Abstract: The fiber tracking algorithm can track the brain fibers through the fiber orientation distribution function. Considering that the dispersion of water molecules is mutual and the fiber orientation distribution function reconstructed from scanning data may have errors, this paper proposes a fiber tracking algorithm optimized by direction correction based on the traditional streamline tracking algorithm combined with the maximum cosine similarity. At the same time, considering the existence of anisotropically dispersed and isotropically dispersed water molecules in the human brain, and that the latter accounts for a larger proportion, the maximum expectation algorithm is used to cluster the seed points with the same properties to reduce the tracking of isotropically dispersed voxel points. Finally, the experiments were conducted using simulated and real data respectively, and the results show that the proposed algorithm takes less time for tracking, the average fiber length is longer compared to the traditional STT tracking algorithm, the number of incorrectly tracked clusters is significantly less than that of the traditional fiber tracking algorithm, the ratio of correctly tracked bundles is significantly higher than that of the traditional fiber tracking algorithm, and there are higher overlap and lower overestimation rates for the tracking of most of the specific bundles, which better reflects the results of the traditional STT tracking algorithm. It also has a higher overlap rate and a lower overestimation rate in most specific fiber bundles, which is more representative of the structural distribution of fibers in the actual situation.

Key words:

"> fiber tracking, maximum expectation clustering, moving least squares, streamline tracking algorithm