Brain network hub node identification method based on dynamic and static feature fusion

Abstract

According to the brain network hub node recognition method based on dynamic and static feature fusion, static topological features and dynamic attribute features of brain network nodes are combined, degree centrality and betweenness centrality are adopted as static feature measurement, weighted superposition fusion is carried out in combination with dynamic features, the features are enhanced, and the recognition efficiency is improved. The accuracy and reliability of an identification result are improved, and the hub node which can perform a finer and longer-period tracing effect on a dynamic process of a neurodegenerative disease similar to an Alzheimer disease can be identified.

Description

Technical Field

[0001] The present invention relates to the field of brain network hub node identification, and more specifically to a brain network hub node identification method based on dynamic and static feature fusion. Background Art

[0002] The human brain is a complex network nervous system composed of tens of billions of neurons and the connections between them. It dominates and coordinates various physiological and psychological activities of our body. Therefore, having a healthy brain is the premise and basis for ensuring that we enjoy a normal life. Although the sophisticated brain nervous system has given us humans infinite wisdom, its complex operating mechanism is still a research topic full of challenges and unsolved mysteries. Studies have shown that there is a close relationship between the pathogenesis of many neurodegenerative brain diseases and the complex network nervous system of the human brain. For example, the hub nodes in the brain network usually show more sensitive characteristics to the attack of brain diseases. Therefore, the accurate identification of the hub nodes of the brain network is particularly important for the advanced diagnosis, prevention and treatment of related brain diseases.

[0003] In recent years, many excellent hub node identification methods for brain networks have emerged, such as hub node identification methods based on graph theory measures such as node degree and betweenness centrality. Although these methods can simply and directly identify some important hub nodes in the brain network, these methods usually only consider the static topological role of each network node in its network during the hub node identification process, but ignore the consideration of its dynamic characteristics. The pathological attacks related to many brain diseases are precisely a process of dynamic erosion. Therefore, the existing methods that only identify hub nodes based on the static characteristics of network nodes are difficult to discover brain network hub nodes with stronger pathological properties, which in turn limits their pathological target properties. Summary of the invention

[0004] In order to solve the problem that the existing static feature identification method of hub nodes is difficult to discover brain network hub nodes with stronger pathological attributes, the present invention proposes a brain network hub node identification method based on dynamic and static feature fusion.

[0005] In order to achieve the above technical effects, the technical solution of the present invention is as follows:

[0006] The brain network hub node identification method based on dynamic and static feature fusion includes:

[0007] Acquire neural imaging sample data, and pre-process the neural imaging sample data;

[0008] Construct mathematical models of sample brain networks;

[0009] Calculate the controllability of the brain network nodes according to the mathematical model, and obtain the dynamic characteristics of the brain network nodes based on the controllability of the brain network nodes; calculate the degree centrality of the brain network nodes, calculate the betweenness centrality of the brain network nodes, and obtain the static characteristics of the brain network nodes based on the degree centrality and the betweenness centrality of the brain network nodes;

[0010] Obtain the static and dynamic fusion characteristics of the brain network nodes of the sample based on the controllability, the degree centrality, and the betweenness centrality of the brain network nodes;

[0011] Obtain the node feature vector of the sample based on the static and dynamic fusion characteristics of the brain network nodes of the sample, and obtain the brain network node feature matrix based on the node feature vector of the sample;

[0012] Obtain the hub attribute score vector of the node based on the node feature vector of the sample and the brain network node feature matrix; output the hub nodes based on the hub attribute score vector of the node.

[0013] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0014] The present invention combines the static topological features and the dynamic attribute features of the brain network nodes, uses the degree centrality and the betweenness centrality as the static feature measures, and combines and fuses them with the dynamic features through weighted superposition, enhancing the features, improving the accuracy and reliability of the recognition results, and being able to identify the hub nodes that can trace the dynamic process of the onset of neurodegenerative diseases similar to Alzheimer's disease more precisely and for a longer period. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the method for identifying brain network hub nodes based on the fusion of static and dynamic features shown in the embodiments of the present invention.

[0016] Figure 2 It is a flowchart of the recognition algorithm shown in the embodiments of the present invention.

