Recently several new multi-slice spiral CT (MSCT) machines have come into the market and they can provide 4D-CT imaging datasets which are useful for the dynamic function analysis of heart and lung. However, the very challenging key step is to find a powerful automatic/semi-automatic method to segment these organs from 4D-CT dataset accurately. A novel approach based on the improved coupled level set ( ICLS) method has been developed to extract both the epicardium and endocardium boundaries of the left ventricle ( LV ) automatically. Based on structure continuity of MSCT slices, the LV cavity coarse region can be extracted from the real datasets by a new automatic localization algorithm, as the initial contour of the improved coupled level set. By incorporating the coarse cavity contours and prior knowledge into the traditional levelset model, the new established coupled level set approach could extract the epicardium and endocardium automatically and accurately. The experimental results of 8 cases of 256-slice 3D cardiac MSCT datasets show that, the average similarity of segmentation results of the LV cavity between ICLS model and manual operation is 95% or more, on average the LV myocardium more than 90%. The segmentation results of 3D surface reconstruction demonstrate the identity and integrity of the LV extracted from real MSCT datasets by ICLS approach.%新型的多层螺旋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、肺功能的分析,但如何从中自动或半自动地精确分割出心脏和肺等器官是研究成功的关键.提出一种基于改进的耦合水平集自动分割方法(ICLS),从心脏MSCT数据集中精确提取左心室腔和心肌.根据层片间的结构连续性,自动定位并获取左心室腔的粗轮廓,作为水平集的初始化轮廓,同时,将腔粗轮廓和左心室先验知识融合到耦合水平集中,自动获取左心室内外膜的精确边缘.在8例256层MSCT三维心脏数据集上的实验表明,ICLS模型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90％以上；分割结果的三维表面重建验证了ICLS方法提取出的左心室具有良好的各向同一性和完整性.
展开▼
