A real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint

Abstract

The present invention discloses a real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint, which belongs to the field of ultrasonic stimulation control technology, and comprises an ultrasonic parameter configuration module and an ultrasonic stimulation generation module connected together, and the ultrasonic stimulation generation module releases ultrasonic stimulation on the Shenmen acupoint of an experimental subject. A signal acquisition module is arranged on the head of the experimental subject, and a data processing and feature extraction module and a feedback control module are connected to the signal acquisition module in sequence, and the feedback control module is then connected back to the ultrasonic parameter configuration module. A data visualization module is connected to the data processing and feature extraction module. The present invention realizes closed-loop control of EEG in the NREM sleep stage and improves sleep problems. Replacing traditional acupuncture with ultrasonic stimulation reduces possible operational accidents that may occur in traditional acupuncture.

Description

Technical Field

[0001] The present invention relates to the technical field of ultrasonic stimulation regulation, in particular to a real-time sleep rhythm closed-loop regulation system based on ultrasonic stimulation of Shenmen acupoint. Background Art

[0002] Sleep is one of the most important and basic physiological activities of human beings. About one-third of human life is spent on sleep. It can help people relieve fatigue, restore body functions, improve immunity, and play an important role in the normal maintenance of various functions of the human body. High-quality sleep is essential. Sleep quality is closely related to sleep rhythm. Studying the EEG characteristics during sleep is an important part of studying sleep rhythm.

[0003] There are many reasons for sleep problems, including: psychological emotions, physiological diseases, environmental changes, drug effects, behavioral habits, etc. According to traditional Chinese medicine, sleep problems are caused by the imbalance of human organs and the imbalance of yin and yang, which leads to deficiency of the heart and spleen, and the lack of nourishment of the spirit; yin deficiency and excessive fire disturb the mind. In traditional Chinese medicine, sleep problems are often treated by acupuncture directly piercing the Shenmen acupoint, inserting the needle about 0.5 inches, and combining it with other treatments. However, objectively, acupuncture therapy also has certain limitations, such as the possibility of patients fainting from the needle, the possibility of the needle breaking due to improper operation by the doctor, and the possibility of cross infection due to failure to strictly follow the principle of one needle per person during operation.

[0004] In recent years, neuromodulation technology has developed rapidly, and many technologies such as transcranial electrical stimulation, transcranial magnetic stimulation, and transcranial ultrasound stimulation have been applied to the field of rehabilitation therapy. The penetration depth of the first two technologies is very limited, and the latter uses ultrasound to stimulate the target points of cranial nerves through the human skull. The sound energy is seriously reduced, so it is difficult to achieve a good therapeutic effect for sleep problems by directly stimulating cranial nerves with ultrasound. When collecting EEG signals, artifacts will be generated due to problems such as incorrect recording settings, good conductivity of the scalp, and blinking during the signal collection process, making the collected EEG signals unable to be directly applied.

[0005] According to the standard sleep staging algorithm, commonly used sleep staging includes wakefulness, non-rapid eye movement N1 sleep (NREM-N1 sleep), non-rapid eye movement N2 sleep (NREM-N2 sleep), non-rapid eye movement N3 sleep (NREM-N3 sleep), and rapid eye movement sleep (REM), that is, the two main stages of sleep staging are NREM and REM. Since the duration of the NREM stage is most of the sleep cycle, accounting for about 65%-85% of the entire sleep cycle, the 4-7Hz theta wave is the most obvious in this stage. And sleep is a long-term and continuous behavior, so there is an urgent need for a system that can use ultrasound instead of acupuncture to stimulate the Shenmen acupoint by guiding the NREM stage EEG signals to achieve real-time closed-loop regulation of EEG during sleep, improve sleep problems, and regulate sleep rhythm. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of the Shenmen acupoint, which uses the EEG signals in the NREM stage as a guide to correct the ultrasonic stimulation parameters in real time, stimulate the Shenmen acupoint in real time, and improve sleep.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: a real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint, comprising:

