A method and device for tracing and reconstructing EEG signals based on microstate analysis
Through the EEG signal traceability reconstruction algorithm based on microstate analysis and spatiotemporal variation Bayesian method, the problem of insufficient spatial and temporal constraints in the existing technology is solved, and efficient and accurate EEG signal traceability reconstruction is achieved.
Patent Information
- Application Number
- CN202211447411.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-11-18
- Publication Date
- 2025-08-15
- Estimated Expiration
- 2042-11-18
AI Technical Summary
The existing EEG traceability algorithm fails to effectively combine spatial and temporal constraints, resulting in the traceability results being susceptible to noise, poor accuracy, high computational complexity and long time.
Using a method based on microstate analysis, the EEG data is divided into multiple segments in the time domain, and the source imaging is performed in combination with the spatial and temporal variation Bayesian method. Through parallel calculations, the traceability reconstruction process is optimized, and the optimal time window is selected to improve accuracy and efficiency.
The accuracy and computing efficiency of the traceability reconstruction results are improved, spatial sparsity and local smoothness are ensured, while reducing the computational complexity, and achieving rapid and stable traceability reconstruction.
Smart Images
Figure CN115990025B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of electroencephalogram (EEG) signal processing, and in particular relates to an EEG signal source tracing and reconstruction method and device based on microstate analysis. Background Art
[0002] Electroencephalography (EEG) is a non-invasive physiological signal monitoring technology that detects the brain's electrophysiological activity through multiple electrodes attached to the surface of the scalp. It can sensitively reflect changes in neural potentials in various brain regions and is an important tool for studying the neurophysiological mechanisms of the human brain. However, due to the volume conduction effect, the spatial resolution of EEG is very limited, and there is an error between scalp EEG signals and actual brain power activity. Therefore, in recent years, researchers have proposed EEG source imaging to solve this problem. Through source tracing methods, EEG signals from the scalp are projected onto the source of intracranial neural activity, and the activity status of each brain region can be analyzed and judged to more accurately control external devices or monitor abnormal brain activity.
[0003] EEG source tracing is a highly ill-posed inverse problem whose solution requires constraints derived from prior physiological and anatomical knowledge. Common source imaging techniques are primarily based on distributed source models, tracing sources by estimating the current density distribution on the cortical surface. Minimum-norm methods based on the least-squares (L2-norm) are among the earliest solutions to distributed source models, including weighted minimum-norm (wMNE), low-resolution electromagnetic tomography (LORETA), and standardized LORETA (sLORETA). Although computationally inexpensive, these methods produce poor spatial specificity in reconstructed sources. Alternative approaches are based on sparse constraints, namely the L1-norm, Lp-norm, and sparse Bayesian learning. However, directly applying sparse constraints to sources can produce overly focused results, and sparse constraints are primarily based on a single time point, ignoring the temporal correlation of EEG signals. In addition to spatial priors, temporal information can be incorporated to better reconstruct the spatiotemporal patterns of source activity. For example, temporal information can be exploited through autoregressive (AR) models, temporal norm constraints, or temporal basis functions (TBFs) to achieve source imaging. While this source imaging method offers improved performance compared to single-time-point-based methods, it is often computationally expensive for high-resolution source space. Furthermore, since the optimal time window length is not considered, the accuracy of source tracing still has room for improvement.
[0004] The shortcomings of existing technologies are summarized as follows: the currently widely used tracing algorithms only adopt spatial constraints and do not consider temporal information, which makes the tracing results susceptible to noise and poor accuracy; although some advanced tracing algorithms use spatial constraints and temporal information, they do not consider the choice of time window length. The assumptions of the tracing algorithms are inconsistent with the brain activity in actual applications, resulting in their performance decline; algorithms that use both spatial constraints and temporal information often have high computational complexity and are time-consuming. Summary of the Invention
[0005] The purpose of the present invention is to address the deficiencies of the existing technology and provide a method and device for tracing and reconstructing EEG signals based on microstate analysis, which is used to solve the following technical problems: 1) Including spatial constraints and temporal constraints in the tracing algorithm to more accurately reconstruct the spatiotemporal characteristics of dynamic brain activities; 2) Selecting the optimal time window length through brain signal microstate analysis to optimize the tracing and reconstruction results; 3) Optimizing the calculation process of the tracing algorithm and using parallel computing technology to improve calculation efficiency.