[0017] Figure 3 It is a statistical analysis chart of the position distribution of the hub nodes identified by the present method and three other traditional methods and the corresponding Amyloid protein level shown in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0019] The terms used in the present invention are for the purpose of describing particular embodiments only and are not intended to limit the present invention. The singular forms "a", "the", and "said" used in the present invention and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0020] It should be understood that although the terms first, second, third, etc. may be used in the present invention to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to a determination".

[0021] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Embodiment 1

[0023] This embodiment proposes a method for identifying hub nodes in a brain network based on the fusion of static and dynamic features, as Figure 1 shown, is a flowchart of the method for identifying hub nodes in a brain network based on the fusion of static and dynamic features of this embodiment.

[0024] The method for identifying hub nodes in a brain network based on the fusion of static and dynamic features proposed in this embodiment includes:

[0025] Obtain neuroimaging sample data and preprocess the neuroimaging sample data;

[0026] Construct a mathematical model of the sample brain network;

[0027] Calculate the controllability of the brain network nodes according to the mathematical model, and obtain the dynamic features of the brain network nodes according to the controllability of the brain network nodes; calculate the degree centrality of the brain network nodes, calculate the betweenness centrality of the brain network nodes, and obtain the static features of the brain network nodes according to the degree centrality of the brain network nodes and the betweenness centrality of the brain network nodes;

[0028] Obtain the dynamic and static fusion features of the brain network nodes of the sample based on the controllability of the brain network nodes, the degree centrality of the brain network nodes, and the betweenness centrality of the brain network nodes;

[0029] Obtain the node feature vector of the sample according to the dynamic and static fusion features of the brain network nodes of the sample, and obtain the brain network node feature matrix according to the node feature vector of the sample;

[0030] Obtain the hub attribute score vector of the node according to the node feature vector of the sample and the brain network node feature matrix; Output the hub node according to the hub attribute score vector of the node.

[0031] The present invention combines static topological features and dynamic attribute features of brain network nodes, uses degree centrality and betweenness centrality as static feature measures, and combines dynamic features for weighted superposition fusion, enhancing the features, improving the accuracy and reliability of the recognition results, and being able to identify hub nodes that can trace the dynamic process of the onset of neurodegenerative diseases similar to Alzheimer's disease more precisely and for a longer period; At present, with the continuous intensification of the global population aging trend, Alzheimer's disease has become one of the key diseases leading to human disability and death. However, no clear and effective treatment methods and drugs for this disease have been found yet. Numerous studies have shown that the pathological process of Alzheimer's disease is significantly associated with hub nodes in the brain network nervous system. Therefore, identifying hub nodes that have a stronger enlightenment effect on the pathological process of Alzheimer's disease not only helps to deeply understand the disease mechanism, but also provides a possible direction for new intervention measures. By accurately locating these hub nodes, the disease progression can be monitored more effectively.

[0032] Embodiment 2

[0033] This embodiment proposes a method for identifying hub nodes of a brain network based on the fusion of dynamic and static features, and makes a detailed description on the basis of Embodiment 1.

[0034] In the field of neuroscience, each brain network can be represented by a graph encoding. The node set V of the brain network = {v1, v2, v3,..., v N}, where v i represents the i-th node (brain region) in the brain network, and N = |V| represents the total number of nodes included in each brain network. The anatomical connectivity matrix is an N×N weighted adjacency matrix, and w ij represents the connection strength between a pair of brain regions (v i , v j ), and its value is determined by the number of white matter fiber bundles that physically connect the corresponding brain regions, while w ii= 0, that is, the constructed brain network does not consider self - connections. The state vector of the brain network nodes at time t where x i (t) is a real scalar value representing the state value at time t at the i - th node in the brain network. By performing thresholding on the weighted adjacency matrix W of the above - mentioned brain network, the corresponding binary adjacency matrix can be obtained where a ij represents the connection relationship between the i - th node and the j - th node. If there is a direct connection between this pair of nodes, then a ij = 1; if there is no direct connection between this pair of nodes, then a ij = 0.

