White matter microstructure calculation system and method based on multi-shell diffusion weighted imaging

Abstract

The invention discloses a white matter microstructure calculation system and method based on multi-shell diffusion-weighted imaging, and the system comprises a data collection module which is used for obtaining diffusion-weighted imaging and structural image magnetic resonance scanning of a testee; the first data processing module is used for obtaining individual fiber orientation distribution and generating a public template of group-specific fiber orientation distribution at the same time; the second data processing module is used for co-registering the individual fiber orientation distribution diagrams to a common template and segmenting the fiber orientation distribution to obtain white matter microstructure indexes of a whole brain level; the third data processing module is used for obtaining a white matter microstructure index of each anatomical fiber bundle level; and the fourth data processing module is used for obtaining a whole brain fiber streamline based on probabilistic fiber bundle tracking and spherical deconvolution information filtering, and generating a white matter microstructure index of a network level in combination with a prior brain map. By adopting the technical scheme of the invention, quantitative calculation and analysis of the white matter microstructure change of the brain are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of neuroimaging data processing, and particularly relates to a white matter microstructure calculation system and method based on multi-shell diffusion weighted imaging. Background Art

[0002] Current research shows that neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, etc. usually involve irreversible and progressive destruction of the brain white matter microstructure and disconnection of brain structural connections. By describing the restricted diffusion of water molecules in nerve fibers, diffusion magnetic resonance imaging (dMRI) greatly improves our understanding of the white matter microstructure changes in neurodegenerative diseases with its objectivity and non-invasiveness advantages. The quantified neuroimaging diffusion characteristics can be comprehensively judged and analyzed by clinicians and researchers, providing an important theoretical basis for abnormal changes in the degenerative central nerve mechanism. Research shows that in patients with Alzheimer's disease and its preclinical mild cognitive impairment stage based on traditional diffusion tensor imaging, white matter degeneration and disruption of the structural connectome occur in the brain, specifically manifested as a decrease in fractional anisotropy and an increase in mean diffusivity.

[0003] Diffusion weighted imaging (DWI) is a dMRI technique that shows the diffusion rate of water molecules in different directions. Diffusion tensor imaging uses a symmetric positive definite tensor field derived from DWI and uses an ellipsoidal tensor model to characterize the anisotropy of water molecule diffusion in brain tissue, thereby reflecting the organization and structure of white matter fibers. It is the most commonly used dMRI quantification method. However, nearly 98% of the brain white matter consists of multiple fibers with cross orientations, which makes it limited for the diffusion tensor model to characterize white matter abnormalities. A research report shows an abnormal increase in fractional anisotropy in the cross-fiber region of the centrum semiovale in Alzheimer's disease, possibly due to the limitation of the tensor model in fitting multiple fibers in different directions within a single voxel. Considering the complexity of brain white matter anatomy, it is very important to quantitatively calculate the fine-grained white matter microstructure, especially in the cross-fiber region, to understand the white matter degeneration related to neurodegeneration. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a white matter microstructure calculation system and method based on multi-shell diffusion weighted imaging to achieve quantitative calculation and analysis of changes in the brain white matter microstructure.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A white matter microstructure calculation system based on multi-shell diffusion weighted imaging, comprising:

[0007] A data acquisition module, which is used to acquire the diffusion-weighted imaging and structural magnetic resonance scans of the subjects and preprocess the original brain images;

[0008] A first data processing module, which is used to obtain the individual fiber orientation distribution based on the multi-shell multi-tissue constrained spherical deconvolution algorithm, and simultaneously generate a common template of the fiber orientation distribution specific to the population;

[0009] A second data processing module, which is used to co-register the individual fiber orientation distribution map to the common template, and segment the fiber orientation distribution to obtain the white matter microstructure index at the whole-brain level;

[0010] A third data processing module, which is used to segment the anatomical fiber bundle mask according to the fiber orientation distribution to obtain the white matter microstructure index at the level of each anatomical fiber bundle;

[0011] A fourth data processing module, which is used to obtain the whole-brain fiber streamlines based on probabilistic fiber tractography and spherical deconvolution information filtering, and generate the white matter microstructure index at the network level by combining the prior brain atlas.

[0012] Preferably, the magnetic resonance scan of the data acquisition module includes multi-shell diffusion-weighted imaging, that is, in addition to the reference image with b = 0, the diffusion-weighted imaging with two or more b values, and the high-resolution structural image of T1-weighted imaging.

