Scale-free attribute brain dynamic effective connection network estimation method

Abstract

The invention is applicable to the technical field of neuroscience, and provides a scale-free attribute brain dynamic effective connection network estimation method, which comprises the following steps of: acquiring time sequence signals of different regions of a brain according to acquired fMRI data, and calculating the dynamic effective connection network of the brain on the basis of the time sequence signals; constructing a time variation vector autoregression model containing group space regularization and time smoothness regularization, and performing multiple regularization optimization on the time variation vector autoregression model by adopting an alternating multiplier method based on an augmented Lagrangian multiplier so as to estimate a model coefficient of the time variation vector autoregression model; and generating the brain dynamic effective connection network according to the optimized model parameters, so that the generated brain dynamic effective connection network keeps the scale invariant characteristic of the brain network, and the optimization capability of the brain dynamic effective connection network is improved.

Description

Technical Field

[0001] The present invention belongs to the field of neuroscience technology, and in particular relates to a method for estimating a scale-free brain dynamic effective connection network. Background Art

[0002] Functional Magnetic Resonance Imaging (fMRI) is a non-invasive technology widely used to study brain activity. fMRI reflects brain neural activity by measuring changes in blood oxygen level dependent (BOLD) signals. However, traditional fMRI analysis methods mainly focus on the study of static brain networks and have difficulty capturing changes in dynamic functional connections in the brain. Dynamic Effective Connectivity (dEC), as a connection network that can reflect the causal relationship between the brain at different time points, has gradually attracted the attention of researchers.

[0003] At present, existing methods often ignore the scale-invariant characteristics of brain networks when estimating dEC, resulting in misestimation of important hub nodes and network graph characteristics. At the same time, the complexity of multiple regularization problems also poses challenges to the efficiency of the algorithm. Therefore, there is an urgent need for a dEC network estimation method that can retain the scale-invariant characteristics of brain networks while having efficient optimization capabilities. Summary of the invention

[0004] The purpose of the present invention is to provide a method, device, equipment and medium for estimating a scale-free brain dynamic effective connection network, aiming to solve the problem that it is difficult to retain the scale-invariant characteristics of the brain network due to the existing technology.

[0005] In one aspect, the present invention provides a method for estimating a scale-free brain dynamic effective connection network, the method comprising the following steps: Based on the acquired fMRI data, time series signals of different brain regions are obtained; Based on the time series signal, construct a time-varying vector autoregressive model including group space regularization and time smoothness regularization; Using an alternating multiplier method based on augmented Lagrange multipliers to perform multiple regularization optimization on the time-varying vector autoregression model to estimate model coefficients of the time-varying vector autoregression model; A brain dynamic effective connection network is generated according to the optimized model parameters.

[0006] Preferably, the step of performing multiple regularization optimization on the time-varying vector autoregressive model by using an alternating multiplier method based on augmented Lagrange multipliers comprises: Constructing an augmented Lagrangian function based on the target loss function of the time-varying vector autoregression model; The augmented Lagrangian function is solved in a cyclic iterative manner using an alternating multiplier method to update the model parameters until a preset iteration end condition is reached.

[0007] Preferably, before the step of constructing an augmented Lagrangian function based on the target loss function of the time-varying vector autoregressive model, the method further comprises: Based on the first, second and third regularization terms and the time series signal, the target loss function is constructed, wherein the first regularization term is used to find highly hub-like nodes in the network, and the second regularization term is used to ensure that the network itself is effectively connected at the time point The third regularization term is used to The norm forces the difference between two temporally adjacent effectively connected networks to become sparse.

[0008] Preferably, the objective loss function is ,in, , For at time point Time series signals measured in brain regions, , represents the total number of measured time points, It's time point The coefficient matrix of the corresponding model parameters, is the first regularization term, is the second regularization term, is the third regularization term, represents the vectorization of the matrix, For at time point The vector of model coefficients to be estimated.

[0009] Preferably, , , ,in, , is the index of the node in the coefficient matrix, , , , , , is the regularization parameter that controls the weight of the corresponding regularization term, is an estimate of the previous point in time. For the matrix A permutation function that arranges the elements corresponding to each node.

