磁性粒子成像重建算法分析-analysis of magnetic particle imaging reconstruction algorithm.docxVIP

AbstractMagnetic Particle Imaging (MPI) is a new type of tomographic Imaging technology. Compared with the predominant imaging technology-Magnetic Resonance Imaging technology, MPI reaches a high level of sensitivity and spatial resolution. A complete set of Magnetic Particle Imaging system includes the excitation fields, the signal pickup assembly and the image processing part in host computer. The image processing part, which is mainly responsible for obtaining the concentration distribution of the magnetic particle sample through the reconstruction algorithm by using the nonlinear response of Magnetic nanoparticles(MNPs), is the core part of the imaging system.Based on the introduction of a study summarization on Magnetic Particle Imaging both at home and abroad, this paper did deep research and analyzed the related problems of image reconstruction in 1D MPI. This paper firstly introduced the common attributes of nanometer materials and the superparamagnetism of MNPs, as well as the basic principle of MPI in details based on the Langevin theory of magnetism. Then, the mathematical model was established in Matlab according to the imaging principle and two mainstream algorithm - SVD algorithm and LSQR algorithm were used to discuss the factors that influence the accuracy of the solution concentration, such as acquisition frequency, size distribution ofMNPs, density of spatial coding points，ambient noise and so on. The condition number ofsystem matrix, which is a mathematical quantity that directly related to the solution precision, was found in the process of simulation. The simulation results showed that the condition number of system matrix not only directly influenced the solution precision in the absence of noise, but also influenced the noise rejection capability of system with the existence of ambient noise. Finally the Tikhonov regularization method was proposed to improve thealgorithm used in MPI, simulation results showed that the improved algori