Date of Award

December 2012

Degree Type


Degree Name

Doctor of Philosophy



First Advisor

Jun Zhang

Committee Members

Leslie (Lei) Ying, Kevin F. King, Yi Hu, Zeyun Yu


Circulant Random Encoding, Compressed Sensing, GRAPPA, MRI, Parallel Imaging, Pulse Sequence


Magnetic Resonance Imaging (MRI) scanners are one of important medical instruments, which can achieve more information of soft issues in human body than other medical instruments, such as Ultrasound, Computed Tomography (CT), Single Photon Emission Computed Tomography (SPECT), Positron Emission Tomography (PET), etc. But MRI's scanning is slow for patience of doctors and patients. In this dissertation, the author proposes some methods of parallel imaging and compressed sensing to accelerate MRI data acquisition. Firstly, a method is proposed to improve the conventional GRAPPA using cross-sampled auto-calibration data. This method use cross-sampled auto-calibration data instead of the conventional parallel-sampled auto-calibration data to estimate the linear kernel model of the conventional GRAPPA. The simulations and experiments show that the cross-sampled GRAPPA can decrease the quantity of ACS lines and reduce the aliasing artifacts comparing to the conventional GRAPPA under same reduction factors. Secondly, a Hybrid encoding method is proposed to accelerate the MRI data acquisition using compressed sensing. This method completely changes the conventional Fourier encoding into Hybrid encoding, which combines the benefits of Fourier and Circulant random encoding, under 2D and 3D situation, through the proposed special hybrid encoding pulse sequences. The simulations and experiments illustrate that the images can be reconstructed by the proposed Hybrid encoding method to reserve more details and resolutions than the conventional Fourier encoding method. Thirdly, a pseudo 2D random sampling method is proposed by dynamically swapping the gradients of x and y axes on pulse sequences, which can be implemented physically as the convention 1D random sampling method. The simulations show that the proposed method can reserve more details than the convention 1D random sampling method. These methods can recover images to achieve better qualities under same situations than the conventional methods. Using these methods, the MRI data acquisitions can be accelerated comparing to the conventional methods.