Application of regularized maximum likelihood algorithm in PET image reconstruction combined with nonlocal fuzzy anisotropic diffusion

Abstract

Although the traditional regularized maximum likelihood (RML) algorithms can obtain the high-quality reconstructed image in positron emission tomography (PET), it still suffers from over-smoothing with the increase of the iteration number. In this paper, we propose a novel regularized maximum likelihood reconstruction algorithm for PET imaging, which uses the fuzzy nonlinear anisotropic diffusion as the regularization to reduce noise and preserve edges. In addition, we introduce the nonlocal means idea which can adequately exploit the global information of the image into the reconstruction process to further improve the quality of the result image. The proposed algorithm not only absorbs the advantages of fuzzy set theory in deal with uncertain problems, but also has the good ability of anisotropic diffusion, namely protecting edges perfectly and suppressing noise. Experimental results and quantitative analysis show that the novel method has the more excellent performance for positron emission tomography imaging.