Comprehensive acquisition, reconstruction, and visualization for diagnostic and interventional time-resolved three-dimensional MR angiography

Magnetic Resonance Imaging (MRI) has become a major technique for the diagnosis of disease processes, especially in the neurological and musculoskeletal system. It is particularly valuable because it provides excellent soft-tissue contrast in arbitrary imaging planes, without the use of ionizing radiation. It is infinitely flexible due to the multitudes of contrast mechanisms and imaging sequence that can be utilized, and has been effectively utilized to provide both morphological and functional information.; Although MR imaging is fundamentally slow and has historically been used mainly as a clinical diagnostic tool, improving hardware has spurred the development of rapid imaging sequences that have enabled vascular imaging and the study of other dynamic processes. These sequences have the potential of enabling interventional procedures to be performed under MR guidance. Interventional Radiology offers the promise of minimally invasive surgery for procedures such as cardiac catheterization and the treatment of liver cancer.; To speed imaging, techniques frequently sample the MR data space using non-Cartesian patterns that more effectively utilize scanner hardware. Extensive research into non-Cartesian imaging over the last fifteen years has yielded techniques which are successful in trials, but inconsistent performance has limited their clinical impact. One reason for the inconsistent performance is that non-Cartesian imaging techniques are sensitive to scanner imperfections that have minimal impact on conventional Cartesian exams. A major focus of this work is an analysis of the impact of gradient imperfections on 3D Projection Reconstruction (3DPR) acquisitions and a discussion of techniques to correct resulting errors. A second major focus of this work is transitioning MR from a diagnostic to interventional tool by developing a real-time system that performs acquisition, reconstruction, and interactive visualization of 3D images. An enhancement of the real-time system that uses distributed computing to improve reconstruction performance is also presented. Also discussed is an enhancement to the time-resolved 3DPR imaging technique that uses multiple echoes to improve data acquisition efficiency and reduce streak artifacts, a technique which uses phase information in the multiple echoes to generate water and fat images from a single acquisition, and techniques for visualization and interactive manipulation of time-resolved 3D volume images.