| Recent advances in imaging technologies have allowed an increased understanding of normal and abnormal brain structures and functions. The huge variations among individual brains make it difficult to compare brains across subjects. This thesis presents a novel and mathematically rigorous computational framework for image registration, which is based on global landmark-constrained conformal mapping. The accuracy of our framework is evaluated and further refined by Mutual Information based spherical domain registration. Our framework provides a powerful tool for cross-subject imaging data analysis, and has been used for inter-subject fiber tracts quantification and intra-subject resection contour mapping. |