[0035] In an alternative embodiment, the pre - processing of the neuroimaging sample data is specifically as follows:

[0036] Step 1: Perform tissue segmentation on the images in the neuroimaging sample data; Step 2: Construct the cerebral cortex surface and calculate the level of Amyloid protein in each voxel; Step 3: Divide the cerebral cortex surface into brain regions; Step 4: Generate an anatomical connectivity matrix, and each element of this matrix is measured by the number of fibers between brain regions; Step 5: Calculate the average level of Amyloid protein in each brain region; Step 6: Select the gray matter of the cerebellum as a reference region, normalize the Amyloid protein levels of the brain regions, and generate a neuropathological load.

[0037] Furthermore, first, use advanced image registration techniques to accurately align all imaging data except T1 - weighted MRI with its corresponding T1 - weighted MRI imaging data. Subsequently, further process the data according to the following steps: Step 1: Use the FreeSurfer software platform to perform tissue segmentation on the images in the neuroimaging sample data; Step 2: Construct the cerebral cortex surface and calculate the level of Amyloid protein in each voxel; Step 3: Use the Destrieux brain atlas to divide the cerebral cortex surface into 148 brain regions; Step 4: Generate a 148×148 anatomical connectivity matrix using surface - based seed - based probabilistic fiber tractography imaging technology, and each element of this matrix is measured by the number of fibers between brain regions; Step 5: Calculate the average level of Amyloid protein in each brain region; Step 6: Select the gray matter of the cerebellum as a reference region, normalize the Amyloid protein levels of the 148 brain regions, and generate a neuropathological load of size 148×1.

[0038] Furthermore, a total of 141 data samples were selected from the ADNI database for the neuroimaging sample data, including 50 normal control data samples (NC), 44 mild cognitive impairment data samples (MCI), and 47 Alzheimer's disease data samples (AD). Each data sample has multiple T1-weighted MRI, DWI, and Amyloid-PET scan data.

[0039] In an alternative embodiment, the mathematical model includes an anatomical connectivity matrix and a binary adjacency matrix; the expression of the anatomical connectivity matrix is:

[0040]

[0041] where W represents an N×N weighted adjacency matrix, and w ij represents the connection strength between a pair of brain regions (v i , v j ), j represents the j-th node, and i represents the i-th node;

[0042] The expression of the binary adjacency matrix is:

[0043]

[0044] where A represents the binary adjacency matrix, and a ij represents the connection relationship between the i-th node and the j-th node.

[0045] The node controllability of the brain network is used to capture the dynamic pathological attribute characteristics of the brain network nodes. To measure the controllability of the brain network nodes, a neural dynamics model is first defined. A large number of dynamic behaviors in neural dynamics can be described and predicted by a linear model. Therefore, a linear discrete-time invariant network model is defined here, and its state equation is specifically as follows:

[0046] x(t + 1) = Wx(t) + Bu(t)

[0047] where W represents the coefficient matrix of the internal state relationship of the system, which is the anatomical connectivity matrix here, represents the input matrix for identifying the control nodes in the brain network, where is a standard vector corresponding to the c i control nodes, i = 1, 2, 3,..., m, is the input vector representing the control strategy.

[0048] In an alternative embodiment, a measurement index for the dynamic characteristics of the hub nodes is obtained according to the controllability of the brain network nodes; its expression is:

[0049]

[0050] Among them, represents the Gram matrix, τ represents the time step in the iterative solution of the Gram matrix, T represents the matrix transpose operation, W represents the anatomical connectivity matrix, and B represents the input matrix.

[0051] Furthermore, according to the controllability study of dynamic systems in control theory, the controllability of the brain network system can be quantitatively evaluated through its corresponding Gram matrix; the elements on the main diagonal of the Gram matrix respectively reflect the contribution of each node in the network to the controllability of the entire network. Therefore, it is used as a measurement index for identifying the dynamic characteristics of hub nodes. The controllability measure of the i-th node is

[0052] In an alternative embodiment, the controllability measure of the i-th node in the Gram matrix is

[0053] Given that the degree centrality and betweenness centrality of network nodes measure the importance of network nodes in the network topology from local and global levels respectively, in order to improve the comprehensiveness of the hub node identification, these two complementary graph theory measures are selected simultaneously as the static feature measures for hub node identification.

