基于改进多尺度ASM和非刚性配准的4D-CT左心室分割

Left ventricle segmentation based on improved multi-scale ASM and non-rigid registration from 4D-CT dataset

64排及以上多层螺旋CT(MSCT)能提供含有时间信息的四维CT成像数据,可用于动态心、肺等功能的分析.为此提出了基于改进的多尺度ASM和非刚性配准的分割方法,从心脏MSCT四维数据集中提取左心室腔和心肌.该方法根据MSCT层片间的特点,采用改进的特征点标记方法对收缩末期的左心室进行形状建模,在基于改进的模型定位和搜索算法基础上,运用多尺度ASM算法实现三维数据集上的分割.通过互信息量的Demons非刚性配准方法,利用膨胀末期的分割结果实现对其余心相上的左心室配准分割.在256排MSCT四维心脏数据集上的实验表明,该方法对血腔的分割结果和手工分割结果的平均相似度在95%以上,心肌的平均相似度在88%以上.

64-slice or more multi-slice spiral CT（MSCT） machines can provide 4D-CT imaging datasets which are useful for the dynamic function analysis of the heart and lung.A novel approach based on the improved multi-scale active shape model（ASM） and non rigid registration method was proposed to extract both the epicardium and endocardium boundaries of the left ventricle（LV） from cardiac 4D-CT datasets.On the basis of characteristic of MSCT slices,an improved method for feature point marks for end-systolic（ES） left ventricular shape modeling,new positioning and improved search algorithm and multi-scale ASM algorithm was applied for the ES 3D datasets segmentation.Based on Demons non-rigid registration incorporating mutual information between corresponding slices,all the remaining phases of left ventricular datasets were achieved successfully.The experimental results of 10 phases of 256-slice 4D cardiac MSCT datasets show that the average similarity of segmentation results of the LV cavity by this approach and manual operation is 95% or more,and that,on average,the LV myocardium is more than 88%.