Research On MRI Image Reconstruction Technology Based On Compressed Sensing

Magnetic resonance imaging(MRI)is an important tool for medical imaging,but the data acquisition process is very slow.Compressed sensing(CS)is a new signal sampling theory.Based on compressed sensing,sparse or compressible signals can accurately be recovered from a small quantity of sampled data,which is far less than the sampling quantity by the traditional sampling methods.So it is a new way to improve the speed of magnetic resonance imaging.Based on the study of compressed sensing and magnetic resonance imaging,how to apply the compressed sensing theory to the magnetic resonance image reconstruction is studied in this paper.And the major work includes the following aspects:(1)MRI and compressed sensing theory are researched.The principle and method of MRI reconstruction based on compressed sensing are expounded.(2)In view of the high coherence between the common sampling matrix and sparse matrix,this paper puts forward an optimization method of sampling matrix.The variable density circle sampling method is proposed with the superposition of the deterministic ring sampling matrix and the common random sampling matrix.Simulation experiments results show that this method has a lower correlation between sampling matrix and sparse matrix,and improve the quality of the reconstructed image on the condition that the sampling quantity is reduced.(3)As the reconstruction algorithms based on l1-norm minimization model need more sampled data,this paper transforms the reconstruction model of CS into the lp-norm minimization model,and applies it to the optimization goal of the CS-MRI.This paper modifies the threshold function and the shrinkage rule.Based on the majorization minimization algorithm,this paper proposes a new iterative threshold shrinkage algorithm for lp-norm non-convex optimal problem.Simulation experiments results show that this modified algorithm can get higher quality reconstruction image,by contrasting with other algorithms on peak signal noise ratio and normalized mean squared error.