[0006] The purpose of the present invention is achieved through the following technical solutions:
[0007] According to a first aspect of the present specification, a method for tracing and reconstructing an EEG signal based on microstate analysis is provided, the method comprising the following steps:
[0008] S1, obtain complete EEG data for microstate analysis, and divide the continuous EEG data into several segments according to the microstate division results;
[0009] S2, solve source imaging for the EEG data B after segmentation based on microstates; the source imaging model is expressed as a linear equation B = LWΦ + LE + ∈, where L is the known forward conduction matrix, Φ represents the time domain basis function, W represents the weighting coefficient of the time domain basis function of each source dipole, and E and ∈ are both error terms; the source imaging model is further expressed as a probability distribution form;
[0010] S3, using the variational Bayesian method to perform a posteriori estimation of the probability distribution of the source imaging model obtained in S2, that is, to maximize the free energy of the model
[0011] S4, update the hyperparameters corresponding to W, E, and Φ;
[0012] S5, according to the methods of S3 and S4, continue to update W, E, Φ and their corresponding hyperparameters until the free energy of the model converges;
[0013] S6, concatenate the traceability reconstruction results of each section of EEG data in chronological order to obtain the complete EEG signal traceability reconstruction result.
[0014] Furthermore, in step S1, the microstate analysis process is based on the EEGLAB microstate toolbox; the microstate analysis process includes bandpass filtering, microstate cluster analysis, microstate number selection, and signal segmentation.
[0015] Furthermore, in step S1, the bandpass filtering uses 4-30 Hz filtering; the microstate cluster analysis uses the improved k-means algorithm clustering to calculate the microstate space topology map; the number of microstates is selected based on the global explained variance and the cross-validation criterion.
[0016] Furthermore, in step S2, the source imaging model is expressed as the following probability distribution form:
[0017]
[0018]
[0019]
[0020]
[0021] Where T is the duration of EEG data B, is the normal distribution function, b t ,φ t ,e t They are the EEG data, time domain basis function, and error value at time t, respectively. I is the unit matrix, diag represents the diagonal matrix function, K is the number of time domain basis functions, and w i is the weighting coefficient of the i-th time-domain basis function, M is a smooth spatial coefficient matrix, T represents the transpose operation, Г = diag (γ) controls the relative distribution of each source dipole, Λ = diag (λ) controls the relative distribution of the error term, Controls the contribution of each time-domain basis function. α, γ, and λ are all hyperparameters.
[0022] Furthermore, in step S2, W adopts a smooth prior value, expressed as M, M = D-τA, where D and A represent the degree matrix and connectivity matrix of the source dipole, respectively, and τ is a user-defined value, which is (0, 1).
[0023] Furthermore, in step S2, the contribution of each source dipole is measured by 1 / γ, and the calculation is accelerated by ignoring the source dipoles whose contribution values are lower than the set threshold; the contribution of each time domain basis function is measured by 1 / α, and the calculation is accelerated by ignoring the time domain basis functions whose contribution values are lower than the set threshold.
[0024] Furthermore, it is characterized in that, in step S3, the probability distribution of the source imaging model is estimated a posteriori using the variational Bayesian method, that is, maximizing the free energy of the model
[0025]
[0026] in and are all probability distribution functions, <·> q(x) =∫·q(x)dx;
[0027] According to the mean field estimation method:
[0028]
[0029] By iteratively updating the parameters, the free energy Maximize, where w k The update method looks like this:
[0030]
[0031]
[0032]
[0033]
[0034] in Represents w in each round of iterative update k ,φ k ,E,φ k Represents the kth time domain basis function, F = LM -1 , represents w k The covariance matrix of
[0035] The update method of E is as follows:
[0036]
[0037]
[0038]
[0039]
[0040] in Represents the e of each round of iterative update t ,W,φ t ,φ t represents the time domain basis function at time t, Σ e represents the covariance matrix of E;
[0041] The update method of Ф is as follows:
[0042]
[0043]
[0044]
[0045] where Σ φ represents the covariance matrix of Φ.
