Advance in Diffusion Tensor Image Registration and Its Evaluation

With the extensive application of diffusion tensor imaging (DTI) in neuroscience, DTI registration has become one of the hottest topics in image processing and analysis. Meanwhile, DTI registration has generated considerable interests and been studied more generally. In this paper, we reviewed three categories of DTI registration methods that are classified on the basis of the processing data, including the scalar image based registration algorithms, the tensor image based registration algorithms and the fiber bundle based registration algorithms. Since there is no unified standard framework for registration evaluation, we also retrospect several evaluation methods, including visual assessment, fiber based evaluation and region based evaluation criteria proposed in recent years. We also mentioned some other common image registration evaluations briefly. In the end, we summarized and analyzed the characteristics of different registration methods. In addition, we prospected the development of DTI registration in the future.