[0008] Ultrasonic parameter configuration module: used to configure the parameters of ultrasonic stimulation, including frequency, intensity and duration;

[0009] Ultrasonic stimulation generation module: generates corresponding ultrasonic stimulation signals according to ultrasonic parameters, and performs ultrasonic stimulation on the Shenmen acupoint of the experimental subject;

[0010] Signal acquisition module: set on the head of the sleeping experimental subject, collecting the sleeping EEG signal of the experimental subject after the Shenmen acupoint is stimulated by ultrasound;

[0011] Data processing and feature extraction module: used to pre-process and extract features of the collected EEG signal data, including connecting the following modules in sequence:

[0012] Neural signal processing module: Receives sleep EEG signals and performs preliminary processing. The preliminary processing steps are to downsample the sleep EEG signals, perform 50Hz notch denoising, and perform 0.1Hz-100Hz filtering to obtain preliminary processed sleep EEG signals.

[0013] Preprocessing module: The preliminarily processed sleep EEG signals are subjected to EEMD+FastICA method to remove artifacts and obtain preprocessed sleep EEG signals.

[0014] Feature extraction module: receives the pre-processed sleep EEG signals and performs sleep stage classification and spectrum analysis, divides different sleep cycles and extracts spectrum features to obtain the θ rhythm EEG signals and sleep θ wave EEG features of the NREM stage;

[0015] The feedback control module is used to compare the sleep θ wave EEG characteristics with the preset expectations, and to adjust the frequency, intensity and duration of ultrasound stimulation to achieve real-time adjustment of ultrasound stimulation parameters, including the following modules connected in sequence:

[0016] Data receiving module: receiving the θ rhythm EEG signals and sleep θ wave EEG characteristics in the monitored NREM stage;

[0017] Parameter adjustment module: manually preset expectations, perform difference calculation on the currently obtained θ rhythm EEG signal and sleep θ wave EEG characteristics and the preset expectations, and calculate the ultrasound parameter adjustment amount based on the difference results and control strategy;

[0018] PID module: Using the PID control algorithm, the error size, change trend and accumulation are comprehensively considered through proportional terms, integral terms and differential terms to achieve precise regulation and optimization of the adjustment amount of ultrasound stimulation parameters.

[0019] The further improvement of the technical solution of the present invention is that: the ultrasonic parameter configuration module, the ultrasonic parameters initially configured are parameters selected empirically or obtained based on effective solutions in previous studies, and the real-time ultrasonic parameter adjustment amount obtained by the feedback control module is calculated in real time on the basis of the initial ultrasonic parameters to obtain subsequent ultrasonic parameters. To ensure human safety, the ultrasonic frequency is selected in the low frequency range of 20KHz to 100KHz; the ultrasonic intensity is set at 0.1mW / cm 2 Up to 3mW / cm 2 between.

[0020] The further improvement of the technical solution of the present invention is that the pre-processing module performs the EEMD+FastICA method to remove artifacts in the following steps:

[0021] Step 1.1: Add an additive white noise with a standard normal distribution to the EEG signal;

[0022] Step 1.2: Use the EMD algorithm to decompose the new single-channel sleep EEG signal into a series of intrinsic mode functions (IMFs);

[0023] Step 1.3: Repeat the first two steps several times to obtain multiple IMFs sets;

[0024] Step 1.4: Take the average of the entire IMSFs set to obtain the average IMFs set;

[0025] Step 1.5: Use the FastICA algorithm on the average IMFs set to obtain the corresponding confusion matrix and de-confusion matrix W and independent components;

[0026] Step 1.6: Use the confusion matrix M to reconstruct the EEG signal source into an IMFs set with the noise source removed;

[0027] Step 1.7: Sum the reconstructed IMFs set to reconstruct the signal of interest, i.e., the denoised single-channel sleep EEG signal;

[0028] Perform independent component decomposition on the collected sleep EEG signal x to obtain an unmixing matrix W, which linearly decomposes the independent sources mixed in the multi-channel EEG signal. The output matrix S = xW, each row of which is the time series of each independent source, the mixing matrix M is the inverse matrix of the unmixing matrix W, and the denoised data matrix x′ = MS′;

