A New Fuzzy Penalized Likelihood Method for PET Image Reconstruction

摘要

In positron emission tomography (PET) image reconstruction, classical regularization methods are usually used to overcome the slow convergence of the expectation maximization (EM) methods and to reduce the noise in reconstructed images. In this paper, the fuzzy set theory was employed into the reconstruction procedure. The observations of emission counts were viewed as Poisson random variables with fuzzy mean values. And the fuzziness of these mean values was modelled through choosing an appropriate fuzzy membership function with several adjustable parameters. Coupled with this fuzzy method, the new fuzzy penalized likelihood expectation maximization (FPL-EM) method was proposed for PET image reconstruction. Simulation results showed that the proposed method might perform better in both the image quality and the convergence rate compared with the classical maximum likelihood expectation-maximization (ML-EM).