基于非扩展熵相似度的三维医学图像配准  

THREE-DIMENSIONAL MEDICAL IMAGE REGISTRATION BASED ON NONEXTENSIVE ENTROPIC SIMILARITY MEASURE

作　　者：李碧草 李润川[2,3] 刘洲峰[1] 李春雷[1] 王宗敏 Li Bicao;Li Runchuan;Liu Zhoufeng;Li Chunlei;Wang Zongmin(School of Electronic and Information Engineering,Zhongyuan University of Technology,Zhengzhou 450007,Henan,China;School of Information Engineering,Zhengzhou University,Zhengzhou 450001,Henan,China;Cooperative Innovation Center of Internet Healthcare,Zhengzhou University,Zhengzhou 450052,Henan,China)

机构地区：[1]中原工学院电子信息学院,河南郑州450007 [2]郑州大学信息工程学院,河南郑州450001 [3]郑州大学互联网医疗与健康服务河南省协同创新中心,河南郑州450052

出　　处：《计算机应用与软件》2020年第11期95-100,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61772576,61872324);国家重点研发计划项目(2017YFB1401200);河南省科技攻关计划项目(192102210127);河南省高校科技创新团队支持计划项目(18IRTSTHN013);河南省博士后基金项目(19030018);河南省高校重点科研项目(17A510006,19B510011)。

摘　　要：针对互信息相似性测度中传统香农熵的扩展性,结合Arimoto非扩展熵的性质,构造一种新的相似性测度,提出基于该相似度的多模态三维医学图像配准算法。构造非扩展熵相似度,建立图像配准框架;利用基于B样条的帕曾窗估计联合概率分布,得到连续的目标函数;采用拟牛顿优化方法对配准模型进行求解。在三维临床医学图像的实验结果表明:与传统的基于互信息相似度的图像配准算法相比,该算法的配准精度较高。For the expansibility of traditional Shannon entropy in the similarity measure based on mutual information,a novel similarity measure is proposed combining the Arimoto entropy,and an algorithm of multi-modal three-dimensional medical image registration based on the similarity is presented.An entropic measure with non-extension was constructed,and a framework of image registration was built;the B-splines Parzen window was employed to estimate the joint probability distribution between the images to be registered,and the continuous objective function was obtained;the registration model was optimized by the quasi-newton method.Experimental results on the 3D clinical medical images show that the proposed method provides higher registration accuracy compared with the traditional image registration methods based on mutual information similarity.

关 键 词：Arimoto熵 非扩展性 帕曾窗估计 拟牛顿法 多模态 图像配准 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]