|
|
A Bayesian Constraint Stochastic Framework for DT-MRI White Matter Fiber Tractography |
Wu Xi①② Zhou Ji-liu③ Xie Ming-yuan① Luo Dai-sheng② |
①(Department of Electronic Engineering, Chengdu University of Information Technology, Chengdu 610225, China)
②(School of Electronics and Information Engineering, Sichuan University, Chengdu 610065, China)
③(School of Computer Science, Sichuan University, Chengdu 610065, China) |
|
|
Abstract Diffusion Tensor Magnetic Resonance Imaging (DT-MRI) can track the brain white matter fiber by tracing the local tensor orientation and reconstruct the three dimensional image noninvasively. The commonly used tracking method is usually based on the local diffusion information and insufficient to consider the geometrical structure and fractional anisotropy which is constrained by anatomical structure and physiological function of human been. Therefore, a novel method of fiber tracking based on Bayesian constrained stochastic framework is proposed. In this method, the correlation of tracking direction to both the diffusion directions of the current voxel and the structure information of the current fiber segment is considered synthetically. Meanwhile, the two components are constrained by the fractional anisotropy and angle of the fiber curve respectively. The probability distributions of the tracking directions of the next voxel is estimated under the Bayesian constrained stochastic framework. Then, according to the probability distributions, the fiber bundle is sampled with Markov Chain Monte Carlo method and the 3D image of its structure is reconstructed under multiply tracking. By the method, imaging simulations using a synthetic diffusion tensor dataset and imaging experiments using an in vivo brain DT-MRI dataset have been done. The results of the simulations and experiments demonstrate that using the method proposed, brain white matter fiber can be reconstructed properly as expected, more reliably and reproducibly compared with the common methods.
|
Received: 25 August 2009
|
|
Corresponding Authors:
Wu Xi
E-mail: wuxi@scu.edu.cn
|
|
|
|
|
|
|