A brain region causal relationship identification method and system based on electroencephalogram data
By constructing a causal framework map of brain regions and a nonlinear function model, and combining independence tests and the maximum clique algorithm, the problem of inaccurate causal relationship identification in existing technologies is solved, and more accurate causal relationship identification and structural understanding of brain regions are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-28
- Publication Date
- 2026-04-07
AI Technical Summary
Existing brain region structure learning methods cannot accurately distinguish between causal and correlational relationships, cannot construct asymmetric causal connection patterns between brain regions, and cannot identify nonlinear effects, resulting in inaccurate brain region structure learning results.
Based on EEG data, a causal relationship skeleton map between brain regions is constructed, a nonlinear function model is established, and the direction of binary and multivariate causal relationships is obtained through independence test analysis. The causal relationship skeleton map is decomposed using the maximum clique algorithm to identify causal relationships between brain regions.
It can more accurately reflect the true working mechanism of the brain, identify causal relationships between brain regions, eliminate redundant relationships, and provide a more reliable understanding of brain region structure and a basis for intervention.
Smart Images
Figure CN116965834B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of brain region structure reasoning, and more specifically, to a method and system for identifying brain region causal relationships based on electroencephalogram (EEG) data. Background Technology
[0002] Existing methods for learning brain region structures include correlation analysis and dynamic Bayesian network learning. These methods have achieved certain results in the field of brain region structure learning, but they still have some shortcomings. Most of these methods infer brain region structures based on linear relationships, but the actual brain region connections often have nonlinear effects. At the same time, symmetrical correlation analysis may not be able to reflect the differences in the actual working mechanisms of the brain and cannot correctly construct asymmetrical causal connection patterns between brain regions. Furthermore, since traditional techniques cannot distinguish between causal relationships and correlation relationships, they cannot guarantee the accuracy of brain region structure learning results.
[0003] Existing technology provides a method for analyzing effective connectivity in brain function. This method addresses the errors that arise in existing brain functional imaging techniques, where the conditional Granger causality method is used to study effective connectivity between regions of interest in functional magnetic resonance imaging (fMRI) and electroencephalogram (EEG) data. These errors are caused by high redundancy among multiple signals. Furthermore, this method fails to identify nonlinear effects in brain region connectivity and cannot distinguish between causal and correlational relationships, thus failing to construct correct asymmetric causal connectivity patterns. Summary of the Invention
[0004] To overcome the inaccuracy of structural learning results in brain region structure estimation in the prior art, this invention provides a brain region causal relationship identification method and system based on EEG data, which finds a more accurate brain region structure by reflecting the differences in brain working mechanisms among different subjects.
[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:
[0006] This invention provides a method for identifying causal relationships in brain regions based on electroencephalogram (EEG) data, the method comprising:
[0007] S1: Acquire EEG data from different brain regions;
[0008] S2: Construct a causal relationship framework diagram between brain regions based on EEG data from different brain regions;
[0009] S3: Based on the causal relationship skeleton diagram, establish a nonlinear function model between different brain regions;
[0010] S4: Based on a nonlinear function model, independence tests are performed on all brain regions to obtain the direction of binary causal relationships between brain regions;
[0011] S5: Determine whether a binary causal relationship exists in all brain regions; if not, proceed to step S6; otherwise, proceed to step S8.
[0012] S6: Perform structural decomposition on the causal relationship skeleton diagram to obtain a local structure diagram;
[0013] S7: Based on the local structure map, independence test analysis is performed on brain regions that do not have binary causal relationship directions to obtain the multivariate causal relationship directions between brain regions that do not have binary causal relationship directions;
[0014] S8: Obtain the recognition result of the causal relationship direction of the brain region based on the binary causal relationship direction of the brain region or the multivariate causal relationship direction of the brain region.
[0015] Preferably, in step S1, the subject's electroencephalogram (EEG) data is acquired using magnetic resonance imaging (MRI) technology, and the blood oxygen level-dependent signals of different brain regions of the subject are sampled to acquire EEG data of different brain regions.
