The Study Of Fast Mri Technique Based On Compressed Sensing

The improvement of people's living usually attributes to science and technology, the field of medical treatment and healthy is no exception. At present, the researches of main medical imaging technologies are springing up, including the Ultrasound(US), Computed Tomography(CT), Nuclear Magnetic Resonance Imaging(NMRI) and Positron Emission Computed Tomography(PET). Nuclear Magnetic Resonance Imaging is a nuclear magnetic resonance based body-imaging method, which could stimulate abundant hydrogen protons in human body and capture energy signal that released from the vibrational state to the steady, thereby the image information of the human tissue and physiological function are obtained. But the acquisition of imaging data needs long time, and the efficiency of imaging is very low due to limitation of the Nyquist sampling theorem in practice. Therefore, how to improve the speed of imaging and ensure the quality of reconstruction has become a study hotspot in the field of magnetic resonance imaging.Compressed Sensing(CS) has been applied in the magnetic resonance imaging by reason of its superior ability to recover the down-sampled data. Specially, one can make use of image's sparsity in one liner transform domain to accurately reconstruct original image by a small amount of non-coherent down-sampled data in larger probability. To solve the problem of the long process of data acquisition and slow imaging speed, this work proposes two kind of fast magnetic resonance imaging methods based on Compressed Sensing frame, which aims to ensure the imaging quality and improve the ratio of undersampling.One method is to use general series(GS) model to make the similar reference image couple to the general expressed series of target image, which is as a new expression of target image, after that one constrains the sparsity of the residual image from the target image. When the accelerating factor is high, the reconstruction method based on CS still attains the accurate results. Here the multiple scan experimental data are used to verify the validity of the proposed method. The relative error of reconstruction is decreased sharply from 23.22% to 12.31% under condition that the accelerating factor is 5.The other method is to use the gradient orientation from similar reference image to constrain the edge reconstruction. Specifically, we regularize the tangent vector in the target image to be enough perpendicular to the corresponding normal vector in the reference image over all spatial locations to make the gradient orientations in the reference and the target image consistent. The experimental results sign that the proposed method further improve the accelerating factor and shorten the imaging time, whether it is for a single contrast image or multiple contrast images. The accelerating factor can reach up to 6 and 8 in the one and two dimensional encoding direction respectively.The proposed methods focus on the two problems of slow imaging speed and the poor quality of recovery. Our methods both provide the better ability of reconstruction compared to other conventional methods by the simulated and in-vivo data experiments, which can be as a new idea and method for the practical application of fast MR imaging based on CS.