Image-Based FMRI Artifacts Suppression Methods

Functional Magnetic Resonance Imaging is a new development of MRI technology in the early 1990s. Because its high spatial resolution, temporal resolution, and absolutely non-invasive characteristics, it has become a significant tool to study brain science and life science. However, there are always kinds of artifacts in fMRI due to defection of technology, limitation of device and patient's motions during scan. These artifacts degrade quality of the result image, make the detecting and locating of the functional active region more complicated.We study EPI ghost artifact and motion artifact in this thesis, which are the most common fMRI artifacts and have gravest influence on image.We have implements several methods to suppress EPI ghost artifact, including odd/even line reconstruction method, projection onto convex sets method, second moment method. We analyze capabilities of these methods through emulation experiments and practical experiments. The results show the most useful method is second moment method. We improve the last two methods by using linear fitting, which can reduce calculation, improve quality of the result image and the ability of noise resistance.As to motion artifact, we have studied some methods for suppressing rigid translational motion artifact, including translational motion in the imaging plane and slice selection axis translational motion. We analyze their speed, suppression performances and noise resistance. By combining phase correction method and magnitude correction method, we successfully correct 3-D translational motion artifact in emulation experiment.We present an automatic ROI extraction method based on Mathematical Morphology. It can extract ROI comparatively precise on the occasion that artifacts and noise are not very strong.The research of this thesis is supported by 973 plan (Code: 2003CB716102).