A sleep state monitoring system and method integrating EEG and neuronal population electrical signals

Abstract

The present invention discloses a sleep state monitoring system and method integrating EEG and neuronal population electrical signals, which relates to the technical field of sleep state monitoring. The technical key points of this solution are as follows: decompose the EEG signal and the neuronal population electrical signal into a number of IMF components respectively, construct the signal features of each IMF component, combine the signal features of each sleep state, identify the IMF components related to each sleep state and reconstruct them, perform Fourier transform on the EEG reconstructed signal and the neuronal population reconstructed electrical signal to obtain the spectra of the EEG reconstructed signal and the neuronal population reconstructed electrical signal, construct a comprehensive feature vector, construct a sleep state recognition model based on historical EEG signal and neuronal population electrical signal data, identify the current comprehensive feature vector to determine the current sleep state, which can capture more comprehensive brain activity information and help to more accurately identify different sleep stages.

Description

Technical Field

[0001] The present invention relates to the technical field of sleep state monitoring, and specifically to a sleep state monitoring system and method that integrates EEG and neuronal population electrical signals. Background Art

[0002] EEG signal is a non-invasive neural signal detection technology, which is widely used in the monitoring and analysis of sleep states. The EEG signal collects the electrical activities of the brain through scalp electrodes, and can reflect the state changes of the brain in different sleep stages. Neuronal population electrical signal refers to the electrical activity generated by the synchronous discharge of a large number of neurons in the cerebral cortex. This electrical signal is closely related to the higher functions of the brain such as cognition, emotion, and behavior. During sleep, the neuronal population electrical signal also undergoes corresponding changes, reflecting different activities and functional states of the brain during sleep.

[0003] In the Chinese invention application with the publication number CN114662530A, a sleep stage staging method based on temporal signal convolution and multi-signal fusion is disclosed, including: obtaining physiological signal data of a subject in a sleep state, including: 4-channel electroencephalogram signal EEG, 1-channel electrocardiogram signal ECG, 2-channel electrooculogram signal EOG, 1-channel electromyogram signal EMG; inputting the physiological signal data of the subject in the sleep state into a trained monitoring model to obtain a classification result; the trained monitoring model includes: a temporal signal convolution network for separately performing one-dimensional convolution on each channel in the input physiological signal data to extract the temporal waveform change characteristics of the signals of each channel; a spatial functional connection network for extracting the relationship between any two channels in the physiological signal data; a fully connected network for classifying the physiological signal data according to the temporal waveform change characteristics of the signals of each channel in the physiological signal data and the relationship between any two channels to obtain a classification result.

[0004] Combined with the above invention, the prior art has the following deficiencies:

[0005] 1. The prior art faces the problem of limited accuracy in identifying different sleep stages, fails to fully analyze and utilize the internal structures of EEG signals and neuronal population electrical signals, lacks the ability to integrate EEG signals and neuronal population electrical signals, making it difficult for the prior art to accurately capture the comprehensive information of brain activities, thus affecting the accuracy of sleep state recognition;

[0006] 2. The prior art lacks effective methods to accurately identify EEG signals highly correlated with each sleep state, lacks means for signal screening and combination, and insufficiently processes complex signals in a refined manner, making it impossible to fully reflect the characteristics of each sleep state, thus affecting the accuracy of sleep state recognition. Summary of the Invention

[0007] (1) Technical problems to be solved

[0008] In view of the deficiencies of the prior art, the present invention provides a sleep state monitoring system and method for integrating EEG and neuronal population electrical signals. By decomposing EEG signals and neuronal population electrical signals into a number of IMF components respectively, constructing the signal features of each IMF component, identifying the IMF components related to each sleep state for reconstruction, constructing a comprehensive feature vector, and determining the current sleep state, the problems mentioned in the background art are solved.

[0009] (2) Technical solutions

[0010] To achieve the above object, the present invention is realized through the following technical solutions: A sleep state monitoring method for integrating EEG and neuronal population electrical signals, comprising the following steps:

[0011] Using EEG technology to collect EEG signals of different brain regions in real time, and synchronously recording the neuronal population electrical signals of different brain regions through the in-vivo multi-channel synchronous recording technology of central neuron discharges;

[0012] Taking the EEG signal and the neuronal population electrical signal as the original signals respectively, decomposing them into a number of IMF components and a residue, performing Hilbert transform on each IMF component, combining the IMF component signal and the signal after Hilbert transform to form an analytic signal, and constructing the signal features of each IMF component;