[0013] Preferably, the data acquisition module uses MRtrix3 software to perform noise removal, Gibbs ring artifact removal, distortion correction caused by eddy currents, head motion correction, and bias field correction on the multi-shell diffusion-weighted imaging, and linearly register the average b0 image to the high-resolution T1 structural image; uses FreeSurfer software to perform recon-all processing on the T1 structural image to obtain five tissue types.

[0014] Preferably, the first data processing module derives and estimates the response function of the spherical deconvolution model based on the co-registered five-tissue type map to obtain the response functions of gray matter, white matter, and cerebrospinal fluid in the individual space; calculates the fiber orientation distribution at the voxel level based on the multi-shell multi-tissue constrained spherical deconvolution algorithm.

[0015] Preferably, the first data processing module performs joint bias field correction and intensity normalization on the obtained fiber orientation distribution map in the individual space, and uses the fiber orientation distribution maps of some subjects to generate a common template of the fiber orientation distribution specific to the research population.

[0016] Preferably, the second data processing module divides the fiber orientation distribution in voxel units into single fiber bundle population units representing different direction fiber populations, and simultaneously calculates and obtains three white matter microstructure indexes at the whole-brain level, namely fiber density, fiber cross-section, and fiber density and cross-section combining the above two indexes.

[0017] Preferably, the third data processing module uses the TractSeg tool to divide the common fiber orientation distribution map into 72 prior anatomical fiber bundles, and then extracts the average values of three white matter microstructure indexes of the fiber bundles.

[0018] Preferably, the fourth data processing module obtains whole-brain fiber tract streamlines based on the probabilistic fiber tract tracking algorithm with anatomical constraints of the fiber orientation distribution map represented by spherical harmonic basis, and uses spherical deconvolution information filtering to optimize the quality of the fiber tract streamlines; meanwhile, the white matter structural connections between brain regions are calculated and defined as the total number of fiber tract streamlines passing through the volume-normalized brain regions.

[0019] The present invention also provides a method for calculating white matter microstructure based on multi-shell diffusion weighted imaging, including:

[0020] Step S1: Obtain diffusion weighted imaging and structural magnetic resonance scans of the subject, and preprocess the original brain images;

[0021] Step S2: Obtain individual fiber orientation distributions based on the multi-shell multi-tissue constrained spherical deconvolution algorithm, and simultaneously generate a common template of the fiber orientation distributions specific to the population;

[0022] Step S3: Co-register the individual fiber orientation distribution maps to the common template, and segment the fiber orientation distributions to obtain white matter microstructure indexes at the whole-brain level;

[0023] Step S4: Divide the anatomical fiber bundle masks according to the fiber orientation distributions to obtain white matter microstructure indexes at the level of each anatomical fiber bundle;

[0024] Step S5: Obtain whole-brain fiber streamlines based on probabilistic fiber tract tracking and spherical deconvolution information filtering, and generate white matter microstructure indexes at the network level in combination with a priori brain atlases.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] (1) For multi-shell diffusion weighted imaging data, the present invention calculates the fiber orientation distribution of each voxel through a constrained spherical deconvolution model, and further divides it into characteristic indexes for describing the white matter microstructure situation of all fiber populations within a single voxel with direction information, overcoming the disadvantage of poor quantification calculation effect in the cross-fiber region of the traditional tensor model, and providing a novel technology for fine quantification of the brain white matter microstructure.

[0027] (2) To obtain more accurate index estimation, the present invention uses multiple optimization algorithms for calculating white matter microstructure indexes, including: estimating fiber orientation distribution based on the multi-shell multi-tissue algorithm, where multiple b-values provide more brain diffusion characteristics to optimize the accurate construction of the spherical harmonic model; using various brain tissues segmented from high-resolution structural images to provide anatomical structure information, providing a better starting point selection for fiber orientation distribution estimation and fiber tract tracing with anatomical constraints; using spherical deconvolution information filtering to optimize fiber tract tracing reconstruction, which can effectively reduce the appearance of false positive fiber streamlines.

