Super-resolution Reconstruction of Orientation Distribution Function from Motion Scattered Diffusion Weighted Image Slices and its Application to Fetal Brain Connectivity Study

Thesis (Master's)--University of Washington, 2015Diffusion-weighted MRI brings new and complimentary information about the development of the brain. It is also actively investigated how diffusion contrast properties correlates with brain diseases. Though promising, this technique faces a number of technical challenges, especially in fetal imaging, where unrestricted movement occurs. Despite the usage of fast imaging sequences, the acquired images are still corrupted by motion artifacts, making them hardly ready to use by doctors or researchers. Therefore, post image processing methods are in great need to estimate the movement of the subject and to reconstruct high quality diffusion volumes for both clinical and research purposes, e.g., the study of fetal brain connectivity. This thesis firstly gives a background of diffusion and its mathematical models, diffusion inside the brain and diffusion-weighted MRI. Then it introduces the source of the motion artifacts in fetal MRI, and presents the current state-of-the-art method for its removal. After that, a novel super-resolution reconstruction framework for recovering a higher order ODF volume from motion scattered DWI image slices is proposed and experimentally compared to the methods reported in the literature. Both human adult data and macaque fetal data are used for evaluation. Finally, a structural connectivity study of the developing macaque fetal brains is carried out using unbiased template free brain parcellation schemes and graph theory based analysis