[0013] Obtaining the signal features of each sleep state, combining the signal features of each IMF component, identifying and marking the IMF components related to each sleep state, and reconstructing the marked IMF components belonging to the EEG signal and the neuronal population electrical signal respectively to obtain the EEG reconstructed signal and the neuronal population reconstructed electrical signal;

[0014] Performing Fourier transform on the EEG reconstructed signal and the neuronal population reconstructed electrical signal respectively to obtain the spectra of the EEG reconstructed signal and the neuronal population reconstructed electrical signal, and constructing a comprehensive feature vector through the spectra of the EEG reconstructed signal and the neuronal population reconstructed electrical signal;

[0015] Based on historical EEG signal and neuronal population electrical signal data, constructing a sleep state recognition model, and using the sleep state recognition model to identify the current comprehensive feature vector to determine the current sleep state.

[0016] Further, taking the EEG signal and the neuronal population electrical signal as the original signals respectively, decomposing them into a number of IMF components and a residue, specifically including:

[0017] Find all the local maximum and minimum points in the original signal. Using the cubic spline interpolation method, fit the local maximum and minimum points to the upper and lower envelope lines of the signal respectively, and calculate the mean of the upper and lower envelope lines to obtain the mean line m1;

[0018] Subtract the mean line m1 from the original signal to obtain a new signal h1. Check whether the new signal h1 meets the IMF conditions. IMF conditions: The number of extreme points is equal to or differs by 1 from the number of zeros, and the mean of the upper and lower envelopes is zero.

[0019] Furthermore, if the new signal h1 does not meet the IMF conditions, then use the new signal h1 as the original signal and repeat the above steps until a component that meets the IMF conditions is obtained;

[0020] If the new signal h1 meets the IMF conditions, then use the new signal h1 as the first IMF component c1;

[0021] Repeat the above steps for the remaining signal until the remaining signal becomes a monotonic function and no longer decomposes into components that meet the IMF conditions. In each iteration, subtract the extracted IMF component from the original signal to obtain the next remaining signal to be decomposed.

[0022] Furthermore, construct the signal features of each IMF component, including:

[0023] By analyzing the real and imaginary parts of the signal, calculate the instantaneous frequency and instantaneous amplitude of each IMF component signal, and combine the instantaneous frequency and instantaneous amplitude of each IMF component signal to construct the signal feature V: , and the calculation formulas for the instantaneous frequency and instantaneous amplitude are as follows:

[0024]

[0025] Among them, represents the instantaneous frequency, represents the instantaneous phase, represents the instantaneous amplitude, represents the real part of the analytic signal, represents the imaginary part of the analytic signal, and t represents the time.

[0026] Furthermore, identify and label the IMF components related to each sleep state, including:

[0027] Calculate the correlation between each IMF component and each sleep state through the Pearson correlation coefficient. Preset the correlation threshold, and compare the absolute value of the Pearson correlation coefficient of each IMF component with the correlation threshold. When the absolute value of the Pearson correlation coefficient is greater than the correlation threshold, label the corresponding IMF component; otherwise, do nothing.

[0028] Further, the IMF components belonging to and labeled for the EEG signal and the neuronal population electrical signal are reconstructed respectively, and the reconstruction process is as follows:

[0029] Calculate the weighting coefficient of each labeled IMF component through the Pearson correlation coefficient of the labeled IMF components. The weighting coefficient: , where represents the weighting coefficient of the k-th labeled IMF component, represents the Pearson correlation coefficient of the k-th labeled IMF component, and reconstruct the labeled IMF components through weighted combination: , where represents the reconstructed signal, represents the k-th labeled IMF component.

[0030] Further, construct the comprehensive feature vector including:

[0031] Calculate the power spectral density of the EEG reconstructed signal and the neuronal population reconstructed electrical signal respectively. The power spectral density: , where represents the power spectral density, represents the Fourier transform of the signal, f represents the frequency, and N represents the length of the signal;

[0032] Calculate the cross-power spectral density of the EEG reconstructed signal and the neuronal population reconstructed electrical signal. The cross-power spectral density: , where represents the cross-power spectral density, 、 represent the Fourier transforms of the EEG reconstructed signal and the neuronal population reconstructed electrical signal respectively;

[0033] Further, calculate the coherence between the EEG reconstructed signal and the neuronal population reconstructed electrical signal. The coherence: , where represents the coherence, represents the cross-power spectral density, and represent the power spectral densities of the EEG reconstructed signal and the neuronal population reconstructed electrical signal respectively;

[0034] Stitch and fuse the power spectral density of the EEG reconstructed signal, the power spectral density of the neuronal population reconstructed electrical signal, and the coherence to construct a comprehensive feature vector.

