通过超松弛加速最大后验期望最大化（MAP-EM）算法

Acceleration of MAP-EM algorithm via over-relaxation.

作者信息

Tsai Yu-Jung, Huang Hsuan-Ming, Fang Yu-Hua Dean, Chang Shi-Ing, Hsiao Ing-Tsung

机构信息

Department of Medical Imaging and Radiological Sciences, College of Medicine, Chang Gung University, Taoyuan, Taiwan.

Medical Physics Research Center, Institute of Radiological Research, Chang Gung University and Chang Gung Memorial Hospital, Taoyuan, Taiwan.

出版信息

Comput Med Imaging Graph. 2015 Mar;40:100-7. doi: 10.1016/j.compmedimag.2014.11.004. Epub 2014 Nov 15.

DOI:10.1016/j.compmedimag.2014.11.004

PMID:25465068

Abstract

To improve the convergence rate of the effective maximum a posteriori expectation-maximization (MAP-EM) algorithm in tomographic reconstructions, this study proposes a modified MAP-EM which uses an over-relaxation factor to accelerate image reconstruction. The proposed method, called MAP-AEM, is evaluated and compared with the results for MAP-EM and for an ordered-subset algorithm, in terms of the convergence rate and noise properties. The results show that the proposed method converges numerically much faster than MAP-EM and with a speed that is comparable to that for an ordered-subset type method. The proposed method is effective in accelerating MAP-EM tomographic reconstruction.

摘要

为提高断层重建中有效最大后验期望最大化（MAP-EM）算法的收敛速度，本研究提出一种改进的MAP-EM算法，该算法使用超松弛因子来加速图像重建。所提出的方法称为MAP-AEM，在收敛速度和噪声特性方面，对其进行了评估，并与MAP-EM算法和有序子集算法的结果进行了比较。结果表明，所提出的方法在数值上比MAP-EM收敛得快得多，且速度与有序子集类型的方法相当。所提出的方法在加速MAP-EM断层重建方面是有效的。