[0028] (3) The present invention calculates and obtains white matter microstructure fiber-specific quantitative indexes at the whole-brain level, fiber bundle level, and network level. By calculating the results of indexes with the same physiological significance at different levels, the potential white matter changes at different scales are clarified. A single anatomical fiber bundle is represented by many voxels, and the structural connections between different cerebral cortices are composed of different secondary fiber bundles. The connection between the calculation of white matter microstructure indexes at the whole-brain level, fiber level, and network level can help in the discussion of the mechanism of the hierarchical organization of the brain white matter. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following-described drawings are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0030] Figure 1 It is a schematic structural diagram of the white matter microstructure calculation system based on multi-shell diffusion weighted imaging of the present invention;

[0031] Figure 2 It is a schematic flow diagram of the white matter microstructure calculation method based on multi-shell diffusion weighted imaging of the present invention;

[0032] Figure 3 It is a schematic diagram of the data processing process of the white matter microstructure calculation method based on multi-shell diffusion weighted imaging of the present invention and the obtained multi-level white matter microstructure indexes;

[0033] Figure 4 It is a schematic diagram of the between-group comparison at the anatomical fiber bundle level, the correlation with cognitive function, and the mediation analysis of multi-level white matter microstructure indexes and cognitive function obtained by the white matter microstructure calculation method based on multi-shell diffusion weighted imaging of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0035] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0036] Embodiment 1:

[0037] As Figure 1 shown, an embodiment of the present invention provides a white matter microstructure calculation system based on multi-shell diffusion weighted imaging, including:

[0038] A data acquisition module for acquiring diffusion weighted imaging and structural magnetic resonance scans of a subject and preprocessing the original brain images;

[0039] A first data processing module for obtaining an individual fiber orientation distribution based on the multi-shell multi-tissue constrained spherical deconvolution algorithm and simultaneously generating a common template of the population-specific fiber orientation distribution;

[0040] A second data processing module for co-registering the individual fiber orientation distribution map to the common template and segmenting the fiber orientation distribution to obtain white matter microstructure indicators at the whole-brain level;

[0041] A third data processing module for obtaining an anatomical fiber bundle mask by segmenting according to the fiber orientation distribution and obtaining white matter microstructure indicators at the level of each anatomical fiber bundle;

[0042] A fourth data processing module for obtaining whole-brain fiber streamlines based on probabilistic fiber tractography and spherical deconvolution information filtering and generating white matter microstructure indicators at the network level in combination with a priori brain atlases.

[0043] As an implementation manner of the embodiment of the present invention, in the data acquisition module, the magnetic resonance scan includes multi-shell diffusion weighted imaging, that is, in addition to the reference image with b = 0, diffusion weighted imaging including two or more b values, and a high-resolution structural image of T1 weighted imaging for registration with the multi-shell diffusion weighted imaging and the MNI standard brain space to solve the problem of poor direct registration effect caused by the low spatial resolution of the diffusion weighted imaging.

[0044] Furthermore, the MRtrix3 software is used for the preprocessing steps of the original brain images, specifically including: removing the noise of the original images, estimating and denoising the diffusion MRI noise level based on the eigenvalue spectrum of the stochastic covariance matrix described by the universal Marchenko-Pastur distribution; removing Gibbs ring artifacts, and using the local sub-voxel displacement method to remove Gibbs ringing artifacts from the MRI images; correcting the distortion caused by eddy currents and head motion correction, by simulating the influence of the Gaussian process of diffusion eddy currents and head motion on the images, for predicting the given brain images defined by the b-value and diffusion gradient direction; bias field correction, based on the N4 algorithm of the ANTs software, by estimating the bias field from the b0 image and applying it to correct all diffusion-weighted volumes. Obtain the average value of all b0 images and linearly register them to the high-resolution T1 structural image, and obtain the transformation matrix for converting the diffusion space to the individual space, and apply it to all diffusion-weighted imaging to achieve accurate registration of the diffusion space to the individual space; use the FreeSurfer software to perform recon-all processing on the T1 structural image, and generate five tissue types of gray matter, white matter, cerebrospinal fluid, subcutaneous nuclei, and pathological tissues for fiber tract tracing and multi-tissue modeling with anatomical constraints using the cortical information of the preprocessed high-resolution structural image, and further generate a mask image suitable for the starting point of the streamline at the gray matter-white matter interface.

[0045] As an implementation manner of the embodiment of the present invention, in the first data processing module, the response functions of the cerebral gray matter, white matter, and cerebrospinal fluid are estimated by the multi-shell five-tissue algorithm. First, the response functions of the spherical deconvolution model are derived and estimated based on the co-registered five-tissue type map to obtain the response functions of gray matter, white matter, and cerebrospinal fluid in the individual space; further, the fiber orientation distribution at the voxel level is calculated based on the multi-shell multi-tissue constrained spherical deconvolution algorithm. The algorithm processing is carried out based on Spherical Harmonics, which provides a smooth representation for the data distributed on the sphere, and the coefficients of the generated spherical harmonics are saved in the form of a fiber orientation distribution image.

