Electroencephalogram information flow interaction identification method and device based on Granger causality analysis

Abstract

The invention discloses an electroencephalogram signal analysis method and device based on Granger causal analysis. The electroencephalogram signal analysis method comprises the following steps: acquiring electroencephalogram signals of different brain regions acquired through a high-density electroencephalogram electrode array; carrying out residual iterative decomposition signal reconstruction on the electroencephalogram signal and then carrying out signal preprocessing; performing time-frequency analysis on the preprocessed electroencephalogram signals by adopting wavelet transform, and extracting local time-frequency characteristics of the electroencephalogram signals; analyzing a Granger causal value between every two electrodes by adopting nonparametric Granger causal analysis and based on local time-frequency characteristics; eEG Granger causal values in different brain regions are visualized and presented to a cortex in a single-point-to-multipoint mean value manner, so that a spatial-temporal mode of frequency-specific EEG information flow transmission in different directions in different brain regions is shown, and the understanding of spatial-temporal dynamic characteristics of EEG signals in information interaction is improved; and an important neuroscience analysis basis is provided for complex recall and cognitive processes and neural activities under other experimental paradigms.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning and brain science, and specifically relates to a method and device for identifying interaction of electroencephalogram (EEG) information flows based on Granger causality analysis. Background Art

[0002] In the field of neuroscience, the analysis of electroencephalogram (EEG) signals is an important means to understand brain function. EEG signals are changes in the electric field generated by neuronal activity in the cerebral cortex, which are recorded by placing electrodes on the cortex. These signals reflect the dynamic processes of the brain in various cognitive tasks, including perception, attention, memory, and decision-making. µECoG is an electrophysiological technique that records local field potentials from an array of electrodes implanted on the surface of the brain. Characteristics of ECoG signals include high spatial resolution (several micrometers for µECoG compared to several millimeters for EEG), wide bandwidth (0-500 Hz compared to 0-40 Hz for EEG), high signal-to-noise ratio, and reduced susceptibility to interference from external factors due to the direct placement of the electrode array on the cortical surface. These characteristics make ECoG particularly suitable for studying detailed spatial patterns of neural oscillations.

[0003] Traditional EEG signal analysis methods mainly focus on the power spectrum characteristics of the signal, such as power spectral density analysis and changes in spectral energy. These methods can reveal the oscillatory activity of the brain in different states, such as EEG activity during sleep, wakefulness, or specific cognitive tasks. However, these analyses usually ignore the timing characteristics and dynamic interactions of EEG signals, limiting the in-depth understanding of the complex network dynamics of the brain.

[0004] As an effective time-frequency analysis tool, wavelet transform has attracted attention because it can provide information on both time and frequency. Through wavelet transform, transient oscillatory activities in EEG signals can be detected, which are closely related to specific neurocognitive processes. However, it mainly focuses on the time-frequency characteristics of the signal itself, and the analysis of how brain regions interact with each other through these time-frequency characteristics is relatively limited.

[0005] Granger Causality (GC) is a method for evaluating the causal relationship between time series. It determines whether there is a causal relationship between the two by analyzing whether one signal can predict the future value of another signal. This method has been widely used in fields such as economics and meteorology, and has gradually been introduced into neuroscience research. The introduction of GC provides a new perspective for EEG signal analysis, allowing researchers to explore the flow of information between brain regions, rather than just local EEG activity. However, traditional GC methods have certain limitations when processing high-dimensional EEG signal data, especially when processing brain networks with complex dynamic characteristics.

[0006] In addition, traditional EEG analysis often obtains event-related potentials (ERPs) by averaging a sufficient number of single-trial EEG responses, assuming that the basic components evoked by the stimulus are locked to the stimulus onset, and that the residual average between stimuli and a given single trial is just noise. However, this dominant assumption in ERPs research has long been questioned. For example, ERPs may consist of several continuous components, and different components vary with respect to the onset of the locked stimulus. In this case, the average ERPs waveform blurs the representation of neural activity in a single trial. Summary of the invention

[0007] In view of the above, the purpose of the present invention is to provide an EEG signal analysis method and device based on Granger causality analysis, which combines the advantages of residual iterative decomposition EEG reconstruction and nonparametric Granger Causality (NPGC) analysis, and visualizes the EEG Granger causal values ​​between different brain regions to the cortex through a single-point to multi-point mean method, thereby showing the spatiotemporal pattern of frequency-specific EEG information flow transmission in different directions of different brain regions, so as to improve the understanding of the spatiotemporal dynamic characteristics of EEG signals in information interaction, and provide important neuroscience analysis basis for complex memory and cognitive processes.

