Continuous Spatial Neural Fiber Tracking Algorithm Based On Flow Field Distribution

Diffusion magnetic resonance fiber tracking imaging is currently the only non-invasive nerve fiber imaging method,which provides a new tool for humans to understand the nerve fiber structure of living brains.Researchers have proposed many tracking imaging algorithms,but they still face two problems in clinical application.On the one hand,how to choose a tracking imaging algorithm and what are the differences between different algorithms,a quantitative analysis of various algorithms is required.On the other hand,existing algorithms reconstruct a large number of "false" fibers,limiting the accuracy of clinical applications.To this end,based on quantitative analysis of existing algorithms,this thesis proposes a new continuous fiber tracking method based on flow field distribution.The main work is as follows:1.Quantitative comparative analysis of current typical fiber tracking imaging algorithms.This part mainly focuses on the following 9 typical tracking imaging algorithms: tensor based streamline algorithm,spherical deconvolution based streamline algorithm,continuous fiber contact tracking,fiber direction distribution tracking,particle filter tracking,traceless Kalman filter tracking,Gibbs tracking,anatomical constraint tracking,and machine learning tracking algorithms.Quantitative comparative analysis was conducted through simulated data,and the imaging results and challenges of these algorithms in clinical data were analyzed.Fiber tracking imaging is of great research value and application value for analyzing nerve fiber connections in human brain.Different types of algorithms have their own advantages and disadvantages.At present,there is no one tracking algorithm that can abandon the disadvantages of other algorithms and combine all the advantages.In addition,there is a certain gap between the results of the current fiber tracking algorithm and the actual situation.How to describe a more accurate fiber trajectory is still a challenging problem.2.The flow field distribution model of fiber direction is reconstructed based on streamline differential equations.The diffusion motion of water molecules causes the change of the magnetic resonance signal.By modeling the magnetic resonance signal,the diffusion motion of water molecules is described to indirectly describe the direction distribution of the fiber.In this thesis,the diffusion motion of water molecules is described as fluid motion,and a continuous flow field distribution model of fiber direction in three-dimensional space is established.The 26-voxel information of the central voxel neighborhood is used to solve the flow field distribution model,and the fiber consistency in the voxel and the fiber consistency between the voxels are used to constrain the solution function to obtain a more accurate flow field distribution model,with the flow field distribution representing the fiber's direction distribution.The model has the following characteristics: field distribution is an asymmetric distribution model,which is different from the case where the traditional fiber direction distribution model is mostly center-symmetric;field distribution is a spatially continuous direction distribution,while the traditional fiber direction distribution is generally a discrete representation about the distribution in the center direction.3.Aiming at the problem that the traditional fiber tracking imaging algorithm ends prematurely when the fiber tracking reaches the boundary and cannot reach the correct area,a continuous spatial nerve fiber tracking algorithm is proposed based on the flow field distribution model in the fiber direction.After obtaining the flow field distribution model in the fiber direction,the fiber direction distribution of each voxel is integrated to reconstruct the complete fiber connection trajectory.Since integration is not easy to solve,tracking and iterating the fiber trajectory through the fourth-order Runge-Kutta numerical integration method reduces the complexity of the algorithm.When the fiber tracking iteration encounters a boundary(white matter area boundary)or a voxel with missing direction information,the traditional tracking algorithm will terminate the iteration step to make the fiber trajectory incomplete.In this thesis,the spatial continuity of the breakpoint is calculated by according to the given threshold,keep the fiber trajectory and continue to follow the iteration to obtain a more complete fiber trajectory.The spatial continuity of the breakpoint is obtained by the difference integral of the three adjacent voxel flow field distribution models where the fiber trajectory is located.The effectiveness of the algorithm is evaluated by comparing the simulation data and clinical actual data with typical tracking algorithms.The results show that the proposed method can obtain more accurate fiber trajectory reconstruction results.