局部更新的分层B样条医学图像非刚性配准算法  

Local Updating Algorithm of Hierarchical B-spline Based Non-rigid Registration for Medical Images

作　　者：秦绪佳[1,2] 肖佳吉 陈珊[1] 王琪[1] 陈胜男[1,2] 

机构地区：[1]浙江工业大学计算机科学与技术学院,杭州310032 [2]浙江省可视媒体智能处理技术研究重点实验室,杭州310023

出　　处：《小型微型计算机系统》2016年第10期2338-2342,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61075118;61303140)资助

摘　　要：针对存在自主运动器官所采集的医学影像,在利用传统的基于B样条的医学图像非刚性配准算法的基础上,提出一种多层B样条非刚性配准的方法,网格控制点由少到多变化,后一次配准在前一次配准的效果上进行,通过多次迭代降低了控制网格密度选取上的不准确性.针对多层次B样条非刚性配准方法计算量相当大导致配准时间过长,无法达到临床上实时配准要求的问题,提出一种力矩主轴法和局部更新策略相结合的分层B样条快速配准方法.该方法首先利用力矩主轴实现两幅待配准图像的粗配准,减少细配准时的计算量;然后采用局部更新策略的B样条非刚性配准方法实现细配准.实验结果表明,本文方法兼顾了配准的精度与速度,获得了较好的医学图像的非刚性配准结果.For the medical images gathered from automatically moved organs, a multilevel B-spline non-rigid medical image registra- tion method is proposed. This method is based on the traditional B-spline method. The number of control points of grid is changed from few to many, and the latter registration effect is based on the previous registration result. It reduces the inaccuracies of choosing the density of control grid through multiple iterations. Since the computation of the multilevel B-spline registration is considerably large, it cannot meet the requirements of real-time clinical registration. Therefore a hierarchical B-spline fast registration method that combines local updating strategy with principal axes method to achieve the final registration is presented. This method first uses princi- pal axes method for rough-registration of the two images to be registered, and this can reduce the amount of computation of the latter fine-registration process. Then it uses the B-spline non-rigid image registration method that is based on local updating strategy to re- fine the final registration process. Experimental results show that our method takes both the registration accuracy and speed into ac- count, and thus enables to obtain good non-rigid medical image registration results.

关 键 词：医学图像 B样条 非刚性配准 局部更新策略 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]