Reconstruction method for gamma-ray coded-aperture imaging based on convolutional neural network

Abstract

Abstract Coded-aperture gamma-ray imaging has great application value in the fields of nuclear security, nuclear facility decommissioning, and decontamination verification. However, conventional reconstruction methods cannot handle the signal-independent noise. In this paper, a coded-aperture imaging reconstruction method based on convolutional neural network (CNN) was proposed to improve the performance of image reconstruction and enhance the source position recognition ability of imaging systems. In addition, a compact gamma camera based on cadmium zinc telluride (CZT) pixel detector and uniformly redundant array (MURA) mask was modeled. Monte Carlo simulation data were used to train CNN and test the performance of this method. Furthermore, the reconstruction of the CNN method and the correlation analysis method with different radioactive sources and measurement conditions were compared. Results show that the proposed method can suppress the reconstructed image noise well. The reconstructed images have higher contrast-to-noise ratio (CNR) than the correlation analysis method in radioactive source location.