Performance of high-resolution PET detectors consisting of long GAGG slabs with different reflectors

Depth of interaction (DOI) encoding detectors are essential for pre-clinical or organ-dedicated PET scanners to simultaneously achieve high spatial resolution and high sensitivity. Semi-monolithic scintillator PET detectors inherently possess the DOI encoding capability that monolithic scintillator detectors have and can also improve the spatial resolution of the detector by reducing the edge effect. Compared to the widely used lutetium yttrium oxyorthosilicate (LYSO) scintillator, the newly developed gadolinium aluminum gallium garnet (GAGG) scintillator has a higher light output, which may lead to a better energy resolution and higher three-dimensional positioning accuracy in semi-monolithic PET detectors.

The semi-monolithic scintillator detectors consisted of 12 long GAGG slabs of 0.96 ⋅ 56 ⋅ 10 mm³, with one using BaSO₄ (detector 1) and the other using ESR (detector 2) as the reflector between the slabs. The front and two end surfaces of the slabs were painted with black ink. The detectors were single-ended read out by a 4×16 silicon photomultiplier (SiPM) array with an active pixel area of 3×3 mm² and a pitch of 3.65 mm. A row and column summing signal readout circuit was used to convert the 64 SiPM signals into 4 row and 16 column signals. Different monolthic (y) and DOI (z) positions of the detector were selectively irradiated by using an electronic collimation plus a mechanical collimation. The electronic collimation is realized by using a 0.3 mm diameter ²²Na point source and a coincidence reference detector consisting of a single LYSO crystal of 1×1×20 mm³. The mechanical collimation is realized by using a 5 mm thick tungsten block with a 1 mm diameter hole drilled. The centroid of gravity (COG) and squared COG algorithms were used to determine the y position, and the inverse standard deviation algorithm was applied to obtain the DOI information.

All slabs are clearly identified from the flood histograms of both detectors. The average peak-to-valley ratios (PVRs), calculated from the line profiles through the center y-axis of the flood histograms, are 3.96 and 3.55 for detectors 1 and 2, respectively. The average energy resolutions are 15.6 ± 1.8% for detector 1 and 14.7 ± 1.2% for detector 2. The COG method provides y-position resolutions of 2.21 ± 0.77 mm for detector 1 and 2.16 ± 0.55 mm for detector 2, whereas the squared COG method provides y-position resolutions of 1.60 ± 0.32 mm and 1.59 ± 0.26 mm, respectively. The average DOI resolutions are 2.33 ± 0.68 mm for detector 1 and 2.37 ± 0.58 mm for detector 2. The average timing resolutions are 4.30 ± 0.57 ns for detector 1 and 4.76 ± 0.56 ns for detector 2. Compared to the detector using LYSO scintillator slabs, the detector using GAGG scintillator slabs provides similar y-position and DOI resolutions, a better energy resolution, a better PVR in the flood histogram and a worse timing resolution.

In this work, the performance of two semi-monolithic scintillator detectors composed of GAGG slabs was investigated for the first time. The detector using the ESR reflector achieved similar y-position and DOI resolutions, slightly better energy resolution, and a poorer flood histogram compared to the detector using the BaSO₄ reflector. Both detectors clearly identified all scintilator slabs with a thickness of 0.96 mm, achieved y-position resolutions of ~ 1.6 mm, DOI resolutions of ~ 2.3 mm, and energy resolutions of ~ 15%. These detectors are suitable for developing small-animal and high-resolution organ-specific PET scanners.