A phase synchronization quantization method for multi-channel EEG signals

Abstract

The present invention discloses a multi-channel EEG signal phase synchronization quantization method, which belongs to the field of multi-channel EEG signal analysis, and includes the following steps: collecting EEG signals of multiple channels; filtering the collected multi-channel EEG signals; calculating the instantaneous phase of each channel at N moments; calculating the phase difference between multi-channel signals; dividing the phase difference into I intervals; obtaining the symbol sequence at N moments; calculating the probability of occurrence of each arrangement pattern; calculating the normalized entropy; obtaining the multi-channel signal phase synchronization quantization index SPDPE; performing Wilcoxo rank sum test analysis on the results; and drawing the ROC curve to evaluate the performance of the method. The present invention combines the symbolization and sorting methods, and can be used to calculate the phase synchronization strength of multi-channel EEG signals, which can accurately reflect the synchronization of multi-channel EEG signals, and the calculation process is simple, the data length requirement is low, and it has strong anti-noise performance.

Description

Technical Field

[0001] The invention belongs to the field of multi-channel electroencephalogram signal analysis, and in particular relates to a method for quantifying the phase synchronization intensity of multi-channel electroencephalogram signals. Background Art

[0002] Studies have shown that human cognitive activities need to integrate functions widely distributed in brain regions and between regions to achieve continuous information interaction. The synchronization of EEG signals plays an important role in large-scale neural interactions. There are many ways to explore the synchronization of EEG signals between different brain units. Traditional EEG synchronization methods cannot effectively separate amplitude and phase. Phase synchronization can effectively distinguish amplitude and phase, and only detect the weak mutual relationship between different signal pairs through the instantaneous phase relationship between signals. It is sensitive to nonlinear coupling and can measure the coupling characteristics between EEG signals without considering the signal amplitude, thereby accurately analyzing EEG signals. However, most of the current phase synchronization quantification methods are aimed at dual channels, focusing on the synchronization between time series pairs. The dual-channel phase synchronization quantification method cannot reflect the overall synchronization situation. In addition, the existing small number of multi-channel phase synchronization quantification methods, such as S estimator and hyperdimensional geometry-based measurement method HTS, do not have ideal quantification effects and cannot accurately reflect the degree of phase synchronization of multi-channel signals.

[0003] Symbolization is to convert the time series into a symbolic sequence while retaining the dynamic characteristics of the original signal, which can improve the calculation speed of the method and save computing resources. Sorting entropy has strong anti-noise performance and low requirements on data length. The above two methods have been widely used in research fields such as EEG function analysis. Summary of the invention

[0004] The present invention aims to provide a method capable of calculating the phase synchronization strength of multi-channel electroencephalogram signals.

[0005] To achieve the above objectives, the following technical solutions are adopted:

[0006] Step S1: Collect EEG signals from multiple channels and capture 40 seconds of data from the non-epileptic seizure period and the seizure period.

[0007] Step S2: Filter the collected multi-channel EEG signals to obtain signals in the δ, θ, α, and β frequency bands respectively.

[0008] Step S3: Obtain the instantaneous phase of each channel at N moments using Hilbert transform M is the total number of channels.

[0009] Step S4: To calculate the multi-channel phase synchronization strength, the phase relationship between multiple channels needs to be considered. The phase difference is defined as the phase of each channel minus the angle mean of other channels:

[0010]

[0011] Where M represents the total number of channels, φ i and φ m represent the instantaneous phase of the ith and mth channels respectively.

[0012] Step S5: segment the phase difference. The magnitude of each instantaneous phase is between -π and π, so the range of the phase difference is -2π to 2π. -2π to 2π is divided into I intervals.

[0013] Step S6: label each interval from small to large as 1, 2, ..., I, and correspond each phase difference to the divided interval according to the numerical value. N Assume that there are W sorting patterns in the symbol sequence, and represent each sorting pattern as s(n), that is, {s(n), n=1,2,…,W}.

[0014] Step S7: Calculate the probability of each sorting pattern occurring:

[0015]

[0016] Wherein, j=1,2,...W, # represents the total number of sorting pattern symbol sequences, and N is the total number of symbol sequences.

