The effects of spacelimiting constraints on statistical image reconstruction for PET

For images that are represented by a finite number of parameters, statistical methods can take advantage of oversampling in the sinogram to reduce the noise in the reconstructed images. In this study, a new statistical angular sinogram filter is introduced for spacelimited images. Algebraic equations along with an efficient method for statistical image reconstruction that accounts for the non-stationary (Poisson) noise are formulated. A computer simulation and thorax phantom study were chosen with the count levels and heart-to-thorax activity ratio comparable to that of a typical whole body PET study. Using multiple independent noise realizations, this new method was found to reduce the noise variance by 5% and 2% for computer simulations and thorax phantom, respectively without introducing any bias.