Registration Algorithm Based On Diffusion Tensor Images

With the development of science and technology, more and more image modeis used to assist in clinical research. Diffusion tensor imaging is emerging as afunctional imaging method, being widely used as a psychiatric disease research anddiagnostics adjunct. Especially, it was irreplaceable in describing potential whitematter lesions. The technology is an in vivo method of detecting water moleculesdiffuse movement. There are a variety of factors may cause the data collectedinformation with the actual situation of human errors in the imaging process. So theregistration of clinical images is important before being used.The general scalar image registration process requires a deformation model, aninterpolation scheme, a similarity measure and a regularization of the program. DTI isone type of functional imaging technology, which data contains the info that couldpresent the trends of body growing. Tensor is rotation invariant, so it also need to beredirected to improve the registration process, to insure the registration results withthe original structure information consistent.Based on the characteristics of DTI images, we study the DTI imaging principlefirst, and then analyze the important parameters in the post-processing of DTI data,and briefly describe the current mainstream DTI registration algorithm. Finally, weproposed some new methods for the deficiencies of the existing algorithms of DTIregistration methods. In this paper, the eigenvalues and eigenvectors of thecoincidence rate as the basis of the results of the evaluation of the registration andcomprehensive analysis of the proposed algorithm robustness compare existingalgorithms and the efficiency of the proposed algorithm and match results, to studythe clinical applications of DTI image registration method.In this paper, the human brain DTI images as the experimental data, simulationexperiments in MATLAB experimental platform. DTI imaging principle, registrationprograms, as well as the results of the registration judge did a thorough investigationand study, on the basis of the existing registration method is proposed to improvethe new algorithms and achieve judge from the results of theoretical and experimental analysis. The main work of this paper can be summarized in thefollowing points:1) proposed scalar image vectorization method to achieve roughregistration of the direction of the images, aims to narrow the search range of thesubsequent registration algorithm to improve the efficiency of the registration;2)surrounding areas of information combined with a single pixel similarity measure,designed to reduce the miss alignment of a single pixel rate;3) multi-featureextraction proposed a tensor auxiliary judgments match point registration algorithm,based on the characteristics of tensor analysis be the approximate location of theregistration point to narrow the scope of the registration and reduce the rate of missalignment;4) the flexibility to select adaptive algorithm based on sparse expressiontheory of marked point, the DTI image registration with TPS deformation correctionalgorithm, DTI images registration. Finally, we select the experiment and analyze theresults of the comparison registration the multilayer DTI data to give an objectiveevaluation of the proposed algorithm.