Magnetic Resonance Image Reconstruction Using Reference Images

Magnetic resonance imaging(MRI)is a technique for human imaging using the spin resonance of hydrogen protons.It has become one of the most important technologies in the field of medical imaging because of its no radiation and high resolution.However,the relatively slow imaging speed has been always a major factor affecting MRI clinical throughput and imaging quality.How to improve the imaging speed and image quality has always been a hot topic in MRI technology.In recent years,with the emergence of Compressed Sensing(CS)theory,CS-MRI imaging methods have drawn much attention.CS-MRI technology effectively reduces the amount of k-space data required by using the sparsity or compressibility of the signal.In MRI applications,it is required to image the same objects many times at different time the similarity among multiple images exist.If the information provided by such images used effectively,the measurement data of CS-MRI method will be further reduced.Based on CS-MRI theory,this paper uses reference image information to establish MR image reconstruction model,enabling high-quality reconstruction from highly undersampled k-space data.This paper proposes and studies the following three MR image reconstruction methods:(1)A new method of MR image reconstruction using differential image gradient restoration is proposed.The method considers the target image as the sum of the reference image and the difference image.Because the gradient image of difference image is more sparse than the difference image itself,this paper first proposes to reconstruct the gradient image of the difference image,so as to get the difference image,and finally get the target image.Experimental results on practical MR images demonstrate that our methods can reduce the data sampling rate further while ensuring the quality of reconstruction.(2)A CS-MRI method based on difference image and low-rank matrix which is constructed by local k-space neighborhoods(LORAKS).The target image is modeled as the summation of the reference image and the difference image.By using the linear correlation indifference image's k-space data,forms a low rank C matrix and a low rank S matrix.By constraining the low rank of C matrix and S matrix,the difference image is reconstructed and the target image is obtained.Several experiments shows that the reconstruction performance of our methods outperforms the compared method,especially have better reconstructed quality at low sampling rate.(3)A CS-MRI method using the the spatial information of the reference image to reconstruct the MR image based on union of subspaces is proposed.This method based on the framework of the Uo S-CS-MRI reconstruction method.The main coefficients of the image are determined by the space information provided by the reference image.The simulation results from real MR image show that the reconstruction quality under different sampling rates of the proposed method is optimal.