[0016] Preferably, the specific method of S2 is as follows: the fully connected relationships between different brain regions are represented as a completely undirected graph. Based on the structure of the completely undirected graph, the independence relationships implied in the EEG data between any two brain regions are searched and traversed. If the independence relationship test result between the EEG data of any two brain regions is valid, the connection between the corresponding brain regions is removed; otherwise, the connection between the corresponding brain regions is retained, thus completing the construction of the causal relationship skeleton graph between brain regions. The causal relationship skeleton between brain regions learned in this stage can reduce the search space for further causal relationship discovery in subsequent stages.
[0017] Preferably, based on the causal relationship framework between brain regions learned by S2, a nonlinear function model is established for different brain regions by assuming that the EEG data under the real brain region structure conforms to the generation assumptions of nonlinear additive functions and independent noise effects:
[0018]
[0019] in, and These represent the EEG data of the i-th and j-th brain regions, respectively. This represents the EEG data of the k-th brain region. and Representing the i-th brain region respectively The observable direct parent set and the latent variable direct parent set, This represents a third-order differentiable nonlinear function mapping of observable variables from the j-th brain region to the i-th brain region. This represents a third-order differentiable nonlinear function mapping of latent variables from the k-th brain region to the i-th brain region. This indicates that the i-th brain region is affected. Random noise, and the random noise in different brain regions is independent of each other.
[0020] Preferably, the specific method of S4 is as follows: for any two different brain regions that have a connection... and Let's assume the causal direction is... and ;right and Bivariate nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and In the causal direction Below, the residuals of EEG data and brain region variables The EEG data were analyzed using a binary independence test; in the case of opposite causal relationships... Below, the residuals of EEG data and brain regions The EEG data were analyzed using a binary independence test; specifically:
[0021] If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the causal direction of the brain region is accepted. This represents the direction of binary causal relationships within brain regions;
[0022] If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the causal direction of the brain region is accepted. This represents the direction of a binary causal relationship within a brain region.
[0023] If the independence test shows that the causal direction is... superior, and Independent, and simultaneously in the opposite causal direction superior, and If it is also independent, then it is believed that the current brain region's EEG data does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis;
[0024] If the independence test shows that the causal direction is... superior, and Not independent, and simultaneously in opposite causal directions. superior, and If it is not independent, then it is considered that the brain region and There is no binary causal relationship direction.
[0025] Preferably, the independence test analysis between the regression residuals of EEG data and the EEG data of the brain region hypothesized as the cause, and the asymmetry existing in the nonlinear causal direction, helps to identify the true brain region structural orientation from the observation data, specifically:
[0026]
[0027] in, Representing the brain region variable of the outcome The latent variable is the nonlinear effect of direct parents. Variables representing the cause of brain regions The latent variable is the nonlinear effect of direct parents. Representing the brain region variable of the outcome All observable direct parents, Representing the brain region variable of the outcome All observable direct parents. The role of nonlinear regression and independence tests based on the generation hypothesis of nonlinear additive functions and independent noise effects provides a strong guarantee for discovering causal relationships in brain region structures from EEG data using a classical structural causal function model. Simultaneously, thanks to the assistance of the brain region structural causal framework in S2, it is possible to more efficiently and accurately infer causal relationships between different brain regions, thereby attempting to locate unobserved brain regions that cause the influence of latent variables, providing a foundation for more robust latent variable causal discovery.
[0028] Preferably, the dependency edges in the brain region causal framework in S2 reflect both the real causal relationships between different brain regions and the spurious relationships formed by the influence of latent variables from unobserved brain region variables. The latter's generation mechanism tends to simultaneously form maximal clique structures in the graphical model, thus providing location information for unobserved brain region variables. Therefore, based on the brain region causal framework, the maximal clique algorithm is used to decompose the framework, further locating unobserved brain region variables while locking down local structures where the direction of causal relationships cannot be determined. The role of maximal clique decomposition based on the brain region causal framework is to avoid a global search of the brain region causal framework.
