摘要
为了在MR图像中精确提取左心室外膜,本文提出一种基于改进动态规划的分割方法。先用Otsu阈值法分割心内膜,再在此基础上设计一种改进的动态规划方法来寻找一条闭合的最小代价路径作为心外膜边界。该方法的关键在于路径代价函数的设计,它包括边界灰度因子、梯度因子和形状因子。3个因子之间的权重系数用混沌粒子群算法优化,同时,为了避免心外膜越过心包脂肪,设置了代价无穷大的心外膜"禁区"。将该方法在138幅图像上的分割结果与金标准及相关算法进行比较,结果表明该方法具有更高的分割精度和鲁棒性。
In order to accurately extract the epicardium of the left ventricle from cardiac magnetic reso-nance images, a method based on developed dynamic programming is proposed. First, the endocardiumis segmented by the Otsu method. Then, the epicardium is derived by designing an improved dynamicprogramming method to find a closed path with minimum local cost. The key to this method is the de-sign of the local cost function, which consists of three factors： boundary gradation, boundary gradientand shape features. The weighting coefficients of the three factors are obtained by a chaos particle swarmoptimization method. A comparison with other segmentation methods and the gold standard is providedbased on 138 images. The experimental results show that the method proposed has high accuracy and ro-bustness.
出处
《广西师范大学学报（自然科学版）》
CAS
北大核心
2014年第2期35-41,共7页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家自然科学基金资助项目(61302192)
湖北省自然科学基金资助项目(2012FFC13401)
中央高校专项基金资助项目(CZQ13010)
关键词
左心室
分割
动态规划
混沌粒子群
left ventricle
segmentation
dynamic programming
chaos particle swarm optimization
