Method for analyzing module integration and separation capability of brain network

Abstract

The invention discloses a method for analyzing the module integration and separation capability of a brain network, and the method comprises the following steps: 1, firstly obtaining the magnetic resonance and functional magnetic resonance image data of a corresponding subject, and obtaining all continuous time sequence Bold signals after the preprocessing of a functional network after the preprocessing; step 2, dividing the obtained Bold signal through a sliding window technology; step 3, enabling each time sequence fragment of the acquired Bold signal to pass through a brain template and a Pearson's correlation coefficient so as to construct each independent network structure; 4, connecting the constructed network structures, and performing community detection on the network; and 5, calculating the modularity and the number of the corresponding communities, and then calculating the integration level, the flexibility, the recruitment coefficient and the hybrid coefficient of each node in the change process of each layer. According to the dynamic change process of the brain module, the integration and separation degree of the brain module in the dynamic process can be analyzed.

Description

Technical Field

[0001] The present invention relates to the research field of human brain functional networks based on multi-layer networks, and specifically to a method for analyzing the module integration and separation capabilities of large-scale brain networks. Background Art

[0002] The brain is the most mysterious and complex system in the world. Our current understanding of the brain is only a small part, and there are still many unknown areas. At the same time, the brain is the most important organ of human beings, which can handle advanced functions in cognition, calculation, emotion, perception, language and decision-making. Various complex functions are realized by neurons, neuron clusters, brain regions, etc. connected and interacted through complex networks. Studying the brain region interaction process of the human brain can effectively explore how each brain region cooperates with other brain regions in the process of realizing the function.

[0003] Brain modules refer to clusters of closely connected brain regions in the brain. They are a form of organization of brain structure and function, specializing in processing specific types of information or performing specific functions. They are a relatively independent and complete whole in the brain. Studying the integration and separation of modules in the brain can analyze the changes in nodes in the brain, thereby analyzing the cognitive functions of the human brain.

[0004] The existing integration and separation capabilities of brain functional network modules are mainly divided into two forms. The static analysis method based on the resting state functional network treats the activity of the entire human brain as a static structure and ignores the dynamic changes in the brain function conversion process. The research method based on the maximum entropy principle adopts the random walk process, collects the original characteristic information of the brain, treats the network as a static structure, and obtains the dynamic process of the brain by simulating the changes of the brain through random walks. Summary of the invention

[0005] Therefore, the technical problem to be solved by the present invention is: based on the above-mentioned extraction and research mainly on the static characteristic state of the brain network, the present invention designs a method for analyzing the integration and separation of modules in the dynamic change process of a large-scale brain network. According to the dynamic change process of the brain modules, this application can analyze the degree of integration and separation of the brain modules in the dynamic process.

[0006] The present invention is implemented by constructing a method for analyzing the modular integration and separation capabilities of brain networks, characterized by comprising the following steps:

[0007] Step 1: First, obtain the magnetic resonance imaging (MRI) and functional magnetic resonance imaging (fMRI) image data of the corresponding subjects, and obtain all continuous time series Bold signals after functional network preprocessing through preprocessing;

[0008] Step 2: Divide the acquired Bold signal into multiple time series segments using the sliding window technology;

[0009] Step 3: Each time series segment of the acquired Bold signal is passed through the brain template and Pearson correlation coefficient to construct each individual network structure;

[0010] Step 4: Connect the above-constructed network structures and perform community detection on the network; first, convert each layer of the network into a Newman-Girvan null model, and then use the GenLouvain algorithm to perform community detection and division on the constructed null model;

[0011] Step 5: Perform the following operations on the divided community structure: first calculate the modularity and number of the corresponding communities; then calculate the integration, flexibility, recruitment coefficient and hybrid coefficient of each node in the change process of each layer, and study the changes in the community of each node during the change process.