[0029] Preferably, the algorithm in S2 is reused on the current local graph structure using all the decomposed local graph structures, and the existing causal relationships are taken into consideration. The bivariate nonlinear regression is extended to multivariate nonlinear regression. Based on the analysis logic of the S2 stage, the independence test analysis is performed on the EEG data residuals and the EEG data in the brain regions assumed to be the cause under the current mechanism.
[0030] Preferably, in step S4, the ANM method is used to determine the connection between any two different brain regions that are connected. and Let's assume the causal direction is... and ;right and Nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and ;right and Nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and In the causal direction Below, the residuals of EEG data and brain region variables The EEG data were analyzed using a binary independence test; in the case of opposite causal relationships... Below, the residuals of EEG data and brain regions The EEG data were analyzed using a binary independence test.
[0031] Preferably, the causal relationship skeleton graph is decomposed using the maximum clique algorithm based on the brain region causal skeleton graph to obtain a local structure graph.
[0032] Preferably, the maximum clique algorithm is the Bron–Kerbosch maximum clique decomposition algorithm.
[0033] Preferably, the specific method of S7 is as follows: for those in the causal direction superior, and Not independent, in the opposite causal direction superior and brain region variables brain regions that are not independent and Perform a multivariate independence test analysis; assume the causal direction is respectively and ,right and Multivariate nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and Specifically:
[0034] If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the multiple regression analysis is considered to accept the current causal direction of the brain region. This indicates a direction of multiple causal relationships, and it is inferred that no causal direction exists. ;
[0035] If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the multiple regression analysis is considered to accept the current causal direction of the brain region. This indicates a direction of multiple causal relationships, and it is inferred that no causal direction exists. ;
[0036] If the independence test shows that the causal direction is... superior, and Independent, and simultaneously in the opposite causal direction. and If it is also independent, then it is believed that the EEG data under the current brain region structure does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis;
[0037] If the independence test shows that the causal direction is... superior, and Not independent, and simultaneously in the opposite causal direction. and If it is not independent, then it is assumed that the EEG data under the current brain region structure does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis.
[0038] Preferably, the role of multivariate nonlinear regression is to eliminate the causal nonlinear dependence of other brain regions on the current local graph structure, thus providing a guarantee for the identification of the remaining multivariate causal relationship directions.
[0039] The present invention also provides a brain region causal relationship recognition system based on electroencephalogram (EEG) data for implementing the above-described method, the system comprising:
[0040] The data acquisition module is used to acquire EEG data from different brain regions;
[0041] The skeleton construction module is used to construct a skeleton diagram of causal relationships between brain regions based on EEG data from different brain regions;
[0042] The function model building module is used to build nonlinear function models between different brain regions based on the causal relationship skeleton diagram;
[0043] The binary causal relationship direction determination module is used to perform independent testing and analysis on all brain regions to obtain the binary causal relationship direction between brain regions;
[0044] The judgment module is used to determine whether a binary causal relationship exists in all brain regions; if not, it proceeds to the structure decomposition module; otherwise, it proceeds to the recognition module.
[0045] The structural decomposition module is used to decompose the causal relationship skeleton diagram to obtain a local structural diagram.
[0046] The module for determining the direction of multiple causal relationships, based on local structural maps, performs independence tests and analyses on brain regions that do not have directions of binary causal relationships, thereby obtaining the direction of multiple causal relationships in brain regions that do not have directions of binary causal relationships.
[0047] The recognition module is used to obtain the recognition results of the causal relationship direction of brain regions based on the binary causal relationship direction or the multivariate causal relationship direction of brain regions.
