A DTI registration algorithm based on spectral representation

Abstract

AbstractDiffusion Tensor Imaging (DTI) is a new technique to characterize the brain structure. It is a special modality for Magnetic Resonance (MRI) which provides the information of fiber orientation. By quantitatively comparing the difference of the fiber information of different image pairs, one can effectively track the status of the nerve diseases. For this purpose, it is necessary to deform the DTI image of the patient (floating image) to the same shape of the template image (target image). This is so called DTI registration, which is usually formulated by a 3D variational problem. However, the numerical implementation for 3D variational problem costs too much CPU consumption. To reduce CPU consumption, it is meaningful to give some research on the structure of the solution for DTI registration. In this paper, the structure of the solution for DTI registration is mainly focused on. For this purpose, an explicit representation of the solution for a DTI registration model is presented. Based on this representation, a new DTI registration algorithm is induced. Several numerical experiments are also performed to show the efficiency of the proposed algorithm in DTI registration.