[0035] Further, construct a sleep state recognition model, including:

[0036] Collect historical EEG signals and neuronal population electro-signal data, extract comprehensive feature vectors, perform clustering analysis on the comprehensive feature vectors using a clustering algorithm, and assign a label to each cluster according to the clustering results to represent different sleep states;

[0037] Divide the clustered comprehensive feature vector data set into a training set and a test set, establish a model using a machine learning algorithm, train the model using the data in the training set, evaluate the trained model using the test set data, and adjust and optimize the model according to the evaluation results to obtain a sleep state recognition model.

[0038] A sleep state monitoring system integrating EEG and neuronal population electro-signals, comprising:

[0039] A signal acquisition module that uses EEG technology to collect EEG signals from different brain regions in real time, and synchronously records the electro-signals of neuronal populations in different brain regions through in-vivo multi-channel synchronous recording technology for central neuron discharges;

[0040] A signal processing module that takes the EEG signal and the neuronal population electro-signal as raw signals respectively, decomposes them into a number of IMF components and a residue, performs Hilbert transform on each IMF component, combines the IMF component signal and the signal after Hilbert transform to form an analytic signal, and constructs the signal features of each IMF component; obtains the signal features of each sleep state, combines the signal features of each IMF component, identifies and marks the IMF components related to each sleep state, and reconstructs the IMF components to which the EEG signal and the neuronal population electro-signal belong and are marked respectively to obtain an EEG reconstructed signal and a neuronal population reconstructed electro-signal;

[0041] A feature extraction module that performs Fourier transform on the EEG reconstructed signal and the neuronal population reconstructed electro-signal respectively to obtain the spectra of the EEG reconstructed signal and the neuronal population reconstructed electro-signal, and constructs a comprehensive feature vector through the spectra of the EEG reconstructed signal and the neuronal population reconstructed electro-signal;

[0042] A sleep state recognition module that constructs a sleep state recognition model based on historical EEG signals and neuronal population electro-signal data, and uses the sleep state recognition model to identify the current comprehensive feature vector to determine the current sleep state.

[0043] (III) Beneficial effects

[0044] The present invention provides a sleep state monitoring system and method integrating EEG and neuronal population electro-signals, having the following beneficial effects:

[0045] (1) By analyzing the IMF components and instantaneous characteristics of EEG signals and neuronal population electrical signals, the refined processing of complex signals is achieved, which helps to better understand and analyze the internal structure of the signals, thereby improving the accuracy of subsequent feature extraction, more precisely capturing the dynamic changes of the signals at different time points, and understanding the generation mechanism of the dynamic patterns of neurons in different brain regions.

[0046] (2) By calculating the Pearson correlation coefficient between each IMF component and each sleep state and setting a correlation threshold for screening, the IMF components highly correlated with each sleep state can be accurately identified. By performing weighted combination reconstruction on all the labeled IMF components, a signal that can better reflect each sleep state can be obtained.

[0047] (3) By performing Fourier transform on the EEG reconstructed signal and the neuronal population reconstructed electrical signal, calculating their power spectral density, cross-power spectral density, and coherence, a comprehensive feature vector containing rich information can be constructed, which can accurately reflect the electrical activity patterns of the brain in different sleep states.

[0048] (4) By integrating EEG signals and neuronal population electrical signals and constructing a comprehensive feature vector, more comprehensive brain activity information can be captured, which helps to more accurately identify different sleep stages. The sleep state recognition model trained using historical data can learn the subtle differences between different sleep stages, thereby improving the accuracy of recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 FIG. is a schematic diagram of the steps of the sleep state monitoring method for integrating EEG and neuronal population electrical signals according to the present invention;

[0050] Figure 2 FIG. is a schematic diagram of the structure of the sleep state monitoring system for integrating EEG and neuronal population electrical signals according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0052] Please refer to Figure 1 , the present invention provides a sleep state monitoring method for integrating EEG and neuronal population electrical signals, including the following steps:

[0053] Step 1: Use EEG technology to collect EEG signals from different brain regions in real time, and synchronously record the electrical signals of neuron clusters in different brain regions through the in-vivo multi-channel synchronous recording technology of central neuron discharges;

[0054] The first step includes the following contents:

[0055] Step 101: Use EEG technology to collect EEG signals from different brain regions in real time. According to the 10-20 international system standard, place electrodes on the target scalp to ensure that the impedance between the electrodes and the scalp is relatively low, usually maintained below 10 kΩ, and even below 3 kΩ, to reduce noise interference;