[0048] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:
[0049] The data foundation of this invention is EEG data between different brain regions. By constructing a causal structure network of brain regions, it can better reflect the true working mechanism of the brain. The conditional independence test method is used to obtain the causal skeleton of brain regions, remove redundant causal relationships between brain regions, and then use the properties of nonlinear functions to determine the causal direction between given brain regions, providing an efficient and powerful explanation for the real causal function influence mechanism between brain regions. Based on the decomposition of the causal skeleton of brain regions, local results are obtained, and unobserved brain region variables are distinguished and located, resulting in a more accurate brain region structure, which provides strong support for subsequent understanding and intervention of brain mechanisms. Attached Figure Description
[0050] Figure 1 This is a flowchart of a brain region causal relationship identification method based on electroencephalogram (EEG) data as described in Example 1.
[0051] Figure 2 This is a structural diagram of a brain region causal relationship recognition system based on electroencephalogram (EEG) data, as described in Example 3. Detailed Implementation
[0052] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent.
[0053] To better illustrate this embodiment, some parts in the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions;
[0054] It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings.
[0055] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.
[0056] Example 1
[0057] This embodiment provides a method for identifying causal relationships in brain regions based on electroencephalogram (EEG) data, such as... Figure 1 As shown, the method includes:
[0058] S1: Acquire EEG data from different brain regions;
[0059] S2: Construct a causal relationship framework diagram between brain regions based on EEG data from different brain regions;
[0060] S3: Based on the causal relationship skeleton diagram, establish a nonlinear function model between different brain regions;
[0061] S4: Based on a nonlinear function model, independence tests are performed on all brain regions to obtain the direction of binary causal relationships between brain regions;
[0062] S5: Determine whether a binary causal relationship exists in all brain regions; if not, proceed to step S6; otherwise, proceed to step S8.
[0063] S6: Perform structural decomposition on the causal relationship skeleton diagram to obtain a local structure diagram;
[0064] S7: Based on the local structure map, independence test analysis is performed on brain regions that do not have binary causal relationship directions to obtain the multivariate causal relationship directions between brain regions that do not have binary causal relationship directions;
[0065] S8: Obtain the recognition result of the causal relationship direction of the brain region based on the binary causal relationship direction of the brain region or the multivariate causal relationship direction of the brain region.
[0066] A causal relationship identification method based on electroencephalogram (EEG) data includes acquiring EEG data from different brain regions, constructing a causal relationship skeleton map between brain regions, establishing a nonlinear function model between different brain regions, performing independence test analysis on all brain regions to obtain the binary causal relationship direction between brain regions, determining whether a binary causal relationship direction exists in all brain regions, and if not, performing structural decomposition on the causal relationship skeleton map to obtain a local structure map, performing independence test analysis on brain regions without a binary causal relationship direction to locate latent variables and identify the multivariate causal relationship direction of brain regions without a binary causal relationship direction, and finally obtaining the identification result of the causal relationship direction of brain regions based on the binary causal relationship direction or the multivariate causal relationship direction of brain regions.
[0067] Example 2
[0068] This embodiment provides a method for identifying causal relationships in brain regions based on electroencephalogram (EEG) data, such as... Figure 1 As shown, the method includes:
[0069] S1: Acquire EEG data from different brain regions;
[0070] S2: Construct a causal relationship framework diagram between brain regions based on EEG data from different brain regions;
[0071] S3: Based on the causal relationship skeleton diagram, establish a nonlinear function model between different brain regions;
[0072] S4: Based on a nonlinear function model, independence tests are performed on all brain regions to obtain the direction of binary causal relationships between brain regions;
[0073] S5: Determine whether a binary causal relationship exists in all brain regions; if not, proceed to step S6; otherwise, proceed to step S8.
[0074] S6: Perform structural decomposition on the causal relationship skeleton diagram to obtain a local structure diagram;
[0075] S7: Based on the local structure map, independence test analysis is performed on brain regions that do not have binary causal relationship directions to obtain the multivariate causal relationship directions between brain regions that do not have binary causal relationship directions;
[0076] S8: Obtain the recognition result of the causal relationship direction of the brain region based on the binary causal relationship direction of the brain region or the multivariate causal relationship direction of the brain region.