[0046] Spherical harmonics are special functions defined on the sphere, forming a complete orthonormal set, and are equivalent to the Fourier series of functions defined in spherical coordinates in many cases, defined as:

[0047]

[0048] where l and m represent the integer order and phase respectively, l≥0 and -l≤m≤l, involving associated Legendre polynomials, the harmonic order l corresponds to the angular frequency of the basis function, and the harmonic phase m corresponds to different orthogonal modes at this frequency.

[0049] Any well-behaved function on the sphere can be represented as a spherical harmonic expansion:

[0050]

[0051] For a smooth function with negligible high angular frequency components, the series can be truncated at an appropriate maximum harmonic order l max and approximated as:

[0052]

[0053] Since the diffusion MRI design data is real and the phase information is usually discarded due to its instability to motion, a real basis without imaginary parts can be used. In addition, the content involved has antipodal symmetry, so all odd-order terms in the series can be ignored.

[0054] The formula used in the embodiment of the present invention is as follows:

[0055]

[0056] Based on spherical harmonics, the fiber orientation distribution coefficient vector is further obtained by the constrained spherical deconvolution method, expressed as a constrained linear least squares problem, which is expressed as:

[0057]

[0058] where x is the unknown fiber orientation distribution function coefficient vector, d is the diffusion-weighted signal intensity vector measured on a single shell in q-space, C is the matrix that relates the fiber orientation distribution function coefficients to the diffusion-weighted signal through spherical convolution, and A is the matrix that relates the fiber orientation distribution function coefficients to the amplitude, and its role is to enforce non-negativity on the domain of the fiber orientation distribution function. This method is called the single-shell single-tissue constrained spherical deconvolution model.

[0059] Furthermore, it is extended to the constrained spherical deconvolution model for multi-shell diffusion-weighted imaging data to support diffusion-weighted imaging with m shells, i.e., m b-values:

[0060]

[0061] where d i is the diffusion-weighted signal intensity vector measured on the i-th shell in q-space; C i is the matrix that relates the fiber orientation distribution function coefficients to the diffusion-weighted signal intensity measured on the i-th shell through spherical convolution.

[0062] Furthermore, it is extended to accommodate n types of brain tissues such as gray matter, white matter, and cerebrospinal fluid:

[0063]

[0064] where x jis the coefficient vector of the fiber orientation distribution function of tissue j, C i,j is the matrix that relates the fiber orientation distribution function coefficients of tissue j to the diffusion-weighted signal intensity measured on the i-th shell by spherical convolution, A j is a tissue-specific matrix related to the amplitude of the fiber orientation distribution function coefficients of tissue j, used to enforce each fiber orientation distribution function to be non-negative.

[0065] Perform joint bias field correction and intensity normalization on the obtained fiber orientation distribution maps in the individual space, further correct the residual intensity inhomogeneity, and improve the comparability of the absolute amplitudes between subjects. Considering the time cost, fiber orientation distribution maps of some (30 - 40) subjects can be used to generate a fiber orientation distribution common template specific to the study population, which needs to include subjects from all groups.

[0066] As an implementation manner of an embodiment of the present invention, in the second data processing module, based on the symmetric difference FOD registration algorithm, first deform the fiber orientation distribution maps in the individual space to the previously generated template space. Further divide the fiber orientation distribution at the voxel unit into single fiber bundle population units representing different direction fiber populations, and further calculate three whole-brain level white matter microstructure indexes: fiber density (FD), fiber cross-section (FC), and fiber density and cross-section (FDC) that combines the above two indexes.

[0067] Based on the fiber orientation distribution map, use apparent fibre density (AFD) to calculate the fiber density parameter. For the fiber orientation distribution function F(θ, φ), where θ and φ are spherical coordinate variables, based on the extreme points of the function peak and valley (θ peak , φ peak ) and (θ trough , φ trough ) to obtain different fiber orientation distribution lobes R i , that is, the single fiber bundle population unit representing different fiber distribution directions. Further, perform non-parametric numerical integration within R i :

[0068]

[0069] where the sampling point set is ω j is the weight size determined by the sampling method.