[0046] Furthermore, in step S4, the hyperparameters γ, λ, and α corresponding to W, E, and Φ are updated. The update formula is as follows:
[0047]
[0048]
[0049]
[0050] where x i represents the i-th element, Respectively represent the i-th element corresponding to The optimal values of ξ, β, and ν are as follows:
[0051]
[0052]
[0053]
[0054] Further, in step S5, when the free energy When the relative change is small, it is considered to have converged; a parallel computing method is used to simultaneously perform the above S2 to S5 analyses on all EEG data segments after microstate segmentation to improve the efficiency of tracing and reconstruction.
[0055] According to the second aspect of this specification, there is provided an EEG signal tracing and reconstruction device based on microstate analysis, comprising a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the executable code, it is used to implement the EEG signal tracing and reconstruction method based on microstate analysis as described in the first aspect.
[0056] The beneficial effects of the present invention are:
[0057] 1. Through traceability reconstruction based on microstate analysis, the accuracy of traceability reconstruction results is improved while improving the interpretability of traceability results at the neurophysiological level.
[0058] 2. The provenance reconstruction method of the present invention has spatial sparsity while ensuring local smoothness, has low computational complexity, can generate stable and accurate provenance reconstruction results, and has a faster convergence speed.
[0059] 3. The source tracing and reconstruction method of the present invention can support parallel computing, which greatly improves the computing speed while ensuring the accuracy of the results. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] Figure 1 is a flow chart of a traceability algorithm method provided by an exemplary embodiment of the present invention;
[0061] Figure 2 An EEG traceability probability model and posterior distribution estimation provided by an exemplary embodiment of the present invention;
[0062] Figure 3 (a) is a schematic diagram of the location of the simulated EEG source activity. The generated complete simulation signal contains three different states, and the location of the EEG source activity in each state remains unchanged. (b) is the result of tracing and reconstructing the complete simulation signal using the traceability algorithm of the present invention.
[0063] Figure 4 This is a schematic diagram comparing the traceability reconstruction effects using the three indicators of AUC, DLE, and SD;
[0064] Figure 5 This is a structural diagram of the EEG signal tracing and reconstruction device based on microstate analysis of the present invention. DETAILED DESCRIPTION
[0065] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present application. Unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention pertains. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.
[0066] The present invention proposes a new source tracing and reconstruction algorithm framework, which optimizes the EEG source tracing and reconstruction results through EEG time domain division based on microstate analysis and source imaging algorithm based on spatiotemporal variational Bayesian. It can also improve the efficiency of the source tracing algorithm by performing parallel calculations on all segmented EEG time periods.
[0067] The method for tracing and reconstructing the EEG signal source based on microstate analysis proposed in the present invention comprises the following steps:
[0068] Step 1: Obtain complete EEG data, perform microstate analysis, and divide the continuous EEG data into several segments according to the microstate division results.
[0069] Specifically, the microstate standard analysis process can be based on the EEGLAB microstate toolbox.
[0070] Specifically, the microstate analysis process includes: 1. bandpass filtering, 2. microstate cluster analysis, 3. microstate number selection, and 4. signal segmentation.
[0071] Optionally, the bandpass filter may use 4-30 Hz filtering to reduce noise and improve the signal-to-noise ratio.
[0072] Optionally, the microstate cluster analysis may use an improved k-means clustering algorithm to calculate a microstate space topology graph.
[0073] Alternatively, the number of microstates can be selected based on the global explained variance (GEV) and the cross-validation criterion (CV). The global explained variance is used to assess the similarity between each EEG data sample and its corresponding microstate space topology. The higher the similarity (the larger the GEV value), the better.
[0074] The second step is to solve the source imaging problem for the EEG data B after segmentation based on microstates. The source imaging model can be expressed as a linear equation B = LWΦ + LE + ∈, where L is the known forward conduction matrix, Φ represents the time domain basis function (TBF), W represents the weighting coefficient of the time domain basis function of each source dipole, and E and ∈ are both error terms. The source imaging model is further expressed as the following probability distribution form:
[0075]
[0076]
[0077]
[0078]
[0079] Where T is the duration of EEG data B, is the normal distribution function, b t ,φ t ,e t They are the EEG data, time domain basis function, and error value at time t, respectively. I is the unit matrix, diag represents the diagonal matrix function, K is the number of time domain basis functions, and w i is the weighting coefficient of the i-th time domain basis function, M is a smooth spatial coefficient matrix, T represents the transpose operation, Г=diag(γ) controls the relative distribution of each source dipole, Λ=diag(λ) controls the relative distribution of the error term, and Controls the contribution of each time-domain basis function. α, γ, and λ are all hyperparameters.