[0029] The further improvement of the technical solution of the present invention is that the EEMD algorithm is as follows:

[0030] Step 2.1: Set the overall average number m;

[0031] Step 2.2: Substitute a white noise n with a standard normal distribution i (t) is added to the sleep EEG signal x(t) to generate a new signal sequence, x i (t) = x(t) + n i (t)(i=1,2,…,m);

[0032] Step 2.3: Find each set of noisy sleep EEG signals x i (t); and fit the upper and lower envelopes e max (t) and e min (t);

[0033] Step 2.4: Calculate the mean m of the upper and lower envelopes i (t), and calculate the intermediate signal h i (t) = x i (t)-m i (t);

[0034] Step 2.5: Determine h i (t) Whether it is IMF, that is, h i (t) Whether the two conditions of the eigenmode function are met;

[0035] Step 2.6: If h i (t) is not IMF, then use h i (t) replaces x i (t), repeat steps 2.2-2.5 until h i(t) satisfying the criteria;

[0036] Step 2.7: If yes, then h i (t) is the selected IMF1, denoted by c i (t), calculate the difference signal r i (t) = x i (t)-c i (t);

[0037] Step 2.8: i (t) replaces x i (t) Repeat the above process m times to decompose m IMF components and obtain the form of their respective IMF sums, that is,

[0038] Step 2.9: Repeat the above steps m times, adding white noise signals with different amplitudes each time to obtain the set of IMFs: c 1,j (t),c 2,j (t),…c M,j (t),j=1,2,…J;

[0039] Step 2.10: Based on the principle that the statistical mean of unrelated sequences is zero, the corresponding IMFs are averaged to obtain the final IMF after EEMD decomposition, namely:

[0040] The further improvement of the technical solution of the present invention is that the FastICA algorithm is as follows:

[0041] Step 3.1: Normalize the sleep EEG signal x, that is, subtract its mean m = E{x} to make it have a zero mean;

[0042] Step 3.2: Whiten the sleep EEG signal, x′ = ED -1 / 2 E T x, the whitening process removes the correlation between signals and obtains a sleep EEG signal sequence that is uncorrelated with each other;

[0043] Step 3.3: Set the number m of independent components to be estimated and the number of iterations p;

[0044] Step 3.4: Initialize the weight vector matrix W p ;

[0045] Step 3.5: According to the maximum negative entropy principle, W p =E{x′g(W p T x′)}-E{g′(W p T x′)}W p ;

[0046] Step 3.6: make

[0047] Step 3.7: Determine W p Whether it converges, if W p If it does not converge, return to step 3.5;

[0048] Step 3.8: Let p = p + 1, repeat the iteration, if p ≤ m, return to step 3.4.

[0049] The further improvement of the technical solution of the present invention is that: the feature extraction module is specifically calculated as follows: using a standard sleep staging algorithm to perform sleep staging, the sleep staging includes the wakefulness stage, the non-rapid eye movement NREM stage and the rapid eye movement REM stage, and marking each time point as the corresponding sleep stage; using deep learning to detect the NREM sleep time period; for the NREM sleep time period, using fast Fourier transform to convert the signal from the time domain to the frequency domain, and obtain the spectrum of the NREM sleep time period; selecting 4Hz-8Hz theta waves to obtain the theta wave frequency band; integrating the theta wave power on the theta wave frequency band to obtain the theta rhythm electroencephalogram signal of the NREM stage and the sleep theta wave electroencephalogram feature, the specific steps are as follows:

[0050] Step 4.1: Let the preprocessed continuous-time signal be X pre (t);

[0051] Step 4.2: Use the standard sleep staging algorithm to perform sleep staging and obtain the continuous time signals of different sleep stages stage(t), including the continuous time signal of wakefulness stage Wakefulness(t), the continuous time signal of rapid eye movement stage REM(t), and the continuous time signal of non-rapid eye movement stage NREM(t);