[0010] Preferably, the step of constructing an augmented Lagrangian function based on the target loss function of the time-varying vector autoregressive model comprises: The augmented Lagrangian function is constructed according to the target loss function and the preset constraints. The constraints are: , the augmented Lagrangian function is ,in, It's at the time The scaled Lagrange multiplier of , is the step size parameter, is a dummy variable.

[0011] In another aspect, the present invention provides a device for estimating a scale-free brain dynamic effective connection network, the device comprising: A data processing unit, used for acquiring time series signals of different brain regions based on the acquired fMRI data; A model building unit, used for building a time-varying vector autoregression model including group space regularization and time smoothness regularization based on the time series signal; A model optimization unit, configured to perform multiple regularization optimization on the time-varying vector autoregression model by using an alternating multiplier method based on augmented Lagrange multipliers, so as to estimate model coefficients of the time-varying vector autoregression model; A network generation unit is used to generate a brain dynamic effective connection network based on the optimized model parameters.

[0012] Preferably, the model optimization unit comprises: A first function construction unit is used to construct an augmented Lagrangian function based on the target loss function of the time-varying vector autoregression model; The model parameter updating unit is used to perform cyclic iterative solution on the augmented Lagrangian function by using an alternating multiplier method to update the model parameters until a preset iteration end condition is reached.

[0013] On the other hand, the present invention also provides a computing device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps described in the above-mentioned method for estimating a scale-free brain dynamic effective connection network are implemented.

[0014] On the other hand, the present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps described in the above-mentioned method for estimating a scale-free brain dynamic effective connection network.

[0015] The present invention obtains time series signals of different brain regions according to the collected fMRI data, constructs a time-varying vector autoregression model including group space regularization and time smoothness regularization based on the time series signals, uses an alternating multiplier method based on augmented Lagrange multipliers to perform multiple regularization optimization on the time-varying vector autoregression model to estimate the model coefficients of the time-varying vector autoregression model, and generates a brain dynamic effective connection network according to the optimized model parameters, so that the generated brain dynamic effective connection network retains the scale-invariant characteristics of the brain network and improves the optimization ability of the brain dynamic effective connection network. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart for implementing a method for estimating a scale-free brain dynamic effective connection network provided in the first embodiment of the present invention; Figure 2 is a schematic diagram of the first experimental result provided by Example 1 of the present invention; Figure 3 is a schematic diagram of the second experimental result provided in Example 1 of the present invention; Figure 4 is a schematic diagram of the third experimental result provided in Example 1 of the present invention; Figure 5 It is a schematic diagram of the structure of a device for estimating a scale-free attribute brain dynamic effective connection network provided by the second embodiment of the present invention; Figure 6 It is a schematic diagram of the structure of a computing device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] The terms "first", "second", "third", etc. in the present invention are used to distinguish similar or similar objects or entities, and do not necessarily mean to limit a specific order or sequence, unless otherwise noted. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, for example, they can be implemented in an order other than those given in the diagrams or descriptions of the embodiments of the present disclosure.

[0019] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments: Embodiment 1: Figure 1 The implementation process of a method for estimating a scale-free brain dynamic effective connection network provided by the first embodiment of the present invention is shown. For the sake of convenience, only the part related to the embodiment of the present invention is shown, which is described in detail as follows: In step S101, time series signals of different brain regions are acquired based on the collected fMRI data.

[0020] The embodiment of the present invention is applicable to computing devices, such as personal computers, servers, etc. In the embodiment of the present invention, fMRI data is obtained by performing functional magnetic resonance imaging (fMRI) of the blood flow changes of local brain areas at different time points in a specific task or resting state by a magnetic resonance imaging device. When acquiring time series signals of different brain areas based on the acquired fMRI data, specifically, the acquired fMRI data is subjected to preprocessing operations including but not limited to data cleaning (such as removing eye movement channels), head movement correction, time layer correction, registration, segmentation, standardization and smoothing to reduce noise and artifacts and improve the accuracy and reliability of the data. The preprocessed data contains the blood oxygen level dependent (BOLD) signal of each voxel or brain area of ​​the brain that changes over time, that is, the time series signal.

[0021] In step S102, based on the time series signal, a time-varying vector autoregression model including group space regularization and time smoothness regularization is constructed.

