基于仿射变换与B样条自由形变的医学图像配准  

An Algorithm for Medical Image Registration Based on Affine Transform and B-spline Free-form Deformation

作　　者：杭利华[1] 蒋佩钊[1] 邓冲[1] 

机构地区：[1]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《兰州交通大学学报》2013年第3期44-48,65,共6页Journal of Lanzhou Jiaotong University

基　　金：国家自然科学基金(60962004);国家自然科学基金(61162016);甘肃省科技支撑计划项目(1104FKCA102);甘肃省科技支撑计划项目(1104GKCA057);金川公司-兰州交通大学联合基金项目资助

摘　　要：针对医学图像配准的人体组织器官的几何形体复杂多样,传统B样条配准方法较难很好地描述组织器官形变,配准结果有较大的误差等问题,使用互信息作为相似性测度,采用仿射变换作为全局配准的粗配准以及B样条自由形变作为局部配准的精确配准,为了提高配准精确度以及降低计算复杂度,引入多分辨率策略,对脑部CT和MRI图像进行配准.实验结果表明,该方法对多尺度的形变图像都有很好的配准效果.In medical image registration process, human tissues and organs are complex, the traditional b- spline registration is difficult to describe the deformation of the tissues and organs, and the results of regis- tration have the error. In this paper, the mutual information is used as the measurement of similarity, the affine transformation is used as global rough registration, as well as b-spline free-form deformation as local registration and the multi-resolution strategy is introduced to improve the accuracy of registration, reduce the computational complexity and complete the registration of brain CT and MRI medical image. The ex- periment results show that the method of registration can complete the registration of multi-scale deforma- tion with prospective effect.

关 键 词：B样条自由形变 仿射变换 多分辨率 L-BFGS 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]