Magnetic resonance imaging (MRI) is a noninvasive clinical imaging modality with very rich contrasts based on the physical properties of the imaged tissues. MRI can be used for quantification of volumetric distributions of various biomolecules and chemical elements - such as triglycerides, calcium and iron - that regarded as participants in normal tissue biochemistry, and whose dysregulations are often manifested in pathologic processes. This dissertation reports optimization steps undertaken to overcome technical challenges in quantitative susceptibility mapping (QSM) in different parts of human body. Often in QSM it is assumed that susceptibility is the only contributor to the observed field inhomogeneity, which may be a valid assumption for neuroimaging applications. However, multiple molecules found in biological tissues (e.g., triglycerides of fat) have a resonance frequency different from that of water, and this resonance frequency offset is referred to as chemical shift. This chemical shift affects the phase of the MRI signal. Although ways to estimate field inhomogeneity in the presence of chemical shift have been proposed, they often rely on the a priori knowledge of the chemical spectrum. Unfortunately, variability of chemical spectra have been reported. In this dissertation, an automated joint estimation of the chemical shift and the susceptibility from an MRI dataset is reported, where the chemical shift is also treated as an unknown variable subject to optimization. QSM may become a useful diagnostic tool for noninvasive assessment of bone health without the use of ionizing radiation, however this application has been a challenging task challenging because QSM requires complete measurements of phase everywhere within the region of interest, and cortical bone typically has very low or no signal at conventional echo times in gradient echo (GRE) imaging. An additional problem arises from intermingling of fat and water protons in the bone marrow, necessitating the application of water–fat separation techniques for field mapping. In this dissertation, a novel signal model is proposed, feasibility of using QSM for measuring bone MRI signal is investigated, and the inherent technical issues involved in this application are highlighted. QSM has been widely applied in neuroimaging. In particular, due to its ability to accurately map iron deposits in deep brain nuclei, QSM promises precise targeting of the subthalamic nucleus (STN) in deep brain stimulation surgery (DBS). This dissertation reports results of comparison between QSM and standard-of-care T2w imaging of the STN, and their performance in high-resolution presugrical anatomic imaging.