Brain structure and function image processing method based on multi-mode magnetic resonance brain image

Abstract

The invention belongs to the field of medical image processing, and provides a brain structure and function image processing method based on a multi-modal magnetic resonance brain image for the multi-modal magnetic resonance brain image, which comprises the following steps: acquiring the multi-modal magnetic resonance brain image of a healthy crowd, preprocessing data, dividing the brain into regions based on a prior template, and obtaining the multi-modal magnetic resonance brain image of the healthy crowd; calculating a function connection matrix, extracting fiber bundle structure connection, obtaining a fiber bundle number matrix, calculating community attributes of the structure connection matrix, calculating a shortest path of the structure connection matrix, calculating an Euclidean distance between node coordinates in the structure connection matrix, and constructing a brain structure and function coupling model. The coupling index of the structure and the function is helpful for reflecting the constraint relation of the structure and the function of the brain and depicting the decoupled condition of brain decoupling and over-coupling.

Description

Technical Field

[0001] The present invention belongs to the field of medical image processing, and particularly relates to a method for processing brain structure and function images based on multi-modal magnetic resonance brain images. Background Art

[0002] Existing neuroscience research shows that there is a complex bidirectional relationship between brain structure and function. Structure determines the potential of function, while functional activities in turn shape and change the structure. This interaction is the key to understanding how the brain works and its changes in healthy and diseased states.

[0003] Magnetic resonance imaging is a means of detecting brain function and activity with high resolution and non-invasiveness. Magnetic resonance images can not only depict the anatomical structure of the brain in detail, but also observe the functional activities of the brain in real time. For example, structural magnetic resonance imaging can observe the anatomical structure of the brain, including gray matter volume, white matter integrity, and morphological characteristics of different brain regions. Through structural magnetic resonance imaging, differences in brain structure between different individuals can be revealed, and how these differences are related to cognitive function and behavioral performance can be understood. For the brain, white matter fiber tracts are its structural basis. Diffusion tensor imaging can be used to detect the integrity and connectivity of white matter fiber tracts. On the other hand, the brain is an efficient and complex system. Different brain regions work together to maintain the functional activities of the brain. Functional magnetic resonance imaging monitors changes in brain activity by measuring blood oxygenation level-dependent signals, and is mainly used to study the functional responses of the brain when performing specific tasks and the functional connectivity between brain regions.

[0004] During development, structural changes such as the formation of brain neurons and synapses and the myelination of white matter fibers promote the improvement of cognitive function. As people age, the brain structure gradually degenerates, and this structural degeneration limits cognitive function, and changes in functional activities also reflect the degree of structural degeneration. However, previous studies often characterized the structure or function of the brain based on single modality, and did not comprehensively understand the relationship between structure and function by combining multi-modal brain image data such as structural magnetic resonance imaging, functional magnetic resonance imaging, and diffusion tensor imaging.

[0005] Therefore, based on multi-modal brain images, the present invention proposes a method for processing images to quantitatively characterize the coupling relationship between brain structure and function. Summary of the Invention

[0006] The present invention aims to solve the problems described in the background art, and provides a method for processing brain structure and function images based on multi-modal magnetic resonance brain images.

