形状约束下活动轮廓模型冠脉血管图像多尺度分割

发布时间：2018-05-27 18:49

本文选题：冠脉分割 + 心脏CT图像　；参考：《中国图象图形学报》2016年07期

【摘要】：目的由于计算机断层血管造影(CTA)图像的复杂性,临床诊断冠脉疾病往往需要经验丰富的医师对冠状动脉进行手动分割,快速、准确自动分割出冠状动脉对提高冠脉疾病诊断效率具有重要意义。针对双源CT图像特点以及传统单一基于区域或边界的活动轮廓模型的不足,研究了心脏冠脉3维分割算法,提出一种基于血管形状约束的活动轮廓模型分割方法。方法首先,利用改进的FCM(fuzzy C-means)对心脏CT图像感兴趣区域初分割,其结果用于初始化C-V模型水平集演化曲线及控制参数,提取感兴趣区域轮廓。接着,由3维心脏图像数据获取多尺度梯度矢量信息构造边界型能量泛函,然后利用基于Hessian矩阵的多尺度血管函数对心脏感兴趣区域3维体数据增强滤波,获取血管先验形状信息用于约束能量泛函。最后融合边界、区域能量泛函并利用变分原理及水平集方法得到适合冠脉血管分割的水平集演化方程。结果由于血管图像的灰度不均匀,血管末端区域更为细小,所以上述算法的实施是面向被划分多个子区域的血管,在缩小的范围内进行轮廓的演化。相比于传统的血管分割方法,该方法充分融合血管图像的先验信息及梯度场信息,能够从灰度及造影剂分布不均匀的冠脉血管图像中准确分割出冠状动脉,对于细小的血管结构亦能获得较好的分割效果。实验结果表明,该方法只需在给定初始轮廓前提下,有效提取3维冠脉血管。结论对多组心脏CT图像进行分割,本文基于血管先验形状约束的活动轮廓模型可以准确分割出冠脉结构完整轮廓,并且人工交互简单。该方法在双源CT冠脉图像自动分割方面具有较好的正确率与优越性。
[Abstract]:Objective because of the complexity of computed tomography angiography (CTAA) images, the clinical diagnosis of coronary artery disease often requires experienced physicians to segment the coronary artery manually and quickly. Accurate and automatic segmentation of coronary artery is of great significance to improve the diagnostic efficiency of coronary artery disease. In view of the characteristics of dual-source CT images and the shortcomings of the traditional single active contour model based on region or boundary, a 3D segmentation algorithm for coronary artery was studied, and a segmentation method of active contour model based on vascular shape constraint was proposed. Methods first, the improved FCM(fuzzy C-means was used to segment the region of interest in the cardiac CT image. The results were used to initialize the evolution curve and control parameters of the level set of the C-V model and extract the contour of the region of interest. Then, the boundary energy functional is constructed from the 3D heart image data, and then the multi-scale vascular function based on the Hessian matrix is used to enhance the filtering of the 3D volume data of the region of cardiac interest. The prior shape information of blood vessel is used to constrain energy functional. Finally, the boundary, the regional energy functional and the evolution equation of the level set suitable for coronary artery segmentation are obtained by using the variational principle and the level set method. Results because the grayscale of blood vessel image is uneven and the end region of blood vessel is smaller, the implementation of the above algorithm is aimed at the vessels divided into several sub-regions, and the contour evolves in a narrow range. Compared with the traditional vascular segmentation method, this method fully integrates the prior information and gradient field information of the vascular image, and can segment the coronary artery accurately from the coronary artery image with uneven distribution of grayscale and contrast media. Good segmentation effect can also be obtained for small vascular structures. The experimental results show that the proposed method only needs to extract 3D coronary vessels effectively under the condition of given initial profile. Conclusion Multi-group cardiac CT images can be segmented. The active contour model based on prior vascular shape constraint can accurately segment the complete contour of coronary artery structure, and the artificial interaction is simple. This method has good accuracy and superiority in automatic segmentation of coronary artery image with dual source CT.
【作者单位】：南京工业大学电气工程与控制科学学院;
【分类号】：TP391.41

【相似文献】