Magnetic Particle Imaging Using Discrete Sampling and Image Reconstruction with Few Orthogonal Bases Obtained by Singular Value Decomposition of Selected Delta Responses

Abstract

The conventional magnetic particle imaging reconstruction methods use observed signals and system functions, which result in an enormous amount of data and long processing time being required to reconstruct large image matrices. We propose a new image reconstruction method that uses less data and a limited number of orthogonal bases obtained via the singular value decomposition (SVD) of selected point spread functions (PSFs). By using the features of the diagonal and nondiagonal elements of a singular value matrix, image blurring and artifacts can be reduced in the reconstructed image. This is because the diagonal components commonly indicate the similarities between each orthogonal basis for the system function and an observed signal, whereas the nondiagonal components indicate the differences of both them. In this paper, we use numerical analyses to demonstrate that image reconstruction is possible by using effective orthogonal bases obtained through the SVD of a limited number of PSFs selected from a general system function. The reconstruction time is reduced to 1/12th to 1/50th of that of the conventional method. Int. J. Mag. Part. Imag. 6(1), 2020, Article ID: 2003003, DOI: 10.18416/IJMPI.2020.2003003