摘要
微分同胚Demons能保证变形场的光滑可逆性,避免不合理形变的产生,但迭代次数需手动设定,且对配准结果影响较大。为解决此问题,提出自适应微分同胚的多分辨率Demons算法。首先利用非刚性配准的优化理论框架和多分辨率的策略,引入基于灰度的相似度能量函数,并设置配准终止条件,最终实现迭代次数的自适应。用check board图像,相同模态和不同模态的磁共振图像进行测试,采用配准评价指标进行定量分析,并分析不同驱动力和参数对配准结果的影响。实验结果表明:对于相同模态的磁共振图像,均方误差为514.796 5,归一化互相关系数为0.999 3,结构相似度为0.994 8;对于不同模态的磁共振图像,均方误差为1 354.1,归一化互相关系数为0.593 5,结构相似度为0.511 6;该算法均方误差最小,归一化相关系数、结构相似度最高;该算法具有高效性和鲁棒性,可用于磁共振图像的非刚性配准。
Diffeomorphic Demons can guarantee the deformation smooth and reversible and avoid producing unreasonable deformation simultaneously. But its iterations need to be set manually and have great impact on registration results. In order to solve this problem,the adaptive diffeomorphic multi-resolution demons was proposed in this paper. Firstly optimization theory framework of non-rigid registration and multi-resolution strategy were used,then similarity energy function based on gray level was designed,and termination condition was set,finally the iteration number was realized adaptively. Check board image,same modality and different modality MRI were tested,quantitative analysis was made using registration evaluation index,and the influence of different driving forces and parameters on registration result were analyzed. Experimental results indicated that,for the same modality of MRI,the mean square error was 514. 7965,normalized cross correlation was0. 9993,structural similarity was 0. 9948 by this method. For the different modality of MRI,the mean square error was 1354. 1,normalized cross correlation was 0. 5935,structural similarity was 0. 5116. The Mean square error ws the lowest,normalized cross correlation and structural similarity was the highest. In conclusion,this method is effective and robust,showing the application potential in the non-rigid registration of MRI images.
作者
王昌
任琼琼
秦鑫
刘艳
李振新
于毅
Wang Chang;Ren Qiongqiong;Qin Xin;Liu Yan;Li Zhenxin;Yu Yi(School of Biomedical Engineering, Xinxiang Medical University, Xinxiang 453003, Henan, China;Key Lab of Neurosense and Control, Xinxiang Medical University, Xinxiang 453003, Henan, China)
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2018年第2期155-162,共8页
Chinese Journal of Biomedical Engineering
基金
河南省青年骨干教师资助项目(2014GGJS-096)
河南省教育厅(17A310005)(14B416010)