Signal processing method and system based on electroencephalogram microstate trajectory and symbolic features

Abstract

The application provides a signal processing method and system based on electroencephalogram microstate trajectory and symbol characteristics, and relates to the technical field of electroencephalogram signal processing.The method comprises the following steps: S1, acquiring electroencephalogram signal data and performing electroencephalogram signal preprocessing; S2, constructing an electroencephalogram microstate symbol sequence, and performing scale unification processing on the discrete electroencephalogram microstate symbol sequence; S3, generating an electroencephalogram microstate chaotic game representation trajectory distance sequence and analyzing the trajectory characteristics; S4, analyzing and verifying the electroencephalogram microstate symbol sequence, performing complexity analysis and substitution verification on the electroencephalogram microstate symbol sequence; and S5, outputting electroencephalogram state dynamics characteristics and performing electroencephalogram dynamics state recognition.The application converts the discrete electroencephalogram microstate symbol sequence into a continuous geometric trajectory through chaotic game, analyzes the spectrum and continuous trajectory complexity, verifies the symbol sequence complexity, realizes multi-dimensional representation of electroencephalogram state dynamics, recognizes various electroencephalogram dynamics states, and provides a new signal processing method for electroencephalogram state recognition.

Description

Technical Field

[0001] This invention relates to the field of electroencephalogram (EEG) signal processing technology, specifically to a signal processing method and system based on EEG microstate trajectories and symbolic features. Background Technology

[0002] Anesthesia status monitoring is a crucial component of modern clinical anesthesia. Stable and accurate anesthesia status assessment can reduce the risk of intraoperative awareness, decrease the overdose of anesthetic drugs, and improve perioperative patient safety. Currently, clinical anesthesia status assessment mainly relies on bispectral index (BPI), vital signs, and the anesthesiologist's experience. However, existing anesthesia monitoring methods still have limitations, such as insufficient adaptability to various anesthetic drugs and difficulty in revealing the mechanisms of dynamic changes in brain status.

[0003] In recent years, EEG microstate analysis has attracted widespread attention due to its ability to reflect the transient functional state switching process of the brain. EEG microstates are typically represented as discrete symbol sequences, with each category arranged chronologically to form a symbolic sequence. Existing EEG microstate research mainly focuses on the statistical characteristics of EEG microstates, such as duration, probability of occurrence, transition probability, and complexity. However, most of this research remains at the level of discrete symbols, lacking a systematic analysis of the continuous dynamic characteristics of EEG microstates.

[0004] Chaotic game theory is a method that maps discrete symbol sequences to continuous geometric trajectories. Previous studies have applied the chaotic game representation of EEG microstates (CGR) to EEG microstate analysis and performed spectral analysis based on trajectory distance sequences to study the periodic characteristics of changes in conscious states. However, existing methods mainly focus on spectral analysis, lacking joint analysis of trajectory complexity, unpredictability, and multi-scale dynamics. Furthermore, the relationship between the trajectory analysis results of the chaotic game representation of EEG microstates (CGR) and the complexity of the original symbol sequence, as well as whether changes in complexity exceed the influence of state proportions and first-order transition structures, still lack effective verification.

[0005] Furthermore, existing studies typically analyze data from various states of consciousness by mixing them together, which can easily lead to the cancellation of dynamic characteristics between states, making it difficult to extract state-specific EEG microstate dynamic characteristics. Therefore, there is an urgent need for a dynamic analysis method for anesthesia detection states that can perform continuous dynamic analysis of EEG microstate switching processes and jointly assess periodicity, complexity, and symbol sequence organization characteristics. Summary of the Invention

[0006] To address the shortcomings of the existing technologies, the present invention aims to provide a signal processing method and system based on EEG microstate trajectory and symbol features. This method converts discrete EEG microstate symbol sequences into continuous geometric trajectories using chaotic game representation, and performs spectral analysis, continuous trajectory complexity analysis, and symbol sequence complexity verification based on trajectory distance sequences, forming a multi-dimensional joint evaluation system that includes periodicity, unpredictability, and pattern richness. It characterizes EEG microstate dynamic changes from both the continuous trajectory and symbol sequence levels, and identifies differentiated EEG microstate dynamic patterns induced by various anesthetic drugs for EEG state assessment.

[0007] Specifically, on the one hand, the present invention provides a signal processing method based on EEG microstate trajectories and symbolic features, which includes the following steps: S1: Obtain the EEG signal dataset to be analyzed, read the drug type label and state category label corresponding to the EEG signal dataset to be analyzed; perform preprocessing and output full-band EEG signal, segmented EEG signal and corresponding drug type label and state category label information; S2: Perform EEG microstate analysis on the full-band and segmented EEG signals from S1, construct EEG microstate symbol sequences, and perform scale unification on the discrete EEG microstate symbol sequences to generate corresponding discrete EEG microstate symbol sequences. ; S3: Obtain the discrete EEG microstate symbol sequence from S2 This is mapped to a chaotic game representation of continuous geometric trajectories. ; Generate distance sequence of EEG microstate feature trajectories ; Distance sequence of EEG microstate feature trajectories Spectral analysis, sample entropy calculation, and multi-scale entropy analysis were performed respectively to obtain spectral features, sample entropy features, and multi-scale entropy features. S4: The discrete EEG microstate symbol sequence in S2 Complexity analysis was performed to obtain the normalized signal mode complexity. Set the number of times the alternative data is generated. Construct randomized replacement sequences and first-order Markov replacement sequences; output the deviation parameters between the complexity of the original sequence and the complexity of the replacement sequence. Deviation parameters of first-order Markov substitution sequences Analyze and verify the symbol sequence of brainwave microstates; S5: Combine the spectral features, sample entropy features, and multi-scale entropy features from S3 with the normalized mode complexity LZC and deviation parameters from S4. and Feature-level fusion is performed to form a multidimensional EEG state dynamic feature vector.

[0008] Preferably, S3 specifically comprises: S31: Constructing the EEG Microstate Feature Point Matrix Generate continuous geometric trajectories of EEG microstates Determining the distance sequence of EEG microstate feature trajectories using Euclidean distance. After performing mean-removal processing, the mean-removed trajectory distance sequence is obtained. ; S32: Distance sequences of EEG microstate feature trajectories across the entire frequency band and each sub-band. Power spectral density analysis was performed to extract peak frequency, peak power, and peak-to-background ratio spectral characteristics to characterize the periodicity of micro-state switching. S33: Distance sequence of full-band EEG microstate feature trajectories Perform sample entropy analysis separately to obtain sample entropy features; analyze the trajectory distance sequence. Coarse-graining is performed to obtain coarse-grained trajectory distance sequences at various scales. Determine the multi-scale entropy characteristics of EEG signals.

[0009] Preferably, the current EEG microstate feature points generated in S31 are: ; ; in, For the first One EEG microstate feature point; This represents the (k-1)th EEG microstate feature point; For the first A matrix of EEG microstate feature points corresponding to each time point; For the first The x-coordinate of a reference EEG microstate feature point; For the first The vertical coordinates of one reference EEG microstate feature point; For the first The x-coordinate of each EEG microstate feature point; For the first The vertical coordinates of the EEG microstate feature points.

[0010] Preferably, the distance sequence of EEG microstate feature trajectories in S31 The method for obtaining it is as follows: ; ; in, For the first The distance sequence of EEG microstate feature trajectories corresponding to each time point; For the first The x-coordinate of the EEG microstate feature points corresponding to each time point; For the first The vertical coordinates of the EEG microstate feature points corresponding to each time point; The mean-removed distance sequence of EEG microstate feature trajectories; For the first A sequence of EEG microstate feature trajectories; N is the scale of the trajectory distance sequence.

[0011] Preferably, the multi-scale entropy features of the EEG signal in S33 are as follows: ; ; ; ; in, This is the average parameter of the sample entropy values ​​within the scale range of 1 to 20; For coarse-grained trajectory distance sequences under scale factor τ The calculated sample entropy value; The integral area of ​​the sample entropy curve within the scale range of 1 to 20; This is the average parameter of the sample entropy values ​​within the scale range of 1 to 5; It is the average parameter of the sample entropy values ​​in the scale range of 6 to 20.

