardium segmentation in that the intensity gradient at the conjunction of the left and the right ventricles is near zero in cardiac magnetic resonance (MR) images. This paper presented two kinds of local information to enhance the weak boundary and built reasonable external force field. Then, taken the deformation results of the traditional snake as the prediction of the new contour position, a priori shape knowledge of the LV was used to correct the predictive results such that the snake contour deformed via two steps. The experimental results validate the performance of this two-step-deformation snake for the LV epicardium segmentation of MR images.