Research On Parallel Magnetic Resonance Imaging By Compressed Sensing

Magnetic resonance imaging(MRI)is an important medical imaging modality,however,its scanning time is too long,which limits its further development and application.Parallel magnetic resonance imaging(pMRI)accelerates the imaging process by using multiple coil during acquisition,but the larger acceleration factor will bring noises and artifacts,which makes the signal to noise ratio of the reconstructed image decrease.Compressed sensing(CS)can be applied to MRI,which can improve the speed of imaging to some extent.Some researches show that the combination method of pMRI and CS can be achieved,which can not only ensure the quality of reconstruction but also reduce the reconstruction time,thereby reducing the time of diagnosis of patients,so this study has great significance in the field of medical imaging.This paper studies the related reconstruction problems of traditional pMRI combined with CS imaging method,and focuses on the combing of the CS method and MRI by analyzing of their advantages and disadvantages.At last,We put forward the improvement method for the three aspects: CS measurement matrix,CS-MRI reconstruction model and reconstruction algorithm.The concrete contents are divided into the following four parts:(1)We firstly focus on the pMRI and CS-MRI reconstruction method,and mainly analyze their advantages and disadvantages.pMRI can accelerate the imaging process by parallel acquisition,but the SNR will decrease obviously with the increase of the acceleration factor,while the CS-MRI method can reconstruct the satisfied spatial resolution image when the reconstruction condition is satisfied.In this paper,we focus on the three concrete reconstruction conditions,which including CS measurement matrix,CS-MRI reconstruction model and reconstruction algorithm.(2)The improvement of multi-channel compressed sensing MRI reconstruction method(MCS-MRI)is developed.Since the incoherence between the encoding matrix Fourier and the sparsity transform matrix wavelet is not optimal,in this paper,we introduce the ±1 matrix Noiselet which satisfies the Bernoulli distribution to the MCS-MRI mode.As a new measurement matrix,the Noiselet matrix has a stronger incoherence with the sparse transform matrix,and more satisfies the CS imagingcondition.Some experiments have proved that the improved algorithm can improve the quality of reconstructed images.(3)A new model SCS-MRI is developed base on the sparse prior information.In the traditional reconstruction model,the prior information of the MRI image in the sparse transform domain is not used.According to the characteristics of the MRI image,the sparse item based on the prior information of the MRI image is added into the CS-MRI image.We use the fast iterative shrinkage thresholding algorithm(FISTA)to solve the whole model.In the process of solving the problem by FISTA,we found that the objective function is not strictly decreasing.In order to solve this problem,we put forward a constrained FISTA algorithm.At last,the experimental results show that the improved model is more stable than the previous model and the reconstruction errors of the reconstructed images are reduced at the same sampling rate.(4)Combing the new model with the SENSE and GRAPPA method respectively.The improved CS-MRI model is respectively combined with SENSE and GRAPPA methods,and we need to design the appropriate sampling mask for the specific method.By using experimental analysis and quantitative calculation,we show that the improved model can improve the reconstruction quality of pMRI and it has good applicability.