Purpose: Propose a novel PET image reconstruction method to improve image resolution and noise property simultaneously in PET/CT studies.Method: The PET/CT scanner provides registered PET and CT images.The CT anatomy is used to guide PET image reconstruction in our modified Maximum-A-Posterior(MAP)reconstruction framework.Firstly,a model based segmentation algorithm is adopted to extract the organ boundary information from the CT image.A spatially variant weighting matrix is then constructed,where the weight is lower at the boundary and higher otherwise.This weighting matrix is plugged in to the MAP reconstruction to adjust spatially variant smoothing.Results: The proposed CT-guided MAP reconstruction method is compared with standard OSEM or MAP reconstruction method.From CRC analysis of IQ phantom,it is found that the image reconstructed with the proposed method has the best noise-resolution tradeoff among the three.CT-guided MAP generally produces image with lower noise than OSEM yet maintain similar resolution,and is sharper than MAP on structure boundaries while almost equally smooth.Conclusions: An innovative MAP reconstruction method incorporating CT anatomical information is proposed.The spatially varying weighting parameter computed from CT image leads to better noise-resolution tradeoff than conventional OSEM and MAP reconstruction method.