Water-fat imaging and general chemical shift imaging with spectrum modeling

Water-fat chemical shift imaging (CSI) has been an active research area in magnetic resonance imaging (MRI) since the early 1980's. There are two main reasons for water-fat imaging. First, water-fat imaging can serve as a fat-suppression method. Removing the usually bright fatty signals not only extends the useful dynamic range of an image, but also allows better visualization of lesions or injected contrast, and removes chemical shift artifacts, which may contribute to improved diagnosis. Second, quantification of water and fat provides useful chemical information for characterizing tissues such as bone marrow, liver, and adrenal masses. A milestone in water-fat imaging is the Dixon method that can produce separate water and fat images with only two data acquisitions. In practice, however, the Dixon method is not always successful due to field inhomogeneity problems. In recent years, many variations of the Dixon method have been proposed to overcome the field inhomogeneity problem. In general, these methods can at best separate water and fat without identifying the two because the water and fat magnetization vectors are sampled symmetrically, only parallel and anti-parallel. Furthermore, these methods usually depend on two-dimensional phase unwrapping which itself is sensitive to noise and artifacts, and becomes unreliable when the images have disconnected tissues in the field-of-view (FOV). We will first introduce the basic principles of nuclear magnetic resonance (NMR) and magnetic resonance imaging (MRI) in chapter 1, and briefly review the existing water-fat imaging techniques in chapter 2. In chapter 3, we will introduce a new method for waterfat imaging. With three image acquisitions, a general direct phase encoding (DPE) of the chemical shift information is achieved, which allows an unambiguous determination of water and fat on a pixel by pixel basis. Details of specific implementations and noise performance will be discussed. Representative results from volunteers and patients in a clinical setting will be presented. In chapter 4, new improvements in the signal-to-noiseratio (SNR) for the DPE method will be introduced and details of noise performance analysis will be discussed. In chapter 5, a special DPE sampling scheme will be introduced. With three-orthogonal phase (TOP) image acquisitions, it allows a correction of image magnitude errors caused by factors such as T2* relaxation. Details of data acquisition and signal processing will be discussed. Representative results from volunteers will be presented. In chapter 6, we will introduce a new two-point water-fat imaging method. By sampling water and fat asymmetrically and minimizing the gradient energy in a phase map, this method determines water and fat without ambiguity and handles disconnected tissues well. Details of data acquisition, signal processing, and noise performances will be discussed. Representative, results from volunteers will be presented. In chapter 7, we will introduce a new general method of chemical shift imaging with spectrum modeling (CSISM). CSISM models a spectrum as several peaks with known resonance frequencies but unknown peak amplitudes which can be resolved from a set of spin-echo images. Details of data acquisition, signal processing and noise performances will be discussed. Representative results from phantom experiments and a clinical scan will be presented. In chapter 8, the general ideas, results and conclusions of all the methods we introduced in this thesis will be discussed, compared, and summarized.Science, Faculty ofPhysics and Astronomy, Department ofGraduat