PET image reconstruction with anatomical prior using multiphase level set method

Acquiring both anatomical and functional images during one scan, PET/CT systems improve the ability to distinguish pathology from normal uptake and to precisely localize abnormal loci. Researchers have investigated using anatomical boundary information in CT to regularize PET images. Here we propose a novel approach to maximum a posteriori (MAP) reconstruction of PET images with CT-based prior information. The image prior models both the smoothness of PET images and the similarity between functional boundaries in PET and anatomical boundaries in CT. Minimal smoothing is applied across functional boundaries to preserve shape edges. Level set functions are use to describe the anatomical boundaries from CT and to track the evolution of the functional boundaries in PET. The proposed method does not assume an exact match between PET and CT boundaries, but rather maximizes the similarity between the two boundaries, while allowing different region definition in the PET image. This is an important feature as mismatches between anatomical and functional boundaries have been observed in clinical images that could be caused either by inherent difference in the contrast mechanisms and/or subject motion. We conducted computer simulations to evaluate the performance of the proposed method. A digital phantom was built based on a PET/CT scan of a mouse. Two anatomical priors are obtained by modifying the segmented CT boundaries. The proposed method is compared with other methods including ML-EM, MAP with spatial-invariant and spatial-variant smoothing. The reconstructed images of the proposed method are locally smooth with sharp functional boundaries, while blurred boundaries appears in the results of other methods. The region of interest quantification study shows that the proposed method achieves less bias at the same noise level compared to the existing methods.