[0012] Preferably, S4 specifically comprises: S41: Symbolic sequences of discrete EEG microstates after scale unification Complexity analysis is performed, using signal pattern complexity to calculate the number of patterns appearing in the symbol sequence. Normalization is performed to obtain the normalized signal mode complexity. ; S42: Set the number of times alternative data is generated. A random rearrangement substitution sequence is constructed based on the discrete microstate symbol sequence. The normalized signal mode complexity of each substitution sequence is calculated, and the mean value of the random rearrangement substitution sequence complexity is obtained by taking the mean value. ; S43: Calculate the first-order transition probability based on the original symbol sequence. A first-order transition probability matrix is ​​formed, and a first-order Markov substitution sequence with the same scale as the original symbol sequence is generated by combining the initial state probability. The normalized signal mode complexity is calculated for each substitution sequence, and the mean value of the first-order Markov substitution sequence complexity is obtained. .

[0013] Preferably, in S41, the signal mode complexity is normalized, specifically as follows: ; in, Normalized signal mode complexity; denoted as k, where k is the number of patterns appearing in the symbol sequence at scale n; k is the number of EEG microstate categories; and n is the symbol sequence scale.

[0014] Preferably, the average complexity of the random rearrangement substitution sequence in S42 is as follows: ; in, Let m be the complexity feature of the random rearrangement substitution sequence; The average complexity of the replacement sequence is that of random rearrangement; This represents the total number of random rearrangement replacement sequences.

[0015] Preferably, the average complexity of the first-order Markov substitution sequence in S43 is as follows: ; in, Let m be the complexity feature of the first-order Markov substitution sequence; This represents the average complexity of a first-order Markov substitution sequence.

[0016] On the other hand, the present invention provides a signal processing system based on a signal processing method for EEG microstate trajectories and symbol features, comprising: an EEG signal data acquisition module, a data preprocessing module, an EEG microstate symbol sequence extraction module, an analysis scale determination module, an EEG microstate trajectory generation module, a trajectory feature extraction module, a symbol sequence verification module, a feature fusion module, and a state discrimination module.

[0017] The EEG signal data acquisition module acquires the EEG signal dataset to be analyzed and reads the corresponding drug type label and state category label of the EEG signal dataset to be analyzed.

[0018] The data preprocessing module preprocesses the EEG signal dataset to be analyzed and obtains full-band EEG signals and multiple sub-band EEG signals based on the preprocessed EEG signals.

[0019] The EEG microstate symbol sequence extraction module performs EEG microstate analysis on full-band EEG signals and multiple sub-band EEG signals to obtain the EEG microstate categories under the corresponding frequency bands and generate the corresponding discrete EEG microstate symbol sequences.

[0020] The analysis scale determination module performs a complexity index convergence test on discrete EEG microstate symbol sequences, determines a unified analysis scale while ensuring the stability of the complexity index estimation, and performs scale unification processing on discrete EEG microstate symbol sequences according to the unified analysis scale.

[0021] The EEG microstate trajectory generation module uses chaotic game representation mapping on the discrete EEG microstate symbol sequence after scale unification to generate continuous geometric trajectory of EEG microstate chaotic game representation, and calculates the distance sequence of EEG microstate feature trajectory based on the continuous geometric trajectory of EEG microstate chaotic game representation.

[0022] The trajectory feature extraction module performs power spectral density analysis, sample entropy analysis, and multi-scale entropy analysis on the trajectory distance sequence of EEG microstate features to obtain spectral features, sample entropy features, and multi-scale entropy features.

[0023] The symbol sequence verification module performs complexity analysis on the scale-unified discrete EEG microstate symbol sequences, constructs alternative sequences based on the discrete EEG microstate symbol sequences for verification, and obtains symbol sequence complexity features and alternative data verification results.

[0024] The feature fusion module fuses spectral features, sample entropy features, multi-scale entropy features, normalized signal pattern complexity, and deviation parameters at the feature level to generate a multidimensional EEG state dynamic feature vector.

[0025] The state discrimination module inputs the multidimensional EEG state dynamic feature vector into the pre-established consciousness state discrimination model to determine the dynamic state category corresponding to the EEG signal to be analyzed. The dynamic state categories include deinhibited active state and frozen state.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention maps discrete EEG microstate symbol sequences into continuous geometric trajectories using a chaotic game representation method, and further constructs EEG microstate feature trajectory distance sequences, thereby realizing the continuous and quantitative expression of discrete microstate switching information and providing a unified data foundation for subsequent dynamic feature analysis.

[0027] (2) This invention performs spectral analysis on the distance sequence of the EEG microstate feature trajectory corresponding to the anesthesia detection state. The full-band spectral analysis and feature extraction are used as the main analysis results, and the sub-band spectral analysis and feature extraction are used as supplementary analysis results. By combining the full-band and sub-band analysis strategies, the frequency band source can be explained when the full-band periodicity is significant, and the EEG frequency band with relatively obvious periodicity can be located when the full-band periodicity is not significant. In this way, the state-specific periodicity of each anesthesia detection state can be extracted, and the influence of the mixture of dynamic features of each state on the analysis results can be reduced.

[0028] (3) By performing sample entropy analysis and multi-scale entropy analysis on the distance sequence of EEG microstate feature trajectories, this invention can characterize the unpredictability and multi-scale complexity changes of EEG microstate switching trajectories in anesthesia detection state, and provides a complexity assessment dimension in addition to spectral periodicity analysis.

[0029] (4) This invention verifies the original EEG microstate symbol sequence through signal pattern complexity analysis, and combines random rearrangement substitution sequence and first-order Markov substitution sequence to determine whether the complexity change exceeds the influence of state ratio and first-order transition structure, thereby improving the reliability and interpretability of the analysis results.

[0030] (5) The present invention integrates spectral features, sample entropy features, multi-scale entropy features, normalized signal pattern complexity and alternative sequence deviation parameters at the feature level to form a multi-dimensional EEG state dynamic feature vector, which can characterize the differences in EEG microstate dynamics under various anesthesia detection states from multiple dimensions such as periodicity, unpredictability, multi-scale complexity and symbol pattern organization.

[0031] (6) The present invention identifies the brain dynamic state based on the multidimensional brain state dynamic feature vector, which can distinguish between the de-inhibited active state and the low-frequency frozen state, providing a basis for brain state monitoring, state recording and subsequent analysis. Attached Figure Description

[0032] Figure 1 This is an overall flowchart of the signal processing method based on EEG microstate trajectory and symbol features of the present invention; Figure 2 This is a schematic diagram of the chaotic game representation mapping of EEG microstates and the generation of EEG microstate feature trajectory distance sequences in this invention; Figure 3 This is a flowchart of the spectral analysis of the distance sequence of EEG microstate feature trajectories in this invention; Figure 4 This is a flowchart of the sample entropy analysis and multi-scale entropy analysis of the distance sequence of EEG microstate feature trajectories in this invention; Figure 5 This is a flowchart of the signal pattern complexity analysis and alternative data verification of EEG microstate symbol sequences in this invention. Figure 6 This is a framework diagram of the three-dimensional joint analysis method in this invention. Detailed Implementation

[0033] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0034] This invention proposes a signal processing method based on EEG microstate trajectories and symbolic features, such as... Figure 1As shown, the procedure involves acquiring EEG signal data and performing EEG signal preprocessing; constructing EEG microstate symbol sequences and performing scale unification processing on discrete EEG microstate symbol sequences; generating trajectory distance sequences and analyzing trajectory features using chaotic game representation of EEG microstates; analyzing and verifying EEG microstate symbol sequences, performing complexity analysis and substitution verification of EEG microstate symbol sequences; and outputting EEG state dynamic features for EEG dynamic state recognition. Specifically, the procedure includes the following steps: S1: Acquire EEG signal data and perform EEG signal preprocessing. Acquire the EEG signal dataset to be analyzed, and read the corresponding drug type label and state category label; preprocess the EEG signal dataset to be analyzed, and obtain full-band EEG signal data and multiple sub-band EEG signal data based on the preprocessed EEG signal data.