[0022] In an embodiment of the present invention, in order to accurately infer the brain dynamic effective connection network (abbreviated as dEC network) from the time series signal, a time-varying vector autoregressive model (abbreviated as T-VAR model) is constructed based on the time series signal, and group space regularization and time smoothness regularization are introduced in the T-VAR model at the same time. The coefficient matrix of the T-VAR model will form the adjacency matrix of the dEC network, and the size of the element value in the adjacency matrix represents the connection strength between different brain regions. Among them, group space regularization is used in the dEC network inference to reduce the connectivity of potential hub nodes by grouping brain regions and applying regularization constraints to make the network more sparse, which helps to identify the nodes and connections that play a key role in the network. Time smoothness regularization is used in the dEC network inference to ensure that the network changes smoothly over time by applying time smoothness constraints to avoid sudden jumps or instability, which not only helps to improve the accuracy and reliability of the network, but also helps to capture the dynamic characteristics of the brain's effective connection network evolving over time.

[0023] In a feasible embodiment, when constructing a time-varying vector autoregression model including group space regularization and time smoothness regularization, a target loss function of the T-VAR model is constructed based on the first, second, and third regularization terms and the time series signal, wherein the group space regularization includes a first regularization term and a second regularization term, the first regularization term is used as a group-level penalty constraint to find highly hub-like nodes in the network (i.e., nodes with a large number of connections in the network), and the second regularization term is used as a spatial sparsity constraint to ensure that the effectively connected network (abbreviated as EC network) itself is at a point in time. The spatial sparsity of , the third regularization term acts as a temporal smoothness constraint, which is used to The norm forces the difference between two temporally adjacent EC networks to become sparse, which means that most EC networks will remain unchanged, but only a few will change in a short period of time, making the adjacent EC network topologies similar and the entire network topology changes smoothly. As a result, the objective loss function takes into account both the fit of the data and the scale-free, sparsity, and temporal smoothness of the dEC network when fitting the T-VAR model, which helps to infer a dEC network that is consistent with the data and satisfies specific network characteristics.

[0024] In another feasible embodiment, the objective loss function is ,in, , For at time point Time series signals measured in brain regions, , represents the total number of measured time points, It's at the time The coefficient matrix of the model parameters of the T-VAR model, is the first regularization term, is the second regularization term, is the third regularization term, represents the vectorization of the matrix, For at time point The vector of model coefficients to be estimated.

[0025] In yet another possible embodiment, , , ,in, , is the coefficient matrix The index of the midpoint (i.e., the connected group), , , , , , is the regularization parameter that controls the weight of the corresponding regularization term, is an estimate of the previous point in time. For the matrix The permutation function is used to arrange the elements corresponding to the nodes. Specifically, for the time point No. nodes, For the reason Extracted from Line A column vector formed by all off-diagonal elements of a column.

[0026] In step S103, the time-varying vector autoregression model is multi-regularized and optimized by using an alternating multiplier method based on augmented Lagrange multipliers to estimate the model coefficients of the time-varying vector autoregression model.

[0027] In an embodiment of the present invention, in the parameter estimation of the T-VAR model, the alternating multiplier method (AMM) based on the augmented Lagrange method (ALM) is used to optimize the objective loss function of the T-VAR model containing multiple regularization terms, thereby reducing the overfitting phenomenon of the model and improving the generalization ability of the model.

[0028] In a feasible embodiment, multiple regularization optimization of the T-VAR model is achieved through the following steps: (1) Constructing the augmented Lagrangian function based on the target loss function of the time-varying vector autoregression model; In an embodiment of the present invention, in the alternating multiplier method, it is necessary to construct an augmented Lagrangian function to handle constraints (such as regularization terms). Here, based on the target loss function, Lagrangian multipliers and auxiliary variables are introduced to construct the augmented Lagrangian function.

[0029] In a feasible embodiment, an augmented Lagrangian function is constructed according to the target loss function and the preset constraints, and the constraints are: , the augmented Lagrangian function is ,in, It's at the time The scaled Lagrange multiplier of , is the step size parameter, usually chosen to be 1. is a dummy variable.

[0030] (2) The augmented Lagrangian function is solved iteratively using the alternating multiplier method to update the model parameters until the preset iteration end condition is reached.

