Performance improvement of GAGG PET by inter-detector and inter-crystal scatter recovery

Objectives: Gadolinium Aluminum Gallium Garnet (GAGG, usually doped with Cerium) is a promising scintillator material for PET detectors because it has high light output and good energy resolution. In addition, Mg²⁺-co-doped GAGG is being developed to improve the timing performance of GAGG. Due to the low density and low effective atomic number (Z_eff) of GAGG, Compton scattering is expected to be more dominant in GAGG PET than in PET with Lutetium-based scintillation crystals such as LSO and LYSO. Inter-detector scatter (IDS) and inter-crystal scatter (ICS) are the major types of Compton scattering events within the detectors or crystals in PET. IDS is a type of triple coincidence caused by the scattering of a photon from one detector block to another, and it is practically not used for imaging. ICS is a Compton scatter from one crystal to another, and it reduces the accuracy of crystal positioning methods that use charge sharing. In this study, we conducted a Monte Carlo simulation to measure the sensitivity and occurrence frequency of IDS and ICS in GAGG PET dedicated to brain imaging, and compared it with LSO PET. In addition, we investigated the effect of IDS and ICS recoveries on image quality in GAGG and LSO PET.

Methods: With GATE v7.0, we simulated our brain PET scanner with two different crystal materials: LSO (Lu₂SiO₅, density = 7.4 g/cm³, Z_eff = 64) and GAGG (Gd₃Al₂Ga₃O₁₂, density = 6.63 g/cm³, Z_eff = 54). The inner diameter and axial length of the scanner were 33 cm and 10 cm, respectively. A block consisted of offset-layered crystals having 2.09 × 2.09 mm² cross section area and lengths of 8 and 12 mm. We measured the sensitivity of the PET with and without IDS events using a 70-cm back-to-back rod source. We also measured the ICS occurrence rate and ICS mispositioning rate of 511 keV photons from the rod source and a uniform source covering the field-of-view. To evaluate the image quality, we simulated a voxelized Hoffman brain phantom and a brain-sized lesion phantom. The lesion phantom was designed for quantitative analysis of the image quality, and it contained four hot and two cold lesions. We used 1 to 5 iterations and 18 subsets for OSEM reconstruction, and measured the background variabilities (BVs) and contrast recovery coefficients (CRCs) of the lesions. We applied IDS and ICS recoveries to observe their effects on image quality for each imaging configuration of crystal material and phantom. To recover the IDS events, we used a proportional scheme that distributes the IDS count proportionally to the double coincidence events. We chose the earliest interacted crystals of the ICS events to simulate an ideal ICS recovery.

Results: Sensitivity of the GAGG PET was 1.74 kcps/MBq, and it increased by 53.7% due to the IDS events. Compared to LSO PET, the ratio of IDS events to double coincidence events was about 2 times higher than that of LSO. In GAGG PET, about 51% of the detected 511 keV photons from the uniform source underwent ICS, and 42% of the photons resulted in mispositioning of the crystals. Detailed structure of the Hoffman phantom was most clearly visible when both ICS and IDS recoveries were applied. ICS recovery yielded better visibility of the lesions, while IDS recovery increased the overall mean pixel intensities of the images. In GAGG PET, the increments of CRC of the hot lesions due to ICS and IDS recovery were higher than 10%. Improvement in BV due to IDS recovery was more significant in GAGG than LSO due to higher proportion of IDS events.

Conclusions: Although GAGG PET performed less sensitively than LSO PET, the utilization of IDS events contributed to considerable improvement in sensitivity. As the occurrence of IDS and ICS events were more frequent in GAGG PET than in LSO PET, the impact of recoveries on image quality was significant in GAGG PET. For PET systems with scintillator materials having low stopping power, IDS and ICS recovery would be the effective solution for sensitivity and image quality improvement.