Study On Artifacts Reduction In SENSE Parallel Mri And Sparse Sampling Reconstructed Algorithm

Magnetic resonance imaging (MRI), based on the principle of nuclear magnetic resonance (NMR), can be used to produce high quality images with the functional, structural and lesion's information of an organism. Compared with other imaging modalities (such as X-ray , CT et al.), MRI has many advantages with higher spatial resolution, arbitrary slice, multi-parameters imaging and free harmful radiation, and has been widely used in clinical diagnosis and scientific research fields. However, conventional MRI is time consuming and thus has limitation on some special applications such as functional brain imaging, real-time imaging of cardiovascular system, and so on. How to improve MRI speed and reconstruction image quality have been always focused on in this area. In this thesis, SENSE parallel imaging and PSF sparse imaging are investigated based on the systematic study of the existing fast MRI methods. The main works and contributions of this thesis are described as follows:Firstly, a weighted-sum-of-squares (WSOS) method based on the mutual information is presented to improving the reconstruction quality of SENSE parallel imaging technique. In Parallel MRI, multiple receiver coils are used to collect data at the same time. It can skip some phase encoding lines to speed the MRI with the same spatial resolution as full sampling method. However, when data from one or more coils are corrupted due to involuntary motion in the duration of sampling, the quality of reconstructed image by conventional Sum-of-Squares (SOS) method will be low with obvious motion artifacts. In this work, an improved SOS algorithm called as Weighted-Sum-of-Squares (WSOS) is presented to increase robustness of SENSE parallel imaging. Homomorphic filtering and B-spline smoothing method are firstly employed to obtain a group of uniform brightness phased array coil images, and then the corrupted data based on the mutual information between coil images are accurately estimated. Finally, the weighting values are adaptively assigned in the reconstruction step according to the estimation results, thus the influence of corrupted data is attenuated in the reconstruction stage and the motion artifacts could be effectively reduced. Experiments for Phantom and brain imaging with involuntary movement show that this method can effectively reduce or eliminate the motion artifacts.Secondly, the combination of partially separable functions (PSF) model and keyhole technique is proposed for further accelerating the MRI speed. The PSF model is an approach that provides high spatial-temporal dynamic images by exploiting the redundancy of frequency and time domain data. However, there are some critical parameters, including temporal basis, spatial basis and motional frequency component in the model. It can't be used to reconstruct any image before obtaining these parameters value. According to the algorithm of the PSF, it is necessary to acquire a large amount of pre-data to accurately calculate the critical parameters. It would cost a long pre-data scan time for imaging. In order to overcome this problem, the combination of PSF model and keyhole technique was proposed in this work to reduce the pre-data scanning time. Simulation experiments for a dynamic ball model which is periodic rapid expansion and contraction demonstrate that, the proposed method can efficiently reduce the pre-scan time while maintaining the same quality image as that by the conventional PSF method.