Study On Super-resolved Reconstruction Of Spatiotemporally Encoded Single-shot UltirafastMagnetic Resonante Imaging

Ultrafast MRI plays an essential role in biomedicine like functional neuroimaging, real-time dynamic imaging and diffusion tensor imaging due to its excellent time resolution. A newborn method based on spatiotemporally encoded single-shot technique develops rapidly in recent years because of its robustness to inhomogeneous magnetic field and chemical shift. Different from the conventional method, which obtains the image by Fourier transform, the spatiotemporally encoded method can produce a raw image by straightforward magnitude-calculation operation. However, the resolution of the raw image is inferior to that of conventional method. The super-resolved (SR) reconstruction is indispensable to improve the image resolution without additional acquisition. In this thesis, the principle of SR reconstruction was introduced in detail and some novel SR algorithms were proposed to efficiently reduce the aliasing artifacts caused by vast under sampling and improve the image quality. The main work is listed below.1. The principles and properties of spatiotemporally encoded single-shot ultrafast method were systematically and comprehensively introduced, including the principles of spatiotemporal encoding and decoding, the inherent resolution of the resulting image and the resistance of the method to inhomogeneous magnetic field and chemical shift. The development and limitations of spatiotemporally encoded method were also discussed. Moreover, its applications in chemical shift imaging, parallel imaging and multi-slice imaging were described in detail.2. The principle of SR reconstruction was introduced in detail. The problems raised from vast under sampling was discussed, such as the huge condition number of coefficient matrix and the low digital resolution of SR image. The singular value decomposition preprocessing, partial Fourier transform and deconvolution algorithm were proposed to overcome these problems. The simulation and experimental results showed that these algorithms could efficiently improve the SR image quality.3. The pseudo-random sampling was introduced to the acquisition process to disperse the aliasing artifacts cause by vast under sampling. The compress sensing based on contourlet transform and total variation was employed to further reduce the aliasing artifacts and improve the resolution and signal-noise ratio of the image. The robustness of this scheme was demonstrated by various simulations and experiments under different parameters, circumstances and models.