Simulated NEMA NU2 Performance of the Ultra-Compact Clinical NeuroLF Brain PET

A new clinical brain positron emission tomography (PET) scanner NeuroLF with octagonal detector arrangement is being finalized. Image reconstruction and data correction methods adapted for non-cylindrical scanner geometry were recently added to the open-source Software for Tomographic Image Reconstruction (STIR). We first evaluate the performance of the NeuroLF following NEMA NU2-2018. Second, the results are compared to our earlier brain PET system prototype BPET. Finally, the newly adapted STIR image reconstruction methods are validated. For this purpose, NEMA tests were performed on NeuroLF GATE simulations. The results were processed with STIR using the "Cylindrical" scanner geometry class and using the new "BlocksOnCylindrical" class. NeuroLF is based on the clinical prototye PET system BPET. The resultring spatial resolution close to the field-of-view center is estimated in simulations to improve from 4.0×3.9×3.5 mm
³
when measured with the BPET prototype scanner to 1.8×2.3× 1.6 mm
³
for the NeuroLF scanner variant with 10-mm-long crystals. Total sensitivity using an energy window of [425 keV, 650 keV] is estimated in simulations to improve from (2.9 cps/kBq / 3.7 cps/kBq) (BPET) to (5.3 cps/kBq / 7.6 cps/kBq) for NeuroLF with 10 mm-long crystals (center / off center). When using 20 mm/long crystals in NeuroLF, this is simulated to improve to (13.2 cps/kBq / 17.0 cps/kBq). The contrast recovery of image quality phantoms is estimated to improve significantly as well. We show further comparisons on the peak noise-equivalent count rate. The obtained results validate the new STIR image reconstruction methods and support the performance evaluation of NeuroLF.