[0080] (1) Because the brain's neural response is spatially continuous and locally consistent, W uses a smooth prior value, which is represented by M in this embodiment. M uses a smooth spatial coefficient matrix (for example, the discrete Laplace operator defined on the surface of the cerebral cortex), M = D - τA, where D and A represent the degree matrix and connectivity matrix of the source dipole, respectively, and τ is a user-defined value, usually (0, 1). In this embodiment, τ = 0.9.
[0081] (2) The contribution of each source dipole is measured by 1 / γ, and the calculation is accelerated by ignoring source dipoles whose contribution values are lower than the set threshold.
[0082] (3) The contribution of each time-domain basis function is measured by 1 / α, and the calculation is accelerated by ignoring the time-domain basis functions whose contribution values are lower than the set threshold.
[0083] The third step is to use the variational Bayesian method to perform a posterior estimation of the probability distribution of the source imaging model obtained in the previous step, that is, to maximize the free energy of the model.
[0084]
[0085] in and are all probability distribution functions, <·> q(x) =∫·q(x)dx;
[0086] According to the mean-field approximation method:
[0087]
[0088] By iteratively updating the parameters, the free energy Maximize, where w k The update method looks like this:
[0089]
[0090]
[0091]
[0092]
[0093] in Represents w in each round of iterative update k ,φ k ,E,φ kRepresents the kth time domain basis function, F = LM -1 , represents w k The covariance matrix of
[0094] Similarly, the update method of E is as follows:
[0095]
[0096]
[0097]
[0098]
[0099] in Represents the e of each round of iterative update t ,W,φ t ,φ t represents the time domain basis function at time t, Σ e represents the covariance matrix of E;
[0100] The update method of Ф is as follows:
[0101]
[0102]
[0103]
[0104] where Σ φ represents the covariance matrix of Φ;
[0105] The fourth step is to update the hyperparameters γ, λ, and α corresponding to W, E, and Φ. The update formula is as follows:
[0106]
[0107]
[0108]
[0109] where x i represents the i-th element of x, Respectively represent the i-th element corresponding to The optimal values of ξ, β, and ν can be obtained directly:
[0110]
[0111]
[0112]
[0113] In the fifth step, according to the methods of the third and fourth steps, continue to update W, E, Φ and their corresponding hyperparameters γ, λ, α until the free energy of the model converges.
[0114] (1) When free energy When the relative change is small (e.g. 10 -3 ), which can be regarded as convergence;
[0115] (2) A parallel computing method can be used to simultaneously perform the above-mentioned second to fifth steps of analysis on all EEG data segments after microstate segmentation to improve the efficiency of tracing and reconstruction.
[0116] The sixth step is to concatenate the traceability and reconstruction results of each segment of EEG data in chronological order to obtain the complete EEG signal traceability and reconstruction results.
[0117] The following simulation of brain power source activity is carried out to test the effectiveness of the traceability algorithm in the present invention. The SEREEGA toolbox is used to generate brain power source activity on the anatomical model of the cerebral cortex. In order to show the characteristics of brain activity state changing over time, three different brain activity states are simulated, such as Figure 3 As shown in (a), the number and location of brain power sources in each state are different. These three states contain 6, 4, and 2 source activities, respectively, with durations of 200ms, 100ms, and 120ms. The source activities of the three states are combined to generate the final simulation signal.
[0118] The method of the present invention is used to trace the source of the complete simulation signal after combination, and the results are as follows: Figure 3 As shown in (b), the method of the present invention distinguishes three different types of brain activity states. In each state, the source position obtained by tracing corresponds to the simulated source position, the area size is basically consistent, and there is no signal confusion between different states. The tracing algorithm provided by the present invention demonstrates excellent tracing accuracy.