[0012] According to the method for analyzing the module integration and separation capabilities of a brain network of the present invention, the feature is that the specific preprocessing process of step 1 is as follows:

[0013] (1) Remove the initial time point data. This is because the equipment is unstable or the state of the subjects is unstable during the initial scanning process. Usually, the data of 10 time points are removed;

[0014] (2) Time correction: In the process of data collection, in order to avoid mutual interference of signals, interlayer scanning is usually used for data collection. It is necessary to redefine the time sequence and combine it into a complete data sequence;

[0015] (3) Head correction: During the data collection process, the subjects may move their heads unconsciously, which will reduce the quality of the collected data. This impact on the data needs to be corrected;

[0016] (4) Adjust the orientation: The x, y, and z directions of the image need to be manually adjusted to facilitate subsequent data processing;

[0017] (5) Registration: Mapping structural image data with functional image data to improve the resolution of functional image data;

[0018] (6) Standardization: Due to the diversity of data, all data need to be mapped to the same module space to facilitate subsequent consistent normalization processing;

[0019] (7) Denoising: During the data collection process, there are noises inside the machine, human breathing, heartbeats, etc., which make the data different from the real data and need to be processed.

[0020] According to the method for analyzing the module integration and separation capabilities of a brain network of the present invention, it is characterized in that: Step 2 is specifically as follows;

[0021] In the process of dividing the time series into several time segments through the sliding window technology, it is necessary to define the window size and the distance of each sliding. The sliding window size set in this article is 50, and the distance of each sliding is 10 time points. The number of layers of the constructed network is shown in Formula 1-1;

[0022]

[0023] Where n represents the number of layers in the network, t represents the time length of the entire time series, size represents the size of the sliding window, and l represents the distance of each sliding.

[0024] According to the method for analyzing the modular integration and separation capabilities of a brain network of the present invention, the method for constructing the functional network for each time series in step 3 is as follows:

[0025] (1) Define nodes in the network: Use the ALL90 template to divide the entire brain and divide the entire time series data into 90 time segments of brain regions;

[0026] (2) Define the edges in the network: The edge strength between any two brain region nodes is calculated using the Pearson correlation coefficient. The specific edge strength calculation formula is shown in Formula 1-2. In order to avoid negative values ​​in the connection strength in the network, the absolute value of the coefficient strength needs to be taken after obtaining the correlation coefficient to ensure that all edge strengths are non-negative.

[0027]

[0028] where ρ x,y represents the linear correlation coefficient of the time series of the Bold signal of brain area x and brain area y, with a value between -1 and 1; T represents the number of values ​​in the time series; x t and t represents the Bold signal intensity of the corresponding brain area at time point t; It represents the mean of the time series signal x. The mean of the time series signal y represented by ; when the absolute value of the correlation coefficient is closer to 1, the correlation is stronger, and when it is closer to 0, the correlation is weaker;

[0029] Through the above operations, a network structure with 90 nodes will be constructed for each time segment.

[0030] According to the method for analyzing the module integration and separation capabilities of a brain network of the present invention, the specific steps of community division in step 4 are as follows:

[0031] (1) First, initialize all nodes in the network G and treat each node in the network as a separate community;

[0032] (2) For each node i in the network G, try to remove node i from its community and add it to the community where one of its neighboring nodes j is located, and compare the changes in modularity during this process; if the modularity increases, move node i from the original community to the new community, otherwise it remains unchanged;

[0033] (3) Repeat step 2 until the communities to which all nodes belong no longer change;

[0034] (4) Reconstruct the original network G, and record the new network as G1. First, each community is regarded as each node in the new network, where the weight of the edge within the community is converted into the internal weight of the new node. If there is an edge between two communities, the weight of the edge between the two communities is converted into the weight of the edge between the two nodes in the new network.

[0035] (5) Take the new network G1 obtained in step 4 as the input of step 1, set G = G1 and repeat steps 2 and 3. If the modularity of the network changes, return to step 2 to further reconstruct the network, otherwise end the algorithm;

[0036] The algorithm was repeated 50 times and all the results were normalized.

[0037] According to the method for analyzing the modular integration and separation capabilities of a brain network of the present invention, it is characterized in that the process of calculating the modularity in step 5 is:

[0038] Step 1: First, sort the edges in the network G = (V, E) by simple walks with a maximum length of k (referred to as k-path edges). The calculation formula is shown in Formula 1-7.

[0039]

[0040] Among them, L k (e) refers to the sum of the percentages of times that all possible source nodes s pass through edge e during the traversal process; is the number of K paths starting from node s and passing through edge e, is the number of K paths starting from node s;

[0041] Step 2: The importance of each edge is obtained through the above calculation, and then the proximity of the connection between any two nodes is calculated. The calculation formula is shown in Formula 1-8;

[0042]

[0043] where r ij represents the proximity between node i and node j; d(k) is the degree of the node; n is the number of nodes in the entire network;

[0044] Step 3: Partition the network and then calculate the modularity of the network. The incremental calculation formula for moving a node i to community C is shown in Equation 1-9.