[0031] In the embodiment of the present invention, under the ALM-AMM framework, , and It is optimized by alternating updates, that is, in each iteration, other variables are fixed and only one variable is updated. Specifically, When and ,right Solve, update When and ,right Perform group Lasso regularization solution and update When and ,right Solve and repeat the update process until the preset iteration end condition is reached (such as the relative change of the loss value of the target loss function is less than a certain threshold (that is, the target loss function converges), or the number of iterations reaches the preset maximum number of iterations), where, The round of the current iteration.

[0032] Through the above steps (1) and (2), multiple regularization optimization of the T-VAR model is achieved, thereby improving the convergence speed and numerical stability, making the estimated model parameters more accurate, stable and reliable.

[0033] In step S104, a brain dynamic effective connection network is generated according to the optimized model parameters.

[0034] In an embodiment of the present invention, the coefficient matrix formed by the optimized model parameters can reflect the connection relationship between different brain regions more accurately and stably. Here, a brain dynamic effective connection network with scale-invariant characteristics is generated according to the optimized model parameters, which helps to reveal the dynamic characteristics of brain activity and the interaction between different regions.

[0035] In order to demonstrate the superior ability of the method proposed in this application in scale-free dEC network reconstruction, corresponding simulation experiments were conducted. In the simulation experiments, 20 nodes were used to randomly generate the network topology. The entire generated time series contained 3 different segments, each with 100 time points (a total of 300 time points). Therefore, at time points 100 and 200, the network topology will immediately change from one to another. The specific experiment is as follows: Performance evaluation experiments and regularization necessity experiments were conducted when using synthetic datasets. In the performance evaluation experiments, the method proposed in this application was compared with several other methods (including: Least absolute shrinkage and selection operator (LASSO), fused lasso regression (Fused LASSO), elastic network (Elastic Net), L21 regularization method, sparse group lasso regression (Sparse Group Lasso, SGL)) when using synthetic datasets. The experimental results are as follows: Figure 2 As shown, the performance evaluation indicators include: Area Under the Curve (AUC), Mean Squared Error (MSE), Precision, Recall, F1-Score and Matthews Correlation Coefficient (MCC). AUC is used to illustrate the topological recognition accuracy of different methods, MSE is used to measure the connection level deviation, precision is used to measure the correct connection between all recognitions, and recall is used to measure the correctness of all real connections. In the experiment on the necessity of regularization terms, the method proposed in this application (including the first, second and third regularization terms, referred to as group-wise spatial-temporal regularization (GSTR)) is compared with three degraded versions that remove one of the three imposed regularization terms to verify the necessity of these three regularization terms. The experimental results are shown in FIG. Figure 3 As shown; afterwards, the GSTR method proposed in this application was compared with several other methods in terms of hub detection capability using a real data set. The experimental results are shown in Figure 4As shown, in addition, Figure 2 , Figure 3 The average values ​​and MSE of different methods are shown in different colored lines and shadows. The network structure has two mutations at time points 100 and 200. Figure 4 The mean and MSE of different methods in are shown as lines of different colors and black error bars, respectively. The asterisk above each bar represents the statistically significant results obtained by comparing each competing method with the method of this application; Since the randomly generated network topology is designed to have two change points at time points 100 and 200, Figure 2 , Figure 3 The experimental results shown show that the performance of all methods decreases at these two time points and at the beginning of the simulation. This is because when dealing with such state changes or starting estimates, the regularization term may introduce inaccurate prior information from previous estimates, which will greatly affect the prediction accuracy. However, a good dEC network inference method should be able to quickly adjust the estimate from Figure 2 , Figure 3 It can be seen from the figure that the method of the present application can trace back to an accurate estimate faster than other methods, which illustrates the superiority of the dynamic tracking capability of the method of the present application; Figure 3 From the experimental results shown, it can be seen that the method of the present application performs better than its downgraded version, which shows that all components of the method of the present application are necessary for performance improvement; in addition, from Figure 4 It can be seen that the method of the present application has better hub detection capability than several other methods.

