Simulation and Image Reconstruction of Fast Neutron Imaging System

Abstract

The traditional X-ray interrogation system cannot detect shielded nuclear materials. Due to the high penetration ability of fast neutrons, the fast neutron imaging can complement the existing x-ray based technology and contribute to the national security and defense. In this paper, a fast neutron imaging system composed of an adjoint α particle detector, a DT neutron source and a fast neutron detector array was designed. Using the simulated cone beam CT projection data from the GEANT4 program of the simplified system, the study on the image reconstruction methods was carried out. The results show that the iterative algorithm MLEM can achieve a better image quality compared with the analytic algorithm FDK. While an only drawback of the MLEM algorithm is its more time-consuming. Under the same conditions, the FDK algorithm takes 5.6s to reconstruct a profile while the single iteration of MLEM algorithm takes about 4s and 10 iterations are required to obtain a good result. When the GPU acceleration is used, the single iteration can be done in 2s. Iterative algorithms are more suitable for neutron imaging while the GPU acceleration technology can shorten the iteration time. Further study about simulations for the complete system and new iterative algorithm suitable for this work and neutron imaging will be carried on.