[0045]

[0046] where ∑ C represents the sum of the weights of all internal edges in community C; represents the sum of the weights of all internal edges in the community after the new node i is moved in; k i represents the sum of the weights of the edges connected to node i, k i C represents the sum of the weights of the edges from node i to all nodes in community C, and m is the sum of the weights of all edges in the network.

[0047] According to the present invention, a method for analyzing the module integration and separation capabilities of a brain network is characterized in that: when step 5 is implemented, after the network is community detected according to the previous step, the quality of the community is calculated, and the quality calculation formula of the community is shown in formula 1-10;

[0048]

[0049] Where m is the number of communities divided, l s is the number of edges between nodes in community s, where d s is the sum of the degrees of all nodes in community s; the number of communities calculated is m.

[0050] The present invention has the following advantages: The present invention is an analysis method for the dynamic integration and separation capabilities of brain modules. By analyzing the changes in brain networks over time, the dynamic changes in brain modules are studied, and the degree of integration and separation in the dynamic process of brain modules can be analyzed. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of brain functional network constructed from magnetic resonance data;

[0052] Figure 2 It is a map of the transformation of the brain community of people with autism;

[0053] Figure 3 It is a map of brain community transformation in healthy people;

[0054] Figure 4 It is a schematic diagram of network flexibility;

[0055] Figure 5 It is a schematic diagram of network integration;

[0056] Figure 6 It is a schematic diagram of network recruitment coefficient;

[0057] Figure 7 It is a schematic diagram of network hybrid coefficient;

[0058] Figure 8 This is a comparison chart of flexibility between healthy people and autistic patients;

[0059] Fig. 9 This is a comparison chart of the integration level of healthy people and autistic patients;

[0060] Fig.10 It is a schematic diagram of recruitment coefficients for healthy people and autistic patients;

[0061] Fig.11 This is a schematic diagram of the confounding coefficient between healthy people and autistic patients;

[0062] Fig.12 It is a schematic diagram of dividing the entire Bold signal time period into multiple time series segments. DETAILED DESCRIPTION

[0063] The following will be combined with the attached Figure 1-Figure 12 The present invention is described in detail, and the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] The present invention provides a method for analyzing the module integration and separation capabilities of a large-scale brain network, comprising the following steps:

[0065] Step 1: First, obtain the corresponding subject's magnetic resonance imaging (MRI) and functional magnetic resonance imaging (fMRI) image data, and obtain all continuous time series Bold signals after functional network preprocessing through preprocessing; compared with directly using functional magnetic resonance imaging data, the purpose of using magnetic resonance imaging data to align functional magnetic resonance imaging data is to improve the resolution of image segmentation and facilitate subsequent data extraction and segmentation. The specific preprocessing process is as follows:

[0066] (1) Remove the initial time point data. This is because the equipment is unstable or the state of the subjects is unstable during the initial scanning process. Usually, the data of 10 time points are removed;

[0067] (2) Time correction: In the process of data collection, in order to avoid mutual interference of signals, interlayer scanning is usually used for data collection. It is necessary to redefine the time sequence and combine it into a complete data sequence;

[0068] (3) Head correction: During the data collection process, the subjects may move their heads unconsciously, which will reduce the quality of the collected data. This impact on the data needs to be corrected;

[0069] (4) Adjust the orientation: The x, y, and z directions of the image need to be manually adjusted to facilitate subsequent data processing;

[0070] (5) Registration: Mapping structural image data with functional image data to improve the resolution of functional image data;

[0071] (6) Standardization: Due to the diversity of data, all data need to be mapped to the same module space to facilitate subsequent consistent normalization processing;

[0072] (7) Denoising: During the data collection process, there are noises inside the machine, human breathing, heartbeats, etc., which make the data different from the real data and need to be processed.

[0073] Step 2: Divide the acquired Bold signal by sliding window technology, and divide the entire Bold signal time period into multiple time series segments, such as Fig.12 shown.

[0074] In the process of dividing the time series into several time segments through the sliding window technology, it is necessary to define the window size and the distance of each sliding. The sliding window size set in this article is 50, and the distance of each sliding is 10 time points. The number of layers of the constructed network is shown in Formula 1-1.