[0054] The degree k of the i-th node in an undirected network i is equal to the number of all edges connecting this node to other nodes in the network. Since the adjacency matrix of an undirected network is a symmetric matrix, the value of the degree centrality of the brain network node is equal to the sum of all elements in the i-th row or the i-th column of the binary adjacency matrix of this undirected network.

[0055] In an alternative embodiment, the degree centrality of the brain network node is the sum of all elements in the i-th row or the i-th column of the binary adjacency matrix; its expression is:

[0056]

[0057] Among them, k i represents the degree centrality of the brain network node, a ij represents the binary adjacency matrix. When all elements on the main diagonal of the binary adjacency matrix have been set to zero, a ii = 0, j represents the j-th node, and i represents the i-th node.

[0058] Based on the assumption that information between two nodes in the network is preferentially propagated through the shortest path, the betweenness centrality of a node measures the role of the node in bridging different groups in the network by the ratio of the shortest paths between all different nodes in the network passing through a certain node.

[0059] In an alternative embodiment, the betweenness centrality of the brain network node is the sum of the proportions of the shortest paths between all nodes passing through node i; its expression is:

[0060]

[0061] where C B (i) represents the betweenness centrality of the i-th node, and g hj represents the total number of the shortest paths between the h-th node and the j-th node, represents the number of the shortest paths passing through the i-th node among the above shortest paths.

[0062] Furthermore, in view of the fact that there may be multiple shortest paths between any two nodes in the network, the betweenness centrality formula is defined in the form of a ratio.

[0063] In an alternative embodiment, the controllability of the brain network node, the degree centrality of the brain network node, and the betweenness centrality of the brain network node are respectively subjected to hierarchical normalization to obtain corresponding features, and the features are weighted and superimposed and fused to obtain the fused feature of the brain network node; its expression is:

[0064]

[0065] where and respectively represent the measured values after hierarchical normalization of the controllability, degree centrality, and betweenness centrality of the brain network node, i = 1, 2, 3,..., N, and λ C , λ D and λ B respectively represent the corresponding weight scalar parameters.

[0066] Furthermore, in order to improve the balance of the contributions of the dynamic features and the static features to the final fused feature, the following constraints are imposed on the above three weight parameters, that is, λ C = λ D + λ B .

[0067] In this embodiment, since there are significant differences in the numerical magnitudes and scale ranges of the feature information measures, in order to retain the uniqueness of different features while effectively enhancing the comparability between different features, and in order to reduce the redundant information in various feature information and reduce the complexity of the subsequent hub node recognition operation, each hub attribute feature of the brain network node is respectively subjected to hierarchical normalization, and then multiple features are weighted and superimposed and fused, so as to generate a fair and complementary feature representation.

[0068] In an alternative embodiment, the sample node feature vector is represented as Input the node feature vector of the sample into the brain network node feature matrix F = [f 1 , f 2 , f 3 , …, f n , where n represents the total number of samples, and the expanded form of F is shown as follows:

[0069]

[0070] In an alternative embodiment, calculate the hub attribute score vector of each node according to the brain network node feature matrix; the set of hub attribute score vectors of the nodes is represented as:

[0071] s = [s1, s2, s3, …, s U

[0072] where s represents the set of hub attribute score vectors of the nodes.

[0073] When the hub attribute score vector of the node exceeds the preset threshold, output all hub nodes that exceed the preset threshold.

[0074] Exemplarily, the program of the entire recognition algorithm is written in MATLAB language, and the flowchart of the recognition algorithm is as Figure 2 shown, where n represents the total number of samples l.

[0075] Exemplarily, use the sorting method to screen out several nodes with the highest equal division as the hub nodes of the brain network, and take 8 according to existing research experience.

[0076] Embodiment 3

[0077] This embodiment proposes to verify the performance of the method for identifying hub nodes in the brain network based on the fusion of static and dynamic features:

[0078] Conduct a comparative analysis with classical traditional methods (such as hub nodes selected based on node controllability, node degree centrality, and betweenness centrality).