[0008] To achieve the above-mentioned purpose of the invention, an embodiment provides an EEG signal analysis method based on Granger causality analysis, comprising the following steps: Acquire EEG signals from different brain regions through a high-density cortical EEG electrode array; The EEG signal is subjected to residual iterative decomposition and signal reconstruction and then signal preprocessing is performed; Wavelet transform is used to perform time-frequency analysis on the preprocessed EEG signals to extract local time-frequency features of the EEG signals. Non-parametric Granger causality analysis was used to analyze the Granger causality between two electrodes based on local time-frequency characteristics. The Granger causal values ​​of brain regions were projected to electrode sites using a single-point to multi-point mean approach, and the spatial and temporal patterns of information flow interactions in different directions between brain regions were visualized.

[0009] The EEG signals were reconstructed by residual iterative decomposition, and ERPs were decomposed into component clusters with different latency variability through an iterative process. The nonparametric Granger causality analysis method of wavelet transform was used to perform time-frequency analysis on the EEG signals and extract the local time-frequency features of the signals. The extracted time-frequency features were mapped to the cortex to visualize the GC information flow and intuitively reflect the information interaction of high-density brain signals. The intuitive representation of the GC information flow pattern of the cortex was achieved, and the analysis results could also be further processed in combination with machine learning algorithms. This study provides a new analytical tool for neuroscience research to better understand the information interaction mechanism of neural oscillations between different brain regions.

[0010] Preferably, performing residual iterative decomposition signal reconstruction on the EEG signal includes: The event-related potentials (ERPs) corresponding to the EEG signals are decomposed into EEG component clusters with different latency variability through an iterative process, including initial latency estimation, decomposition based on stimulus occurrence, reaction time and single trial latency, and re-estimation of latency using specified components as templates. The above steps are iterated until convergence.

[0011] Preferably, the signal preprocessing includes signal segmentation, bandpass filtering, denoising, re-referencing, and standardization.

[0012] Preferably, when non-parametric Granger causality analysis is used, the analysis is divided into two types of information flow: bottom-up and top-down, corresponding to the two directions from the inferior temporal cortex to the prefrontal cortex and from the prefrontal cortex to the inferior temporal cortex, respectively.

[0013] Preferably, visualizing the information flow interaction includes: mapping the non-parametric Granger causality value back to the cerebral cortex using a single-point to multi-point mean to visualize the spatiotemporal patterns of information flow interaction in different directions.

[0014] Preferably, visualizing the information flow interaction also includes: constructing a visualization diagram of the directional information flow interaction between brain regions by calculating the causal index between each electrode in a certain brain region relative to all electrodes in other brain regions and mapping it back to the cerebral cortex.

[0015] Preferably, complex wavelet transform is used to perform time-frequency decomposition on the preprocessed signal.

[0016] To achieve the above-mentioned object of the invention, an embodiment of the present invention provides an electroencephalogram signal analysis device based on Granger causality analysis, comprising: A signal acquisition module, which is used to acquire EEG signals from different brain regions through a high-density cortical EEG electrode array; A signal processing module, which is used to perform signal preprocessing after performing residual iterative decomposition and signal reconstruction on the EEG signal; A time-frequency analysis module is used to perform time-frequency analysis on the preprocessed EEG signal using wavelet transform to extract local time-frequency features of the EEG signal; A causal analysis module, which is used to use non-parametric Granger causal analysis and analyze the Granger causal value between two electrodes based on local time-frequency characteristics; The visualization module is used to project the Granger causal values ​​of brain regions to electrode sites by using a single-point to multi-point mean method and to visualize the spatial and temporal patterns of information flow interactions in different directions between brain regions.

[0017] To achieve the above-mentioned purpose of the invention, an embodiment further provides a computing device, including a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the above-mentioned EEG signal analysis method based on Granger causality analysis.

