Research On White Matter Fiber Tracking In Diffusion Tensor Imaging

Diffusion Tensor Imaging (DTI), imaging with diffusing information of watermolecules in the brain, is a new Magnetic Resonance Imaging technique based onDiffusion Weighted Imaging（DWI）.White matter tractography based on DTI is the uniquemethod that can non-invasively reconstruct nerve fibers in vivo. It’s important toreconstruct the fibers accurately and rapidly. It’s used for learning the mechanism of someclinical diseases, providing the reliable datum for scheme selection of brain surgery andsurgical navigation. It also provides feasible methods for learning brain function, analyzingcognitive function, and revealing cranial never conduction mechanism. The main task ofthis thesis is to research fiber tracking algorithm by diffusion information of watermolecules.Firstly, this paper summarizes line propagation algorithms which are commonly usedfor tracking fibers through a region of interest （ROI） and discusses their shortcomings.Variable-step algorithm is proposed on account of multi-resolution concept. This methodtakes fractional anisotropy （FA） of voxel as step because of the diversity of differentvoxel’s anisotropic degree, and calculates the new diffusing matrix with a tri-linearinterpolation algorithm to get diffusing information for fiber tracking. More details offibers are reserved in this method, and longer fibers can be tracked.Secondly, this paper summarizes global energy minimization algorithms which arecommonly adopted to track fibers between two given ROIs. Genetic and simulatedannealing algorithm （GA-SA） for fiber tracking is proposed in this thesis, overcomingthe drawbacks of traditional genetic algorithm（GA）,such as prematurity and convergenceto the local optimum solution. Simulated annealing algorithm（SA） is introduced forbreeding operator of GA, and new fiber paths are generated by mutation and cross operator in this method. The results prove that it overcomes the shortcomings of traditionalalgorithm, and shows the fibers with smaller average energy and more in line with thedistribution of tensor field.Finally, according to speediness of line propagation algorithms and globaloptimization of global energy minimization, ant group behavior and these advantages arecombined in trial. Two-way ant colony algorithm for fiber tracking is presented in thisthesis. The results show that it is of high time efficiency. However, it makes the fiber pathnot smooth enough, because of ants shifting voxel by voxel. Suggestions and improvementmeasures are provided for subsequent work.