Performance analysis of different PSF shapes for the quad-HIDAC PET submillimetre resolution recovery

In pre-clinical applications, it is quite important to preserve the image resolution because it is necessary to show the details of structures of small animals. Therefore, small animal PET scanners require high spatial resolution and good sensitivity. For the quad-HIDAC PET scanner, which has virtually continuous spatial sampling; improvements in resolution, noise and contrast are obtained as a result of avoiding artifacts introduced by binning the data into sampled projections used during the reconstruction process. In order to reconstruct high-resolution images in 3D-PET, background correction and resolution recovery are included within the Maximum Likelihood list-mode Expectation Maximization reconstruction model. This paper, introduces the performance analysis of the Gaussian, Laplacian and Triangular kernels. The Full-Width Half-Maximum used for each kernel was varied from 0.8 to 1.6 mm. For each quality compartment within the phantom, transaxial middle slices from the 3D reconstructed images are shown. Results show that, according to the quantitative measures, the triangular kernel has the best performance.