[0079] The statistical analysis diagrams of the position distribution of hub nodes identified by this method and three other traditional methods and the corresponding Amyloid protein levels are as Figure 3 shown, Figure 3 Subgraph (a) shows the projection diagrams of a total of 18 hub nodes screened by four hub node identification methods on the brain image, where the dot size represents the frequency of the corresponding node being selected in multiple identification algorithms. From Figure 3 ​The bar chart in sub - figure (b) specifically depicts the frequency of each hub node being selected. Among them, the eight dark - gray bar charts correspond to the eight hub nodes selected by the brain network hub node recognition method of dynamic - static feature fusion (DSFF) proposed in the present invention. Seven of these nodes are among the top eight nodes with the highest frequency, which verifies the rationality of the hub node recognition algorithm proposed in the present invention and shows the effectiveness of the hub nodes identified by this algorithm. In sub - figures (c), (d), (e), and (f) of Figure (3), they respectively depict the performance of amyloid protein pathological burden on the hub nodes identified by four hub node recognition algorithms. The average levels of amyloid protein burden at the four groups of hub nodes identified by the four methods are 0.2533, 0.2968, 0.3269, and 0.3289 respectively. This result is the calculated value after data unified standardization processing. Obviously, the hub nodes identified by the algorithm proposed in the present method have a relatively higher amyloid protein burden level, which is consistent with the pathological mechanism that Alzheimer's disease preferentially attacks hub nodes in the brain network found in current research, and this also proves the effectiveness and rationality of the algorithm proposed in the present method. At the same time, the box plots in sub - figures (c), (d), (e), and (f) of Figure (3) use statistical methods to group and analyze the comparison of amyloid protein distribution between hub nodes (Hub) and non - hub nodes (Non - hub) based on different disease conditions of the samples. The pathological burden levels between the hub node group and the non - hub node group identified by the algorithm proposed in the present method show significant differences in the sample populations of three disease conditions, and the performance in the AD population is also better than that in the NC and MCI populations. While the method based on node controllability shows non - significant differences in the NC and MCI populations and the significance of the differences of the other two comparison algorithms in the AD population relatively decreases. This indicates that the hub nodes identified by the algorithm of the present method have stronger pathological consistency and enlightenment, and can capture the dynamic process of Alzheimer's disease more carefully.

[0080] Each embodiment in the present invention is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For the relevant parts, reference can be made to the partial description of the method embodiment. The device embodiments described above are merely exemplary. The modules described as separate components may or may not be physically separated. When implementing the solution of the present invention, the functions of each module can be realized in the same or multiple software and / or hardware. It is also possible to select some or all of the modules according to actual needs to achieve the purpose of the solution of this embodiment.

[0081] Obviously, the above embodiments of the present invention are merely examples for clearly explaining the present invention, rather than limitations on the implementation manners of the present invention. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to enumerate all the implementation manners here. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the claims of the present invention.

Claims

1. A method for identifying hub nodes in a brain network based on the fusion of dynamic and static features, characterized in that Including: Obtain neuroimaging sample data and preprocess the neuroimaging sample data; Construct a mathematical model of the sample brain network; Calculate the controllability of the brain network nodes according to the mathematical model, and obtain the dynamic characteristics of the brain network nodes according to the controllability of the brain network nodes; Calculate the degree centrality of the brain network nodes, calculate the betweenness centrality of the brain network nodes, and obtain the static characteristics of the brain network nodes according to the degree centrality of the brain network nodes and the betweenness centrality of the brain network nodes; Obtain the dynamic and static fusion characteristics of the brain network nodes of the sample according to the controllability of the brain network nodes, the degree centrality of the brain network nodes, and the betweenness centrality of the brain network nodes; Obtain the node feature vector of the sample according to the dynamic and static fusion characteristics of the brain network nodes of the sample, and obtain the brain network node feature matrix according to the node feature vector of the sample; Obtain the hub attribute score vector of the nodes according to the node feature vector of the sample and the brain network node feature matrix; output the hub nodes according to the hub attribute score vector of the nodes.

