Investigating phase and artifacts in MRI : phase unwrapping, motional deghosting and fast dynamic flow imaging

In this thesis, the phase information contained in Magnetic Resonance (MR) images is studied while investigating the phase wrap and motional ghosting artifacts. The product of the work consists of novel processing techniques and data acquisition schemes. The work can be described in terms of three separate projects. The first project consists of studying the phase wrap artifact in MRI. A better characterisation of wrapped phase maps and the associated pole structures has led to the development of a new type of quality map, called a score map. Based on this, a new pole-guided-cutline method for phase unwrapping is realised. The second project consists of studying the motional ghosting artifact. An existing method of ghost suppression (lpGEM), relying on a gradient energy minimization (GEM) criterion, is improved. An analytical method of obtaining G EM is derived, leading to a new, more general method of ghost suppression: 2pGEM. The gradient energy of an image is proposed as a good quantitative evaluator of ghost suppression methods. The 2pGEM method of ghost suppression relies on the input of two images with significantly differing ghosts. A more efficient method, requiring less data acquisition, is presented. Although the proposed technique has not yet been implemented, due to some practical limitations, this work helps support the extended use of 2pGEM on ghosting artifacts caused by factors other than motion. The 2pGEM method is also tested on a new type of imaging sequence, called the OKscan, which produces dissimilar ghosting patterns. The successful results emphasize the more versatile applicability of the 2pGEM method relative to the original, lpGEM. The last project, while in a preliminary stage, demonstrates how a dynamic flow profile can be extracted from the motional ghosts. Such profiles are clinically available by using cardiac gated or triggered procedures. The benefit of the proposed method is its speed and simplicity of implementation since no sophisticated hardware or patient preparation is necessary. Instead, the velocity-induced artifacts in ungated images are used to predict the sought profile. The results are promising since the profiles are comparable to those obtained using a clinical standard which relies on gating.Science, Faculty ofPhysics and Astronomy, Department ofGraduat