[0018] To achieve the above-mentioned purpose of the invention, an embodiment further provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, the above-mentioned EEG signal analysis method based on Granger causality analysis is implemented.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The present invention uses nonparametric Granger causality analysis (NPGC) based on RIDE reconstruction to evaluate the causal relationship in the EEG signals of macaques in passive observation and color perception tasks. Based on the innovative mapping of GC to the cortex and visualization of the interaction effect of information flow, the EEG signals corrected by RIDE can more effectively prevent distortion and can effectively extract component components based on the combination of known event markers and estimated delays. Compared with the traditional parametric GC method, NPGC does not require linear assumptions on the data, can better handle nonlinear and non-Gaussian data, and is more efficient in calculation, which is suitable for processing large-scale EEG signal data.

[0020] In terms of the construction of Granger causality distribution maps, by calculating the causal indicators between each brain region and all other brain regions, a visualization of the information flow between brain regions was constructed. These distribution maps reveal the spatiotemporal patterns of information flow and dynamic interaction between brain regions during visual perception and memory retrieval and maintenance. By analyzing these distribution maps, we can further explore the spatiotemporal distribution of information flow in different brain regions in different frequency bands and directions, and we can also combine pattern recognition to deeply explore neural mechanisms or explore the neural information interaction patterns between different brain regions from the perspective of neuroscience. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 is a flow chart of an electroencephalogram signal analysis method based on Granger causality analysis provided in an embodiment; Figure 2 is a behavioral experiment design diagram provided in the embodiment; Figure 3 is the corresponding cerebral cortex position where the ECoG electrode is placed in the macaque experiment provided in the embodiment; Figure 4 is a Granger causal interaction timing diagram between two single electrodes in brain regions based on preprocessing and EEG signal reconstruction provided by an embodiment; Figure 5 It is a Granger causal analysis cortical mapping diagram of two brain regions and two directions in the theta frequency band under different task paradigms of two macaques provided in the embodiment; Figure 6 It is a structural schematic diagram of an electroencephalogram signal analysis device based on Granger causality analysis provided in an embodiment. DETAILED DESCRIPTION

[0023] 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 implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0024] like Figure 1 As shown, the embodiment provides an EEG signal analysis method based on Granger causality analysis, comprising the following steps: S1, obtain EEG signals from different brain regions through a high-density cortical EEG electrode array.

[0025] Since the human visual cortex is very similar to non-human primates such as macaques, non-human primates were used as experimental subjects. The collaboration between the inferior temporal cortex (ITC) and prefrontal cortex (PFC) of primates plays an important role in visual processing, information retrieval, and maintenance of working memory. The ITC is a key area for processing visual information of objects, while the PFC is responsible for encoding this information according to behavioral needs. Both are involved in the memory representation related to visual objects. The PFC obtains bottom-up visual input from the ITC and regulates the neural activity of the ITC through top-down feedback. Although neural oscillations are considered to be a mechanism for information transmission between different brain regions, how the information flow between the ITC and PFC is transmitted through oscillations during visual perception, memory retrieval and maintenance remains to be further studied.

[0026] For this reason, healthy Japanese macaques (Macaca fuscata) were selected as experimental subjects in the embodiment, with a body weight of 5.6 to 7.1 kg and an age of 4 to 7 years old. All experimental macaques came from the NBRP "Japanese Monkey" project and lived in standard primate cages. The environment was set to a cycle of 14 hours of light (bright) and 10 hours of darkness (dark) per day, and the temperature and humidity were controlled within an appropriate range. In addition to normal pellet feed, the feed of the experimental animals also contained fruits and vegetables to ensure balanced nutrition. Before the experiment, all experimental animals were given a health check to ensure that the animals were in a healthy state and received strict behavioral task training to ensure the reliability of the experimental data. All experiments were reviewed by the Animal Ethics Committee to ensure compliance with animal welfare and ethical standards, and the operation strictly abided by relevant laws and regulations.

[0027] Experimental design, such as Figure 2 As shown in the figure, macaques were trained to perform two tasks to explore the information flow interaction patterns between brain areas under visual perception, including a picture passive viewing (PPV) task and a picture-color association (PCA) task, and EEG signals were recorded while they were performing the facial and color visual stimulus gaze and picture-color association memory tasks.

