Investigation of a clinical PET detector module design that employs large-area avalanche photodetectors

Abstract

We investigated the feasibility of designing an Anger-logic PET detector module using large-area high-gain avalanche photodiodes (APDs) for a brain-dedicated PET/MRI system. Using Monte Carlo simulations, we systematically optimized the detector design with regard to the scintillation crystal, optical diffuser, surface treatment, layout of large-area APDs, and signal-to-noise ratio (SNR, defined as the 511 keV photopeak position divided by the standard deviation of noise floor in an energy spectrum) of the APD devices. A detector prototype was built comprising an 8 × 8 array of 2.75 × 3.00 × 20.0 mm3 LYSO (lutetium-yttrium-oxyorthosilicate) crystals and a 22.0 × 24.0 × 9.0 mm3 optical diffuser. From the four designs of the optical diffuser tested, two designs employing a slotted diffuser are able to resolve all 64 crystals within the block with good uniformity and peak-to-valley ratio. Good agreement was found between the simulation and experimental results. For the detector employing a slotted optical diffuser, the energy resolution of the global energy spectrum after normalization is 13.4 ± 0.4%. The energy resolution of individual crystals varies between 11.3 ± 0.3% and 17.3 ± 0.4%. The time resolution varies between 4.85 ± 0.04 (center crystal), 5.17 ± 0.06 (edge crystal), and 5.18 ± 0.07 ns (corner crystal). The generalized framework proposed in this work helps to guide the design of detector modules for selected PET system configurations, including scaling the design down to a preclinical PET system, scaling up to a whole-body clinical scanner, as well as replacing APDs with other novel photodetectors that have higher gain or SNR such as silicon photomultipliers.

Design study of dedicated brain PET with polyhedron geometry

“…To speed up the simulation, the crystal surface was set as polished so that only mirror reflections were allowed (reflection coefficient: 98.5%). Further, the crystal's attenuation length and light output were set to 1.16 cm and 26,000 ph/MeV respectively in this simulation [19]. The thicknesses of the light guides varied from 5 mm to 15 mm in the simulation.…”

Relevancies of multiple-interaction events and signal-to-noise ratio for Anger-logic based PET detector designs

“…For comparison, the light output was also studied using DETECT2000, a Monte-Carlo simulation tool for studying transport of scintillation photons [21]. Parameters used in the simulation have been reported in our previous work [7] and the surface treatment was selected to be polish for all studies. The light output profile derived from the DETECT2000 was also fitted with a Gaussian function and compared to the results of the analytical method, for two crystal sizes (3 Â 3 Â 20 mm 3 , 1 Â 1 Â 20 mm 3 ) and four diffuser heights (1 mm, 3 mm, 6 mm, and 9 mm).…”

“…While no crosstalk between adjacent crystal elements occurs, one serious challenge is the large number of readout channels required as well as high cost of electronics. The second method utilizes optical and/or electrical multiplexing in order to reduce the number of readout channels while maintaining satisfactory detector performances [3][4][5][6][7][8][9]. In particular, the Anger-logic scheme combined with light sharing has been widely used in the development of PET block detectors as shown in Fig.…”

“…In such a design, each crystal element has a unique light-sharing ratio among four photodetectors and their output signals are used to derive spatial information with the Anger-logic positioning algorithm [5]. Such Anger-logic scheme was initially based upon photomultipliers (PMTs), it has recently been used for solid-state photodetectors, such as avalanche photodiodes (APDs) and silicon photomultipliers (SiPMs) [6,7]. To avoid any confusion between the terms Anger-logic and block detectors, it should be pointed out that Anger-logic is a generic term referring to a weighted combination of the signals of multiple photodetectors as first implemented in gamma camera [5], only a special case as illustrated in Fig.…”

“…It depends on a number of parameters such as crystal geometry/configuration, light sharing, inter-crystal scattering, photodetector signal-to-noise (SNR), as well as any electrical multiplexing applied. Peak-to-valley ratio (PVR) is a figure-of-merit widely used to study the block effect [7,9,18,19]. A higher PVR indicates that crystals are better distinguished and suffer less from decoding uncertainty.…”