Edge Artifacts in Point Spread Function-based PET Reconstruction in Relation to Object Size and Reconstruction Parameters

OBJECTIVE(S): We evaluated edge artifacts in relation to phantom diameter and reconstruction parameters in point spread function (PSF)-based positron emission tomography (PET) image reconstruction. METHODS: PET data were acquired from an original cone-shaped phantom filled with (18)F solution (21.9 kBq/mL) for 10 min using a Biograph mCT scanner. The images were reconstructed using the baseline ordered subsets expectation maximization (OSEM) algorithm and the OSEM with PSF correction model. The reconstruction parameters included a pixel size of 1.0, 2.0, or 3.0 mm, 1-12 iterations, 24 subsets, and a full width at half maximum (FWHM) of the post-filter Gaussian filter of 1.0, 2.0, or 3.0 mm. We compared both the maximum recovery coefficient (RC(max)) and the mean recovery coefficient (RC(mean)) in the phantom at different diameters. RESULTS: The OSEM images had no edge artifacts, but the OSEM with PSF images had a dense edge delineating the hot phantom at diameters 10 mm or more and a dense spot at the center at diameters of 8 mm or less. The dense edge was clearly observed on images with a small pixel size, a Gaussian filter with a small FWHM, and a high number of iterations. At a phantom diameter of 6-7 mm, the RC(max) for the OSEM and OSEM with PSF images was 60% and 140%, respectively (pixel size: 1.0 mm; FWHM of the Gaussian filter: 2.0 mm; iterations: 2). The RC(mean) of the OSEM with PSF images did not exceed 100%. CONCLUSION: PSF-based image reconstruction resulted in edge artifacts, the degree of which depends on the pixel size, number of iterations, FWHM of the Gaussian filter, and object size.