[0052] Step 4.3: Extract the time domain features of each stage of stage(t), apply the trained classifier to the sleep data, and determine which sleep stage each time point is in based on the features. Define a binary sequence nrem(t), when nrem(t) = 1, it means that the time point is in the NREM sleep stage, when nrem(t) = 0, it means that the time point is in other sleep periods;

[0053] Step 4.4: Let the EEG signal of NREM sleep stage be X θ (t), X θ (t) = X pre (t)*nrem(t), perform fast Fourier transform X θ (f) = FFT(X θ (t)), then the spectrum is P θ (f) = |Xθ (f)| 2 ;

[0054] Step 4.5: The frequency range of theta waves is [4Hz, 8Hz], i.e. f min =4Hz,f max =8Hz. Then the θ wave power is

[0055] Step 4.6: In real time, the θ wave power of each NREM sleep stage is integrated into the form of a feature vector, which is θ = [P1, P2, P3, …, P i ,…,P n ], where P i is the theta wave power of the i-th NREM sleep stage.

[0056] A further improvement of the technical solution of the present invention is that a data visualization module is connected to the data processing and feature extraction module to present the processed data and features in a visual manner, and to help users intuitively understand and analyze NREM sleep stage data by drawing θ waveform graphs and spectrum graphs.

[0057] A further improvement of the technical solution of the present invention is that in the parameter adjustment module, the preset expectation is the average power of the theta wave in the NREM stage set based on previous studies and literature reports, which is vector data.

[0058] A further improvement of the technical solution of the present invention is that: the control strategy in the parameter adjustment module is to adjust the ultrasonic parameters according to the difference between the preset expectation and the actually observed sleep theta wave EEG characteristics. For the ultrasonic frequency, the frequency offset is adjusted according to the positive and negative direction and size of the difference. If the difference is positive, it means that the actually observed sleep theta wave EEG characteristics are lower than the preset expectation, and the ultrasonic frequency is increased; if the difference is negative, it means that the actually observed sleep theta wave EEG characteristics are higher than the preset expectation, and the ultrasonic frequency is reduced; for the ultrasonic intensity and stimulation duration, the intensity or duration is adjusted according to the size of the difference. If the difference is large, the ultrasonic intensity or duration is increased; if the difference is small, the ultrasonic intensity or duration is reduced.

[0059] A further improvement of the technical solution of the present invention is that the ultrasonic stimulation generation module includes the following modules connected in sequence:

[0060] Signal generation module: generates corresponding ultrasonic stimulation signal output according to the configured ultrasonic parameters;

[0061] Collimation module: limits the output ultrasonic stimulation signal within a set spatial range and transmits it to the Shenmen acupoint of the experimental subject.

[0062] Due to the adoption of the above technical scheme, the technical progress achieved by the present invention is as follows: the present invention collects and analyzes the EEG signals of the experimental subjects in the NREM stage in real time, uses closed-loop ultrasonic stimulation technology to act on the Shenmen acupoint of the experimental subjects, and compares the EEG characteristics of the sleep θ wave calculated by the θ rhythm in the EEG signals with the preset expectations, and configures ultrasonic stimulation in real time to achieve closed-loop regulation of EEG in the NREM sleep stage and improve sleep problems; replaces traditional acupuncture with ultrasonic stimulation, that is, low-intensity focused ultrasound can stimulate the medium affected by the wave to produce mechanical vibration effects and thermal effects during the propagation of sound waves, such as stimulating the cell membrane to produce mechanical vibrations and stretch the thin film composed of lipid bilayers when acting on nerve cells, and mechanical vibrations generate heat, stimulating the activity of nerve cells, and causing changes in neurotransmitters, thereby simulating the effect of acupuncture. It reduces the possible operational accidents of traditional acupuncture, and the operation is simple, hygienic and safe. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

[0064] Figure 1 It is a schematic diagram of the structure of the present invention;

[0065] Figure 2 is a real-time sleep regulation flow chart of the present invention; DETAILED DESCRIPTION

[0066] The present invention is further described in detail below in conjunction with embodiments:

[0067] like Figure 1 The figure shows a schematic diagram of the structure of a real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of the Shenmen acupoint, which includes an ultrasonic parameter configuration module and an ultrasonic stimulation generation module connected together. The ultrasonic stimulation generation module releases ultrasonic stimulation on the Shenmen acupoint of the experimental subject. A signal acquisition module is set on the head of the experimental subject, and the signal acquisition module is connected to the data processing and feature extraction module and the feedback control module in turn, and the feedback control module is connected back to the ultrasonic parameter configuration module. A data visualization module is connected to the data processing and feature extraction module.