[0075]

[0076] Where n represents the number of layers in the network, t represents the time length of the entire time series, size represents the size of the sliding window, and l represents the distance of each sliding;

[0077] Step 3: Construct a separate network structure for each time series segment of the acquired Bold signal through the brain template and Pearson correlation coefficient; the functional network construction method for each time series is as follows:

[0078] (1) Define nodes in the network: Use the ALL90 template to divide the entire brain and divide the entire time series data into 90 time segments of brain regions.

[0079] (2) Define the edges in the network: The Pearson correlation coefficient is used to calculate the edge strength between any two brain region nodes. The specific edge strength calculation formula is shown in Formula 1-2. In order to avoid negative values ​​in the connection strength in the network, after obtaining the correlation coefficient, the absolute value of the coefficient strength needs to be taken to ensure that all edge strengths are non-negative.

[0080]

[0081] where ρ x,y represents the linear correlation coefficient of the time series of the Bold signal of brain area x and brain area y, with a value between -1 and 1; T represents the number of values ​​in the time series; x t and t represents the Bold signal intensity of the corresponding brain area at time point t; It represents the mean of the time series signal x. The mean of the time series signal y. When the absolute value of the correlation coefficient is closer to 1, the correlation is stronger, and when it is closer to 0, the correlation is weaker.

[0082] Through the above operations, a network structure with 90 nodes will be constructed for each time segment.

[0083] Step 4: Connect the above constructed network structures and perform community detection on the network. First, convert each layer of the network into a Newman-Girvan null model, and then use the GenLouvain algorithm to perform community detection and division on the constructed null model;

[0084] First, define the structure of the entire multi-layer network. Assume that there is a multi-layer network with K layers, which can be expressed mathematically by the hyper-adjacency matrix G = (A, O). [1] ,A [2] ,...,A [K]}, used to represent the set of adjacency matrices of each layer in the multi-layer network, and then the αth layer can be obtained through A [α] = {V [α] ,E [α]}, where V [α] Represents the set of all nodes in the αth layer (the set is represented by represents node i), E [α] represents the set of all the edges in the αth layer (the edge between node i and node j is denoted as ). When there is an edge between node i and node j in the αth layer, If there is no edge connection, then The inter-layer network adjacency matrix can be obtained by O = {O [α,β] |α,β∈{1,2,3...,K} and α≠β}, the mathematical expression of the adjacency matrix between the αth layer and the βth layer is expressed by O [α,β] = {V [α] ,V [β] ,E [α,β]} is represented by [α] 、V [β] Represent the set of nodes in the αth layer and the βth layer respectively. The edge relationship between the two layers is expressed as And i,j∈{1,2,...} and α≠β}, where when there is an edge between node i in the αth layer and node j in the βth layer, If there is no edge connection, then The expression of the super adjacency matrix of a multi-layer network is shown in Formula 1-3.

[0085]

[0086] There are many ways to connect each layer of the network. This article uses the hidden Markov model to build network connections. Based on the model theory, each state is only related to the previous state, so the relationship between layers during the connection process can be defined as shown in formula 1-4.

[0087]

[0088] For the node connections in adjacent time series, the connections between the same brain regions are recorded as e, and the connections between different brain regions are recorded as e1. In the preprocessing process, the covariate correlation is removed. It is assumed that the changes in the brain region at the current time point are only related to the same brain region at the previous time point, and have no correlation with other brain regions. This assumption is related to the parameters when constructing the model later. Then e=1 and e1=0. Substituting e and e1 into the matrix, it can be seen that T2=E 90 , the matrix expression is shown in formula 1-5.

[0089]

[0090] Note A [i] A i , then the super-adjacency matrix of the fMRI-based temporal functional network is expressed as shown in Formula 1-6.