[0035] The inputs for this step are the EEG signal dataset to be analyzed, drug type labels, and state category labels. In this embodiment of the invention, the drug type labels include: propofol labels and dexmedetomidine labels. For the EEG signal data corresponding to the propofol label, the state category labels include: baseline signal state, mild signal inhibition state, moderate signal inhibition state, and signal recovery state; for the EEG signal data corresponding to the dexmedetomidine label, the state category labels include: baseline signal state, deep signal inhibition state, and signal recovery state.

[0036] In this embodiment of the invention, the data corresponding to the propofol label includes four state category labels, each containing 20 EEG signal data sequences, for a total of 80 EEG signal data sequences. The data corresponding to the dexmedetomidine label includes three state category labels, each containing 20 EEG signal data sequences, for a total of 60 EEG signal data sequences. The EEG signal data sequences are derived from multiple independent EEG signal data samples and are used to verify the applicability of the invention under each drug state category label.

[0037] In this embodiment of the invention, EEG signal preprocessing includes one or more of the following: filtering, denoising, artifact removal, rereference, and segmentation, to reduce the impact of noise, bioelectrical interference, and electrode artifacts on subsequent EEG microstate symbol sequence extraction. Preprocessing parameters are determined based on the parameters of the EEG signal data acquisition device, sampling rate, data segmentation scale, and data quality. EEG microstates are transient, quasi-stable, and globally synchronized EEG field topologies obtained from scalp EEG data acquisition devices. Their temporal characteristics are: single duration of 60-120 ms, rapid switching, and no intermediate transition states; spatial characteristics are: a fixed and unique voltage distribution topology formed by electrodes across the entire scalp. The term "micro" here does not refer to spatial insignificance but has two meanings: firstly, it is small and transient in terms of time scale; the rapid millisecond-level switching of the global electric field configuration and the extremely short window for maintaining stable EEG microstates are called micro-time segments. Secondly, it refers to the microscopic building blocks of the analysis unit; in academia, microstates are called the microscopic atoms of thought, the smallest microscopic units for dissecting spontaneous brain activity, and the basic micro-modules constituting complex brain activity.

[0038] In this embodiment of the invention, preprocessing yields a full-band EEG signal and multiple sub-band EEG signals. The full-band EEG signal is used to subsequently generate a full-band EEG microstate symbol sequence and the corresponding EEG microstate feature trajectory distance sequence. Multiple frequency-segmented EEG signals are used to subsequently generate corresponding frequency-segmented EEG microstate symbol sequences and EEG microstate feature trajectory distance sequences. .

[0039] In this embodiment of the invention, the multiple frequency-segmented EEG signals include one or more of the following: ultra-low frequency characteristic signal band, low frequency characteristic signal band, mid-frequency characteristic signal band, high frequency characteristic signal band, and ultra-high frequency characteristic signal band; wherein, the ultra-low frequency characteristic signal band corresponds to EEG signals with a frequency range of 1Hz to 4Hz; the low frequency characteristic signal band corresponds to EEG signals with a frequency range of 4Hz to 8Hz; the mid-frequency characteristic signal band corresponds to EEG signals with a frequency range of 8Hz to 13Hz; the high frequency characteristic signal band corresponds to EEG signals with a frequency range of 13Hz to 30Hz; and the ultra-high frequency characteristic signal band corresponds to EEG signals with a frequency range of 30Hz to 45Hz.

[0040] Frequency band division is based on preprocessed EEG signals. Feature signal components within corresponding frequency ranges are extracted using bandpass filtering to obtain multiple frequency-band EEG signals. These multiple frequency-band EEG signals are used for subsequent extraction of frequency-band EEG microstate symbol sequences, EEG microstate chaotic game representation (CGR mapping), and EEG microstate feature trajectory distance sequences. generate.

[0041] The output of this step is the preprocessed full-band EEG signal, multiple sub-band EEG signals, and corresponding drug type labels and status category labels.

[0042] S2: Constructing EEG microstate symbol sequences and performing scale unification processing on discrete EEG microstate symbol sequences. EEG microstate analysis is performed on the full-band EEG signal and multiple sub-band EEG signals output from S1 to obtain the EEG microstate categories under the corresponding frequency bands and generate corresponding discrete EEG microstate symbol sequences. The complexity index convergence test is performed on the discrete EEG microstate symbol sequences. A unified analysis scale is determined while ensuring the stability of the complexity index estimation, and the discrete EEG microstate symbol sequences are then scale-unified according to the unified analysis scale.

[0043] In this embodiment of the invention, there are four categories of EEG microstates, each corresponding to one of four EEG topological patterns. For the full-band EEG signal and the EEG signals of each sub-band, corresponding symbolic labels are assigned according to the EEG microstate category to which the EEG signal belongs at each moment. The symbolic labels are arranged in chronological order to form a discrete EEG microstate symbol sequence under the corresponding frequency band.

[0044] Discrete EEG microstate symbolic sequence is ;in, It is a discrete EEG microstate symbol sequence; For the first The EEG microstate category label corresponding to each time point; The scale is based on the symbolic sequence of EEG microstates; Indexed by time point.

[0045] In this embodiment of the invention, the four types of EEG microstates are represented by numerical labels 1, 2, 3, and 4 in the storage file, where numerical labels 1, 2, 3, and 4 correspond to the four types of EEG microstates A, B, C, and D, respectively. When reading the sequence, the numerical labels in the EEG microstate symbol sequence file are converted into a one-dimensional sequence, and invalid or missing values ​​are removed.

[0046] In this embodiment of the invention, a full-band EEG signal corresponds to a full-band discrete EEG microstate symbol sequence, and multiple sub-band EEG signals correspond to multiple sub-band discrete EEG microstate symbol sequences. These discrete EEG microstate symbol sequences are used to characterize the switching process of EEG microstates over time in each frequency band.

[0047] In this embodiment of the invention, Lempel-Ziv complexity is used as the convergence test index. Specifically, discrete EEG microstate symbol sequences are truncated at multiple preset truncation scales, and the normalized signal pattern complexity (Lempel-Ziv complexity) at each truncation scale is determined. Then, a unified analysis scale is determined based on the variation range of the complexity index at each truncation scale. In this embodiment of the invention, the preset truncation scales include 500, 1000, 2000, 3000, 5000, 8000, 10000, 15000, 20000, 30000, 40000, and 45000.

[0048] Several EEG signal data sequences were randomly selected under each state category label. The normalized signal pattern complexity (Lempel-Ziv complexity) at each truncation scale was determined, and the average convergence curve of the multiple sequences was obtained. When the difference in the mean complexity corresponding to adjacent truncation scales is less than a preset threshold, the complexity index is considered to have reached a stable estimate; the preset threshold is 0.005.

[0049] In this embodiment, after convergence testing, when the truncation scale reaches 30,000, the difference in normalized signal pattern complexity (Lempel-Ziv complexity) between adjacent scales is less than a preset threshold, and the complexity change tends to stabilize. To unify the input scale for subsequent EEG microstate chaotic game representation CGR mapping, spectrum analysis, sample entropy analysis, multi-scale entropy analysis, and signal pattern complexity (Lempel-Ziv complexity) analysis, and to reduce computational load, the discrete EEG microstate symbol sequences corresponding to the entire frequency band and each sub-band are uniformly truncated into the first 30,000 label points as input for subsequent analysis.

[0050] The output of this step is a scale-unified full-band discrete EEG microstate symbol sequence and a scale-unified multi-band discrete EEG microstate symbol sequence.

[0051] S3: Generation and trajectory feature analysis of CGR trajectory distance sequence based on chaotic game representation of EEG microstates. Obtain the discrete EEG microstate symbol sequence obtained in S2. And it is mapped to a chaotic game representation of the CGR continuous geometric trajectory. ; Generate distance sequence of EEG microstate feature trajectories ; Distance sequence of EEG microstate feature trajectories Spectral analysis, sample entropy calculation, and multi-scale entropy analysis were performed to obtain spectral features, sample entropy features, and multi-scale entropy features. Spectral features were used to determine the macroscopic energy migration trend of the signal under anesthesia; sample entropy features were used to determine the unpredictability of microscopic state switching; and multi-scale entropy features were used to determine the degree of breakage of spatiotemporal coupling across time scales under the action of anesthetic drugs. These features were then input into a pre-defined consciousness state discrimination model to distinguish whether the EEG signal under analysis was in a de-inhibited active state or a frozen state.