[0056] It should be noted that the 10-20 system electrode placement method is the standard electrode placement method stipulated by the International EEG Society. The distance from the midpoint of the frontal pole to the bridge of the nose and the distance from the occipital point to the external occipital protuberance each account for 10% of the total length of this connection line, and the remaining points are separated by 20% of the total length of this connection line, so it is named the 10-20 system;

[0057] Use an EEG amplifier to amplify the EEG signals. The amplifier includes multiple channels and simultaneously records the signals of multiple brain regions. The brain regions include the frontal lobe, parietal lobe, temporal lobe, and occipital lobe regions, etc. The EEG frequency band ranges include: δ band: 0.5 - 5 Hz, θ band: 4 - 7.5 Hz, α band: 8 - 12.5 Hz, β band: 13 - 30 Hz, γ band: > 30 Hz;

[0058] Step 102: Synchronously record the electrical signals of neuron clusters in different brain regions through the in-vivo multi-channel synchronous recording technology of central neuron discharges;

[0059] It should be noted that the in-vivo multi-channel synchronous recording technology of central neuron discharges is a technology for monitoring the synchronous action potentials of neuron groups. Applying this method can synchronously record the electrical signals of a large number of neurons in multiple brain regions. Sleep and wakefulness can selectively activate the corresponding neuron groups, and the emergence pattern of the electrical signals of the targeted neuron clusters is directly related to the generation of sleep and wakefulness;

[0060] Step 103: Preprocess the EEG signal and neuron cluster electrical signal data, use a band-pass filter or a notch filter for filtering to remove DC drift and high-frequency noise. For example, in offline preprocessing analysis, use a band-pass filter between 0.5 Hz and 75 Hz for the EEG signal, and add a 50 Hz notch filter to filter the power frequency signal; process the artifacts caused by electrode failure or EEG drift through bad channel detection and interpolation processing;

[0061] When in use, combine the contents of Step 101 to Step 103:

[0062] EEG technology can directly reflect the electrical activity of the brain, and the electrical signals of neuron clusters are the direct manifestation of brain activity. Through multi-channel synchronous recording technology, the changes in electrical activity in different brain regions during sleep can be captured, thereby more accurately judging the sleep stage and depth.

[0063] Step 2: Respectively take the EEG signal and the electrical signal of neuron clusters as the original signals, decompose them into several IMF components and a residue, perform Hilbert transform on each IMF component, combine the IMF component signal and the signal after Hilbert transform to form an analytic signal, calculate the instantaneous frequency and instantaneous amplitude of each IMF component signal using the analytic signal, and construct signal features;

[0064] The said Step 2 includes the following contents:

[0065] Step 201: Obtain the EEG signal and the electrical signal of neuron clusters. Respectively take the EEG signal and the electrical signal of neuron clusters as the original signals, and decompose them into several IMF components and a residue, specifically including:

[0066] Step 2011: Find all the local maximum points and minimum points in the original signal, use cubic spline interpolation or other interpolation methods to fit the local maximum points and minimum points into the upper and lower envelope lines of the signal respectively, and calculate the mean value of the upper and lower envelope lines to obtain the mean line m1: =(Mean value of the upper envelope line + Mean value of the lower envelope line) / 2;

[0067] Step 2012: Subtract the mean line m1 from the original signal to obtain a new signal h1, and check whether the new signal h1 meets the IMF conditions. IMF conditions: The number of extreme points is equal to or differs by 1 from the number of zero points, and the mean value of the upper and lower envelopes is zero;

[0068] Step 2013: If the new signal h1 does not meet the IMF conditions, then take the new signal h1 as the original signal and repeat the above Steps 2011 - 2012 until a component that meets the IMF conditions is obtained;

[0069] If the new signal h1 meets the IMF conditions, then take the new signal h1 as the first IMF component c1. The calculation formula for the IMF component: , where represents the mean line obtained in the i-th iteration, represents the signal before the i-th iteration, represents the IMF component in the i-th iteration;

[0070] Step 2014: Repeat the above Steps 2011 - 2013 for the remaining signal until the remaining signal becomes a monotonic function (i.e., the residual) and no more components satisfying the IMF conditions can be decomposed. In each iteration, subtract the extracted IMF component from the original signal to obtain the next remaining signal to be decomposed;

[0071] Step 2015: Record all the extracted IMF components and the final residual (i.e., the monotonic function);

