| dc.description.abstracteng | The conventional multi-echo radial fast flow angle shot (FLASH) utilizes bipolar readout gradients with the same amplitude and duration to sample echoes in a fly-back-and-forth manner. This strategy is inefficient with respect to k-space coverage and temporal resolution. Therefore, in this thesis, a multi-echo multi-spoke radial FLASH sequence, i.e., capable of acquiring multiple echoes with different spatial encodings per radio-frequency excitation, is developed for the purposes of faster k-space sampling and better temporal and/or spatial resolution. This sequence, most importantly, can advance various applications of real-time MRI, e.g., dynamic quantitative T2* mapping, water-fat separation, functional imaging, and quantitative susceptibility mapping. On the other hand, a model-based image reconstruction, especially for real-time phase-contrast flow MRI, is developed in this thesis. Model-based image reconstruction emerges as a novel reconstruction technique that estimates maps of interest in MR signal models directly from the acquired data. In phase-contrast flow MRI, two reconstruction steps are typically involved. First, the measurements with different velocity encodings are treated as independent streams and reconstructed separately via parallel imaging reconstruction methods. Second, the phase-difference calculation between the reconstructed images is used to obtain phase-contrast velocity maps. The second step, however, induces severe random phase noise in no or low MR signal areas (e.g. air and lung), which can hamper the lumen definition, especially in the case of highly undersampled acquisitions. To ameliorate the situation, a novel model-based reconstruction technique, which directly estimates and regularizes the phase-contrast map from the measured datasets based on a proper signal modeling and hence ensures zero phase in the areas without MR signals, is developed. | de |