[0077] A causal relationship identification method based on electroencephalogram (EEG) data includes acquiring EEG data from different brain regions, constructing a causal relationship skeleton map between brain regions, establishing a nonlinear function model between different brain regions, performing independence test analysis on all brain regions to obtain the binary causal relationship direction between brain regions, determining whether a binary causal relationship direction exists in all brain regions, and if not, performing structural decomposition on the causal relationship skeleton map to obtain a local structure map, performing independence test analysis on brain regions without a binary causal relationship direction to locate latent variables and identify the multivariate causal relationship direction of brain regions without a binary causal relationship direction, and finally obtaining the identification result of the causal relationship direction of brain regions based on the binary causal relationship direction or the multivariate causal relationship direction of brain regions.
[0078] fMRI was used to acquire electroencephalogram (EEG) data from subjects. Blood oxygenation level-dependent signals from different brain regions were sampled, and the EEG data were assumed to follow an independent and identically distributed (ICD) model. The EEG data were randomly ordered to eliminate prior dependence on temporal order, providing a reasonable data foundation for subsequent causal relationship analysis of brain regions.
[0079] A PC algorithm based on conditional independence testing is used to learn the causal framework between brain regions from observed EEG data. The PC algorithm represents the fully connected relationships between different brain regions as a completely undirected graph and searches and traverses the (conditional) independence relationships implied in the EEG data between any two brain regions based on the graph structure. If the independence test is accepted, the edges between the corresponding brain regions are removed. In actual research, the actual structural connections of brain regions are often sparse. Therefore, the causal framework between brain regions learned at this stage can reduce the search space for further causal relationship discovery in subsequent stages.
[0080] Based on the causal framework between brain regions learned by S2, and by assuming that EEG data under real brain region structures conforms to the generation assumptions of nonlinear additive functions and independent noise influence, and by assuming the causal direction between any two connected brain regions, the asymmetric relationship of causal functions between brain regions can be analyzed using an ANM model based on nonlinear causal functions. Specifically, by assuming that EEG data under the anterior brain region structures conforms to the generation assumptions of nonlinear additive functions and independent noise influence, a nonlinear function model is established for different brain regions.
[0081]
[0082] in, and These represent the EEG data of the i-th and j-th brain regions, respectively. This represents the EEG data of the k-th brain region. and Representing the i-th brain region respectively The observable direct parent set and the latent variable direct parent set, This represents a third-order differentiable nonlinear function mapping of observable variables from the j-th brain region to the i-th brain region. This represents a third-order differentiable nonlinear function mapping of latent variables from the k-th brain region to the i-th brain region. This indicates that the i-th brain region is affected. Random noise, and the random noise in different brain regions is independent of each other.
[0083] Using the ANM model algorithm, for any two different brain regions that are connected, and Let's assume the causal direction is... and ;right and Nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and ;right and Nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and In the causal direction Below, the residuals of EEG data and brain region variables The EEG data were analyzed using a binary independence test; in the case of opposite causal relationships... Below, the residuals of EEG data and brain regions The EEG data were analyzed using a binary independence test; specifically:
[0084] If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the causal direction of the brain region is accepted. This represents the direction of binary causal relationships within brain regions;
[0085] If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the causal direction of the brain region is accepted. This represents the direction of a binary causal relationship within a brain region.
[0086] If the independence test shows that the causal direction is... superior, and Independent, and simultaneously in the opposite causal direction superior, and If it is also independent, then it is believed that the current brain region's EEG data does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis;
[0087] If the independence test shows that the causal direction is... superior, and Not independent, and simultaneously in opposite causal directions. superior, and If it is not independent, then it is considered that the brain region and There is no binary causal relationship direction.
