Application of Penalized Least-squares Algorithm in PET Image Reconstruction Based a Nonlocal Quadratic Prior

摘要：

In this paper, we present a novel image reconstruction method based on penalized least squares (PLS) objective function for positron emission tomography (PET). Unlike usual PLS algorithm, the proposed method, which is called NLPLS, combines a novel nonlocal quadratic prior with the classical least squares algorithm. The novel prior can not only solve the unfavorable oversmoothing effect produced by the simple quadratic membrane (QM) smoothing prior, but also partly eliminate blocky piecewise regions or so-called staircase artifacts produced by edge-preserving nonquadratic priors. Whats more, we can easily confirm the convergence of the NL-PLS as the objective function quadratic characteristic. The performance of the proposed NL-PLS method is evaluated in experiments using simulated data. The results show that the method is advantageous, compared with the Filter Back Projection (FBP) reconstruction and Maximum Likelihood (MLEM) reconstruction, and Bayesian constructions using the normal local priors.

关键词： image reconstruction penalized least squares positron emission tomography (PET) nonlocal quadratic prior

作者: Zhiguo Gui Jiawei He Xiaobo Ma

作者单位: National Key Laboratory For Electronic Measurement Technology, North University of China Taiyuan, China

会议类型: 国际会议

会议名称: 2010 International Conference on Signal and Information Processing(2010年IEEE信号与信息处理国际会议 ICSIP2010)

会议地点: 长沙

会议语种:英文

页码: 52-57

在线出版日期: 2010-12-14（万方平台首次上网日期，不代表论文的发表时间）