[0052] S31: Using the chaotic game representation of EEG microstates to generate CGR trajectory, a chaotic game representation mapping is performed on the scale-unified discrete EEG microstate symbol sequence to generate the continuous geometric trajectory of EEG microstate chaotic game representation CGR. Then, the distance sequence of EEG microstate feature trajectory across the entire frequency band is calculated based on the continuous geometric trajectory of EEG microstate chaotic game representation CGR. Specifically, the discrete EEG microstate symbol sequences generated from full-band EEG signals are subjected to EEG microstate chaotic game representation (CGR) mapping to obtain the full-band EEG microstate feature trajectory distance sequence. The discrete EEG microstate symbol sequences generated from EEG signals of each frequency band are subjected to EEG microstate chaotic game representation (CGR) mapping to obtain the EEG microstate feature trajectory distance sequences for each frequency band. .

[0053] Distance sequence of EEG microstate features across the entire frequency band Frequency band EEG microstate feature trajectory distance sequence Spectral analysis was performed separately to obtain full-band and sub-band spectral characteristics; the distance sequence of EEG microstate characteristic trajectories across the full-band frequency range was analyzed. Sample entropy analysis and multi-scale entropy analysis were performed to obtain sample entropy features and multi-scale entropy features of EEG signals. Spectral features were used to characterize the periodicity of EEG microstate switching, sample entropy features were used to characterize the unpredictability of EEG microstate switching trajectories, and multi-scale entropy features of EEG signals were used to characterize the multi-scale complexity of EEG microstate switching trajectories. Figure 2 As shown, a corresponding number of feature points are set in the two-dimensional geometric space according to the number of EEG microstate categories. In this embodiment of the invention, when there are four EEG microstate categories, four feature points are set in the two-dimensional geometric space. The correspondence between the four EEG microstate categories and the feature points is as follows: The first type of EEG microstate corresponds to feature point A(-1,2); the second type of EEG microstate corresponds to feature point B(1,2); the third type of EEG microstate corresponds to feature point C(1,0); the fourth type of EEG microstate corresponds to feature point D(-1,0); the EEG microstate feature point matrix is ​​constructed as follows: ; in, The EEG microstate feature point matrix is ​​a feature point matrix composed of feature points corresponding to each EEG microstate category. Each row of the matrix represents the two-dimensional coordinates corresponding to an EEG microstate category.

[0054] The above EEG microstate feature point matrix This is used to map discrete EEG microstate categories to a two-dimensional geometric space, enabling changes in EEG microstate categories to be converted into changes in geometric trajectories. If the... The EEG microstate category label for each time point is: Then the coordinates of its corresponding feature point are expressed as: ; in, For the first A matrix of EEG microstate feature points corresponding to each time point; For the first The EEG microstate category label corresponding to each time point; The x-coordinate of the EEG microstate feature point; The vertical coordinate of the EEG microstate feature point; For the x-axis parameter; This is the parameter for the ordinate.

[0055] Each EEG microstate symbol is read sequentially according to the chronological order of the discrete EEG microstate symbol sequence, and the current EEG microstate feature point is generated iteratively based on the feature point position corresponding to the current EEG microstate symbol and the feature point position of the previous EEG microstate. In this embodiment of the invention, the reference EEG microstate feature point is represented as follows: ; in, For reference EEG microstate feature points; The x-coordinate of the reference EEG microstate feature point; The vertical coordinates of the reference EEG microstate feature points.

[0056] The current EEG microstate feature points are generated as follows: ; ; in, For the first One EEG microstate feature point; This represents the (k-1)th EEG microstate feature point; For the first A matrix of EEG microstate feature points corresponding to each time point; For the first The x-coordinate of a reference EEG microstate feature point; For the first The vertical coordinates of one reference EEG microstate feature point; For the first The x-coordinate of each EEG microstate feature point; For the first The vertical coordinates of the EEG microstate feature points.

[0057] The above iterative process transforms the discrete EEG microstate symbol sequence into the EEG microstate chaotic game representation CGR continuous geometric trajectory, converting the discrete switching process of EEG microstates into a continuous change process of spatial trajectory. Arranging all EEG microstate feature points in chronological order, the EEG microstate chaotic game representation CGR continuous geometric trajectory is obtained as follows: ; in, The continuous geometric trajectory is composed of EEG microstate feature points, which is composed of all EEG microstate chaotic game representations (CGR) EEG microstate feature points.

[0058] S32: Calculate the trajectory distance sequence based on the CGR continuous geometric trajectory represented by the brainwave microstate chaotic game. In this embodiment of the invention, the distance sequence of EEG microstate feature trajectories... It is composed of the Euclidean distances from each EEG microstate feature point to the reference feature point, specifically: ; in, For the first The distance sequence of EEG microstate feature trajectories corresponding to each time point is composed of the distance from the CGR EEG microstate feature point to the reference feature point P0, represented by the EEG microstate chaotic game. For the first The chaotic game representation of the brainwave microstate feature points corresponding to each time point in CGR brainwave microstate is the x-coordinate of the feature points of the brainwave microstate. For the first The chaotic game representation of the EEG microstate feature points at each time point is the ordinate of the CGR EEG microstate feature points.

[0059] In this embodiment of the invention, the distance sequence of EEG microstate feature trajectories... Compressing the two-dimensional trajectories of EEG microstate categories over time into one-dimensional continuous distance signals facilitates subsequent spectral analysis, sample entropy analysis, and multi-scale entropy analysis. This is achieved by calculating the distance sequence of the characteristic trajectories of EEG microstates. Subsequently, the distance sequence of EEG microstate feature trajectories was analyzed. Perform mean-removal processing to obtain the mean-removed trajectory distance sequence, and still denote the mean-removed sequence as... Specifically: ; in, The mean-removed distance sequence of EEG microstate feature trajectories; For the first A sequence of EEG microstate feature trajectories; N is the scale of the trajectory distance sequence.

[0060] By analyzing the distance sequence of brainwave microstate feature trajectories Mean-reduction processing can reduce the impact of differences in the average distance between sequences on spectral and complexity features, allowing subsequent features to more effectively reflect the dynamic changes in the EEG microstate switching trajectory. For example... Figure 3 As shown, the distance sequence of EEG microstate feature trajectories Spectral analysis is performed. In this embodiment of the invention, the Welch method is used to estimate the distance sequence of EEG microstate feature trajectories. The power spectrum was calculated. Parameters were set as follows: sampling rate Fs = 250 Hz, Welch window length of 250 sampling points, window overlap scale nooverlap = 0, and FFT point count nfft = 4096. The distance sequence of each EEG microstate feature trajectory was calculated. After performing Welch power spectrum estimation, the power spectrum is converted to dB form, specifically as follows: ; in, The power spectral density of the distance sequence of EEG microstate feature trajectories at frequency f; This is the power spectral density value after conversion to dB; It is a logarithmic function; This is a frequency parameter.

[0061] In this embodiment of the invention, the wideband range corresponding to the full-band EEG signal is 1-45 Hz; the multiple sub-band EEG signals include one or more of the following: ultra-low frequency characteristic signal band, low frequency characteristic signal band, mid-frequency characteristic signal band, high frequency characteristic signal band, and ultra-high frequency characteristic signal band; wherein, the ultra-low frequency characteristic signal band corresponds to EEG signals with a frequency range of 1Hz to 4Hz; the low frequency characteristic signal band corresponds to EEG signals with a frequency range of 4Hz to 8Hz; the mid-frequency characteristic signal band corresponds to EEG signals with a frequency range of 8Hz to 13Hz; the high frequency characteristic signal band corresponds to EEG signals with a frequency range of 13Hz to 30Hz; and the ultra-high frequency characteristic signal band corresponds to EEG signals with a frequency range of 30Hz to 45Hz. The distance sequence of the EEG microstate feature trajectory generated from the EEG signals of each frequency band is then considered. Set the corresponding peak search range in its power spectrum. For example, the full-band EEG signal corresponds to 6-80 Hz, the ultra-low frequency characteristic signal band corresponds to 6-20 Hz, the low frequency characteristic signal band corresponds to 6-20 Hz, the mid-frequency characteristic signal band corresponds to 10-30 Hz, the high frequency characteristic signal band corresponds to 20-50 Hz, and the high-frequency characteristic signal band corresponds to 50-80 Hz.

