Bi-modal Non-rigid Registration of Brain MRI Data Based on Deconvolution of Joint Statistics

摘要

Images of different contrasts in MRI can contain complementary information and can highlight different tissue types. Such datasets often need to be co-registered for any further processing. A novel and effective non-rigid registration method based on the restoration of the joint statistics of pairs of such images is proposed. The registration is performed with the deconvolution of the joint statistics and then with the enforcement of the deconvolved statistics back to the spatial domain to form a preliminary registration. The spatial transformation is also regularized with Gaussian spatial smoothing. The registration method has been compared to B-Splines and validated with a simulated Shepp-Logan phantom, with the BrainWeb phantom, and with real datasets. Improved results have been obtained for both accuracy as well as efficiency.