[0036] In an embodiment of the present invention, time series signals of different brain regions are obtained based on the collected fMRI data, and a time-varying vector autoregression model including group space regularization and time smoothness regularization is constructed based on the time series signals. The time-varying vector autoregression model is subjected to multiple regularization optimization by an alternating multiplier method based on augmented Lagrange multipliers to estimate the model coefficients of the time-varying vector autoregression model, and a brain dynamic effective connection network is generated based on the optimized model parameters, so that the generated brain dynamic effective connection network retains the scale-invariant characteristics of the brain network and improves the optimization capability of the brain dynamic effective connection network.

[0037] Embodiment 2: Figure 5 The structure of a scale-free attribute brain dynamic effective connection network estimation device provided by the second embodiment of the present invention is shown. For the convenience of description, only the part related to the embodiment of the present invention is shown, including: The data processing unit 51 is used to obtain time series signals of different brain regions based on the collected fMRI data.

[0038] The model building unit 52 is used to build a time-varying vector autoregression model including group space regularization and time smoothness regularization based on the time series signal.

[0039] The model optimization unit 53 is used to perform multiple regularization optimization on the time-varying vector autoregression model by using an alternating multiplier method based on augmented Lagrange multipliers to estimate the model coefficients of the time-varying vector autoregression model.

[0040] The network generation unit 54 is used to generate a brain dynamic effective connection network according to the optimized model parameters.

[0041] Preferably, the model optimization unit 53 includes: A first function construction unit is used to construct an augmented Lagrangian function based on a target loss function of a time-varying vector autoregressive model; The model parameter updating unit is used to perform cyclic iterative solution on the augmented Lagrangian function by using the alternating multiplier method to update the model parameters until a preset iteration end condition is reached.

[0042] Preferably, a scale-free attribute brain dynamic effective connection network estimation device according to an embodiment of the present invention further includes: The second function building unit is used to build a target loss function based on the first, second, and third regularization terms and the time series signal, wherein the first regularization term is used to find highly hub-like nodes in the network, and the second regularization term is used to ensure that the network itself is effectively connected at the time point The third regularization term is used to The norm forces the difference between two temporally adjacent effectively connected networks to become sparse.

[0043] In the embodiment of the present invention, each unit of a scale-free attribute brain dynamic effective connection network estimation device can be implemented by a corresponding hardware or software unit, and each unit can be an independent software or hardware unit, or can be integrated into a software or hardware unit, which is not intended to limit the present invention. Specifically, the implementation of each unit can refer to the description of the aforementioned embodiment 1, which will not be repeated here.

[0044] Embodiment three: Figure 6 The structure of a computing device provided in the third embodiment of the present invention is shown. For the convenience of description, only the part related to the embodiment of the present invention is shown.

[0045] The computing device 6 of the embodiment of the present invention includes a processor 60, a memory 61, and a computer program 62 stored in the memory 61 and executable on the processor 60. When the processor 60 executes the computer program 62, the steps in the embodiment of the above-mentioned method for estimating a scale-free brain dynamic effective connection network are implemented, such as Figure 1 Alternatively, when the processor 60 executes the computer program 62, the functions of each unit in the above-mentioned device embodiments are realized, for example Figure 5 Function of the unit shown.

[0046] In an embodiment of the present invention, time series signals of different brain regions are obtained based on the collected fMRI data, and a time-varying vector autoregression model including group space regularization and time smoothness regularization is constructed based on the time series signals. The time-varying vector autoregression model is subjected to multiple regularization optimization by an alternating multiplier method based on augmented Lagrange multipliers to estimate the model coefficients of the time-varying vector autoregression model, and a brain dynamic effective connection network is generated based on the optimized model parameters, so that the generated brain dynamic effective connection network retains the scale-invariant characteristics of the brain network and improves the optimization capability of the brain dynamic effective connection network.

[0047] The computing device of the embodiment of the present invention can be a personal computer or a server. The steps implemented when the processor 60 in the computing device 6 executes the computer program 62 to implement a method for estimating a scale-free brain dynamic effective connection network can refer to the description of the aforementioned method embodiment, which will not be repeated here.

[0048] Embodiment 4: In an embodiment of the present invention, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method for estimating a scale-free brain dynamic effective connection network are implemented, for example, Figure 1 Alternatively, when the computer program is executed by a processor, the functions of each unit in the above-mentioned device embodiments are realized, for example Figure 5 Function of the unit shown.