[0062] The peak frequency and peak power of the EEG signal are extracted within the peak search range; the peak frequency of the EEG signal is the frequency corresponding to the maximum power in the power spectrum, and the peak power of the EEG signal is the power value corresponding to that frequency point; the peak-to-background ratio of the EEG signal is further determined as follows: ; in, The peak-to-background ratio of the EEG signal; This represents the peak power of the EEG signal; The average power of the background frequency band after excluding frequencies near the peak.

[0063] In an embodiment of the present invention, The calculations were performed within a background frequency range of 2-80 Hz, excluding frequency bands near the peak frequency. Peak frequency, peak power, and peak-to-background ratio together constitute the spectral characteristics used to characterize the periodic changes during EEG microstate switching. In this embodiment, the distance sequence of the EEG microstate feature trajectory corresponding to each signal state is calculated. Independent spectral analysis is performed, rather than a unified analysis of data from all signal states. Distance sequences of EEG microstate feature trajectories generated from full-band EEG signals are used. The full-band spectral characteristics were obtained as the main analysis result; the distance sequence of EEG microstate feature trajectories generated from EEG signals in each frequency band was also analyzed. The spectral characteristics of each frequency band are obtained as supplementary analysis results to aid in interpreting the full-band spectral analysis results and identifying EEG frequency bands with relatively obvious periodic characteristics. For example... Figure 4 As shown, the distance sequence of EEG microstate feature trajectories Sample entropy analysis and multi-scale entropy analysis are performed. Sample entropy is used to characterize the unpredictability of trajectory distance sequences. It measures the probability that similar local patterns in a trajectory distance sequence will remain similar at subsequent time points. The larger the value, the more irregular the local patterns of the trajectory distance sequence at that scale, and the stronger the unpredictability of the EEG microstate switching trajectory; the smaller the value, the more regular the local patterns of the trajectory distance sequence at that scale, and the higher the stability of the EEG microstate switching trajectory. The embedding dimension and tolerance parameters used in the sample entropy calculation are consistent with those in the aforementioned sample entropy analysis. In this embodiment of the invention, the sample entropy analysis parameters are set as follows: ; Where m is the embedding dimension; r is the tolerance parameter; Distance sequence of EEG microstate feature trajectories The standard deviation.

[0064] In specific calculations, the distance sequence of EEG microstate feature trajectories is used. The data is truncated to a uniform scale of 30,000 and then mean-removed; if the distance sequence of EEG microstate feature trajectories... If the standard deviation is less than or equal to 0, then the sequence is not included in the sample entropy analysis.

[0065] Multiscale entropy is used to characterize the trajectory distance sequence of EEG microstate features. Complexity across multiple time scales. Specifically, the distance sequence of EEG microstate feature trajectories. Coarse-graining is performed. Under a scale factor τ, the original sequence is divided into several continuous non-overlapping segments, and the average value of each segment is calculated to obtain the coarse-grained sequence at scale τ, specifically: ; in, scale factor The first in the coarse-grained sequence One feature point; Scale factor; The first in the original trajectory distance sequence One feature point; For feature point indexes.

[0066] At each scale, sample entropy is calculated for the distance sequence of coarse-grained EEG microstate feature trajectories. For the scale factor... The distance sequence of the coarse-grained EEG microstate feature trajectory is denoted as: ; in, The distance sequence of the trajectory of the coarse-grained EEG microstate feature with scale factor τ; For the first coarse-grained sequence One feature point; scale factor The scale of the coarse-grained sequence.

[0067] S33: In this embodiment, using Represents the coarsened trajectory distance sequence under scale factor τ The calculated sample entropy value. The original EEG microstate feature trajectory distance sequence was used. The standard deviation determines the tolerance parameters at each scale, enabling comparability of sample entropy results across different scales; multi-scale entropy characteristics of EEG signals include: ; ; ; ; in, The average parameter of the sample entropy values ​​in the scale range of 1 to 20 is used to characterize the average complexity of the trajectory distance sequence across the entire scale range; For coarse-grained trajectory distance sequences under scale factor τ The calculated sample entropy value; The integral area of ​​the sample entropy curve within the scale range of 1 to 20 is used to comprehensively characterize the overall cumulative complexity of the trajectory distance sequence across multiple time scales. The average parameter of the sample entropy values ​​within the scale range of 1 to 5 is used to characterize the local complexity of the trajectory distance sequence at a short time scale. It is the average parameter of the sample entropy values ​​in the scale range of 6 to 20, used to characterize the complexity of the trajectory distance sequence over a longer time scale.

[0068] The aforementioned sample entropy features and EEG signal multi-scale entropy features together constitute trajectory complexity features, which are used to characterize the dynamic changes of EEG microstate switching trajectories from two aspects: unpredictability and multi-scale complexity.

[0069] The output of this step is a full-band EEG microstate feature trajectory distance sequence. Frequency-segmented EEG microstate feature trajectory distance sequence The features include spectral features, sample entropy features, and multi-scale entropy features of EEG signals; among them, spectral features are used to characterize the periodic changes during EEG microstate switching; sample entropy features are used to characterize the distance sequence of EEG microstate feature trajectories. Unpredictability; Multi-scale entropy features of EEG signals are used to characterize the trajectory distance sequence of EEG microstate features. Complexity changes at different time scales.

[0070] S4: Analyze and verify the EEG microstate symbol sequence, performing complexity analysis and alternative verification of the EEG microstate symbol sequence. The input for this step is the discrete EEG microstate symbol sequence after scale unification processing.

[0071] S41: As Figure 5As shown, Lempel-Ziv complexity analysis was performed on the scale-unified discrete EEG microstate symbolic sequences output by S2. Alternate sequences were constructed based on these discrete EEG microstate symbolic sequences for validation, yielding symbolic sequence complexity features and validation results using alternative data. The symbolic sequence complexity features characterize the richness of EEG microstate switching patterns, while the validation results using alternative data determine whether the complexity features exceed the influence of state proportions and first-order transition structures.

[0072] In this embodiment of the invention, the discrete EEG microstate category label sequence after scaling is converted into a symbol sequence. For example, the numerical labels 1, 2, 3, and 4 are converted into letter symbols A, B, C, and D, respectively, resulting in the EEG microstate category symbol sequence. ;in, This is a symbolic sequence for EEG microstate categories, obtained by converting EEG microstate category labels. is the symbol for the nth EEG microstate category; n is the symbol sequence scale, in this embodiment, n=30000.

[0073] The Lempel-Ziv complexity algorithm is used to calculate the number of patterns c(n) in a symbol sequence. To reduce the impact of sequence scale and the number of symbol classes on the complexity result, the Lempel-Ziv complexity is normalized, specifically as follows: ; in, This is the normalized signal pattern complexity, also known as Lempel-Ziv complexity. is the number of patterns appearing in the symbol sequence at scale n; k is the number of EEG microstate categories. In this embodiment, k=4, corresponding to four EEG microstate categories; n is the symbol sequence scale.

[0074] The above normalization process is used to reduce the impact of sequence scale and the number of symbol categories on the complexity estimation results, making the complexity results obtained under each state category label comparable. A higher normalized signal pattern complexity (Lempel-Ziv complexity) indicates more patterns appearing during EEG microstate switching, and a richer variety of EEG microstate switching patterns; a lower normalized signal pattern complexity (Lempel-Ziv complexity) indicates more regular EEG microstate switching patterns and more obvious repetitive structures.