[0070] Since the volume change perpendicular to the plane of the single fiber bundle population unit direction is related to the axon number difference, the calculation of the fiber cross-sectional area is based on the Jacobian matrix in the non-linear transformation. In the non-linear transformation of spatial conversion, the determinant of the Jacobian matrix reflects the local volume difference. Therefore, the formula for calculating the fiber cross-section is as follows:

[0071]

[0072] where det(J) is the determinant of the Jacobian matrix, reflecting the local volume difference; is the unit direction vector of the single fiber bundle population unit f, and J is the Jacobian matrix of the voxel position of the single fiber bundle population unit.

[0073]

[0074] The fiber density and cross-section are defined as the product of the fiber density and the fiber cross-section, which combines the information of the microscopic fiber density and the macroscopic fiber bundle cross-sectional area and can more comprehensively reflect the characteristics of white matter.

[0075] As an implementation manner of an embodiment of the present invention, in the third data processing module, the TractSeg tool is used to divide the common fiber orientation distribution map into 72 prior anatomical fiber bundle masks including the arcuate fasciculus, anterior thalamic radiation, cingulum bundle, corticospinal tract, medial longitudinal fasciculus, frontopontine tract, fornix, inferior cerebellar peduncle, inferior occipitofrontal fasciculus, inferior longitudinal fasciculus, optic radiation, parieto-occipitopontine tract, superior cerebellar peduncle, three superior longitudinal fasciculus partitions, superior thalamic radiation, uncinate fasciculus, six thalamus-related fiber bundles, seven striatum-related fiber bundles, and the anterior commissure between hemispheres, seven corpus callosum partitions, middle cerebellar peduncle, etc., and then the average values of three white matter microstructural indexes of the above fiber bundles are extracted.

[0076] As an implementation manner of an embodiment of the present invention, in the fourth data processing module, the probabilistic fiber bundle tracking algorithm based on the spherical harmonic basis representation of the fiber orientation distribution map is used to obtain the whole-brain fiber bundle streamlines, and the spherical-deconvolution informed filtering of tractograms (SIFT) is used to optimize the quality of the fiber bundle streamlines and improve the biological accuracy of fiber bundle tracking. Further, the white matter structural connections between brain regions are calculated and defined as the total number of fiber bundle streamlines passing through the volume-normalized brain regions between brain regions.

[0077] The spherical deconvolution information filtering algorithm evaluates the reconstruction accuracy by correlating the streamline reconstruction results with the diffusion signal. Assuming that each streamline represents a certain volume of white matter axons, it assesses the matching degree of the fiber density calculated based on the streamline density and the fiber orientation distribution to evaluate the reconstruction accuracy. The reconstruction results are optimized by removing the streamlines that are unfavorable to the reconstruction, reducing the known reconstruction bias, and enhancing the biological rationality.

[0078] A proportionality coefficient μ is introduced to compare the correlation degree between the fiber density and the streamline density, and the calculation formula is:

[0079]

[0080] where FOD V,l is the integral of the l-th fiber orientation distribution lobe in voxel V, that is, the fiber density of this fiber population, TD is the streamline density of the corresponding fiber tractography, and L V is the total number of fiber orientation distribution lobes in voxel V, and PM V is the processed mask of voxel V.

[0081] A cost function is constructed to quantify the fitting degree between the streamline reconstruction and the potential diffusion data:

[0082]

[0083] To efficiently and accurately estimate the change in the cost function when removing a certain streamline, the fiber orientation distribution lobes in other directions that are not crossed by this streamline are given priority, and the approximate calculation is:

[0084]

[0085] Example 2:

[0086] As Figure 2 shown, the embodiment of the present invention also provides a white matter microstructure calculation method based on multi-shell diffusion weighted imaging, including:

[0087] Step S1, acquiring the diffusion weighted imaging and structural image magnetic resonance scans of the subject, and preprocessing the original brain images;

[0088] Step S2, obtaining the individual fiber orientation distribution based on the multi-shell multi-tissue constrained spherical deconvolution algorithm, and simultaneously generating a common template of the population-specific fiber orientation distribution;

[0089] Step S3, co-registering the individual fiber orientation distribution maps to the common template, and segmenting the fiber orientation distribution to obtain the white matter microstructure indexes at the whole brain level;

[0090] Step S4, segmenting the anatomical fiber bundle mask according to the fiber orientation distribution to obtain the white matter microstructure indexes at the level of each anatomical fiber bundle;

[0091] Step S5: Based on probabilistic fiber tractography and spherical deconvolution information filtering, whole-brain fiber streamlines are obtained, and white matter microstructure indices at the network level are generated by combining with a priori brain atlases.