[0007] To achieve the above object, the present invention adopts the following technical solutions: A method for processing brain structure and function images based on multi-modal magnetic resonance brain images, starting from two modalities of brain structure and functional structure, characterizing the structural and functional coupling of the brain, including the following steps: Step 1, obtain multi-modal magnetic resonance brain images of healthy people, including diffusion tensor imaging and resting-state functional images; Step 2, remove the first N time points of the resting-state functional image data, where N is a natural number, and perform spatial correction, normalization, detrending, and filtering on the resting-state functional image data; Step 3, reduce the artifacts in the diffusion tensor imaging data through eddy current correction, perform linear least squares fitting estimation on each voxel, and perform whole-brain fiber tractography on the brain of each subject in the original space; Step 4: Divide the brain into multiple regions of interest based on a priori templates; extract the signals of multiple regions of interest based on the resting-state functional images, and calculate the correlation coefficients between pairwise regions of interest: ; where X and Y respectively represent the time series of the first region of interest and the second region of interest, i represents the i-th time point, n represents the total number of time points, X i represents the signal intensity of the first region of interest at the i-th time point, Y i represents the signal intensity of the second region of interest at the i-th time point, and thus a functional connectivity matrix is obtained, represents the mean of the time series of the first region of interest, represents the mean of the time series of the second region of interest, the dimension of the functional connectivity matrix is the number of regions of interest × the number of regions of interest, and the correlation coefficients between pairwise regions of interest are the elements constituting the functional connectivity matrix; Step 5: Based on the diffusion tensor imaging data, extract the fiber bundle structural connectivity matrix; Step 6: Calculate the weighted sum of all paths and walks between the network nodes of the fiber bundle structural connectivity matrix to obtain the community attribute matrix of the fiber bundle structural connectivity matrix; Step 7: Calculate the shortest path between the nodes of the fiber bundle structural connectivity matrix to obtain the shortest path matrix of the fiber bundle structural connectivity matrix; Step 8: Calculate the Euclidean distance between the coordinates of multiple regions of interest; Step 9: Construct a structural and functional coupling model: ; where y is a column of the functional connectivity matrix, Eu is a column of the Euclidean distance between the nodes of the fiber bundle structural connectivity matrix, Pl is a column of the shortest path matrix of the fiber bundle structural connectivity matrix, and Co is a column of the community attribute matrix of the fiber bundle structural connectivity matrix, , , , are the coefficients of the structural-functional coupling model; Step Ten: Use partial least squares method to obtain the optimal coefficients of the generalized linear model, and use the goodness of fit of the linear model to characterize the structural-functional coupling relationship of each network node.

[0008] In one implementation, the termination criterion for whole-brain fiber tractography is that the fractional anisotropy value is less than 0.2 or the crossing angle threshold is 60°.

[0009] In one implementation, Step Three further includes transforming the standard template onto the original space of each subject, inversely transforming the standard template into the individual space of each subject, and obtaining the standard template projected onto the individual space for each subject.

[0010] In one implementation, after calculating the correlation coefficient r, the correlation coefficient r of the functional connectivity matrix is transformed as follows: , to obtain a new functional connectivity matrix for constructing the brain structural-functional coupling model, where Z is the element of the new functional connectivity matrix.

[0011] In one implementation, for a binary network A, the weighted sum of all paths and walks between network nodes is calculated as follows: , where e is the natural constant.

[0012] In one implementation, the process of spatial transformation of the standard template includes: first, performing a linear transformation on the brain structural image of each subject and registering it with the non-diffusion weighted image corresponding to the brain structural image of each subject; then, through affine transformation and non-linear transformation with multiple degrees of freedom, mapping the registered brain structural image onto the stereoscopic space of the standard template to obtain the corresponding transformation parameters.

[0013] A method for processing brain structural and functional images based on multi-modal magnetic resonance brain imaging provided by the present invention has at least the following beneficial effects compared with the prior art: 1) Compared with the existing measurement structural-functional coupling models, such as the correlation model of structural indicators and functional indicators, the structural features and functional features proposed by the present invention characterize brain information from multiple dimensions, and the model is more accurate; 2) It provides a lightweight and quantitative measurement framework for revealing the coupling relationship between the structure and function of the human brain. The entire model algorithm has low complexity and consumes little computing power resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1Schematic diagram of the process of the present invention; Figure 2 Schematic diagram of brain functional connectivity, structural connectivity, Euclidean distance between nodes, shortest structural path, and structural community attributes calculated using the prior Diez2514 template based on a single healthy subject; Figure 3 Schematic diagram of the brain structure - function coupling model; Figure 4 Schematic diagram of the structural - functional coupling index of 2514 regions of interest in the whole brain of a single healthy subject and the goodness - of - fit R of the linear model 2 , including cross - section, coronal plane, and sagittal plane; Figure 5 Schematic diagram of the structural - functional coupling index of 2514 regions of interest in the whole brain of a single healthy subject and the goodness - of - fit R of the linear model 2 Distribution map. Specific implementation manners

[0015] To make the objectives and technical implementations of the present invention clearer, the embodiments of the present invention are described in detail below. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0016] In the description of the present invention, unless otherwise specified and defined, it should be noted that terms such as "region of interest", "structure", "function", "community", "fitting", "functional connectivity", "structural connectivity", etc. should be understood in a broad sense. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific situations.

