Experimental Evaluation of System Models for PET with Block Detectors

Modern PET scanners use block detectors. The accuracy of the system models greatly affects the quality of reconstructed images. This work evaluates two approaches to the modeling of PET system response. The first one is a factored system matrix approach that combines analytical calculation and Monte Carlo simulation, while the second stems from purely analytical calculations of solid angle and crystal penetration. Both of them model the block structure. We conduct measurements on the microPET II scanner, which has 90 detector blocks forming 3 axial rings. Each detector block consists of 14times14 equally spaced scintillation crystals with a pitch of 1.15 mm. A Na-22 point source was scanned inside the field of view at every 2.0 mm along the horizontal and vertical directions, and at every 2.8 mm along the 45-degree trajectory within the central axial plane. Sinogram profiles were taken from the projection angle that is perpendicular to the corresponding trajectories. We compared the full width at half maximum (FWHM), peak location, peak intensity, and area under the curve of the measured projection and model-predicted profiles. The results clearly show the effect of detector blocks with the FWHM being piecewise flat as the point source moves away from the radial center. The discontinuous points are consistent with the actual geometry of the scanner. Both system models show reasonably good agreement with experimental measurements.