[0049] In an embodiment of the present invention, time series signals of different brain regions are obtained based on the collected fMRI data, and a time-varying vector autoregression model including group space regularization and time smoothness regularization is constructed based on the time series signals. The time-varying vector autoregression model is subjected to multiple regularization optimization by an alternating multiplier method based on augmented Lagrange multipliers to estimate the model coefficients of the time-varying vector autoregression model, and a brain dynamic effective connection network is generated based on the optimized model parameters, so that the generated brain dynamic effective connection network retains the scale-invariant characteristics of the brain network and improves the optimization capability of the brain dynamic effective connection network.

[0050] The computer-readable storage medium of the embodiment of the present invention may include any entity or device or recording medium capable of carrying computer program code, for example, ROM / RAM, magnetic disk, optical disk, flash memory and other memories.

[0051] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for estimating a scale-free brain dynamic effective connection network, characterized in that: The method comprises the following steps: Based on the acquired fMRI data, time series signals of different brain regions are obtained; Based on the time series signal, construct a time-varying vector autoregressive model including group space regularization and time smoothness regularization; Using an alternating multiplier method based on augmented Lagrange multipliers to perform multiple regularization optimization on the time-varying vector autoregression model to estimate model coefficients of the time-varying vector autoregression model; A brain dynamic effective connection network is generated according to the optimized model parameters.

2. The method according to claim 1, characterized in that The step of performing multiple regularization optimization on the time-varying vector autoregressive model by using an alternating multiplier method based on augmented Lagrange multipliers comprises: Constructing an augmented Lagrangian function based on the target loss function of the time-varying vector autoregression model; The augmented Lagrangian function is solved in a cyclic iterative manner using an alternating multiplier method to update the model parameters until a preset iteration end condition is reached.

3. The method according to claim 2, characterized in that Before the step of constructing an augmented Lagrangian function based on the target loss function of the time-varying vector autoregressive model, the method further includes: Based on the first, second and third regularization terms and the time series signal, the target loss function is constructed, wherein the first regularization term is used to find highly hub-like nodes in the network, and the second regularization term is used to ensure that the network itself is effectively connected at the time point The third regularization term is used to The norm forces the difference between two temporally adjacent effectively connected networks to become sparse.

4. The method according to claim 3, characterized in that The objective loss function is ,in, , For at time point Time series signals measured in brain regions, , represents the total number of measured time points, It's time point The coefficient matrix of the corresponding model parameters, is the first regularization term, is the second regularization term, is the third regularization term, represents the vectorization of the matrix, For at time point The vector of model coefficients to be estimated.

5. The method according to claim 4, characterized in that , , ,in, , is the index of the node in the coefficient matrix, , , , , , is the regularization parameter that controls the weight of the corresponding regularization term, is an estimate of the previous point in time. For the matrix A permutation function that arranges the elements corresponding to each node.

6. The method according to claim 5, characterized in that The step of constructing an augmented Lagrangian function based on the target loss function of the time-varying vector autoregression model comprises: The augmented Lagrangian function is constructed according to the target loss function and the preset constraints. The constraints are: , the augmented Lagrangian function is ,in, It's at the time The scaled Lagrange multiplier of , is the step size parameter, is a dummy variable.

7. A scale-free brain dynamic effective connection network estimation device, characterized in that: The device comprises: A data processing unit, used for acquiring time series signals of different brain regions based on the acquired fMRI data; A model building unit, used for building a time-varying vector autoregression model including group space regularization and time smoothness regularization based on the time series signal; A model optimization unit, configured to perform multiple regularization optimization on the time-varying vector autoregression model by using an alternating multiplier method based on augmented Lagrange multipliers, so as to estimate model coefficients of the time-varying vector autoregression model; A network generation unit is used to generate a brain dynamic effective connection network based on the optimized model parameters.

8. The device according to claim 7, characterized in that The model optimization unit comprises: A first function construction unit is used to construct an augmented Lagrangian function based on the target loss function of the time-varying vector autoregression model; The model parameter updating unit is used to perform cyclic iterative solution on the augmented Lagrangian function by using an alternating multiplier method to update the model parameters until a preset iteration end condition is reached.

9. A computing device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images