Comparison of gamma ray localization using system matrixes obtained by either MCNP simulations or ray-driven calculations for a coded-aperture imaging system

A coded-aperture system based on a large-area silicon photo-multiplier (SiPM) coupled with inorganic scintillator was developed for the gamma ray localization in the field of nuclear safety and security. Monte Carlo simulations of the performance were conducted to verify its performance. For the coded-aperture imaging (CAI) system, the mask was designed with an 11-rank modified uniformly redundant array (MURA), and the SiPM readout consisted of 12 x 12 pixels. A two-centimeter-thick tungsten mask was used to encode the gamma ray field. The 144 pixels are read-out with a resistor-based charge-division circuit that reduces the readout outputs from 144 to four signals per module, from which the deposited energy and interaction position can be extracted. For image localization, maximum-likelihood expectation maximization (MLEM) and compress-sensing (CS) methods are used with either: (1) a system matrix generated both the Monte-Carlo N-Particle (MCNP) code, or (2) a mathematical system-matrix model utilizing a ray-driven method. In this paper, reconstructed images of gamma ray sources with various positions and activities were simulated and measured with the simulation tools and the physical system in order to compare and evaluate the relative strengths of the two different system matrix formulations. Both system matrixes generated by MCNP and ray-driven methods are effective in localizing isolated point sources; however, there are critical differences when several gamma sources with different strengths are in the field of view. Although the MCNP-based system matrix requires more processing time to generate, its more accurate incorporation of the competing gamma-ray interaction processes results in effective localization of multiple point sources. In contrast, the rapidly produced ray-driven reconstruction matrix demonstrates reduced accuracy in localizing extending distributions due to its simplified treatment of the particle transport, the magnitude of which is quantified in the paper.