[0088] Independence testing analysis between the regression residuals of EEG data and EEG data from the hypothesized causal brain region reveals asymmetry in nonlinear causal directions, which helps identify the true orientation of brain region structures from the observational data. Specifically:
[0089]
[0090] in, Representing the brain region variable of the outcome The nonlinear effects of unobservable (latent variables) direct parents. Variables representing the cause of brain regions The nonlinear effects of unobservable (latent variables) direct parents. Representing the brain region variable of the outcome All observable direct parents, Representing the brain region variable of the outcome All observable direct parents.
[0091] The nonlinear regression and independence test method based on the generation hypothesis of nonlinear additive functions and independent noise effects provides a strong guarantee for the discovery of causal relationships in brain region structures based on classical structural causal function models in EEG data. At the same time, thanks to the assistance of the brain region structural causal framework in S2, it is possible to infer causal relationships between different brain regions more efficiently and accurately, thereby attempting to locate unobserved brain regions that cause the influence of latent variables, and providing a foundation for more robust latent variable causal discovery.
[0092] This example constructs a causal relationship identification model, applying methods based on conditional independence testing and nonlinear causal functions. Its foundation lies in Bayesian network structure learning under causal connotation constraints and nonlinear causal function models. These two causal discovery methods are combined through a hybrid framework, incorporating the strengths of classical algorithms and ensuring the effectiveness of subsequent algorithms' divide-and-conquer strategies.
[0093] The dependency edges that make up the causal framework of brain regions reflect both the real causal relationships between different brain regions and the spurious relationships formed by the influence of latent variables from unobserved brain region variables. The latter's generation mechanism tends to simultaneously form maximal clique structures in the graphical model, thus providing localization information for unobserved brain region variables. Therefore, based on the causal framework of brain regions, the maximal clique algorithm is used to decompose the framework, further locating unobserved brain region variables while locking down local structures where the direction of causal relationships cannot be determined.
[0094] The purpose of maximal clique decomposition based on the brain region causal framework is to avoid a global search of the brain region causal framework. The maximal clique decomposition algorithm is the Bron–Kerbosch maximal clique decomposition algorithm.
[0095] Utilizing the local graph structures obtained from all decompositions, the algorithm in S2 is reused for the current local graph structure, and existing causal relationships are taken into consideration. This extends the bivariate nonlinear regression to a multivariate nonlinear regression. Based on the analytical logic of the S2 stage, a multivariate independence test analysis is performed on the EEG data residuals and the EEG data from brain regions assumed to be the cause under the current mechanism. Specifically, the method involves: for data in the causal direction... superior, and Not independent, in the opposite causal direction superior and brain region variables brain regions that are not independent and Perform a multivariate independence test analysis; assume the causal direction is respectively and ,right and Multivariate nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and .
[0096] Specifically:
[0097] If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the multiple regression analysis is considered to accept the current causal direction of the brain region. This indicates a direction of multiple causal relationships, and it is inferred that no causal direction exists. ;
[0098] If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the multiple regression analysis is considered to accept the current causal direction of the brain region. This indicates a direction of multiple causal relationships, and it is inferred that no causal direction exists. ;
[0099] If the independence test shows that the causal direction is... superior, and Independent, and simultaneously in the opposite causal direction. and If it is also independent, then it is believed that the EEG data under the current brain region structure does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis;
[0100] If the independence test shows that the causal direction is... superior, and Not independent, and simultaneously in the opposite causal direction. and If it is not independent, then it is assumed that the EEG data under the current brain region structure does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis.
[0101] The advantage of multivariate nonlinear regression lies in eliminating the causal nonlinear dependence of other brain regions on the current local graph structure, thus ensuring the identification of the remaining multivariate causal relationships. The regression method is either a generalized additive model regression or a Gaussian process regression, and the independence test method is based on the nuclear independence criterion.