[0028] In the PPV task, the task is performed by fixating on different types of visual stimuli, such as colors. The experimental subject, the macaque, needs to fixate on a point in the center of the screen, and then the screen presents different types of visual stimuli, such as faces, colors, and achromatic images. The experimental process begins with the monkey first starting the task by holding the joystick, requiring it to fixate for 1.5 seconds before the visual stimulus is presented. In the delay phase after the visual stimulus disappears, the macaque needs to keep fixating until the fixation point becomes dark, release the joystick, complete the task, and receive a reward. Under different task conditions, the visual stimulus presentation time and delay time are different. The PPV task provides control data for the subsequent PCA task, that is, the data of the PPV task is used to compare with the subsequent picture-color association (PCA) task to help evaluate the differences in neural activity.

[0029] The PCA task tests the memory ability of macaques by having them learn the association between the cue stimulus and the target color. The cue stimulus is a Fourier operator image, and the target color is a red, green, and blue figure of different shapes. The macaques complete the task by looking at the cue stimulus and selecting the correct color figure corresponding to the cue stimulus after a delay period, and receive a juice reward. Error trials and pseudo-random sequences are also designed in the task to ensure the diversity and complexity of the task and to ensure that the memory ability of the macaques is fully tested.

[0030] These tasks can effectively stimulate the visual perception, short-term memory and response behavior of macaques through the design of fixation points, the presentation of visual stimuli and the addition of delay stages. These behavioral paradigms can not only effectively stimulate the interactive activities between ITC and PFC, but also provide a reliable neural activity basis for subsequent ECoG signal recording.

[0031] After the experimental animals completed basic task training, a high-density ECoG electrode array was implanted in the monkey's cerebral cortex. First, the macaque's head was fixed under general anesthesia, and the operation was performed under sterile conditions. Under the assistance of a microscope, specific cortical areas of the macaque's brain were exposed, mainly including the inferotemporal cortex (ITC) and prefrontal cortex (PFC). Through precise positioning, the customized ECoG electrode array was implanted in the following areas: Figure 3 The target cortical areas are shown. The center-to-center spacing of each electrode array was 2.5 mm, covering an area of ​​37.5 × 17.5 mm in the ITC region and 17.5 × 17.5 mm in the PFC region. After surgery, the electrodes were connected to microconnectors outside the skull for data collection. After surgery, the macaques needed to receive three days of pain medication and a week of antibiotic injections to prevent infection. After surgery, magnetic resonance imaging (MRI) was used to confirm the precise implantation position of the electrodes to ensure that the electrodes covered the target areas.

[0032] The EEG signal is collected by connecting the ECoG electrode to a high-precision differential amplifier, which can amplify the signal and perform bandpass filtering in the range of 0.7 to 300 Hz. The signal acquisition frequency is set to 1 kHz to ensure that various types of neural oscillation activities in the brain can be captured. The obtained LFP signal undergoes rigorous preprocessing, including denoising and re-reference processing, to eliminate possible machine noise and system errors, providing a high-quality data basis for subsequent time-frequency analysis and causal analysis, and storing the data in real time on the hard disk for subsequent analysis. During the acquisition process, the signal source mainly includes the cortical electrical activity of the macaque at different task stages (visual stimulation and three delay periods).

[0033] S2, perform signal preprocessing after performing residual iterative decomposition and signal reconstruction on the EEG signal.

[0034] In the embodiment, the EEG signal is reconstructed by residual iterative decomposition (RIDE) signal to solve the problems caused by the latency variability between trials in the traditional ERPs analysis. The ERPs are analyzed by decomposition, reconstruction and single trial to more accurately reflect the combined activities of the single trial ERPs and provide richer information about the ERPs subcomponents. The specific signal reconstruction process includes: decomposing the event-related potential ERP corresponding to the EEG signal into EEG component clusters with different latency variability through an iterative process, including initial latency estimation, decomposition based on stimulus occurrence, reaction time and single trial latency, and re-estimating the latency using the specified component as a template, and iterating the above steps until convergence.

[0035] After signal reconstruction, the reconstructed EEG component clusters are subjected to signal preprocessing, including signal segmentation, bandpass filtering, denoising, re-referencing, and standardization to ensure the quality of the data. For signal segmentation, the time period from 4.5 seconds before the visual stimulation to 5 seconds after the visual stimulation is extracted for each trial, which is divided into a prompt stage and three delay stages; for bandpass filtering, a bandpass filter is used to limit the signal to between 0.7 Hz and 250 Hz to remove low-frequency drift and high-frequency noise; for denoising, an independent component analysis (ICA) method is used to remove noise-containing components and abnormal trials; for re-referencing, the average signal of adjacent electrodes is subtracted to reduce the interference of machine noise; for standardization, each electrode signal is Z-score standardized to ensure the comparability of signal data between different electrodes. In the embodiment, the Granger causal time series diagrams between single electrodes in different brain regions after preprocessing and signal reconstruction are Figure 4 , it can be seen that there are interactions in different directions in different periods.