[0092] Using multi-shell diffusion-weighted imaging data from the publicly available ADNI dataset, white matter microstructure indices of 75 healthy subjects were calculated, and the associations between white matter microstructure indices and related cognitive functions were further analyzed.

[0093] As Figure 3 shown, the white matter microstructure calculation process includes the following:

[0094] 1. All MRI data were acquired on a 3T Siemens Magnetom Prisma / PrismaFit scanner with a 64-channel head coil. A multi-band echo planar imaging multi-shell diffusion-weighted sequence was used to obtain diffusion MRI, with parameters: repetition time 3400 ms, echo time 71 ms, diffusion encoding directions 48×b = 1000 s / mm2 and 60×b = 2000 s / mm2 and 13×b = 0 images, flip angle 90°. An isotropic T1-weighted magnetization-prepared rapid acquisition gradient echo image was obtained, with repetition time 2300 ms, echo time 2.98 ms, inversion time 900 ms, and flip angle 9°.

[0095] 2. The diffusion-weighted images were denoised, Gibbs ring artifacts removed, eddy current distortions corrected, head motion corrected, and bias field corrected. The mean b0 image was linearly registered to the high-resolution T1w space, and 5 tissue types of the brain were obtained using FreeSurfer software.

[0096] 3. The diffusion-weighted imaging response function was calculated using a multi-shell multi-tissue algorithm, and further the fiber orientation distribution function of diffusion-weighted imaging in individual space was calculated based on the multi-shell multi-tissue constrained spherical deconvolution algorithm. A study-specific group template was generated using 15 healthy subjects, and the individual space fiber orientation distribution function was registered to the common space, and fiber density, fiber cross-section, and fiber density and cross-section white matter microstructure indices at the whole-brain level were calculated.

[0097] 4. A deep learning tool for automatic white matter tract segmentation based on TractSeg divides the fiber orientation distribution function template into 72 anatomical fiber tracts, including the arcuate fasciculus of the left and right hemispheres, anterior thalamic radiation, cingulum bundle, corticospinal tract, medial longitudinal fasciculus, frontopontine tract, fornix, inferior cerebellar peduncle, inferior occipitofrontal fasciculus, inferior longitudinal fasciculus, optic radiation, parieto-occipitopontine tract, superior cerebellar peduncle, three divisions of the superior longitudinal fasciculus, superior thalamic radiation, uncinate fasciculus, six thalamus-related fiber tracts, seven striatum-related fiber tracts, and the anterior commissure between hemispheres, seven divisions of the corpus callosum, and middle cerebellar peduncle. To reduce comparisons, the FBA measurements of the left and right brains were averaged, and tracts located in the cerebellum and certain regions were excluded. Finally, 24 anatomical fiber tracts were selected to evaluate fiber-specific metrics and their relationships with specific cognitive functions, and fiber bundle-level white matter microstructure metrics of the anatomical fiber tracts were extracted.

[0098] 5. Probabilistic fiber tractography based on fiber orientation distribution maps was used to reconstruct the white matter fiber streamlines of the brain. First, the cerebral cortex of all subjects was divided into 7 large-scale brain networks based on the Yeo atlas, 20 million fiber streamlines were generated in the individual space of the subjects, and then 2 million fiber streamlines were filtered using the spherical deconvolution information filtering algorithm. The network structural connectome was defined as the number of streamlines between different functional networks, and the mediating role of the structural connection between large-scale functional regions in the white matter microstructure and cognitive function at the fiber bundle level was further explored.

[0099] As Figure 4 shown, the final results showed that correlations between specific cognitive functions (such as language function) and white matter microstructure characteristics at the fiber bundle level were observed, and it was also found that the structural connectome between functional regions had a potential mediating role in the impact of fiber bundle integrity on specific cognitive functions.

