| Resolution is an attribute of magnetic resonance imaging that is important to optimize. In diffusion-weighted magnetic resonance imaging, not only good spatial resolution is highly desirable, but also good resolution of the measured probability distribution function of underlying spin diffusion. Higher resolution of the probability distribution function requires more accurate physical models, while higher image spatial resolution requires more advanced data acquisition techniques and better system hardware.; The first part of this thesis presents attempts to improve the resolution of the probability distribution function by introducing a generalized diffusion tensor imaging method. This method derives the relationship between the magnetic resonance signal and the higher order statistics of a spin's random displacement. By measuring the higher order statistics, not only Gaussian diffusion but also non-Gaussian diffusion can be characterized. Complex biological structures can be potentially inferred from the measured diffusion properties.; The second part of this thesis develops two techniques to improve the spatial resolution of diffusion-weighted images: multi-shot technique and parallel imaging. A method is introduced to acquire data using self-navigated interleaved spiral readout trajectory. Two basic techniques are developed for image reconstruction in the presence of motion-induced phase errors, namely a direct phase subtraction method and a conjugate gradient method. Multi-shot technique is also combined with parallel imaging to further improve the image resolution or shorten the acquisition time. A technique for simultaneous phase correction and parallel imaging reconstruction is proposed and demonstrated with both computer simulation and in vivo studies. |
