摘要
为了从医学图像中获取准确的脑血管信息,提出了一种新颖的基于Hessian矩阵和聚类思想的脑血管分割方法。利用非局部均值滤波方法对原始医学图像数据进行预处理,减少了成像过程中产生的噪声对血管分割的干扰。利用多尺度邻域信息来计算各像素点的Hessian矩阵。求取其特征值并构造为一个向量。对各像素点的特征值组成的向量利用k-means方法进行聚类并最终得到血管类的像素点。实验结果表明:基于Hessian矩阵特征值聚类的方法分割得到的结果能够包含所有的脑血管点,在之后的工作中可在此分割的基础上再进行精细加工,得到更为精确的血管数据,这将对基于Hessian矩阵的脑血管分割方法研究有着深远的意义。
In order to extract the cerebral vessels accurately from medical images, a novel segmentation algorithm based on Hessian matrix and clustering method was proposed. Nonlocal means filtering was used on the original image data to reduce the interference of noise generated in the imaging process. The Hessian matrix of each pixel was computed, which considered the spatial neighborhood information of pixels. Besides, the vectors were constructed by eigenvalues of each Hessian matrix. The vectors above were clustered through the k-means method to get the vessel class. Experimental results indicate that all the vessels pixels are included in the pixels segmented by the proposed method. So other segmentation methods can be done on the pre-segmentation results and more accurate vessels can be obtained, which has the far-reaching significance to the research of cerebrovascular segmentation method based on Hessian matrix.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2014年第9期2104-2109,2115,共7页
Journal of System Simulation
基金
国家自然科学基金(61271366
61170170
61170203)
首都科技条件平台专项科学仪器开发培育项目(Z131110000613062)
中央高校基本科研业务费专项基金项目(2012LYB49)