[0075] In this embodiment of the invention, the Lempel-Ziv complexity calculation method is used to calculate the complexity of the scale-unified discrete EEG microstate symbol sequences, thereby improving computational efficiency. The obtained Lempel-Ziv complexity result is used as the distance sequence of EEG microstate feature trajectories. Symbol sequence level comparison index of complexity analysis results.

[0076] S42: Construct random rearrangement substitution sequences and first-order Markov substitution sequences based on the original discrete EEG microstate symbol sequences, and calculate their normalized signal pattern complexity (Lempel-Ziv complexity) respectively. In this embodiment of the invention, the number of times the substitution data is repeatedly generated... The randomized replacement sequence is obtained by randomly rearranging the original discrete EEG microstate symbol sequence. This replacement sequence retains the proportion of each EEG microstate category in the original sequence but disrupts the original temporal order. For each original symbol sequence, 20 randomized replacement sequences are generated, and their normalized signal pattern complexity (Lempel-Ziv complexity) is calculated for each. Finally, the mean is calculated. ; in, Let be the complexity feature of the m-th random rearrangement substitution sequence, and be the normalized complexity of the m-th random rearrangement substitution sequence. The average complexity of the replacement sequence is that of random rearrangement; This represents the total number of random rearrangement replacement sequences.

[0077] Randomized rearrangement of replacement sequences is used to assess whether changes in original complexity are determined solely by the proportion of EEG microstate categories in EEG signals.

[0078] S43: For a first-order Markov substitution sequence, the first-order transition probability is calculated based on the original symbol sequence, specifically as follows: ; in, The current EEG microstate category is The next moment shifts to the category of brainwave microstates. The probability of; For the original symbol sequence, the brainwave microstate categories Transfer to EEG microstate category The number of times.

[0079] Subsequently, the probability of the initial state is determined based on the frequency of occurrence of each EEG microstate category in the original symbol sequence, and the starting state is randomly selected; then, a first-order Markov substitution sequence with the same scale as the original symbol sequence is generated step by step according to the first-order transition probability matrix.

[0080] For each original symbol sequence, the example also generates 20 first-order Markov substitution sequences, calculates their normalized signal mode complexity (Lempel-Ziv complexity) for each, and finally calculates the mean, specifically: ; in, Let m be the complexity feature of the m-th first-order Markov substitution sequence, and let m be the normalized signal mode complexity (Lempel-Ziv complexity) of the m-th first-order Markov substitution sequence. This represents the average complexity of a first-order Markov substitution sequence.

[0081] Further calculations: ; in, The normalized complexity of the original discrete EEG microstate symbol sequence; This is a deviation parameter between the original sequence complexity and the rearranged replacement sequence, used to evaluate the degree of deviation of the original sequence complexity from the random rearranged replacement sequence; The deviation parameter is the original sequence complexity from the first-order Markov substitution sequence, used to evaluate the degree of deviation of the original sequence complexity from the first-order Markov substitution sequence.

[0082] Through the above-mentioned alternative data verification, it can be determined whether the change in the complexity of the EEG microstate symbol sequence is caused solely by the state ratio or first-order transition structure, and whether there are higher-order temporal organizational features beyond the first-order transition structure. In this embodiment of the invention, the labels corresponding to each state category are further analyzed. A Friedman test is performed to assess whether the complexity difference between the original symbol sequence and the first-order Markov substitution sequence changes with the state category label. The Friedman test is a nonparametric statistical method used to compare whether there are significant differences in the medians of three or more related samples. Significant changes in state category labels indicate that the higher-order temporal organization of the EEG microstate symbol sequence is affected by state changes.

[0083] The output is the complexity feature of the original symbol sequence. Features of the complexity of random rearrangement substitution sequences First-order Markov substitution sequence complexity characteristics Alternative data comparison indicators and Lempel-Ziv complexity features are used to characterize the pattern richness of discrete EEG microstate symbol sequences; alternative data validation results are used to determine whether the complexity variation exceeds the influence of state proportions and first-order transition structures.

[0084] S5: Outputs EEG state dynamics characteristics and performs state recognition. For example... Figure 6 As shown, feature-level fusion is performed based on the spectral features, sample entropy features, and multi-scale entropy features of EEG signals obtained in S3, as well as the symbol sequence complexity features and alternative data verification results obtained in S4, to form a multidimensional EEG state dynamic feature vector.

[0085] The feature vector is input into a pre-established consciousness state discrimination model, which determines the dynamic state category corresponding to the current EEG signal to be analyzed. The dynamic state categories include deinhibited active state and frozen state; when the small-scale entropy value corresponding to the multi-scale entropy feature in the multi-dimensional feature vector shows a synchronous trend with the sample entropy feature, it is determined to be a deinhibited active state; when the small-scale entropy value of the multi-scale entropy feature shows a decreasing trend and the large-scale entropy value of the multi-scale entropy feature shows an increasing trend, it is determined to be a frozen state.

[0086] Output dynamic state categories and their corresponding multidimensional EEG state dynamic feature vectors for EEG state identification, recording, and comparative analysis.

[0087] The aforementioned indicators collectively form a three-dimensional joint analysis method encompassing periodicity, unpredictability, and pattern richness. Specifically, for each type of characteristic indicator, statistical analysis is performed on the corresponding indicator value according to the state category label to obtain the variation differences of the indicator among various state category labels. In this embodiment of the invention, for each indicator in peak frequency, peak power, peak-to-background ratio, sample entropy, EEG signal multi-scale entropy characteristics, signal pattern complexity (Lempel-Ziv complexity sequence characteristics), and surrogate data validation results, the Friedman Test is used to assess the overall difference among various state category labels, and Kendall's W (Kendall's Coefficient of Conformity) is used as the effect size indicator. When the overall difference is significant, the Wilcoxon Sign-Rank Test is further used for pairwise comparisons, and the False Discovery Rate (FDR) method is used to correct the multiple comparison results. Thus, the results of changes in EEG microstate dynamic characteristics are obtained from the three perspectives of periodicity, unpredictability, and pattern richness.

[0088] In this embodiment of the invention, the data corresponding to the propofol label includes four state category labels: baseline signal state, mild signal inhibition state, moderate signal inhibition state, and signal recovery state; the data corresponding to the dexmedetomidine label includes three state category labels: baseline signal state, signal inhibition state, and signal recovery state. By comparing the spectral characteristics, sample entropy characteristics, EEG signal multi-scale entropy characteristics, signal pattern complexity (Lempel-Zivcomplexity) characteristics, and alternative data verification results under each state category label, the EEG microstate dynamics pattern corresponding to each drug type label can be obtained; among them, the spectral analysis results are used to reflect the periodic changes in EEG microstate switching under each state category label. The distance sequence of EEG microstate feature trajectories generated from full-band EEG signals is also provided. The spectrum is used to obtain overall periodic analysis results; the distance sequence of EEG microstate feature trajectories generated from frequency-band EEG signals. The spectrum was used as a supplementary analysis result to help interpret the frequency band source of the periodicity. The differences in the periodicity of EEG microstate switching with state category labels were characterized by comparing the peak frequency, peak power, and peak-to-background ratio under each state category label.

[0089] Sample entropy features and EEG signal multi-scale entropy features are used to reflect the unpredictability and multi-scale complexity changes of EEG microstate switching trajectories; signal pattern complexity Lempel-Ziv complexity features are used to reflect the pattern richness of discrete EEG microstate symbol sequences; the verification results of random rearrangement substitution sequences and first-order Markov substitution sequences are used to determine whether the symbol sequence complexity changes are caused solely by state proportions or first-order transition structures.

[0090] In this embodiment of the invention, the sample entropy, multi-scale entropy short-scale index, and signal pattern complexity (Lempel-Ziv complexity) of the propofol-labeled data show an increasing trend under the signal inhibition state category label, indicating that the EEG microstate switching patterns are richer and more unpredictable; the sample entropy, multi-scale entropy, and signal pattern complexity (Lempel-Ziv complexity) of the dexmedetomidine-labeled data show a decreasing trend under the signal inhibition state category label, indicating that the EEG microstate switching patterns are more regular and predictable. The output of this step is the EEG state dynamics characteristic result.