[0091]

[0092] Set the inter-layer resolution gamma to 1, the inter-layer coupling degree omega to 1, and then use the Newman-Girvan algorithm to calculate the network and convert the entire network into a zero model. The algorithm mainly randomizes the original graph G and generates edges by random rewiring. The expected degree of each vertex in the generated new graph is the same as the expectation in the original graph G. The matrix representation of the constructed zero model network is a symmetric, semi-positive matrix. The network is used as the input of the subsequent community. The specific steps of community division are as follows:

[0093] (1) First, initialize all nodes in the network G and treat each node in the network as a separate community;

[0094] (2) For each node i in the network G, try to remove node i from its community and add it to the community where one of its neighboring nodes j is located, and compare the changes in modularity during this process. If the modularity increases, node i is moved from the original community to the new community, otherwise it remains unchanged;

[0095] (3) Repeat step 2 until the communities to which all nodes belong no longer change;

[0096] (4) Reconstruct the original network G, and record the new network as G1. First, each community is regarded as each node in the new network, where the weight of the edge within the community is converted into the internal weight of the new node. If there is an edge between two communities, the weight of the edge between the two communities is converted into the weight of the edge between the two nodes in the new network.

[0097] (5) Take the new network G1 obtained in step 4 as the input of step 1, set G = G1 and then repeat steps 2 and 3. If the modularity of the network changes, return to step 2 to further reconstruct the network, otherwise terminate the algorithm.

[0098] The algorithm was repeated 50 times and all the results were normalized.

[0099] The following describes the process of modularity calculation:

[0100] Step 1: First, sort the edges in the network G = (V, E) by simple walks with a maximum length of k (denoted as k-path edge). The calculation formula is shown in Formula 1-7.

[0101]

[0102] Among them, L k (e) refers to the sum of the percentages of times that all possible source nodes s pass through edge e during the traversal process; is the number of K paths starting from node s and passing through edge e, is the number of K paths starting from node s.

[0103] Step 2: The importance of each edge is obtained through the above calculation, and then the proximity of the connection between any two nodes is calculated. The calculation formula is shown in Formula 1-8.

[0104]

[0105] where r ij represents the proximity between node i and node j; d(k) is the degree of the node; n is the number of nodes in the entire network.

[0106] Step 3: Partition the network and then calculate the modularity of the network. The incremental calculation formula for moving a node i to community C is shown in Equation 1-9.

[0107]

[0108] where ∑ C represents the sum of the weights of all internal edges in community C; represents the sum of the weights of all internal edges in the community after the new node i is moved in; k i represents the sum of the weights of the edges connected to node i, k i C represents the sum of the weights of the edges from node i to all nodes in community C, and m is the sum of the weights of all edges in the network.

[0109] Step 5: Perform the following operations on the divided community structure: first calculate the modularity and number of the corresponding communities; then calculate the integration, flexibility, recruitment coefficient and hybrid coefficient of each node in the change process of each layer, and study the changes of each node in the process of change.

[0110] After performing community detection on the network according to the previous step, the community quality calculation formula is shown in Formula 1-10.

[0111]

[0112] Where m is the number of communities divided, l s is the number of edges between nodes in community s, where d s is the sum of the degrees of all nodes in community s. The number of communities calculated is m.

[0113] The module allegiance matrix is ​​a matrix used to describe the dynamic reconstruction of brain network nodes. It describes the degree of node consistency from a multi-level network perspective and is often used in temporal network structures where the number of network nodes does not change. The matrix is ​​specifically expressed as a square matrix of size N×N, where N is the number of nodes in the network. The element P in the matrix ij It indicates the proportion of network layers in which nodes i and j in the network belong to the same community, reflecting the consistency of community assignments of different nodes. ij Calculated using formula 1-11.

[0114]

[0115] Where O represents the number of repetitions of the multi-layer network community detection algorithm, and T represents the number of network layers of the multi-layer temporal network. The calculation formula for each community detection and the corresponding layer is shown in 1-12:

[0116]

[0117] Flexibility is used to measure the probability of changes in the community distribution of each node in the network over time. The indicator is often used to describe the changes in nodes in a temporal network and can be used to evaluate the frequency of changes in brain regions in a multi-layer temporal brain network. The more flexible the node, the more active the node interaction is during the network evolution, and the more important the information interaction is in the system. The flexibility calculation of node i in the network is shown in formula 1-13. M represents the number of layers in the temporal network, g represents the number of layers in the temporal network, and g represents the number of layers in the temporal network. i Indicates the number of times the node changes in the sequential network.

[0118]

[0119] Integration is calculated based on the module consistency matrix. This index quantifies the probability that a node in the network is in the same community as other nodes at different network levels. It is often used to study dynamic brain networks. The network nodes with higher integration are more likely to become core nodes for information exchange between different communities, which has an important impact on the interaction of global information in the network, the evolution of the network, and the stability of the network structure. The integration calculation formula for node i in the network is shown in formula 1-14.