[0102] Example 3
[0103] This embodiment provides a brain region causal relationship recognition system based on electroencephalogram (EEG) data, used to implement the method described in Embodiment 1 or 2, such as... Figure 2 As shown, the system includes:
[0104] The data acquisition module is used to acquire EEG data from different brain regions;
[0105] The skeleton construction module is used to construct a skeleton diagram of causal relationships between brain regions based on EEG data from different brain regions;
[0106] The function model building module is used to build nonlinear function models between different brain regions based on the causal relationship skeleton diagram;
[0107] The binary causal relationship direction determination module is used to perform independent testing and analysis on all brain regions to obtain the binary causal relationship direction between brain regions;
[0108] The judgment module is used to determine whether a binary causal relationship exists in all brain regions; if not, it proceeds to the structure decomposition module; otherwise, it proceeds to the recognition module.
[0109] The structural decomposition module is used to decompose the causal relationship skeleton diagram to obtain a local structural diagram.
[0110] The module for determining the direction of multiple causal relationships, based on local structural maps, performs independence tests and analyses on brain regions that do not have directions of binary causal relationships, thereby obtaining the direction of multiple causal relationships in brain regions that do not have directions of binary causal relationships.
[0111] The recognition module is used to obtain the recognition results of the causal relationship direction of brain regions based on the binary causal relationship direction or the multivariate causal relationship direction of brain regions.
[0112] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.
Claims
1. A method for identifying causal relationships in brain regions based on electroencephalogram (EEG) data, characterized in that, The method includes: S1: Acquire EEG data from different brain regions; S2: Construct a causal relationship framework diagram between brain regions based on EEG data from different brain regions; S3: Based on the causal relationship skeleton diagram, establish a nonlinear function model between different brain regions; S4: Based on a nonlinear function model, independence tests are performed on all brain regions to obtain the direction of binary causal relationships between brain regions; S5: Determine whether a binary causal relationship exists in all brain regions; if not, proceed to step S6; otherwise, proceed to step S8. S6: Perform structural decomposition on the causal relationship skeleton diagram to obtain a local structure diagram; S7: Based on the local structure map, independence test analysis is performed on brain regions that do not have binary causal relationship directions to obtain the multivariate causal relationship directions between brain regions that do not have binary causal relationship directions; S8: Obtain the recognition result of the causal relationship direction of the brain region based on the binary causal relationship direction of the brain region or the multivariate causal relationship direction of the brain region. The specific method of S2 is as follows: the fully connected relationship between different brain regions is represented as a completely undirected graph. Based on the structure of the completely undirected graph, the independence relationship contained in the EEG data between any two brain regions is searched and traversed. If the independence relationship test result between the EEG data of any two brain regions is valid, the connection between the corresponding brain regions is removed. Otherwise, the connection between the corresponding brain regions is retained, and the causal relationship skeleton graph between brain regions is completed. Assuming that the EEG data of the current brain region conforms to the generation assumptions of a nonlinear additive function and the influence of independent noise, a nonlinear function model is established for the interaction between different brain regions: in, and These represent the EEG data of the i-th and j-th brain regions, respectively. This represents the EEG data of the k-th brain region. and Representing the i-th brain region respectively The observable direct parent set and the latent variable direct parent set, This represents a third-order differentiable nonlinear function mapping of observable variables from the j-th brain region to the i-th brain region. This represents a third-order differentiable nonlinear function mapping of latent variables from the k-th brain region to the i-th brain region. This indicates that the i-th brain region is affected. Random noise, and the random noise in different brain regions is independent of each other.
2. The brain region causal relationship identification method based on EEG data according to claim 1, characterized in that, In step S1, magnetic resonance imaging technology is used to acquire the subject's electroencephalogram (EEG) data, and the blood oxygen level-dependent signals of different brain regions in the subject's EEG data are sampled to acquire EEG data of different brain regions.
