| Brain white matter tractography is an important application of revealing the complex neural connectivity structure in the human brain. Diffusion MR imaging enables the measurement of water diffusion in vivo, and Diffusion Tensor Imaging technique makes it possible to detect more detailed diffusion information within the brain white matter. Since the neural fiber orientation is represented by water diffusion properties, but not direct observed, an inverse problem of computing fiber structure from the diffusion information should be solved. In regard to neural fiber tracking, two kinds of models are widely used, either deterministic or probabilistic. The former one has its advantages of easy implementation and low computation requirement, but suffers from the problem of noise accumulation and early termination. The later one presents more accurate fiber tracking results by using probability density functions to reflect the uncertainty within diffusion tensor, but has the problem of computation burdensome. In this thesis, a combined tracking method was explored that contains the properties of both deterministic and probabilistic model. In tensor estimation process, a log-Euclidean metrics method was applied. |
