Abstract
160
Objectives Investigate the impact of time-of-flight (TOF) and point-spread-function (PSF) reconstruction as compared to standard filtered back-projection (FBP) and iterative (OSEM) reconstruction in a Siemens Biograph mCT PET scanner.
Methods A NEMA body phantom with 6 spheres (diameters ranging 10.1 to 37.6 mm) was filled with 3 signal-to-background ratios (8:1, 4:1 and 2:1) and with 3 different background concentrations (5.3, 2.65 and 1.32 kBq/cc). For each of these 9 combinations, 3 acquisitions of 3 minutes were performed. Images were reconstructed with 6 different procedures: FBP, OSEM with 2 iterations and 8 subsets, OSEM with 3 iterations and 24 subsets, OSEM with TOF (3 iterations and 21 subsets), OSEM with PSF (3 iterations and 24 subsets) and OSEM with both TOF and PSF (3 iterations and 21 subsets). Using a spherical volume of interest (VOI) of the same size as the target sphere, maximum SUV (SUVmax), mean SUV within the whole VOI (SUVmean) and mean SUV within an isocontour at 50% of the maximum voxel value (SUV50) were assessed. Relative errors (%) were calculated respect to the expected SUV values and expressed as median [minimum, maximum].
Results Measured errors are presented in the Table. Statistics showed that all these quantification procedures provided an error which was statistically different from zero (t-test, p<0.05). Removing the smallest sphere from statistics, which might present some partial volume effect in all reconstructions, the quantification with SUV50 over the OSEM+TOF+PSF reconstructed images provided no statistical differences with the expected value (t-test, p=0.11).
Conclusions The most accurate quantification of SUV values is obtained using OSEM+TOF+PSF reconstruction and measuring mean SUV within a 3D isocontour at 50% of the maximum voxel value