2. The method for identifying brain network hub nodes based on the fusion of dynamic and static features according to claim 1, wherein The preprocessing of the neuroimaging sample data is specifically as follows: Step 1: Perform tissue segmentation on the images in the neuroimaging sample data; Step 2: Construct the brain cortical surface and calculate the level of Amyloid protein in each voxel; Step 3: Divide the brain cortical surface into brain regions; Step 4: Generate an anatomical connectivity matrix, and each element of this matrix is measured by the number of fibers between brain regions; Step 5: Calculate the average level of Amyloid protein in each brain region; Step 6: Select the gray matter of the cerebellum as the reference region, normalize the Amyloid protein level of the brain region, and generate the neuropathological load.

3. The method for identifying brain network hub nodes based on the fusion of dynamic and static features according to claim 1, wherein The mathematical model includes an anatomical connectivity matrix and a binary adjacency matrix; the expression of the anatomical connectivity matrix is: Among them, W represents an N×N weighted adjacency matrix, and w ij represents the connection strength between a pair of brain regions (v i , v j ), j represents the j-th node, and i represents the i-th node; The expression of the binary adjacency matrix is: Among them, A represents a binary adjacency matrix, and a ij represents the connection relationship between the i-th node and the j-th node.

4. The method for identifying brain network hub nodes based on the fusion of dynamic and static features according to claim 1, wherein Obtain the measurement index of the dynamic characteristics of the hub nodes according to the controllability of the brain network nodes; its expression is: Among them, represents the Gram matrix, τ represents the time step in the iterative solution of the Gram matrix, T represents the transpose operation of the matrix, W represents the anatomical connectivity matrix, and B represents the input matrix.

5. The method for identifying brain network hub nodes based on the fusion of dynamic and static features according to claim 4, wherein The controllability measure of the $i$-th node in the Gram matrix is 6. The method for identifying hub nodes of a brain network based on the fusion of dynamic and static features according to claim 1, characterized in that The degree centrality of the brain network nodes is the sum of all elements in the i-th row or the i-th column of the binary adjacency matrix; its expression is: Among them, k i represents the degree centrality of the brain network nodes, and a ij represents the binary adjacency matrix. When all the elements on the main diagonal of the binary adjacency matrix have been set to zero, a ii = 0, j represents the j-th node, and i represents the i-th node.

7. The method for identifying hub nodes of a brain network based on the fusion of dynamic and static features according to claim 1, wherein The betweenness centrality of the brain network nodes is the sum of the proportions of the shortest paths passing through node i between all nodes; Its expression is: Among them, C B (i) represents the betweenness centrality of the i-th node, and g hj represents the total number of the shortest paths between the h-th node and the j-th node, represents the number of the shortest paths passing through the i-th node among the above shortest paths.

8. The method for identifying brain network hub nodes based on the fusion of dynamic and static features according to claim 1, wherein Perform hierarchical normalization on the controllability of the brain network nodes, the degree centrality of the brain network nodes, and the betweenness centrality of the brain network nodes to obtain the corresponding characteristics respectively, and perform weighted superposition fusion on the characteristics to obtain the brain network node fusion characteristics; Its expression is: Among them, and respectively represent the measured values after hierarchical normalization of the controllability, degree centrality, and betweenness centrality of the brain network nodes. i = 1, 2, 3, …, N, λ C , λ D and λ B respectively represent the corresponding weight scalar parameters.

9. The method for identifying brain network hub nodes based on the fusion of dynamic and static features according to claim 1, wherein The node feature vector of the sample is represented as Input the node feature vector of the sample into the brain network node feature matrix F = [f 1 , f 2 , f 3 , …, f n , where n represents the total number of samples, and the expanded form of F is shown as the following formula:

10. The method for identifying hub nodes of a brain network based on the fusion of dynamic and static features according to claim 9, wherein, Calculate the hub attribute score vector of each node according to the brain network node feature matrix; the set of the hub attribute score vectors of the nodes is expressed as: s = [s1, s2, s3, …, s N ​ where s represents the set of hub attribute score vectors of the nodes; When the hub attribute score vector of the node exceeds the preset threshold, output all hub nodes that exceed the preset threshold.

Images