[0017] As Figure 1 shown, the present application provides a method for processing brain structure and function images based on multi - modal magnetic resonance brain images and tests it on the magnetic resonance data of a healthy subject, including the following steps: Step 1: Obtain multi - modal magnetic resonance brain images of a healthy population, including diffusion tensor imaging and resting - state functional images; Step 2: Remove the first 10 time points of the resting - state functional image data, and perform spatial correction, normalization, smoothing, detrending, regression of white matter and cerebrospinal fluid signals, and filtering on the resting - state functional image data; Step 3: Reduce the artifacts in the diffusion tensor imaging data through eddy current correction. The artifacts in the diffusion tensor imaging data include the artifacts caused by the change of the gradient field direction of the scanning instrument and the artifacts brought by the head movement of the subject. Linear least squares fitting estimation is performed on each voxel, and the continuous tracking algorithm is used to perform whole-brain fiber tract tracking on the brain of each subject in the original space. The tracking termination criterion is that the fractional anisotropy value is less than 0.2 or the crossing angle threshold is 60°. The Montreal Neurological Institute standard template is spatially transformed to the original space of each subject. In this transformation process, first, the structural image of each subject is linearly transformed into the standard space and registered with its corresponding non-diffusion weighted image. Then, through an affine transformation with 12 degrees of freedom and a series of non-linear transformations mainly based on 7×8×7 basis functions, the registered structural image is mapped to the stereoscopic space of the standard template to obtain the corresponding transformation parameters. Using the obtained corresponding transformation parameters, the standard template is inversely transformed into the individual space of each subject to obtain the standard template projected onto the individual space for each subject; Step 4: Divide the brain into 2514 regions of interest based on the prior template; Step 5: Extract the signals of 2514 regions of interest based on the resting-state functional images and calculate the functional connectivity matrix 2514×2514; the method for calculating functional connectivity is the correlation coefficient between every two regions of interest. After calculating the correlation coefficient r, in order to increase the statistical power, the correlation coefficient r is further transformed, , where Z is the element of the new functional connectivity matrix; Step 6: Based on the diffusion tensor imaging data, extract the fiber bundle structural connectivity matrix. The fiber bundle structural connectivity matrix is composed of the number of fiber bundles. Therefore, the fiber bundle number matrix 2514×2514 is used here, as Figure 2 shown; Step 7: Calculate the community attributes of the structural connectivity matrix, as Figure 2 shown. In the schematic diagram, the depth of color represents the strength of functional and structural connectivity or the level of structural indicators; the diagonal line is the diagonal element of the matrix, which is all 1 in the functional connectivity matrix and the structural connectivity matrix, and all 0 in the Euclidean distance and the structural shortest path matrix; the diagonal elements of the functional connectivity matrix and the structural connectivity matrix are 1, indicating the connection of the node to itself; the diagonal elements of the Euclidean distance matrix are 0 because the distance from itself to itself is 0; the diagonal elements of the structural shortest path matrix are 0 because the shortest path from itself to itself is 0; Step 8: Calculate the shortest path between nodes of the structural connection matrix; First, convert the number of fiber bundles in the structural connection matrix into connection distance. The more the number of bundles, the stronger the connection and the shorter the connection distance. Calculating the shortest path includes four steps: 1) Initialization: Set two sets A and B. A stores the points for which the shortest path has been found, and B stores the points for which the shortest path has not been calculated. Initially, A only contains the source point, B contains all other points, and the distance from the source point to itself is 0. 2) Select the shortest path: Select a point in set B that is closest to the source point and add it to set A, and update the distances of its adjacent points. 3) Update adjacent points: For the point newly added to set A, update the shortest paths of all its adjacent points. 4) Repeat steps 2 and 3: Until all points are added to set A, the algorithm ends. Step 9: Calculate the Euclidean distance between the coordinates of 2514 regions of interest, as Figure 2 shown; Step 10: Construct a structural-functional coupling model: ; As Figure 3 shown, in the schematic diagram of the brain structural-functional coupling model, y is a certain column of the functional connection matrix FC in Figure 2 , Eu is a certain column of the Euclidean distance matrix in Figure 2 , Pl is a certain column of the structural shortest path matrix in Figure 2 , and Co is a certain column of the structural community attribute matrix in Figure 2 . After obtaining the best generalized linear model coefficients using partial least squares method, use the goodness of fit R 2 of the linear model to characterize the coupling relationship between the structure and function of each node, as Figure 3 , Figure 4 , Figure 5 shown.

