Resolution modelling in PET image reconstruction usually shows an enhancement of the contrast and an improvement in the resolution of the reconstructed images. The scanner resolution can be modelled in sinogram space more accurately but at the same time this method is more challenging to implement. In addition, the point spread function (PSF) is approximately shift-invariant in the region of the centre of the field of view. For that reason, a shift-invariant PSF implemented in image space can be used in brain imaging. In this work, we compare the performance of the PSF in sinogram space available in the Biograph mMR scanner to a shift-invariant image space PSF with different FWHM values. A long scan of a NEMA IQ phantom was used to evaluate quantitatively the different reconstruction methods. Additionally, patient data of a brain study was also used to evaluate the algorithms. A shift-invariant PSF of 4.5 mm matched the performance of the sinogram space PSF and in both cases Gibbs artifacts were visible.
|Title of host publication
|Accepted/In press - 2017