Component characterization and commissioning of a gamma-PET prototype detector system

Hybrid imaging systems, comprising PET and Compton camera modules, have recently gained in interest, due to their capability to simultaneously detect positron annihilation photons and γ-rays from single-photon emitting sources as also used in SPECT. A unique feature of such systems, however, is the capability to also be operated in a so called γ-PET mode. Here, specific β + - emitting radioisotopes (such as 44 Sc, 1 °C or 14 O) are used to detect triple-coincidences between two annihilation γ-rays (in PET imaging) and a third, prompt photon (in Compton imaging), that is emitted by the deexcitation of the decay’s daughter nucleus. Consequently, an intersection between the line-of-response (LOR) and the Compton cone can be determined, which (in principle) allows to localize the photons’ emission vertices on a single decay basis. In practice, however, a few tens of events are required to localize a point source, which still results in a considerable sensitivity improvement compared to conventional PET imaging. For a proof-of-principle study, we used a pixelated GAGG crystal array (16 × 16 crystals; 1.45 × 1.45 × 6 mm 3 crystal volume; 25 μm SPAD SiPMs as readout) as Compton camera scatterer and PET detectors, and a three-layered LYSO crystal array (1.2 × 1.2 × 6.66 mm 3 crystal volume; 50 μm SPAD SiPMs as readout) as Compton camera absorber. We characterized the individual detector components with regard to their energy resolution and the capability to identify the various scintillator array’s individual crystals. Our first γ-PET prototype was tested in PET-only and Compton-only imaging mode, in which spatial resolutions of 3.2–3.5 mm FWHM (PET-only mode) and 14.4–19.3 mm FWHM (Compton-only mode at 1,274 keV) were achieved, respectively, using a 22 Na point source and 10 iterations of an ML-EM reconstruction algorithm. By using triple-coincidences in a γ-PET mode (event-wise intersection of the LOR and the Compton cone), we could demonstrate the capability of the prototype to perform a full 3D point source reconstruction using only 77 events.