[0017] Step S8: Calculate the normalized entropy:

[0018]

[0019] Step S9: Obtain the phase synchronization quantization formula of multi-channel EEG signal:

[0020]

[0021] The SPDPE value ranges from 0 to 1, where 0 indicates that there is no phase synchronization phenomenon among the M channel signals, and 1 indicates that the phases of the M channels are completely synchronized.

[0022] Step S10: Calculate the SPDPE of non-seizure and seizure data, and perform Wilcoxo rank sum test analysis on the results.

[0023] Step S11: Calculate the sensitivity and specificity based on the experimental results and draw the ROC curve to evaluate the performance of the method.

[0024] Due to the adoption of the above technical solution, the technical effects achieved by the present invention are as follows:

[0025] The present invention combines the symbolization and sorting methods and can be used to calculate the phase synchronization strength of multi-channel EEG signals. It can accurately reflect the synchronization of multi-channel EEG signals, and the calculation process is simple, the data length requirement is low, and it has strong anti-noise performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the method of the present invention;

[0027] Figure 2 This is the Wilcoxo rank sum test analysis chart;

[0028] Figure 3 It is the ROC curve diagram. DETAILED DESCRIPTION

[0029] The present invention will be further described below in conjunction with the accompanying drawings:

[0030] like Figure 1 As shown, the present invention comprises the following steps:

[0031] Step S1: Collect EEG signals from multiple channels and capture 40s of data during non-epileptic seizures and seizures.

[0032] Step S2: Filter the collected multi-channel EEG signals to obtain signals in the δ, θ, α, and β frequency bands respectively.

[0033] Step S3: Assuming the EEG signal is x(n), perform Hilbert transform on x(n) to obtain

[0034]

[0035] according to Construct the analytical signal y(n):

[0036]

[0037] The instantaneous phase φ can be calculated from the analytical signal i (n):

[0038]

[0039] This method can be used to obtain the instantaneous phase of each channel at N moments M is the number of channels.

[0040] Step S4: To calculate the phase synchronization strength of multi-channel EEG signals, the phase relationship between multiple channels needs to be considered. The phase difference is defined as the phase of each channel minus the angle mean of other channels:

[0041]

[0042] Where M represents the total number of channels, φ i and φ m represent the instantaneous phase of the ith and mth channels respectively.

[0043] Step S5: segment the phase difference. The magnitude of each instantaneous phase is between -π and π, so the range of the phase difference is -2π to 2π. -2π to 2π is divided into I intervals.

[0044] Step S6: label each interval from small to large as 1, 2, ... I, and correspond each phase difference to the divided interval according to the numerical value. According to the interval label, the symbol sequence at each moment is obtained, and N symbol sequences z1, z2, ..., z are obtained at N moments. N , assuming that there are W sorting patterns in the symbol sequence, each sorting pattern is represented as s(n), that is, {s(n), n=1,2,…,W}.

[0045] Step S7: There are a total of W sorting modes for N symbol sequences, and the probability of each sorting mode appearing can be calculated:

[0046]

[0047] Wherein, j=1,2,...W, # represents the total number of sorting pattern symbol sequences, and N is the total number of symbol sequences.

[0048] Step S8: The entropy of the N symbol sequences can be calculated based on the probability of each arrangement pattern and normalized to:

[0049]

[0050] Step S9: Adjust the range of normalized entropy to 0-1, and obtain the multi-channel EEG signal phase synchronization quantization formula:

[0051]

[0052] The SPDPE value ranges from 0 to 1, where 0 indicates that there is no phase synchronization phenomenon among the M channel signals, and 1 indicates that the phases of the M channels are completely synchronized.

[0053] Step S10: Calculate the SPDPE of the non-ictal and ictal data, and perform a Wilcoxo rank sum test on the results. If this method can accurately quantify phase synchronization, then the SPDPE calculated from the non-ictal and ictal data must be significantly different.

[0054] Figure 2 The Wilcoxo rank sum test analysis diagram shows the significance test results of SPDPE in ten epilepsy patients with and without seizures. The significance analysis shows that the results have significant differences (** represents p<0.01).