[0072] Step 202: Perform the Hilbert transform on each IMF component, and combine the IMF component signal and the signal after the Hilbert transform to form the analytic signal z: , where j is the imaginary unit, and the Hilbert transform formula is as follows:

[0073]

[0074] where, is the IMF component signal, i.e., the real part of the analytic signal, is the signal after the Hilbert transform, i.e., the imaginary part of the analytic signal, t represents the time instant, represents the time shift, is the value of the IMF component signal at the time shift ;

[0075] Step 203: Calculate the instantaneous frequency and instantaneous amplitude of each IMF component signal through the real and imaginary parts of the analytic signal, and combine the instantaneous frequency and instantaneous amplitude of each IMF component signal to construct the signal feature V: , and the calculation formulas for the instantaneous frequency and instantaneous amplitude are as follows:

[0076]

[0077] where, represents the instantaneous frequency, represents the instantaneous phase, represents the instantaneous amplitude;

[0078] It should be noted that the instantaneous frequency reflects the frequency characteristics of the signal at a certain moment, and the instantaneous amplitude represents the intensity or magnitude of the signal at a certain moment. The instantaneous frequency and instantaneous amplitude of each IMF component signal together constitute the time - frequency characteristics of the signal, which can comprehensively reflect the dynamic changes of EEG signals and neuronal population electrical signals in the time - frequency domain;

[0079] When in use, combine the content of Steps 201 to 203:

[0080] By analyzing the IMF components and instantaneous characteristics of EEG signals and neuronal population electrical signals, refined processing of complex signals is achieved, which helps to better understand and analyze the internal structure of the signals, thereby improving the accuracy of subsequent feature extraction, more precisely capturing the dynamic changes of the signals at different time points, and understanding the generation mechanism of the dynamic patterns of neurons in different brain regions.

[0081] Step 3: Obtain the signal features of each sleep state, combine the signal features of each IMF component, identify and label the IMF components related to each sleep state, and respectively reconstruct the EEG signal and the neuronal population reconstructed electrical signal for the IMF components to which the EEG signal and the neuronal population electrical signal belong and are labeled;

[0082] The said Step 3 includes the following contents:

[0083] Step 301: Obtain the signal features of each sleep state, combine the signal features of each IMF component, calculate the correlation between each IMF component and each sleep state through the Pearson correlation coefficient, and the calculation formula of the Pearson correlation coefficient: , where r represents the correlation coefficient, 、 respectively represent the nth data point of the IMF component and the signal features of each sleep state, 、 respectively represent and the means of;

[0084] Step 302: Preset a correlation threshold, compare the absolute value of the Pearson correlation coefficient of each IMF component with the correlation threshold. When the absolute value of the Pearson correlation coefficient is greater than the correlation threshold, label the corresponding IMF component; otherwise, no operation is performed;

[0085] Step 303: Respectively reconstruct the EEG signal and the neuronal population reconstructed electrical signal for the IMF components to which the EEG signal and the neuronal population electrical signal belong and are labeled, specifically including:

[0086] Calculate the weighting coefficient of each labeled IMF component through the Pearson correlation coefficient of the labeled IMF component. The weighting coefficient: , where, represents the weighting coefficient of the kth labeled IMF component, represents the Pearson correlation coefficient of the kth labeled IMF component. Respectively reconstruct the EEG signal and the neuronal population electrical signal for the labeled IMF components through weighted combination: , where, represents the reconstructed signal, Denote the k-th labeled IMF component;

[0087] It should be noted that the above reconstruction is performed separately on the EEG signal and the electrical signal of the neuron cluster, and the operations on their respective IMF components are carried out separately. For example, in the calculation of the weighting coefficient, it only involves one of the EEG signal and the electrical signal of the neuron cluster. Similarly, so does;

[0088] It should be noted that the correlation threshold is determined according to the actual application scenario and the data analysis requirements to ensure that the selected IMF components can not only reflect the main characteristics of each sleep state but also do not introduce too much noise or irrelevant information, with the aim of screening out the IMF components highly correlated with each sleep state and using the IMF components highly correlated with each sleep state to reconstruct a signal that can better reflect each sleep state;

[0089] When in use, combine the content of steps 301 to 303:

[0090] By calculating the Pearson correlation coefficient between each IMF component and each sleep state and setting a correlation threshold for screening, the IMF components highly correlated with each sleep state can be accurately identified. By performing weighted combination reconstruction on all labeled IMF components, a signal that can better reflect each sleep state can be obtained.

