| Magnetic Resonance Imaging (MRI) is a powerful imaging modality that has been used for diagnosing abnormalities in virtually every part of the body. One of the major challenges in clinical MR imaging is to maximize the diagnostic value of images in the presence of off-resonance conditions, such as field inhomogeneities and chemical shift. Unless properly addressed, field inhomogeneities lead to failure in water-fat separation and severe image distortions, while chemical-shift-induced artifacts degrade the image quality. This thesis focuses on correcting these off-resonance conditions in MR imaging.;The success of phase-sensitive water-fat separation depends on an accurate estimation of field inhomogeneities from the acquisition of multiple images. While efficient image acquisition can be obtained with a multi-echo sequence, the echo-spacings are necessarily long, which impose great challenges on estimating field inhomogeneities. The first part of this thesis describes a robust field map estimation method to handle such long echo-spacings.;The efficiency of multi-echo sequences can be further improved by replacing fly-back gradients with alternating readout gradients. This results in highly efficient bipolar multi-echo sequences. Nevertheless, the alternating readout gradients also lead to a number of technical problems, including phase errors and misregistrations. These problems are addressed in the second part of this thesis to enable reliable water-fat separation with bipolar multi-echo sequences.;The final part of this thesis deals with field inhomogeneities caused by metal. In comparison with field inhomogeneities expected from most MR imaging environments, metallic implants induce such severe off-resonance that the resultant images cannot be used for diagnosis. This work proposes an effective technique towards distortion-free MR imaging near metallic implants with clinically feasible scan times. |