Rapid MR Imaging Methods Based On Parallel Reconstruction And Compressed Sensing

As a non-destructive method of medical imaging diagnosis, magnetic resonance imagingcan be used to obtain accurate information about physiological functions and anatomicalstructure of human organs and tissues, and it has been widely used in modern medicaldiagnosis. However, it requires relatively long time in conventional MR signal scanning andimage reconstruction, which affects real-time imaging and important applications such ascardiac imaging or fMRI. Therefore, fast MR imaging techniques have become a majorconcern for most researchers. In this paper, we investigate the SENSE (sensitivity encoding)reconstruction method and introduce the zero-padding reconstruction in K-space signalacquisition. Moreover, we introduce compressed sensing algorithm. The work in this papercan be mainly described as below:1. The SENSE reconstruction. The SENSE method is a parallel reconstruction algorithmbased on image space, which uses multichannel phased-array coil to receive MR image, andunfolds phase-wrap artifacts caused by under-sampling according to different sensitivity mapof each channel. The key issue of SENSE reconstruction is the estimation of coil sensitivitymap. In this paper, we use reference scanning method to acquire coil sensitivity map andtestify different reduced-sampling factor and ACS.2. Based on the existing zero-padding reconstruction, we consider the K-space conjugatesymmetry and propose a new K-space signal acquisition and reconstruction method called"Advance zero-padding&PFFT(Partial Fast Fourier Transformation, PFFT)". The advanceZero-padding&PFFT method can not only increase speed of MR scan, but also restore morehigh frequency signal in K-pace, which reveals detail information. Consequently, theimproved scheme may reduce truncation artifacts or Gibbs phenomenon in reconstructionimage.3. Compressed sensing (CS) theory in MR fast imaging. In this section, we describe thebasic principles of compressed sensing, and introduce three critical steps: sparsityrepresentation, random measurement and nonlinear reconstruction. Both FFT and Wavelettransformation of MR image could meet the requirement of sparsity, and we testify andcompare different ways of random under-sampling of K-space signal.