Claims

1. A brain structure and function image processing method based on multimodal magnetic resonance imaging, characterized in that: Starting from the two modes of brain structure and functional structure, the structural and functional coupling of the brain is described, including the following steps: Step 1: Obtain multimodal magnetic resonance brain images of healthy people, including diffusion tensor imaging and resting-state functional imaging; Step 2, removing the first N time points of the resting-state functional imaging data, where N is a natural number, and performing spatial correction, standardization, linear drift removal and filtering on the resting-state functional imaging data; Step 3: Reduce artifacts in diffusion tensor imaging data through eddy current correction, perform linear least squares fitting estimation on each voxel, and achieve whole-brain fiber tract tracking for each subject in the original space; Step 4: Divide the brain into multiple regions of interest based on the prior template; extract the signals of multiple regions of interest based on resting-state functional imaging, and calculate the correlation coefficients between the two regions of interest: ； Where X and Y represent the time series of the first and second regions of interest, respectively, i represents the i-th time point, n represents the total number of time points, and X i represents the signal intensity of the first region of interest at the i-th time point, Y i represents the signal intensity of the second region of interest at the i-th time point, thereby obtaining the functional connectivity matrix, represents the mean of the time series of the first area of ​​interest, represents the mean of the time series of the second region of interest. The dimension of the functional connectivity matrix is ​​the number of regions of interest × the number of regions of interest. The correlation coefficient between any two regions of interest is the element that constitutes the functional connectivity matrix. Step 5: Extract the fiber bundle structure connection matrix based on diffusion tensor imaging data; Step 6: Calculate the weighted sum of all paths and walks between the nodes of the fiber bundle structure connection matrix network to obtain the community attribute matrix of the fiber bundle structure connection matrix; Step 7: Calculate the shortest paths between nodes of the fiber bundle structure connection matrix to obtain the shortest path matrix of the fiber bundle structure connection matrix; Step 8: Calculate the Euclidean distance between the coordinates of multiple regions of interest; Step 9: Construct a brain structure-function coupling model: ； Among them, y is a column of the functional connection matrix, Eu is a column of the Euclidean distance between nodes in the fiber bundle structure connection matrix, Pl is a column of the shortest path matrix of the fiber bundle structure connection matrix, and Co is a column of the community attribute matrix of the fiber bundle structure connection matrix. , , , is the coefficient of the structure-function coupling model; Step 10: Use partial least squares method to obtain the optimal generalized linear model coefficients, and use the goodness of fit of the linear model to characterize the structural and functional coupling relationship of each network node.

2. The method for processing brain structure and function images based on multimodal magnetic resonance imaging according to claim 1, characterized in that: The termination criteria for whole-brain fiber tract tracing were anisotropy fraction less than 0.2 or a crossing angle threshold of 60°.

3. The brain structure and function image processing method based on multimodal magnetic resonance imaging according to claim 1, characterized in that: Step three also includes transforming the standard template to the original space of each subject, inversely transforming the standard template to the individual space of each subject, and obtaining the standard template projected to the individual space for each subject.

4. The brain structure and function image processing method based on multimodal magnetic resonance imaging according to claim 1, characterized in that: After calculating the correlation coefficient r, the correlation coefficient r of the functional connectivity matrix is ​​transformed as follows: , a new functional connection matrix is ​​obtained, which is used to construct a brain structure-function coupling model, where Z is an element of the new functional connection matrix.

5. The brain structure and function image processing method based on multimodal magnetic resonance imaging according to claim 3, characterized in that: The process of spatial transformation of the standard template includes: first, linearly transforming the brain structure image of each subject and aligning it with the non-diffusion weighted image corresponding to the brain structure image of each subject; then, through affine transformation and nonlinear transformation with multiple degrees of freedom, mapping the aligned brain structure image to the three-dimensional space of the standard template to obtain the corresponding transformation parameters.

Images