[0091] Step Four: Perform Fourier transforms on the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction respectively to obtain the spectra of the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction. Construct a comprehensive feature vector through the spectra of the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction;

[0092] The said Step Four includes the following content:

[0093] Step 401: Obtain the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction, and perform Fourier transforms on the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction respectively to obtain the spectra of the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction;

[0094] Step 402: Calculate the power spectral density of the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction respectively through the spectra of the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction. The power spectral density: , where, denotes the power spectral density, denotes the Fourier transform of the signal (i.e., the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction), f denotes the frequency, N denotes the length of the signal, calculate the cross-power spectral density of the EEG reconstructed signal and the electrical signal of the neuron cluster reconstruction, and calculate the coherence. The coherence: , where represents coherence, represents the cross-power spectral density, and represent the power spectral density of the EEG reconstructed signal and the power spectral density of the reconstructed electrical signal of the neuronal population respectively, and the cross-power spectral density: , where and represent the Fourier transforms of the EEG reconstructed signal and the reconstructed electrical signal of the neuronal population respectively;

[0095] Step 403: Concatenate and fuse the power spectral density of the EEG reconstructed signal, the power spectral density of the reconstructed electrical signal of the neuronal population, and the coherence to construct a comprehensive feature vector P: ;

[0096] It should be noted that concatenation is one of the most direct feature fusion methods. It simply concatenates different feature vectors together to form a longer feature vector. This method retains the information of all original features, but may increase the dimension of the feature vector, thereby increasing the complexity of the model;

[0097] When used, combine the content of steps 401 to 403:

[0098] By performing Fourier transforms on the EEG reconstructed signal and the reconstructed electrical signal of the neuronal population, and calculating their power spectral density, cross-power spectral density, and coherence, a comprehensive feature vector containing rich information can be constructed, which can accurately reflect the electrical activity patterns of the brain in different sleep states.

[0099] Step Five: Based on historical EEG signal and neuronal population electrical signal data, construct a sleep state recognition model, and use the sleep state recognition model to recognize the current comprehensive feature vector to determine the current sleep state.

[0100] The said Step Five includes the following content:

[0101] Step 501: Collect historical EEG signal and neuronal population electrical signal data, extract comprehensive feature vectors, and use clustering algorithms (such as K-means, hierarchical clustering, etc.) to perform clustering analysis on the comprehensive feature vectors. According to the clustering results, assign a label to each cluster to represent different sleep states;

[0102] Step 502: Divide the clustered comprehensive feature vector dataset into a training set and a test set. Use a machine learning algorithm to build a model, train the model using the data in the training set, and evaluate the trained model using the test set data. Calculate evaluation metrics such as the accuracy, recall rate, and F1 score of the model. According to the evaluation results, adjust and optimize the model, such as adjusting hyperparameters, selecting new features, and trying different algorithms, to obtain the final sleep state recognition model;

[0103] Step 503: Use the sleep state recognition model to identify the current comprehensive feature vector, determine the current sleep state, and preset a normal threshold range. Compare the current comprehensive feature vector with the normal threshold range. If any component of the current comprehensive feature vector exceeds the normal threshold range, trigger an alarm; otherwise, do nothing;

[0104] It should be noted that the setting of the threshold range is based on the distribution range of the feature vectors in the normal sleep state. One or more thresholds are set, and these thresholds can be based on statistics (such as mean, standard deviation) or based on data distribution (such as quantiles);

[0105] When in use, combine the content of Steps 501 to 503:

[0106] By integrating EEG signals and neuronal population electrical signals and constructing comprehensive feature vectors, more comprehensive brain activity information can be captured, which helps to more accurately identify different sleep stages. The sleep state recognition model trained using historical data can learn the subtle differences between different sleep stages, thereby improving the accuracy of recognition.

[0107] Please refer to Figure 2 , the present invention also provides a sleep state monitoring system that integrates EEG and neuronal population electrical signals, including: a signal acquisition module, a signal processing module, a feature extraction module, and a sleep state recognition module; wherein,

[0108] The signal acquisition module uses EEG technology to collect EEG signals from different brain regions of the brain in real time, and synchronously records the neuronal population electrical signals of different brain regions through the in vivo multi-channel synchronous recording technology of central neuron discharges;

[0109] The signal processing module takes the EEG signal and the neuronal population electrical signal as the original signals respectively, decomposes them into a number of IMF components and a residue, performs Hilbert transform on each IMF component, combines the IMF component signal and the signal after Hilbert transform to form an analytic signal, and constructs the signal features of each IMF component; obtains the signal features of each sleep state, combines the signal features of each IMF component, identifies and marks the IMF components related to each sleep state, and reconstructs the marked IMF components belonging to the EEG signal and the neuronal population electrical signal respectively to obtain the EEG reconstructed signal and the neuronal population reconstructed electrical signal;