[0055] Step S11: Calculate the sensitivity and specificity based on the experimental results and draw the ROC curve to evaluate the performance of the method.

[0056] Figure 3 The ROC characteristic curves of the three methods of SPDPE, HTS and COC. Sensitivity is defined as the number of detected epileptic seizures divided by the actual total number of epileptic seizures. Specificity is defined as the number of detected non-epilepsy seizures divided by the actual total number of non-epilepsy seizures. The area under the ROC curve (AUC) can be used to characterize the classification ability of the method. The larger the AUC value, the better the classification ability. The AUC values ​​of the three methods are AUC coc =0.68, AUC HTS =0.71, AUC SPDPE =0.79. Therefore, this method has the best performance compared with other methods.

[0057] Based on the Wilcoxo rank sum test and ROC curve analysis of the method, we concluded that this method can better evaluate the phase synchronization strength of multi-channel EEG signals, which is conducive to further understanding the synchronization mechanism of EEG signals.

[0058] The embodiments described above are merely descriptions of preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A multi-channel EEG signal phase synchronization quantization method, characterized in that The following steps are involved: Step S1, collecting EEG signals of multiple channels; Step S2, filtering the collected multi-channel EEG signals; Step S3, calculating the instantaneous phase of each channel at N moments; Step S4, calculating the phase difference between the multi-channel signals; The phase difference is defined as the phase of each channel minus the angular mean of the other channels: Where M represents the total number of channels, φ i and φ m Represent the instantaneous phase of the ith and mth channels respectively; Step S5, dividing the phase difference into I intervals; Step S6, obtaining a symbol sequence at N moments; Label each interval from small to large as 1, 2, ..., I, and obtain the symbol sequence at each moment according to the interval label. N symbol sequences z1, z2, ..., z are obtained at N moments. N , assuming that there are W sorting patterns in the symbol sequence, each sorting pattern is represented by s(n), that is, {s(n), n=1,2,…,W}; Step S7, calculating the probability of each arrangement pattern appearing; The probability of the permutation pattern is: Where, j = 1, 2, ... W, # represents the total number of sorting pattern symbol sequences, and N is the total number of symbol sequences; Step S8, calculating normalized entropy; The formula for normalized entropy is: Step S9, obtaining a multi-channel signal phase synchronization quantization index SPDPE; Step S10, performing Wilcoxo rank sum test analysis on the results; Step S11, drawing a ROC curve to evaluate the performance of the method.

2. A multi-channel EEG signal phase synchronization quantification method according to claim 1, characterized in that: In step S1, the EEG signals of multiple channels are collected, and 40s of data of the non-seizure period and the seizure period are respectively intercepted; in step S2, the collected multi-channel EEG signals are filtered to obtain signals of the δ, θ, α, and β frequency bands respectively; in step S3, the instantaneous phase of each channel at N moments is obtained by using the Hilbert transform M is the total number of channels.

3. A multi-channel EEG signal phase synchronization quantification method according to claim 1, characterized in that: In step S5, the magnitude of each instantaneous phase is between -π and π, the range of the phase difference is between -2π and 2π, and -2π to 2π is divided into I intervals.

4. A multi-channel EEG signal phase synchronization quantification method according to claim 1, characterized in that: In step S9, the SPDPE formula is obtained: The SPDPE value ranges from 0 to 1, where 0 indicates that there is no phase synchronization phenomenon among the M channel signals, and 1 indicates that the phases of the M channels are completely synchronized.

5. A multi-channel EEG signal phase synchronization quantification method according to claim 1, characterized in that: In step S10, the SPDPE of the 40s data of the non-ictal period and the ictal period is calculated, and the results are analyzed by Wilcoxo rank sum test. If this method can quantify phase synchronization more accurately, the SPDPE calculated from the epileptic non-ictal period and the ictal period data must have significant differences.

6. A multi-channel EEG signal phase synchronization quantification method according to claim 1, characterized in that: In step S11, the sensitivity and specificity are calculated according to the experimental results and the ROC curve is drawn. The area under the ROC curve AUC can be used to characterize the classification ability of the method. The larger the AUC value, the better the classification ability. The performance of the method is evaluated according to the size of the AUC value.

Images