[0120]

[0121] Where n represents the number of nodes in the network, and S represents a layer of the network.

[0122] The calculation of the recruitment coefficient is also based on the module consistency matrix. This indicator is used to describe the probability that nodes in the network are located in the same community module. It is a reference indicator for describing the degree of brain network reconstruction. The size of the recruitment coefficient is related to the quality and formation of the community. The recruitment coefficient of network S can be calculated using formula 1-15:

[0123]

[0124] Promiscuity refers to the ratio of the change in the community that the nodes in the network participate in during the time series change of the network to all the communities that appear in the network during the whole process. This indicator is used to describe the degree to which the node participates in the change of the community over time. The calculation of the promiscuity of node i is shown in formula 1-16.

[0125]

[0126] where k i represents the number of different communities that node i participates in over time, and all represents the number of communities divided by the entire network. i The value range of is 0 to 1, so the initial community of the node is generally not considered. If the community where node i is located does not change during the whole process, then p i The value is 0. If all communities are participated, then p i The value is 1.

[0127] Application Description:

[0128] The present invention proposes an analytical method for the brain's integration and separation capabilities, which can effectively have important applications in neuroscience and clinical research.

[0129] 1. This invention can study the evolution of brain modules in dynamic time series, provide theoretical guidance for further use in building brain function evolution models, explore the effects of different connection patterns on the integration and separation capabilities of brain modules, and provide guidance for new experiments.

[0130] 2. By analyzing the resting-state functional connectivity patterns between different brain regions, the integration and separation patterns of the brain can be revealed, and the degree of integration (the strength of connections within the network) and separation (the weakness of connections between modules) can be assessed.

[0131] 3. By comparing the functional connectivity patterns during task activation with those during resting state, the plasticity and information processing strategies of brain networks under task conditions can be assessed.

[0132] 4. By comparing the changes in integration and separation abilities of various groups of people with those of healthy people under the same state, it is possible to assess whether the group has a decline in cognitive function and provide further treatment; or by analyzing the changes in the integration and separation abilities of the brain modules of the subjects after taking a certain drug or receiving a certain treatment, the effectiveness and degree of effectiveness of the drug or treatment can be assessed.

[0133] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for analyzing the modular integration and separation capabilities of brain networks, characterized in that; The following steps are involved: Step 1: First, obtain the magnetic resonance imaging (MRI) and functional magnetic resonance imaging (fMRI) image data of the corresponding subjects, and obtain all continuous time series Bold signals after functional network preprocessing through preprocessing; Step 2: Divide the acquired Bold signal into multiple time series segments using the sliding window technology; Step 3: Each time series segment of the acquired Bold signal is passed through the brain template and Pearson correlation coefficient to construct each individual network structure; Step 4: Connect the above-constructed network structures and perform community detection on the network; first, convert each layer of the network into a Newman-Girvan null model, and then use the GenLouvain algorithm to perform community detection and division on the constructed null model; Step 5: Perform the following operations on the divided community structure: first calculate the modularity and number of the corresponding communities; then calculate the integration, flexibility, recruitment coefficient and hybrid coefficient of each node in the change process of each layer, and study the changes in the community of each node during the change process.

2. A method for analyzing the modular integration and separation capabilities of a brain network according to claim 1, characterized in that; Step 1 The specific preprocessing process is as follows: (1) Remove the initial time point data. This is because the equipment is unstable or the state of the subjects is unstable during the initial scanning process. Usually, the data of 10 time points are removed; (2) Time correction: In the process of data collection, in order to avoid mutual interference of signals, interlayer scanning is usually used for data collection. It is necessary to redefine the time sequence and combine it into a complete data sequence; (3) Head correction: During the data collection process, the subjects may move their heads unconsciously, which will reduce the quality of the collected data. This impact on the data needs to be corrected; (4) Adjust the orientation: The x, y, and z directions of the image need to be manually adjusted to facilitate subsequent data processing; (5) Registration: Mapping structural image data with functional image data to improve the resolution of functional image data; (6) Standardization: Due to the diversity of data, all data need to be mapped to the same module space to facilitate subsequent consistent normalization processing; (7) Denoising: During the data collection process, there are noises inside the machine, human breathing, heartbeats, etc., which make the data different from the real data and need to be processed.

