医学图像标志点的自动配准  

AUTOMATIC REGISTRATION OF MARKERS IN MEDICAL IMAGE

作　　者：周振环[1] 陈思平[2] 陶笃纯 陈新海[4] 

机构地区：[1]南昌航空工业学院 [2]浙江大学,杭州310027 [3]安科公司,深圳518067 [4]西北工业大学,西安710072

出　　处：《中国生物医学工程学报》2003年第5期438-442,共5页Chinese Journal of Biomedical Engineering

基　　金：深圳企业博士后基金;江西省测试技术与控制工程研究中心基金 ( 0 2 0 0 3 )联合资助

摘　　要：提出一种医学图像标志点自动识别方法 ,构造一个标志点的模板 ,利用区域相关景像匹配法 ,能快速、精确地将图像上的标志点识别出来。同时提出一种标志点自动配准算法 :(1)定义标志点的多维距离向量 ,在图像空间或手术空间中 ,用某一标志点到其余点有序距离分量组成的向量 ,作为每个标志点的多维距离向量 ;(2 )由于刚体变换保持距离不变 ,同一标志点在一个空间中的多维距离向量 ,与它在另一个空间的多维距离向量具有最大的相近性 ,比较每两个点 (分属不同空间 )多维距离向量的相关系数 ,就可以找出图像空间标志点与手术空间标志点的对应关系 ;(3)将所有点对坐标值代入刚体变换线性方程组 ,用最小二乘法求出从图像空间到手术空间的刚体变换矩阵 ,最终实现医学图像标志点的自动识别和配准。We proposed a method to recognize markers automatically in medical image. A template of markers which recognized the markers quickly and exactly from medical image based on digital scene matching area correlation constructed had been constructed. Furthermore, we proposed an algorithm of markers automatic registeration. 1, There are some distances between a marker and rest of them in image space or surgical space, these distances construct a vector, we defined the vector as the marker's multidimensional distance vector (MDV). 2, Because of rigid transform keeping distances unchangeable, one marker's MDV in one space has maximum similarity with the same MDV in another space. For every possible pair of points in different spaces, one can find correct pair of markers in image space and surgical space by comparing with the correlation coefficient of their MDVs. 3, Substituting coordinates of all pairs of points into rigid transform equations, one could use least squares to get the rigid transform matrix from image space to surgical space.

关 键 词：计算机辅助外科 手术导航 图像识别 相关匹配 图像配准 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]