[0091] In summary, this invention constructs EEG microstate symbol sequences, determines a unified scale, generates CGR trajectories using chaotic game representation of EEG microstates, and establishes distance sequences for EEG microstate feature trajectories. Through computation, spectral analysis, sample entropy analysis, multi-scale entropy analysis, Lempel-Ziv complexity analysis of signal patterns, and validation with alternative data, a joint analysis of three dimensions—periodicity, unpredictability, and pattern richness—was achieved to characterize the changing patterns of EEG microstate dynamics under various conditions.

[0092] The data analysis results of dexmedetomidine were used in the examples. The meanings of each statistical indicator in Table 1 are as follows: χ² 2 The statistic represents the Friedman test statistic, used to evaluate whether there are overall differences among the state category labels. p represents the statistical significance level; when p < 0.05, a significant difference is considered to exist among the state category labels. Kendall's W represents the effect size index, used to characterize the degree of influence of state factors on the index changes, ranging from 0 to 1; a larger value indicates a higher proportion of variance explained by the state factor. N represents the number of data sequences involved in the statistical analysis. The indicators in Table 1 are as follows: Peak Frequency represents the peak frequency in the power spectrum of the trajectory distance sequence, used to characterize the dominant cycle during the EEG microstate switching process. PeakPower represents the power value corresponding to the peak frequency, used to characterize the intensity of periodic components; Peak-to-BackgroundRatio (PBR) represents the ratio of peak power to background power, used to characterize the prominence of periodic peak structures; SampEn represents the sample entropy of the trajectory distance sequence, used to characterize the unpredictability of EEG microstate switching trajectories; MSEmean represents the average entropy across multiple scales from scale 1 to 20, used to characterize the overall complexity level; MSEsmall represents the average entropy across multiple scales from scale 1 to 5, used to characterize short-term temporal complexity; MSElarge represents the average entropy across multiple scales from scale 6 to 20, used to characterize long-term temporal complexity; Original LZC represents the normalized complexity of the original EEG microstate symbol sequence, used to characterize the richness of EEG microstate switching patterns; Delta Markov represents the difference between the Lempel-Ziv complexity of the original sequence and the mean Lempel-Ziv complexity of the first-order Markov substitution sequence, used to assess whether there are higher-order temporal organizational features beyond the first-order transition structure in the EEG microstate symbol sequence.

[0093] Table 1. Statistical table of kinetic characteristics of dexmedetomidine under various states.

[0094] For the interpretation of Results Table 1, the method of this invention was validated using the dataset corresponding to the dexmedetomidine label. This dataset contains three state category labels: baseline signal state, signal suppression state, and signal recovery state. Each state category label corresponds to 20 EEG microstate symbol sequences. Following the method of this invention, the following steps were performed sequentially: EEG microstate symbol sequence construction, unified scale processing, EEG microstate chaotic game representation CGR trajectory generation, spectrum analysis, sample entropy analysis, multi-scale entropy analysis, Lempel-Ziv complexity analysis, and alternative data validation analysis.

[0095] The statistical results are shown in Table 1. The results indicate that the Peak Frequency index differs significantly among the three states, χ² 2 =37.14, p=8.60×10 -9 W=0.929, indicating that the periodicity of EEG microstate switching changes significantly with state changes. The Peak-to-Background Ratio also showed a significant difference, χ² 2 =20.80, p=3.04×10 -5 This indicates that the periodic peak structure in the trajectory distance sequence is affected by state changes. Sample entropy SampEn, χ 2 =33.60, p=5.06×10 -8 And the multi-scale entropy short-scale index MSEsmall, χ 2 =36.40, p=1.25×10 -8 All showed large effect sizes W>0.8, indicating significant differences in the unpredictability and complexity of the trajectory distance sequences across different states. The complexity of the original symbol sequence Lempel-Ziv, χ² 2 =33.60, p=5.06×10 -8 Delta Markov index, χ² 2 =18.90, p=7.87×10 -5 All values ​​reached a significant level, indicating that the changes in the complexity of the EEG microstate symbol sequence not only come from changes in state proportion and first-order transition probability, but also include changes in higher-order temporal organization.

[0096] The above results verify that the proposed method for joint analysis of CGR trajectory and symbolic features based on EEG microstate chaotic game representation can simultaneously characterize the dynamic features of EEG states from three dimensions: periodicity, unpredictability, and pattern richness, and has good state discrimination ability.

[0097] The second aspect of this invention proposes a signal processing system based on a signal processing method for EEG microstate trajectories and symbol features, which includes: an EEG signal data acquisition module, a data preprocessing module, an EEG microstate symbol sequence extraction module, an analysis scale determination module, an EEG microstate trajectory generation module, a trajectory feature extraction module, a symbol sequence verification module, a feature fusion module, and a state discrimination module.

[0098] The EEG signal data acquisition module acquires the EEG signal dataset to be analyzed and reads the corresponding drug type label and state category label of the EEG signal dataset to be analyzed.

[0099] The data preprocessing module preprocesses the EEG signal dataset to be analyzed and obtains full-band EEG signals and multiple sub-band EEG signals based on the preprocessed EEG signals.

[0100] The EEG microstate symbol sequence extraction module performs EEG microstate analysis on full-band EEG signals and multiple sub-band EEG signals to obtain the EEG microstate categories under the corresponding frequency bands and generate the corresponding discrete EEG microstate symbol sequences.

[0101] The analysis scale determination module performs a complexity index convergence test on discrete EEG microstate symbol sequences, determines a unified analysis scale while ensuring the stability of the complexity index estimation, and performs scale unification processing on discrete EEG microstate symbol sequences according to the unified analysis scale.

[0102] The EEG microstate trajectory generation module uses chaotic game representation mapping on the discrete EEG microstate symbol sequence after scale unification to generate continuous geometric trajectory of EEG microstate chaotic game representation, and calculates the distance sequence of EEG microstate feature trajectory based on the continuous geometric trajectory of EEG microstate chaotic game representation.

[0103] The trajectory feature extraction module performs power spectral density analysis, sample entropy analysis, and multi-scale entropy analysis on the trajectory distance sequence of EEG microstate features to obtain spectral features, sample entropy features, and multi-scale entropy features.

[0104] The symbol sequence verification module performs complexity analysis on the scale-unified discrete EEG microstate symbol sequences, constructs alternative sequences based on the discrete EEG microstate symbol sequences for verification, and obtains symbol sequence complexity features and alternative data verification results. The feature fusion module fuses spectral features, sample entropy features, multi-scale entropy features, normalized signal pattern complexity, and deviation parameters at the feature level to generate a multidimensional EEG state dynamic feature vector.

[0105] The state discrimination module inputs the multidimensional EEG state dynamic feature vector into the pre-established consciousness state discrimination model to determine the dynamic state category corresponding to the EEG signal to be analyzed. The dynamic state categories include deinhibited active state and frozen state.

[0106] The beneficial effects of the embodiments of this invention are as follows: The signal processing method and system based on EEG microstate trajectory and symbol features proposed in the embodiments of this invention extract the periodic features of microstate switching under different EEG states through spectrum analysis, and determine the main frequency band sources by combining full-band and sub-band analysis; characterize the unpredictability of EEG microstate switching trajectory and the complexity changes at different time scales through sample entropy and multi-scale entropy analysis; improve the reliability and interpretability of the analysis results through normalized signal pattern complexity analysis, combined with random rearrangement substitution sequences and first-order Markov substitution sequences for verification. This invention can characterize the dynamic features of EEG microstates from multiple dimensions such as periodicity, unpredictability, multi-scale complexity, and symbol temporal organization, and realize the identification and differentiation of de-inhibited active state and frozen state.