[0036] S3, using wavelet transform to perform time-frequency analysis on the preprocessed EEG signal to extract the local time-frequency features of the EEG signal.

[0037] In the embodiment, the preprocessed signal is decomposed into time and frequency using Continuous Wavelet Transform (CWT). CWT has a high time-frequency localization characteristic and can reveal the energy distribution of the signal that changes with time at different frequencies. Assume that the input signal is x ( t ), its wavelet transform formula is: ; in, ψ For the mother wavelet, a is the scale factor, b is the time shift factor, t The time-frequency diagram formed by the local time-frequency characteristics obtained by wavelet transform is used for further Granger causality analysis.

[0038] S4, non-parametric Granger causality analysis was used to analyze the Granger causality values ​​between pairs of electrodes based on local time-frequency characteristics.

[0039] In the embodiment, the local time-frequency features obtained by wavelet transform are further used for Granger Causality (GC) analysis to reveal the information flow between cortical regions. The purpose of GC analysis is to determine a signal X( t ) can Granger causally affect another signal Y( t ). Traditional GC analysis can be expressed by the following set of equations: ; in, Y t is the dependent variable in time t The value of Y t−i and X t−j is a lag term, indicating that the variable Y and X At a point in the past t−i and t−j The value of is used to build a regression model to evaluate the influence of past information on current values. Y t predictive ability. a j and β j is the influence coefficient, which indicates Y and X The lagged terms of Y tThe degree of influence of the value. e t represents the error term. β j Directly reflects X right Y Potential causal impact. β j ≠0, it means X Past values ​​of help predict Y The future value of X yes Y The assumption of Granger causation.

[0040] In order to adapt to the non-stationary characteristics of brain signals, the present invention adopts a non-parametric Granger causality analysis method, which converts the time domain to the frequency domain by wavelet transform, extracts the noise covariance and transfer function to evaluate the frequency domain Granger causality effect. The non-parametric Granger causality analysis calculation formula is: ;

[0041] in S xx ( f ) is the spectral density, which reflects the power distribution of the signal x(t) in the frequency domain. H xy ( f ) is the transfer function, which represents the dynamic response characteristics of the system output y(t) and input x(t) in the frequency domain, and ∑ is the noise covariance.

[0042] In the embodiment, when performing non-parametric Granger causality analysis, the analysis is divided into two information flows: bottom-up (BU) and top-down (TD), corresponding to the two directions of ITC to PFC (BU) and PFC to ITC (TD), respectively.

[0043] S5, using the single-point to multi-point mean method to project the Granger causal values ​​of brain regions to electrode sites and visualize the spatial and temporal patterns of information flow interactions in different directions between brain regions.

[0044] In the embodiment, the non-parametric Granger causality value single point to multi-point mean is mapped back to the cerebral cortex for visualization to obtain Figure 5 The cortical GC results shown in the figure intuitively demonstrate the information flow interaction pattern based on RIDE's non-parametric Granger causality, and provide an intuitive visualization of how the cerebral cortical network works together, and can be combined with machine learning algorithms to further process the analysis results. By comparing the BU and TD information flows at different task stages (such as visual stimulation and delay period), the information transmission pattern between cortices under different task conditions can be revealed.

[0045] After the experiment, the electrode implantation position was confirmed by histological methods. The exact position of the electrode was confirmed by fixing the animal's brain and taking pictures, and the results were verified by referring to the position markers of the electrode relative to the cortical sulci to ensure that the electrode covered the expected cortical area.

[0046] In terms of data analysis, custom codes written in MATLAB were used in combination with the RIDE toolbox and the FieldTrip toolbox to perform advanced processing such as signal reconstruction, time-frequency analysis, and Granger causality analysis to analyze the neural activity patterns of cortical areas under different task conditions. Comparative analysis of changes in BU and TD information flow can reveal the neural network patterns of the cerebral cortex in cognitive tasks, and gain an in-depth understanding of the information transmission methods between various cortical areas, thereby obtaining the collaborative working methods of different cortical areas in cognitive tasks.