[0110] The feature extraction module performs Fourier transform on the EEG reconstructed signal and the neuronal population reconstructed electrical signal respectively to obtain the spectra of the EEG reconstructed signal and the neuronal population reconstructed electrical signal, and constructs a comprehensive feature vector through the spectra of the EEG reconstructed signal and the neuronal population reconstructed electrical signal;

[0111] The sleep state recognition module constructs a sleep state recognition model based on the historical EEG signal and neuronal population electrical signal data, and uses the sleep state recognition model to recognize the current comprehensive feature vector to determine the current sleep state.

[0112] In the application, several formulas involved are numerically calculated after dimensionless, and the formula is a formula obtained by software simulation of a large amount of collected data to approximate the real situation as much as possible. The coefficients in the formula are set by those skilled in the art according to the actual situation.

[0113] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by a combination of electronic hardware, computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution.

[0114] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0115] The above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application.

Claims

1. A sleep state monitoring method integrating EEG and neuron cluster electrical signals, characterized in that: The following steps are involved: Use EEG technology to collect EEG signals from different brain regions in real time, and use in vivo multi-channel synchronous recording technology of central neuron discharge to synchronously record the electrical signals of neuron clusters in different brain regions; The EEG signal and the neuron cluster electrical signal are respectively taken as the original signal, decomposed into several IMF components and a residual, each IMF component is Hilbert transformed, the IMF component signal and the signal after Hilbert transformation are combined to form an analytical signal, and the signal characteristics of each IMF component are constructed; Acquire the signal characteristics of each sleep state, combine the signal characteristics of each IMF component, identify and mark the IMF components related to each sleep state, reconstruct the IMF components to which the EEG signal and the neuron cluster electrical signal belong and are marked, and obtain the EEG reconstructed signal and the neuron cluster reconstructed electrical signal; Fourier transform is performed on the EEG reconstructed signal and the neuron cluster reconstructed electrical signal respectively to obtain the frequency spectrum of the EEG reconstructed signal and the neuron cluster reconstructed electrical signal, and a comprehensive feature vector is constructed through the frequency spectrum of the EEG reconstructed signal and the neuron cluster reconstructed electrical signal; Based on historical EEG signals and neuron cluster electrical signal data, a sleep state recognition model is constructed. The sleep state recognition model is used to identify the current comprehensive feature vector to determine the current sleep state.

2. A sleep state monitoring method integrating EEG and neuron cluster electrical signals according to claim 1, characterized in that: The EEG signal and the neuron cluster electrical signal are respectively taken as the original signal and decomposed into several IMF components and a residual, including: Find all the local maximum and minimum points in the original signal, use the cubic spline interpolation method to fit the local maximum and minimum points into the upper and lower envelopes of the signal, and calculate the mean of the upper and lower envelopes to obtain the mean line m1; Subtract the mean line m1 from the original signal to obtain a new signal h1, and check whether the new signal h1 meets the IMF condition: the number of extreme points and the number of zero points are equal or differ by 1, and the mean of the upper and lower envelopes is zero; If the new signal h1 does not meet the IMF condition, take the new signal h1 as the original signal and repeat the above steps until a component that meets the IMF condition is obtained; If the new signal h1 satisfies the IMF condition, the new signal h1 is used as the first IMF component c1; Repeat the above steps for the residual signal until the residual signal becomes a monotonic function and can no longer be decomposed into components that meet the IMF conditions. In each iteration, the extracted IMF component is subtracted from the original signal to obtain the next residual signal to be decomposed.

3. The sleep state monitoring method integrating EEG and neuron cluster electrical signals according to claim 2, characterized in that: Construct the signal characteristics of each IMF component, including: By analyzing the real and imaginary parts of the signal, the instantaneous frequency and instantaneous amplitude of each IMF component signal are calculated, and the instantaneous frequency and instantaneous amplitude of each IMF component signal are combined to construct the signal feature V: , the calculation formulas of instantaneous frequency and instantaneous amplitude are as follows: in, represents the instantaneous frequency, represents the instantaneous phase, represents the instantaneous amplitude, represents the real part of the analytical signal, represents the imaginary part of the analytical signal, and t represents the time.

