Accurate Diffusion Magnetic Resonance Imaging And Brain Connectome Analysis Algorithms

Since the 20 th century,people have been hoping to unlock the secrets of functional connectivity pattern in the human brain.Diffusion magnetic resonance imaging(dMRI)based tractography is the only technology for non-invasively investigating the fiber pathway,and is an important.With the development of non-invasive neuroimaging technology,it is possible to construct a human brain with a fully connected network connection.Brain neurofibrillary remodeling based on diffusion magnetic resonance imaging(dMRI)is the only method for non-invasive display of nerve fiber orientation in vivo and is an important approach to construct the connection network in the brain.It plays an important role in the research of brain surgery navigation and disease analysis.The basic idea is to first estimate the fiber orientation distribution from the dMRI signal per voxel,then find the optimal path connection these orientations,and finally identify the fiber tracts with anatomically meaningful.These fiber tracts will be useful for the study on the anatomical connection in the brain.In fact,there are three key issues of tractography,estimation of fiber orientation distribution,optimization of the fiber pathway and identification of fiber tract.These issues can be attributed to the challenges in complex optimization modeling and algorithm design.With the improvement of high-resolution MRI scanner,the sampling accuracy of the diffusion signal has been greatly improved,which provides more accurate dMRI data for tractography imaging and also leads the highly complex as the tractography.Obviously,existing tractography imaging technologies have been difficult to adjust to the requirements of highresolution diffusion imaging.In this thesis,a series of studies on the shortcomings of existing methods in estimation of fiber orientation distribution,optimization of fiber pathway and identification of fiber tract are carried out.The main works are as follows:1)Aiming at the noise influence with existing methods under low angle resolution diffusion signal,the method based on sparse imaging is proposed to estimate the fiber orientation distribution with high-angle resolution.This method introduces the prior knowledge of the sparsity distribution along the fiber orientation voxel-wise and considers the white matter structure relationship between the neighborhood voxels to establish a spatial regularization optimization model in the sparsity space.The experiments using simulation and in-vivo human brain data show that the proposed method can accurately reconstruct the high-resolution microstructure from the low-angle resolution dMRI data,then improving the robustness of estimation of the complex fiber structure.2)To solve the problem of gyral bias in tractography,the method in estimation of multitissue asymmetric fiber orientation distribution and reconstruction of global geodesic fiber pathway are proposed respectively.The multi-tissue asymmetric fiber orientation distribution is globally modeled with adaptively neighborhood information to construct fiber continuity regularization.This allows the fiber streamline to enter the cortical region to mitigate the gyral bias.The global geodesic tractography aims to construct the fiber pathway model from a fiber orientation field,allows overcomes the one-way approximation tracking as the traditional method,thereby mitigating the gyral bias due to the low-resolution diffusion signal.The experiments using in-vivo human brain data show that the proposed method effectively mitigate the gyral bias.3)Aiming at the problem of automatic tractography-based parcellation in clinical analysis,a group-wise tractography-based parcellation was proposed to automatic classification of whole brain tractography.In clinical analysis,it is important to investigate the group comparison of tractof-interest.However,the traditional methods depend on manual labeling to segment fiber tract from whole brain tractography.This method is inefficient and unstable.In this work,the fiber tracts were automatically identified,this is based on the similarity and specificity between the fiber clusters,and the anatomical features of each fiber cluster connected between two same regions.Thereby each fiber with the anatomical meaning is automatically identified.The experimental results show that the method can automatically detect with high-consistency and be able to apply to investigate into local white matter abnormality in clinical studies.