[0047] Through the above experimental process, the neural activities of the ITC and PFC regions of macaques in visual cognitive tasks were successfully recorded and analyzed, and the changing patterns of information flow patterns between cortices were discovered. These patterns are closely related to the task type and stage. At different stages of the task, the intensity and direction of BU and TD information flows also changed significantly, reflecting the dynamic coordination mechanism of the cerebral cortex in cognitive tasks. This provides a new tool for exploring the interaction of information flows between different brain regions and provides new insights into understanding the working mechanism of the brain's cortical network in the cognitive process.

[0048] Studies have found that during visual perception and memory retrieval and maintenance, information flow between brain regions shows enhanced oscillations at specific frequencies. These information flows are transmitted across brain regions in more spatially clustered units in a specific combination of directions and frequency bands. The EEG signal analysis method based on RIDE and non-parametric Granger causal analysis proposed in the present invention visualizes the information flow interaction effect to the cerebral cortex, providing new tools and perspectives for a deeper understanding of the functional interaction of the brain. By combining RIDE EEG reconstruction and NPGC causal relationship evaluation, the dynamic characteristics of EEG signals and the information flow between brain regions can be more accurately revealed, providing strong support for neuroscience research, including: Providing analysis tools: By combining non-parametric Granger causality analysis with RIDE signal reconstruction, the present invention can more accurately analyze the spatiotemporal characteristic patterns of EEG signals during information interaction. Based on cognitive experimental design, the present invention can combine different brain regions, different directions, and different frequency bands to conduct in-depth exploration and analysis of the spatiotemporal patterns of information flow interaction.

[0049] Experimental theoretical contribution: Using macaques as research subjects, this method deeply studied the synchronization phenomenon mediated by neural oscillations between the prefrontal cortex and the visual cortex during working memory, aiming to reveal the information maintenance pattern specific to visual attributes in the visual cortex. This method provides a new explanation for the sensory recruitment hypothesis in working memory.

[0050] Expanded Scope of Application: It has important potential applications in clinical settings and brain-computer interfaces (BMIs). Understanding the neural mechanisms of working memory and the role of neural oscillations may lead to new therapeutic strategies for treating cognitive disorders associated with working memory impairment, such as ADHD, schizophrenia, and depression. In addition, insights into the cortical networks supporting working memory may help develop BMIs that enhance cognitive function or compensate for cognitive deficits, thereby extending the impact of this research from basic neuroscience to clinical and technological advances.

[0051] Promote research discovery: By gaining a deeper understanding of brain function and information interaction between brain regions, this invention is expected to promote new discoveries in the field of neuroscience and promote the development of related research.

[0052] like Figure 6 As shown, the embodiment also provides an EEG signal analysis device 60 based on Granger causality analysis, including a signal acquisition module 61, a signal processing module 62, a time-frequency analysis module 63, a causal analysis module 64, and a visualization module 65, wherein the signal acquisition module 61 is used to acquire EEG signals of different brain regions collected by a high-density cortical EEG electrode array; the signal processing module 62 is used to perform signal preprocessing after residual iterative decomposition and signal reconstruction of the EEG signal; the time-frequency analysis module 63 is used to perform time-frequency analysis on the preprocessed EEG signal using wavelet transform to extract local time-frequency features of the EEG signal; the causal analysis module 64 is used to use non-parametric Granger causality analysis and analyze the Granger causality values ​​between pairs of electrodes based on local time-frequency features; the visualization module 65 is used to project the Granger causality values ​​between brain regions to electrode sites using a single-point to multi-point mean method and visualize the spatial-temporal pattern of information flow interaction in different directions between brain regions.

[0053] The embodiment further provides a computing device, including a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, the method for analyzing EEG signals based on Granger causality analysis is implemented, including the following steps: S1, obtains EEG signals from different brain regions through a high-density cortical EEG electrode array; S2, performing signal preprocessing after performing residual iterative decomposition and signal reconstruction on the EEG signal; S3, using wavelet transform to perform time-frequency analysis on the preprocessed EEG signal to extract the local time-frequency features of the EEG signal; S4, non-parametric Granger causality analysis was used to analyze the Granger causality value between two electrodes based on local time-frequency characteristics; S5, using the single-point to multi-point mean method to project the Granger causal values ​​of brain regions to electrode sites and visualize the spatial and temporal patterns of information flow interactions in different directions between brain regions.