[0107] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A signal processing method based on EEG microstate trajectory and symbol features, characterized in that: It includes: S1: Obtain the EEG signal dataset to be analyzed, and read the drug type label and state category label corresponding to the EEG signal dataset to be analyzed; The system performs preprocessing and outputs full-band EEG signals, segmented EEG signals, and corresponding drug type and status category labels. S2: Perform EEG microstate analysis on the full-band and segmented EEG signals from S1, construct EEG microstate symbol sequences, and perform scale unification on the discrete EEG microstate symbol sequences to generate corresponding discrete EEG microstate symbol sequences. ; S3: Obtain the discrete EEG microstate symbol sequence from S2 This is mapped to a chaotic game representation of continuous geometric trajectories. ; Generate distance sequence of EEG microstate feature trajectories Distance sequence of EEG microstate features Spectral analysis, sample entropy calculation, and multi-scale entropy analysis were performed respectively to obtain spectral features, sample entropy features, and multi-scale entropy features. S4: The discrete EEG microstate symbol sequence in S2 Complexity analysis was performed to obtain the normalized signal mode complexity. ; Set the number of times the alternative data is generated. Construct random rearrangement substitution sequences and first-order Markov substitution sequences; Output the deviation parameter between the complexity of the original sequence and the rearranged replacement sequence. Deviation parameters of first-order Markov substitution sequences Analyze and verify the symbol sequence of brainwave microstates; S5: Combine the spectral features, sample entropy features, and multi-scale entropy features from S3 with the normalized mode complexity LZC and deviation parameters from S4. and Feature-level fusion is performed to form a multidimensional EEG state dynamic feature vector.

2. The signal processing method based on EEG microstate trajectory and symbol features according to claim 1, characterized in that: S3 specifically refers to: S31: Constructing the EEG Microstate Feature Point Matrix Generate continuous geometric trajectories of EEG microstates Determining the distance sequence of EEG microstate feature trajectories using Euclidean distance. After performing mean-removal processing, the mean-removed trajectory distance sequence is obtained. ; S32: Distance sequences of EEG microstate feature trajectories across the entire frequency band and each sub-band. Power spectral density analysis was performed to extract peak frequency, peak power, and peak-to-background ratio spectral characteristics to characterize the periodicity of micro-state switching. S33: Distance sequence of full-band EEG microstate feature trajectories Perform sample entropy analysis separately to obtain sample entropy characteristics; For trajectory distance sequence Coarse-graining is performed to obtain coarse-grained trajectory distance sequences at various scales. Determine the multi-scale entropy characteristics of EEG signals.

3. The signal processing method based on EEG microstate trajectory and symbol features according to claim 2, characterized in that: The current EEG microstate feature points generated in S31 are: ； ； in, For the first One EEG microstate feature point; This represents the (k-1)th EEG microstate feature point; For the first A matrix of EEG microstate feature points corresponding to each time point; For the first The x-coordinate of a reference EEG microstate feature point; For the first The vertical coordinates of one reference EEG microstate feature point; For the first The x-coordinate of each EEG microstate feature point; For the first The vertical coordinates of the EEG microstate feature points.

4. The signal processing method based on EEG microstate trajectory and symbol features according to claim 2, characterized in that: S31 EEG Microstate Feature Trajectory Distance Sequence The method for obtaining it is as follows: ； ； in, For the first The distance sequence of EEG microstate feature trajectories corresponding to each time point; For the first The x-coordinate of the EEG microstate feature points corresponding to each time point; For the first The vertical coordinates of the EEG microstate feature points corresponding to each time point; The mean-removed distance sequence of EEG microstate feature trajectories; For the first A sequence of EEG microstate feature trajectories; N is the scale of the trajectory distance sequence.

5. The signal processing method based on EEG microstate trajectory and symbol features according to claim 2, characterized in that: The multi-scale entropy features of EEG signals in S33 are as follows: ； ； ； ； in, This is the average parameter of the sample entropy values ​​within the scale range of 1 to 20; For coarse-grained trajectory distance sequences under scale factor τ The calculated sample entropy value; The integral area of ​​the sample entropy curve within the scale range of 1 to 20; This is the average parameter of the sample entropy values ​​within the scale range of 1 to 5; It is the average parameter of the sample entropy values ​​in the scale range of 6 to 20.

6. The signal processing method based on EEG microstate trajectory and symbol features according to claim 1, characterized in that: S4 specifically refers to: S41: Symbolic sequences of discrete EEG microstates after scale unification Complexity analysis is performed, using signal pattern complexity to calculate the number of patterns appearing in the symbol sequence. Normalization is performed to obtain the normalized signal mode complexity. ; S42: Set the number of times alternative data is generated. A random rearrangement substitution sequence is constructed based on the discrete microstate symbol sequence. The normalized signal mode complexity of each substitution sequence is calculated, and the mean value of the random rearrangement substitution sequence complexity is obtained by taking the mean value. ; S43: Calculate the first-order transition probability based on the original symbol sequence. A first-order transition probability matrix is ​​formed, and a first-order Markov substitution sequence with the same scale as the original symbol sequence is generated by combining the initial state probability. The normalized signal mode complexity is calculated for each substitution sequence, and the mean value of the first-order Markov substitution sequence complexity is obtained. .

7. The signal processing method based on EEG microstate trajectory and symbol features according to claim 6, characterized in that: In S41, the signal mode complexity is normalized as follows: ； in, To normalize the signal mode complexity; denoted by , k represents the number of patterns appearing in the symbol sequence at scale n; k represents the number of EEG microstate categories; and n represents the symbol sequence scale.

8. The signal processing method based on EEG microstate trajectory and symbol features according to claim 6, characterized in that: The average complexity of the random rearrangement substitution sequence in S42 is as follows: ； in, Let m be the complexity feature of the random rearrangement substitution sequence; The average complexity of random rearrangement is the alternative sequence. This represents the total number of random rearrangement substitution sequences.

9. The signal processing method based on EEG microstate trajectory and symbol features according to claim 6, characterized in that: The average complexity of the first-order Markov substitution sequence in S43 is as follows: ； in, Let m be the complexity feature of the first-order Markov substitution sequence; This represents the average complexity of a first-order Markov substitution sequence.

10. A signal processing system for the signal processing method based on EEG microstate trajectory and symbol features as described in any one of claims 1 to 9, characterized in that: It includes: The system includes a brainwave signal data acquisition module, a data preprocessing module, a brainwave microstate symbol sequence extraction module, an analysis scale determination module, a brainwave microstate trajectory generation module, a trajectory feature extraction module, a symbol sequence verification module, a feature fusion module, and a state discrimination module. The EEG signal data acquisition module acquires the EEG signal dataset to be analyzed and reads the drug type label and state category label corresponding to the EEG signal dataset to be analyzed; The data preprocessing module preprocesses the EEG signal dataset to be analyzed and obtains full-band EEG signals and multiple sub-band EEG signals based on the preprocessed EEG signals. The EEG microstate symbol sequence extraction module performs EEG microstate analysis on full-band EEG signals and multiple sub-band EEG signals to obtain the EEG microstate category under the corresponding frequency band and generate the corresponding discrete EEG microstate symbol sequence. The analysis scale determination module performs a complexity index convergence test on discrete EEG microstate symbol sequences, determines a unified analysis scale while ensuring the stability of the complexity index estimation, and performs scale unification processing on discrete EEG microstate symbol sequences according to the unified analysis scale. The EEG microstate trajectory generation module uses chaotic game representation mapping on the discrete EEG microstate symbol sequence after scale unification to generate continuous geometric trajectory of EEG microstate chaotic game representation, and calculates the distance sequence of EEG microstate feature trajectory based on the continuous geometric trajectory of EEG microstate chaotic game representation. The trajectory feature extraction module performs power spectral density analysis, sample entropy analysis, and multi-scale entropy analysis on the trajectory distance sequence of EEG microstate features to obtain spectral features, sample entropy features, and multi-scale entropy features. The symbol sequence verification module performs complexity analysis on the scale-unified discrete EEG microstate symbol sequences, constructs alternative sequences based on the discrete EEG microstate symbol sequences for verification, and obtains symbol sequence complexity features and alternative data verification results. The feature fusion module fuses spectral features, sample entropy features, multi-scale entropy features, normalized signal pattern complexity, and deviation parameters at the feature level to generate a multidimensional EEG state dynamic feature vector. The state discrimination module inputs the multidimensional EEG state dynamic feature vector into the pre-established consciousness state discrimination model to determine the dynamic state category corresponding to the EEG signal to be analyzed. The dynamic state categories include deinhibited active state and frozen state.

Images