[0119] For the above-mentioned simulated EEG signals, the tracing algorithm of the present invention is compared with the commonly used advanced EEG tracing algorithms BESTIES (K.Liu, ZLYu, W.Wu, Z.Gu, Y.Li, and S.Nagarajan, “Bayesian electromagnetic spatio-temporal imaging of extended sources with markov random field and temporal basis expansion,” NeuroImage, vol. 139, pp. 385–404, 2016.) and SI-STBF (K.Liu, ZLYu, W.Wu, Z.Gu, J.Zhang, L.Cen, S.Nagarajan, and Y.Li, “Bayesianelectromagnetic spatio-temporal imaging of extended sources based on matrix factorization,” IEEE Transactions on Biomedical Engineering, vol. 66, no. 9, pp. 2457–2469, 2019.), and the results are compared using three evaluation indicators of tracing effect: area under the ROC curve (AUC, area under The receiver operating characteristic curve) is used to evaluate the sensitivity and specificity of tracing the source of brain power activity (the larger the better); spatial dispersion (SD) is used to evaluate the spatial dispersion of the reconstructed brain power source relative to the real brain power source (the smaller the better); the distance of localization error (DLE) is used to evaluate the distance error of the reconstructed brain power source relative to the real brain power source position (the smaller the better). The evaluation results are as follows: Figure 4 As shown in the figure, the traceability reconstruction method of the present invention has the highest AUC value, indicating that it has the best sensitivity and specificity for EEG source identification. At the same time, the SD and DLE values are the lowest, indicating that the EEG source reconstructed by the present invention has the smallest degree of dispersion and distance error compared to the real EEG source, demonstrating excellent traceability accuracy.
[0120] The present invention innovatively integrates EEG microstate analysis into the tracing algorithm, segments the EEG data in the time domain based on the microstate analysis results, and performs tracing and reconstruction on each segmented EEG signal. The tracing algorithm of the present invention includes both spatial and temporal constraints, uses the spatiotemporal variational Bayesian method to calculate the posterior distribution of the tracing probability model, determines the time domain basis function (TBF) by a data-driven method, and defines the spatial prior indicator in advance. The tracing algorithm of the present invention can support the use of parallel computing to optimize the efficiency of the algorithm while ensuring accuracy.
[0121] Corresponding to the aforementioned embodiment of the method for tracing and reconstructing the source of EEG signals based on microstate analysis, the present invention also provides an embodiment of an apparatus for tracing and reconstructing the source of EEG signals based on microstate analysis.
[0122] See also Figure 5 An embodiment of the present invention provides an EEG signal tracing and reconstruction device based on microstate analysis, which includes a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement the EEG signal tracing and reconstruction method based on microstate analysis in the above embodiment.
[0123] The embodiment of the EEG signal tracing and reconstruction device based on microstate analysis of the present invention can be applied to any device with data processing capabilities, and the device with data processing capabilities can be a device or apparatus such as a computer. The device embodiment can be implemented through software, or through hardware or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by the processor of any device with data processing capabilities in which it is located reading the corresponding computer program instructions in the non-volatile memory into the memory for execution. From the hardware level, if Figure 5 The figure shows a hardware structure diagram of any device with data processing capability where the EEG signal tracing and reconstruction device based on microstate analysis of the present invention is located. Figure 5 In addition to the processor, memory, network interface, and non-volatile memory shown, any device with data processing capabilities in which the apparatus in the embodiment is located may also include other hardware, generally based on the actual functions of the device with data processing capabilities, which will not be described in detail.
[0124] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.
[0125] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present invention. A person of ordinary skill in the art can understand and implement the present invention without inventive work.
[0126] An embodiment of the present invention further provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the method for tracing and reconstructing the EEG signal based on microstate analysis in the above embodiment is implemented.
[0127] The computer-readable storage medium may be an internal storage unit of any device with data processing capabilities described in any of the aforementioned embodiments, such as a hard disk or memory. The computer-readable storage medium may also be an external storage device of any device with data processing capabilities, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc. equipped on the device. Furthermore, the computer-readable storage medium may also include both an internal storage unit and an external storage device of any device with data processing capabilities. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and may also be used to temporarily store data that has been output or is to be output.
[0128] The above description is merely a preferred embodiment of one or more embodiments of this specification and is not intended to limit one or more embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of this specification shall be included in the scope of protection of one or more embodiments of this specification.