[0100] The embodiments described above are only descriptions of the preferred embodiments of the present invention and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A white matter microstructure calculation system based on multi-shell diffusion weighted imaging, characterized in that Including: A data acquisition module, which is used to acquire the diffusion weighted imaging and structural magnetic resonance scans of the subject, and preprocess the original brain images; A first data processing module, which is used to obtain the individual fiber orientation distribution based on the multi-shell multi-tissue constrained spherical deconvolution algorithm, and generate a common template of the fiber orientation distribution specific to the population; A second data processing module, which is used to co-register the individual fiber orientation distribution map to the common template, and segment the fiber orientation distribution to obtain white matter microstructure indexes at the whole-brain level; A third data processing module, which is used to segment the anatomical fiber bundle mask according to the fiber orientation distribution, and obtain white matter microstructure indexes at the level of each anatomical fiber bundle; A fourth data processing module, which is used to obtain the whole-brain fiber streamlines based on probabilistic fiber tractography and spherical deconvolution information filtering, and generate white matter microstructure indexes at the network level in combination with the prior brain atlas.

2. The white matter microstructure calculation system based on multi-shell diffusion weighted imaging according to claim 1, wherein The magnetic resonance scan of the data acquisition module includes multi-shell diffusion weighted imaging, that is, in addition to the reference image with b = 0, diffusion weighted imaging with two or more b values, and high-resolution structural images of T1 weighted imaging.

3. The white matter microstructure calculation system based on multi-shell diffusion weighted imaging according to claim 1, wherein The data acquisition module uses MRtrix3 software to perform noise removal, Gibbs ring artifact removal, distortion correction caused by eddy current, head motion correction, and bias field correction on the multi-shell diffusion weighted imaging, and linearly register the average b0 image to the high-resolution T1 structural image; uses FreeSurfer software to perform recon-all processing on the T1 structural image to obtain five tissue types.

4. The white matter microstructure calculation system based on multi-shell diffusion weighted imaging according to claim 3, wherein, The first data processing module derives and estimates the response function of the spherical deconvolution model based on the co-registered five-tissue type map, and obtains the response functions of gray matter, white matter, and cerebrospinal fluid in the individual space; Calculate the fiber orientation distribution at the voxel level based on the multi-shell multi-tissue constrained spherical deconvolution algorithm.

5. The white matter microstructure calculation system based on multi-shell diffusion weighted imaging according to claim 4, wherein The first data processing module performs joint bias field correction and intensity normalization on the obtained fiber orientation distribution map in the individual space, and uses the fiber orientation distribution maps of some subjects to generate a common template of the fiber orientation distribution specific to the research population.

6. The white matter microstructure calculation system based on multi-shell diffusion weighted imaging according to claim 5, wherein The second data processing module divides the fiber orientation distribution at the voxel unit into single fiber bundle population units representing different direction fiber populations, and simultaneously calculates and obtains three white matter microstructure indexes at the whole-brain level, namely fiber density, fiber cross-section, and fiber density and cross-section combining the above two indexes.

7. The white matter microstructure calculation system based on multi-shell diffusion weighted imaging according to claim 6, characterized in that, The third data processing module uses the TractSeg tool to divide the common fiber orientation distribution map into 72 prior anatomical fiber bundles, and then extracts the average values of the three white matter microstructure indexes of the fiber bundles.

8. The white matter microstructure calculation system based on multi-shell diffusion weighted imaging according to claim 7, characterized in that, The fourth data processing module obtains the whole-brain fiber bundle streamlines based on the anatomically constrained probabilistic fiber tractography algorithm of the fiber orientation distribution map represented by spherical harmonics, and optimizes the quality of the fiber bundle streamlines using spherical deconvolution information filtering; at the same time, calculates and obtains the white matter structure connections between brain regions, which are defined as the total number of fiber bundle streamlines passing through the volume-normalized brain regions.

9. A method for calculating white matter microstructure based on multi-shell diffusion weighted imaging, characterized in that, Including: Step S1, acquire the diffusion weighted imaging and structural magnetic resonance scans of the subject, and preprocess the original brain images; Step S2: Obtain the individual fiber orientation distribution based on the multi-shell multi-tissue constrained spherical deconvolution algorithm, and simultaneously generate a common template of the population-specific fiber orientation distribution; Step S3: Co-register the individual fiber orientation distribution map to the common template, and segment the fiber orientation distribution to obtain white matter microstructure indexes at the whole-brain level; Step S4: Obtain the anatomical fiber bundle mask by segmenting the fiber orientation distribution, and obtain white matter microstructure indexes at the level of each anatomical fiber bundle; Step S5: Obtain the whole-brain fiber streamlines based on probabilistic fiber tractography and spherical deconvolution information filtering, and generate white matter microstructure indexes at the network level in combination with the prior brain atlas.

Images