Positioning technique of coded aperture radiation imaging

Abstract

With the widespread application of nuclear technology and radiation protection, the demand for radioactive sources imaging is increasing. As a high-precision imaging and positioning device for radioactive sources, the coded aperture imaging positioning system can accurately determine the location of radioactive sources and reconstruct their rough shape. This study explores the comparison of the reconstruction effects of various reconstruction algorithms in coded aperture imaging positioning on the position and shape reconstruction of radioactive sources with continuous energy spectra, to determine the advantages and disadvantages of different reconstruction algorithms and their applicable scenarios. Geant4 software was used to simulate the encoded aperture imaging positioning system, and the relevant data were obtained. Thereafter, the δ decoding, fine sampling balance decoding, and convolutional neural network (CNN) algorithms, along with the maximum likelihood maximum expected value method (MLEM) were used to program and reconstruct the location of the radioactive source. The results demonstrate that the four reconstruction algorithms can locate the radioactive source clearly; the δ decoding and fine sampling balance decoding algorithms have artifacts to reconstruct the image; and the CNN algorithm has a poor effect on the reconstruction of line and surface sources, which can be addressed by an extended training set; the contrast to noise ratio (CNR) value of the MLEM algorithm is high, and the reconstruction effect is good, however, some details of the line and surface sources reconstruction will be lost.

Keywords

• radiation imaging localization
• collimator
• monte carlo simulation
• image reconstruction
• neural network