4. The sleep state monitoring method integrating EEG and neuron cluster electrical signals according to claim 1, characterized in that: Identify and label the IMF components associated with each sleep state, including: The correlation between each IMF component and each sleep state is calculated by the Pearson correlation coefficient. The correlation threshold is set in advance, and the absolute value of the Pearson correlation coefficient of each IMF component is compared with the correlation threshold. When the absolute value of the Pearson correlation coefficient is greater than the correlation threshold, the corresponding IMF component is marked, otherwise, no operation is performed.

5. The sleep state monitoring method integrating EEG and neuron cluster electrical signals according to claim 4, characterized in that: The IMF components to which the EEG signal and the neuron cluster electrical signal belong and are marked are reconstructed respectively. The reconstruction process is as follows: The weighting coefficient of each marked IMF component is calculated by the Pearson correlation coefficient of the marked IMF component. The weighting coefficient is: ,in, represents the weight coefficient of the k-th marked IMF component, represents the Pearson correlation coefficient of the kth marked IMF component, and the marked IMF components are reconstructed by weighted combination: ,in, represents the reconstructed signal, represents the IMF component of the kth marker.

6. The sleep state monitoring method integrating EEG and neuron cluster electrical signals according to claim 1, characterized in that: Constructing a comprehensive feature vector includes: Calculate the power spectral density of EEG reconstructed signal and neuron cluster reconstructed electrical signal respectively. The power spectral density is: ,in, represents the power spectral density, represents the Fourier transform of the signal, f represents the frequency, and N represents the length of the signal.

7. The sleep state monitoring method integrating EEG and neuron cluster electrical signals according to claim 6, characterized in that: Calculate the cross power spectral density of EEG reconstructed signals and neuron cluster reconstructed electrical signals, cross power spectral density: ,in, represents the cross power spectral density, , They represent the Fourier transform of EEG reconstructed signal and neuron cluster reconstructed electrical signal respectively.

8. The sleep state monitoring method integrating EEG and neuron cluster electrical signals according to claim 7, characterized in that: Calculate the coherence of EEG reconstructed signals and neuron cluster reconstructed electrical signals. Coherence: ,in, Indicates coherence, represents the cross power spectral density, and They represent the power spectral density of EEG reconstructed signals and neuron cluster reconstructed electrical signals respectively; The power spectral density of EEG reconstructed signals, the power spectral density of neuron cluster reconstructed electrical signals, and coherence are spliced ​​and fused to construct a comprehensive feature vector.

9. The sleep state monitoring method integrating EEG and neuron cluster electrical signals according to claim 1, characterized in that: Build a sleep state recognition model, including: Collect historical EEG signals and neuron cluster electrical signal data, extract comprehensive feature vectors, use clustering algorithms to perform cluster analysis on the comprehensive feature vectors, and assign a label to each cluster based on the clustering results to represent different sleep states; The clustered comprehensive feature vector data set is divided into a training set and a test set. The model is established using a machine learning algorithm. The model is trained using the data in the training set and the trained model is evaluated using the test set data. Based on the evaluation results, the model is adjusted and optimized to obtain a sleep state recognition model.

10. A sleep state monitoring system integrating EEG and neuron cluster electrical signals, used to implement the method according to any one of claims 1 to 9, characterized in that: include: The signal acquisition module uses EEG technology to collect EEG signals from different brain regions in real time, and synchronously records the electrical signals of neuron clusters in different brain regions through in vivo multi-channel synchronous recording technology of central neuron discharge; The signal processing module takes the EEG signal and the neuron cluster electrical signal as the original signal, decomposes them into several IMF components and a residual, performs Hilbert transform on each IMF component, combines the IMF component signal and the signal after Hilbert transform to form an analytical signal, and constructs the signal characteristics of each IMF component; obtains the signal characteristics of each sleep state, combines the signal characteristics of each IMF component, identifies and marks the IMF components related to each sleep state, and reconstructs the marked IMF components to which the EEG signal and the neuron cluster electrical signal belong, respectively, to obtain the EEG reconstructed signal and the neuron cluster reconstructed electrical signal; The feature extraction module performs Fourier transform on the EEG reconstructed signal and the neuron cluster reconstructed electrical signal respectively to obtain the frequency spectrum of the EEG reconstructed signal and the neuron cluster reconstructed electrical signal, and constructs a comprehensive feature vector through the frequency spectrum of the EEG reconstructed signal and the neuron cluster reconstructed electrical signal; The sleep state recognition module builds a sleep state recognition model based on historical EEG signals and neuron cluster electrical signal data, and uses the sleep state recognition model to identify the current comprehensive feature vector to determine the current sleep state.

Images