Using the image data of Diffusion tensor magnetic resonance imaging(DT-MRI,DTI) can implement the tracking and visual display of nerve fiber bundles in brain,and can provide effective support for something as the diagnosis of related disease,surgical consultation and so on.The main research contents of the fiber tracing based on DTI image data containsthe calculation of diffusion tensor matrix, the extraction of diffusion anisotropicparameters, the visualization of diffusion tensor matrix and the tracing algorithm ofnerve fiber bundles. Firstly, we can calculate the diffusion tensor matrix through theoriginal DTI image. The diffusion tensor matrix is a3*3symmetrical positivedetermined matrix, the three eigenvalues and eigenvectors of which represent thedirections and magnitude of the diffusive motion of hydrone in the organism structure.Through the eigenvalues, we can calculate some diffusion anisotropic parameters,which can be used as rules in the tracking process to judge if one voxel is on nervefiber bundles. Secondly, visualize the diffusion matrix, which is used to show theanisotropy information more intuitively, as well as the relationship of theneighborhood voxels. There are many methods of visualization, such as color coding,glyph, white matter tractography, volume rendering and so on. Finally, we implementthe fiber tracing through the tensor matrix. Currently, the mainstream algorithm is allbased on the tensor field. Through the information of diffusion tensor matrix, estimatethe trend of fiber. This paper present a new fiber tracing algorithm based on the tensorline algorithm. The improved position in the new algorithm is as follows: select thedense fibrous connective tissue as the seed point, such as the white matter; add newrules to select a better fiber point, such as neighborhood filter and backward tracing; add a interpolation step to get the fiber point more accurately; use curve-fitting to letthe tracing fiber stretch more smoothly.Researches shows that the tracing fiber bundles using the improved algorithm isless error rate, more smooth, and more in line with the actual situation. |