3. The brain region causal relationship identification method based on EEG data according to claim 1, characterized in that, The specific method of S4 is as follows: for any two different brain regions that have a connection... and Let's assume the causal direction is... and ;right and Bivariate nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and In the causal direction Below, the residuals of EEG data and brain region variables The EEG data were analyzed using a binary independence test; in the case of opposite causal relationships... Below, the residuals of EEG data and brain regions The EEG data were analyzed using a binary independence test; specifically: If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the causal direction of the brain region is accepted. This represents the direction of binary causal relationships within brain regions; If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the causal direction of the brain region is accepted. This represents the direction of binary causal relationships within brain regions; If the independence test shows that the causal direction is... superior, and Independent, and simultaneously in the opposite causal direction superior, and If it is also independent, then it is considered that the EEG data of the current brain region does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis; If the independence test shows that the causal direction is... superior, and Not independent, and simultaneously in opposite causal directions. superior, and If it is not independent, then it is considered that the brain region and There is no binary causal relationship direction.
4. The brain region causal relationship identification method based on EEG data according to claim 3, characterized in that, In S4, the ANM method is used to determine the connection between any two different brain regions that are connected. and Let's assume the causal direction is... and ;right and Nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and ;right and Nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and In the causal direction Below, the residuals of EEG data and brain region variables The EEG data were analyzed using a binary independence test; in the case of opposite causal relationships... Below, the residuals of EEG data and brain regions The EEG data were analyzed using a binary independence test.
5. The brain region causal relationship identification method based on EEG data according to claim 1, characterized in that, Based on the causal framework of brain regions, the maximum clique algorithm is used to decompose the causal framework map to obtain the local structure map.
6. The brain region causal relationship identification method based on EEG data according to claim 5, characterized in that, The maximum clique algorithm mentioned is the Bron–Kerbosch maximum clique decomposition algorithm.
7. The brain region causal relationship identification method based on EEG data according to claim 3, characterized in that, The specific method of S7 is as follows: for those in the causal direction superior, and Not independent, in the opposite causal direction superior and brain region variables Brain regions that are not independent and Perform a multivariate independence test analysis; assume the causal direction is respectively and ,right and Multivariate nonlinear regression was performed on the EEG data to obtain the corresponding EEG data residuals. and Specifically: If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the multiple regression analysis is considered to accept the current causal direction of the brain region. This indicates a direction of multiple causal relationships, and it is inferred that no causal direction exists. ; If the independence test shows that the causal direction is... superior, and Independent, but in the opposite causal direction superior, and If not independent, then the multiple regression analysis is considered to accept the current causal direction of the brain region. This indicates a direction of multiple causal relationships, and it is inferred that no causal direction exists. ; If the independence test shows that the causal direction is... superior, and Independent, and simultaneously in the opposite causal direction. and If it is also independent, then it is believed that the EEG data under the current brain region structure does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis; If the independence test shows that the causal direction is... superior, and Not independent, and simultaneously in the opposite causal direction. and If it is not independent, then it is assumed that the EEG data under the current brain region structure does not conform to the generation hypothesis of nonlinear additive function and independent noise influence, and there is no identifiable causal direction of brain region under the current hypothesis.
8. A brain region causal relationship recognition system based on electroencephalogram (EEG) data, used to implement the method described in any one of claims 1-7, characterized in that, The system includes: The data acquisition module is used to acquire EEG data from different brain regions; The skeleton construction module is used to construct a skeleton diagram of causal relationships between brain regions based on EEG data from different brain regions; The function model building module is used to build nonlinear function models between different brain regions based on the causal relationship skeleton diagram; The binary causal relationship direction determination module is used to perform independent testing and analysis on all brain regions to obtain the binary causal relationship direction between brain regions; The judgment module is used to determine whether a binary causal relationship exists in all brain regions; if not, it proceeds to the structure decomposition module; otherwise, it proceeds to the recognition module. The structural decomposition module is used to decompose the causal relationship skeleton diagram to obtain a local structural diagram. The module for determining the direction of multiple causal relationships, based on local structural maps, performs independence tests and analyses on brain regions that do not have directions of binary causal relationships, thereby obtaining the direction of multiple causal relationships in brain regions that do not have directions of binary causal relationships. The recognition module is used to obtain the recognition results of the causal relationship direction of brain regions based on the binary causal relationship direction or the multivariate causal relationship direction of brain regions.