Clustering Algorithm Of White Matter Fiber Based On Neighborhood Asymmetric Structure Tractography

The white matter imaging technology based on diffusion magnetic resonance imaging is the basis for studying and analyzing the structure of white matter,and is currently a hotspot in brain science.Traditional tractography and analysis methods have been difficult to meet the accuracy of white matter fiber reconstruction and clinical application requirements.On the one hand,the existing high-angle-resolution imaging scheme has improved the recognition ability of complex fiber structures,but it only considers the fiber structure information of single voxel,neglects the local fiber structure,and there exist problems such as low accuracy of fiber reconstruction and poor noise immunity.At the same time,the distribution of fiber direction within the voxel is an asymmetrical structure.For tractography,different fiber structures(such as cross,divergent or bending)cannot be accurately described.On the other hand,the analysis of tractography usually divides the whole brain fiber into a series of continuous fiber bundles with a certain structure and shape.This technology currently stays in fiber macroscopic analysis and it is difficult to combine anatomical information.Due to the lack of information on the edge structure of fiber bundles and the high degree of variability of complex white matter structures in individual samples,fiber clustering techniques based on anatomical information are still an open topic.In response to the above issues,the main tasks of the study are as follows:1.The tractography algorithm utilizes neighbor voxel fiber direction information to establish a spatially continuous field to describe the structure of the fiber in tracking process.Then the fiber streamline is produced by the numerical integration method to obtain fiber bundle,for reducing the inaccuracy of the direction estimation caused by the noise of the original image.2.A new fiber clustering technique is proposed,which combines spatial features of whole-brain fibers,prior anatomical information,features of nerve fiber communication pathways as fiber similarity matching and feature extraction.This work matches the anatomical features with a highly consistent fiber bundle coverage in the white matter structure.The method is based on multiple tests of simulated data and actual human brain data.The tractography algorithm based on neighborhood information is tested on simulated data and in vivo data,and is compared by other existing algorithms.The experimental results show that the method has a great improvement in noise immunity and fiber continuity.The experimental of fiber clustering results show that this method not only improves the highly consistent coverage of fiber bundles and anatomical prior knowledge,but also simplifies the fiber data space to improve the fiber clustering.