Efficient Methodology for 3D Statistical Reconstruction of High Resolution Coplanar PET/CT Scanner

摘要

A fully 3D statistical image reconstruction algorithm has been developed for a high-resolution coplanar PET/CT scanner based on rotating planar PET detectors. The system matrix has been modeled with custom Monte Carlo techniques optimized for the specific scanner architecture. The system model includes positron range, non-colinearity of gamma rays and crystal interaction modelling with attenuation and Compton scattering effects. Only 0.21% of the system matrix columns are modeled in detail, obtaining the rest of the values with axial and transaxial voxel-driven symmetries. The iterative algorithm is a fully 3D approach, regularized with the anatomical registered image using a novel version of the minimum cross entropy (MXE) scheme, and accelerated employing ordered subsets. The proposed method has been shown to produce images with superior quality than 3D hybrid (FORE+2D-OSEM) algorithms applied on synthetic GATE data, as well as on real small animal acquisitions.