[0054] The embodiment further provides a computer-readable storage medium, characterized in that a program is stored thereon, and when the program is executed by a processor, the above-mentioned EEG signal analysis method based on Granger causality analysis is implemented, including the following steps: S1, obtains EEG signals from different brain regions through a high-density cortical EEG electrode array; S2, performing signal preprocessing after performing residual iterative decomposition and signal reconstruction on the EEG signal; S3, using wavelet transform to perform time-frequency analysis on the preprocessed EEG signal to extract the local time-frequency features of the EEG signal; S4, non-parametric Granger causality analysis was used to analyze the Granger causality value between two electrodes based on local time-frequency characteristics; S5, using the single-point to multi-point mean method to project the Granger causal values ​​of brain regions to electrode sites and visualize the spatial and temporal patterns of information flow interactions in different directions between brain regions.

[0055] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for analyzing electroencephalogram signals based on Granger causality analysis, characterized in that: The following steps are involved: Acquire EEG signals from different brain regions through a high-density cortical EEG electrode array; The EEG signal is subjected to residual iterative decomposition and signal reconstruction and then signal preprocessing is performed; Wavelet transform is used to perform time-frequency analysis on the preprocessed EEG signals to extract local time-frequency features of the EEG signals. Non-parametric Granger causality analysis was used to analyze the Granger causality between two electrodes based on local time-frequency characteristics. The Granger causal values ​​of brain regions were projected to electrode sites using a single-point to multi-point mean approach, and the spatial and temporal patterns of information flow interactions in different directions between brain regions were visualized.

2. The method for analyzing electroencephalogram signals based on Granger causality analysis according to claim 1, characterized in that: Perform residual iterative decomposition and signal reconstruction on the EEG signal, including: The event-related potentials (ERPs) corresponding to the EEG signals are decomposed into EEG component clusters with different latency variability through an iterative process, including initial latency estimation, decomposition based on stimulus occurrence, reaction time and single trial latency, and re-estimation of latency using specified components as templates. The above steps are iterated until convergence.

3. The method for analyzing electroencephalogram signals based on Granger causality analysis according to claim 1, characterized in that: Signal preprocessing includes signal segmentation, bandpass filtering, denoising, re-reference processing, and normalization processing.

4. The method for analyzing electroencephalogram signals based on Granger causality analysis according to claim 1, characterized in that: When using non-parametric Granger causality analysis, the analysis is divided into two types of information flow: bottom-up and top-down, corresponding to the directions from inferior temporal cortex to prefrontal cortex and from prefrontal cortex to inferior temporal cortex, respectively.

5. The method for analyzing electroencephalogram signals based on Granger causality analysis according to claim 1, characterized in that: Visualizing the interactive information flow includes mapping the non-parametric Granger causal value back to the cerebral cortex using the single-point to multi-point mean to visualize the spatiotemporal pattern of the interactive information flow.

6. The method for analyzing electroencephalogram signals based on Granger causality analysis according to claim 1, characterized in that: Visualizing the interactive information flow also includes: constructing a visualization of information flow between brain regions by calculating the causal indicators between each brain region and all other brain regions.

7. The method for analyzing electroencephalogram signals based on Granger causality analysis according to claim 1, characterized in that: Complex wavelet transform is used to perform time-frequency decomposition on the preprocessed signal.

8. An electroencephalogram signal analysis device based on Granger causality analysis, characterized in that: include: A signal acquisition module, which is used to acquire EEG signals from different brain regions through a high-density cortical EEG electrode array; A signal processing module, which is used to perform signal preprocessing after performing residual iterative decomposition and signal reconstruction on the EEG signal; A time-frequency analysis module is used to perform time-frequency analysis on the preprocessed EEG signal using wavelet transform to extract local time-frequency features of the EEG signal; A causal analysis module, which is used to use non-parametric Granger causal analysis and analyze the Granger causal value between two electrodes based on local time-frequency characteristics; The visualization module is used to project the Granger causal values ​​of brain regions to electrode sites by using a single-point to multi-point mean method and to visualize the spatial and temporal patterns of information flow interactions in different directions between brain regions.

9. A computing device comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that: When the one or more processors execute the executable code, they are used to implement the EEG signal analysis method based on Granger causality analysis described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the electroencephalogram signal analysis method based on Granger causality analysis described in any one of claims 1 to 7 is implemented.

Images