Design and performance evaluation of a self-collimating SPECT system for small animal imaging

316 Objectives: The performance of conventional SPECT system suffers a fundamental limitation that the mechanical collimation forces a compromise between detection efficiency and spatial resolution. We propose and evaluate a novel concept, namely self-collimating, that uses spatially separated scintillation detectors (SSSD) for photon collimation to reach a new level of high spatial resolution and high detection efficiency. Methods: In a system that consists of multiple layers of SSSD, the SSSD on one layer plays both the function of collimation - allowing photons passing through the detector separations to reach SSSDs on the outer layers - and the function of detection. Only the innermost SSSD uses a conventional high efficiency collimator. There is no gap in between the detectors in the outermost layer. We evaluate two highly correlated self-collimating SPECT designs, one through Monte-Carlo simulation and one through prototype experiments. In the Monte Carlo simulation, we simulate a full-ring self-collimating SPECT with 4 layers. The ring diameters corresponding to the four layers are 180, 220, 260 and 300 mm. Each layer is composed of 570 x 11, 690 x 11, 810 x 11, and 1890 x 11 0.5 mm (tangential) x 21 mm (axial) x 6 mm (radial) GAGG scintillators with 6 mm axial gap. In the middle three SSSD layers, the adjacent scintillators have a 0.5 mm gap in between in the tangential direction. We assume that the intrinsic resolution in the axial direction is 1.5 mm. The conventional collimator is a 40-mm-diameter tungsten cylinder with 105 (tangential) x 11 mm (axial) 0.4-mm-diameter pinholes spaced 1.2 mm (tangential) and 6 mm (axial) center-to-center evenly on the cylinder, all pinholes focus at the center of FOV. The experimental prototype is partial ring that consists of a 2.5-mm thick tungsten slab with multiple 1-mm slits spaced 3 mm center-to-center and 3 layers. Each layer consists of 7 detector modules, each module consists of two 24 × 12 scintillator arrays. The scintillators on the inner and middle layer SSSDs are 1.35 mm x 2.7 mm x 4 mm YSO and LYSO crystals, on the outer layer are 1.35 mm x 2.7 mm x 7 mm GAGG and LYSO crystals, all interleaved in tangential and axial directions. Each detector module has one end-side optically coupled to an 8 × 8 SiPM array. System performance was evaluated with hot rod phantom, disc phantom and contrast phantom in the simulated system. Spatial resolution was measured experimentally with three 3x3 Tc99m point source phantoms at 0.3 mm, 0.4 mm and 0.6 mm. OS-EM algorithm was used to reconstruct the image. Results: In the simulated system, the peak and average detection efficiency in a 10 mm x 10 mm FOV is 1.37% and 0.46%, respectively. A 2D hot rod phantom study shows that the best achievable image resolution is 0.05 mm transaxially and 0.2 mm axially in nearly noise-free condition. With a 3D hot rod phantom that consists of 15 0.1 mm hot rods, we demonstrate that with 1 μCi Tc99m total activity and 10 min scan time, all the rods are clearly separable in the reconstructed image. The contrast phantom studies show that with a total activity of 0.2 mCi in a 8 mm (φ) x 3 mm(L) warm cylinder with 6 hot rods at 8:1 contrast, the smallest visible hot rod is 0.15 mm with a 20 min scan. With the prototype system, the point sources with a 0.3 mm spacing can be distinguished. Conclusions: The simulations show astonishingly outstanding imaging performance. The spatial resolution demonstrated in the experiment is slightly better than the best reported value to our knowledge. We conclude that the self-collimating concept opens a door to develop a new generation of SPECT systems and indicates a potential paradigm shift in emission tomography. The research was supported by the National Natural Science Foundation of China (No. 81727807, No.11575096, No. 11605008) and National Key Research and Development (R&D) Plan of China (Grant ID. 2019YFF0302503 and 2016YFC0105405).