Brain Fiber Tracking Algorithm With Flow Line Differential Equations Based On Neighborhood Dictionary Basis Model

The brain white matter fiber imaging technologies based on diffusion magnetic resonance imaging is currently the only non-invasive way to show the direction of nerve fibers in vivo.Among them,the fiber micro-structure reconstruction methods and fiber tracking technologies enable people to study the brain structure from microscopic and macroscopic levels,playing an important role in brain neurosurgery navigation,psychiatric diagnosis and brain network construction.However,as the traditional fiber reconstruction models only consider the information of a single voxel without the integrity of fiber structure during the reconstruction procedure,the accuracy and noise immunity of them are poor.On the other hand,the traditional deterministic and probabilistic fiber tracking methods are over-reliant on the distribution of the fiber directions in local voxels,and are subject to some uncertainties in the tracking process.In global tracking methods,as the model and the solution process are too complicated,the time and space complexity is far from being able to meet the needs of practical use.In this work,the above problems were studied separately,and new methods were proposed to improve the results.Specific work as follows:1)Based on the framework of the dictionary basis,a neighborhood dictionary basis based fiber reconstruction model was proposed.By introducing the fiber information of the adjacent voxels,each voxel will be constrained by the correlation of the fiber distribution and signals between neighbour voxels,thus a model with regional structure constraints can be established.Meanwhile,a sparse dictionary was adaptively learned from neighborhood and built for model reconstruction,help to reduce the computational complexity of the solution and improve the computational efficiency.The experimental results on simulated data and clinical data showed that the fiber micro-structure reconstruction model proposed in this work can maintain the integrity and smoothness of the fiber well,and has high reconstruction precision and stability.2)A brain fiber tracking algorithm based on the streamline differential equations was proposed.The technique searches for the directions of the fibers which are close to the current travel orientation among the neighborhood voxels in the tracking process,and a continuous flow field in a three-dimensional space is then be used to characterize the distribution of the fiber streamlines and reduce the influence of the direction estimation error caused by inaccurate fiber reconstruction.By solving the differential equation of the stream line with runge-kutta numerical integration method,uniform and smooth fiber bundles in an arbitrary space region can be obtained.The proposed method was compared with deterministic,probabilistic and global tracking methods on simulated data and real human brain data.The results showed that the proposed method can obtain more accurate results of white matter fiber reconstruction.