Correcting Susceptibility Artifacts in High-resolution Brain Magnetic Resonance Imaging

Echo planar imaging (EPI) is a fast and non-invasive magnetic resonance imaging (MRI) technique that supports data acquisition at high spatial and temporal resolutions. Thus, EPI is widely used for human brain studies in both clinical diagnosis and scientific investigation. However, susceptibility artifacts, which cause the misalignment to the underlying structural image, are unavoidable distortions in EPI. These distortions are especially severe in high spatial-resolution images and can lead to misrepresentation of the human brain functions. Many susceptibility artifact correction (SAC) methods have been developed to address these challenges, but they require high computational resources, modified scanner hardware, or a modified acquisition protocol. This thesis investigates existing SAC methods and develops new alternatives for high-resolution brain EPI images. The aims of developing new SAC methods are to reduce the computational cost and improve the correction accuracy. Three novel SAC methods, which are from two main categories: traditional iterativeoptimization and deep learning, are proposed and tested in this research.