A pre-clinical PET scanner based on a monolithic annulus of scintillator (AnnPET): construction and NU4-2008 performance testing

<i>Objective.</i>In the past several decades, numerous positron emission tomography (PET) scanners of various designs have been constructed for use in pre-clinical studies. Our group is investigating use of a monolithic annulus of scintillator, instead of the traditional arrays of discrete scintillator elements or individual detectors that utilize continuous blocks of scintillator, to construct a novel pre-clinical PET scanner.<i>Approach.</i>This scanner, called AnnPET, is based on a fourteen-faceted annulus of lutetium yttrium orthosilicate with an inner diameter of 6 cm and length of 7.2 cm. Each facet is populated with four specially constructed 4 × 4 arrays of 4 mm × 4 mm multi-pixel photon counters .To cool and temperature stabilize these devices, the scanner gantry is immersed in dielectric fluid. Positioning of events in the scintillator is accomplished with the application of deep-residual convolutional neural network. The scanner's performance was assessed using the NEMA NU4-2008 protocols.<i>Results.</i>Full-width-at-half-maximum (FWHM) of the images of a point source reconstructed with the single slice rebinned filtered backprojection (SSRB-FBP) algorithm at 5 mm from the center of the scanner are: 1.40 mm (radial), 1.38 mm (tangential) and 1.40 mm (axial). At 18 mm from scanner center (edge of the scanner's inner bore) the FWHMs are: 1.62 mm (radial), 1.43 mm (tangential) and 1.48 mm (axial) FWHM. Peak detection sensitivity is 9.5% (0.086 cps Bq^-1). Peak noise equivalent count rate is 234 kcps at 14.4 MBq.<i>Significance.</i>Overall, testing of the AnnPET system demonstrated very promising performance results for a pre-clinical PET scanner based on a single, cooled annulus of monolithic scintillator used with neural networks. Continued development of the system is planned.