中文摘要：

为了解决脑白质纤维交叉分叉问题，在传统算法的启发下，提出一种基于相邻体素选择的盘状张量分解算法。首先，选择合适的起点进行非分叉纤维的追踪，建立拟合函数数据集，得到拟合函数；其次，在纤维追踪出现交叉分叉问题时，建立该交叉分叉点及周围区域体素所对应的棋盘图；然后，计算以交叉点为中心的相邻张量的夹角，结合夹角的大小并利用得到的拟合函数进行纤维整体走行方向的估计，实现盘状张量的分解。算法既保证了局部信息的合理适用，又考虑了整体信息的影响，能够更加精确完整地跟踪纤维路径，解决纤维交叉分叉问题。与传统方法相比．，该算法可以更有效地解决纤维分又及交叉处的跟踪问题，从而使得到的纤维路径更加真实。

英文摘要：

In order to solve the problem of white matter fiber crossing and bifurcating, inspired by the traditional algorithm, this paper proposed the solving method that based on a neighboring voxel selection for the decomposition of disk tensor. First- ly, the algorithm chose the right point to fiber track and established data set to obtain fitting function. Secondly, when the problem of white matter crossing and fiber bifurcating appeared, it would draw the checkerboard chart for neighboring voxels of the point, calculated the angle between the point and other neighboring voxels. In combination with the angle, it used fitting function according to the data set of fitting function to estimate global fiber direction, thus finished the decomposition of the disk tensor. The method not only used the local information reasonably but also considered the global information effecting. And it can finish the fiber tracking more accurately and completely. Compared with the traditional algorithm, the method is a- ble to solve white matter crossing and fiber bifurcating more effectively, so the fiber tracking is closer to the truth.