3. A method for analyzing the modular integration and separation capabilities of a brain network according to claim 1, characterized in that; The details of step 2 are as follows: In the process of dividing the time series into several time segments through the sliding window technology, it is necessary to define the window size and the distance of each sliding. The sliding window size set in this article is 50, and the distance of each sliding is 10 time points. The number of layers of the constructed network is shown in Formula 1-1; Where n represents the number of layers in the network, t represents the time length of the entire time series, size represents the size of the sliding window, and l represents the distance of each sliding.

4. A method for analyzing the modular integration and separation capabilities of a brain network according to claim 1, characterized in that; In step 3, the functional network for each time series is constructed as follows: (1) Define nodes in the network: Use the ALL90 template to divide the entire brain and divide the entire time series data into 90 time segments of brain regions; (2) Define the edges in the network: The edge strength between any two brain region nodes is calculated using the Pearson correlation coefficient. The specific edge strength calculation formula is shown in Formula 1-2. In order to avoid negative values ​​in the connection strength in the network, the absolute value of the coefficient strength needs to be taken after obtaining the correlation coefficient to ensure that all edge strengths are non-negative. where ρ x,y represents the linear correlation coefficient of the time series of the Bold signal of brain area x and brain area y, with a value between -1 and 1; T represents the number of values ​​in the time series; x t and t represents the Bold signal intensity of the corresponding brain area at time point t; It represents the mean of the time series signal x. The mean of the time series signal y represented by ; when the absolute value of the correlation coefficient is closer to 1, the correlation is stronger, and when it is closer to 0, the correlation is weaker; Through the above operations, a network structure with 90 nodes will be constructed for each time segment.

5. The method for analyzing the modular integration and separation capabilities of a brain network according to claim 1, characterized in that: The specific steps of community division in step 4 are as follows: (1) First, initialize all nodes in the network G and treat each node in the network as a separate community; (2) For each node i in the network G, try to remove node i from its community and add it to the community where one of its neighboring nodes j is located, and compare the changes in modularity during this process; if the modularity increases, move node i from the original community to the new community, otherwise it remains unchanged; (3) Repeat step 2 until the communities to which all nodes belong no longer change; (4) Reconstruct the original network G, and record the new network as G1. First, each community is regarded as each node in the new network, where the weight of the edge within the community is converted into the internal weight of the new node. If there is an edge between two communities, the weight of the edge between the two communities is converted into the weight of the edge between the two nodes in the new network. (5) Take the new network G1 obtained in step 4 as the input of step 1, set G = G1 and repeat steps 2 and 3. If the modularity of the network changes, return to step 2 to further reconstruct the network, otherwise end the algorithm; The algorithm was repeated 50 times and all the results were normalized.

6. A method for analyzing the modular integration and separation capabilities of a brain network according to claim 1, characterized in that; The process of modularity calculation in step 5 is: Step 1: First, sort the edges in the network G = (V, E) by simple walks with a maximum length of k (referred to as k-pathedge). The calculation formula is shown in Formula 1-7. Among them, L k (e) refers to the sum of the percentages of times that all possible source nodes s pass through edge e during the traversal process; is the number of K paths starting from node s and passing through edge e, is the number of K paths starting from node s; Step 2: The importance of each edge is obtained through the above calculation, and then the proximity of the connection between any two nodes is calculated. The calculation formula is shown in Formula 1-8; where r ij represents the proximity between node i and node j; d(k) is the degree of the node; n is the number of nodes in the entire network; Step 3: Partition the network and then calculate the modularity of the network. The incremental calculation formula for moving a node i to community C is shown in Equation 1-9. where ∑ C represents the sum of the weights of all internal edges in community C; represents the sum of the weights of all internal edges in the community after the new node i is moved in; k i represents the sum of the weights of the edges connected to node i, k i C represents the sum of the weights of the edges from node i to all nodes in community C, and m is the sum of the weights of all edges in the network.

7. A method for analyzing the modular integration and separation capabilities of a brain network according to claim 1, characterized in that; When implementing step 5, after the network is community detected according to the previous step, the quality of the community is calculated. The quality calculation formula of the community is shown in formula 1-10; Where m is the number of communities divided, l s is the number of edges between nodes in community s, where d s is the sum of the degrees of all nodes in community s; the number of communities calculated is m.

Images