Claims
1. A method for tracing and reconstructing EEG signals based on microstate analysis, characterized in that: The following steps are involved: S1, obtain complete EEG data for microstate analysis, and divide the continuous EEG data into several segments according to the microstate division results; S2, performing source imaging on the EEG data B segmented according to microstates; The source imaging model is expressed as a linear equation B = LWΦ + LE + ∈, where L is the known forward conduction matrix, Φ represents the time domain basis function, W represents the weighting coefficient of the time domain basis function of each source dipole, and E and ∈ are both error terms. The source imaging model is expressed as the following probability distribution form: Where T is the duration of EEG data B, is the normal distribution function, b t ,φ t ,e t They are the EEG data, time domain basis function, and error value at time t, respectively. I is the unit matrix, diag represents the diagonal matrix function, K is the number of time domain basis functions, and w i is the weighting coefficient of the i-th time-domain basis function, M is a smooth spatial coefficient matrix, T represents the transpose operation, Г = diag (γ) controls the relative distribution of each source dipole, Λ = diag (λ) controls the relative distribution of the error term, Control the contribution of each time domain basis function, α, γ, λ are all hyperparameters; S3, using the variational Bayesian method to perform a posteriori estimation of the probability distribution of the source imaging model obtained in S2, that is, to maximize the free energy of the model in and are all probability distribution functions, <·> q(x) =∫·q(x)dx; According to the mean field estimation method: By iteratively updating the parameters, the free energy Maximize, where w k The update method looks like this: in Represents w in each round of iterative update k ,φ k ,E,φ k Represents the kth time domain basis function, F = LM -1 , represents w k The covariance matrix of The update method of E is as follows: in Represents the e of each round of iterative update t ,W,φ t ,φ t represents the time domain basis function at time t, Σ e represents the covariance matrix of E; The update method of Ф is as follows: where Σ φ represents the covariance matrix of Φ; S4, update the hyperparameters corresponding to W, E, and Φ; S5, according to the methods of S3 and S4, continue to update W, E, Φ and their corresponding hyperparameters until the free energy of the model converges; S6, concatenate the traceability reconstruction results of each section of EEG data in chronological order to obtain the complete EEG signal traceability reconstruction result.
2. The method for tracing and reconstructing EEG signals based on microstate analysis according to claim 1, characterized in that: In step S1, the microstate analysis process is based on the EEGLAB microstate toolbox; the microstate analysis process includes bandpass filtering, microstate cluster analysis, microstate number selection, and signal segmentation.
3. The method for tracing and reconstructing EEG signals based on microstate analysis according to claim 2, characterized in that: In step S1, the bandpass filtering is performed using 4-30 Hz filtering; the microstate clustering analysis is performed using an improved k-means algorithm to calculate a microstate space topology map; The number of microstates is selected based on the global explained variance and the cross-validation criterion.
4. The method for tracing and reconstructing EEG signals based on microstate analysis according to claim 1, characterized in that: In step S2, W uses a smooth prior value, expressed as M, M = D-τA, where D and A represent the degree matrix and connectivity matrix of the source dipole, respectively, and τ is a user-defined value, which is (0, 1).
5. The method for tracing and reconstructing EEG signals based on microstate analysis according to claim 1, characterized in that: In step S2, the contribution of each source dipole is measured by 1 / γ, and the calculation is accelerated by ignoring the source dipoles whose contribution values are lower than the set threshold; the contribution of each time domain basis function is measured by 1 / α, and the calculation is accelerated by ignoring the time domain basis functions whose contribution values are lower than the set threshold.
6. The method for tracing and reconstructing EEG signals based on microstate analysis according to claim 1, characterized in that: In step S4, the hyperparameters γ, λ, and α corresponding to W, E, and Φ are updated. The update formula is as follows: where x i represents the i-th element of x, Respectively represent the i-th element corresponding to The optimal values of ξ, β, and ν are as follows:
7. The method for tracing and reconstructing EEG signals based on microstate analysis according to claim 1, characterized in that: In step S5, when the free energy When the relative change is small, it is considered to have converged; a parallel computing method is used to simultaneously perform the above S2 to S5 analyses on all EEG data segments after microstate segmentation to improve the efficiency of tracing and reconstruction.
8. A device for tracing and reconstructing EEG signals based on microstate analysis, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that: When the processor executes the executable code, it is used to implement the EEG signal tracing and reconstruction method based on microstate analysis as described in any one of claims 1 to 7.