Efficient Schemes for MRI Reconstruction Using Compressive Sensing

Abstract

MRI is the most effective diagnostic imaging modality used for most of the clinical newlineapplications as it provides superior visualization of soft tissues in human body. The newlinenon-invasive nature of the imaging technique makes it more appreciable. The scan newlinetime is the major factor of concern in MR imaging scheme as the patient has to spend newlinemore time under the influence of magnetic field. There were many efforts were put in newlinethe direction of improving the speed of MR imaging process without compromising newlinewith the quality of image reconstructed. Traditional MR imaging techniques still rely newlineon the classic Nyquist-Shannon criterion for signal sampling. But reducing the newlinenumber of samples always lead to aliasing artifacts which intern degrade the image newlinequality. Compressive sensing (CS) has emerged an eminent signal processing newlineapproach and is widely applied in several fields like wireless communications, image newlineprocessing, magnetic resonance imaging, remote sensing imaging etc. as it can newlineperform simultaneous sampling and compression. CS can reconstruct the MR image newlinewith very few k-space coefficients. However major challenges in this field are the newlinequality of reconstruction and hardware limitations. In this thesis several techniques newlineare proposed for efficient application of CS on MR imaging techniques. newlineSampling trajectory play a major role in compressive sensing. There are few standard newlinetrajectories such as Cartesian, random and radial trajectories. The sampling trajectory newlinecan be designed to sample the k-space in an energy efficient way. As the first newlineobjective of this work an optimum sampling pattern for sampling the k-space was newlinedeveloped using the PDF. The white lines in the pattern generated are the locations newlinefrom where the samples are collected. The location of white lines is decided with the newlineproposed PDF. The sampling pattern generated so, was applied on the MR k-space to newlinecollect coefficients. The sampling trajectory samples the k-space to acquire more newlinesamples from the origin than the peripheral portion of MR k-space. The MR im