[0068] Ultrasonic parameter configuration module: used to configure the parameters of ultrasonic stimulation, including frequency, intensity and duration;

[0069] Ultrasonic stimulation generation module: Generates corresponding ultrasonic stimulation signals according to ultrasonic parameters to perform ultrasonic stimulation on the Shenmen acupoint of the experimental subject, including the following modules connected in sequence:

[0070] Signal generation module: generates corresponding ultrasonic stimulation signal output according to the configured ultrasonic parameters;

[0071] Collimation module: confines the output ultrasonic stimulation signal within a set spatial range and transmits it to the Shenmen acupoint of the experimental subject;

[0072] Signal acquisition module: set on the head of the sleeping experimental subject, collecting the sleeping EEG signal of the experimental subject after the Shenmen acupoint is stimulated by ultrasound;

[0073] Data processing and feature extraction module: used to pre-process and extract features of the collected EEG signal data, including connecting the following modules in sequence:

[0074] Neural signal processing module: Receives sleep EEG signals and performs preliminary processing. The preliminary processing steps are to downsample the sleep EEG signals, perform 50Hz notch denoising, and perform 0.1Hz-100Hz filtering to obtain preliminary processed sleep EEG signals.

[0075] Preprocessing module: The preliminarily processed sleep EEG signals are subjected to EEMD+FastICA method to remove artifacts and obtain preprocessed sleep EEG signals.

[0076] Feature extraction module: receives the pre-processed sleep EEG signals and performs sleep stage classification and spectrum analysis, divides different sleep cycles and extracts spectrum features to obtain the θ rhythm EEG signals and sleep θ wave EEG features of the NREM stage;

[0077] The feedback control module is used to compare the sleep θ wave EEG characteristics with the preset expectations, and to adjust the frequency, intensity and duration of ultrasound stimulation to achieve real-time adjustment of ultrasound stimulation parameters, including the following modules connected in sequence:

[0078] Data receiving module: receiving the θ rhythm EEG signals and sleep θ wave EEG characteristics in the monitored NREM stage;

[0079] Parameter adjustment module: manually preset expectations, perform difference calculation on the currently obtained θ rhythm EEG signal and sleep θ wave EEG characteristics and the preset expectations, and calculate the ultrasound parameter adjustment amount based on the difference results and control strategy;

[0080] PID module: Using the PID control algorithm, the error size, change trend and accumulation are comprehensively considered through proportional terms, integral terms and differential terms to achieve precise regulation and optimization of the adjustment amount of ultrasound stimulation parameters.

[0081] Data visualization module: presents the processed data and features in a visual way, and helps users intuitively understand and analyze NREM sleep stage data by drawing theta waveform graphs and spectrum graphs.

[0082] Ultrasonic parameter configuration module The ultrasonic parameter configuration module is a MATLAB program in the PC. The initially configured ultrasonic parameters are empirically selected or based on the effective schemes in previous studies. The subsequent ultrasonic parameters are calculated in real time based on the initial ultrasonic parameters and the real-time ultrasonic parameter adjustment obtained by the feedback control module. To ensure human safety, the ultrasonic frequency is selected in the low frequency range of 20KHz to 100KHz; the ultrasonic intensity is set at 0.1mW / cm 2 Up to 3mW / cm 2 between.

[0083] In the ultrasonic stimulation generation module, the signal generation module is GE Healthcare's ultrasonic generator. The ultrasonic generator is a device for generating ultrasonic waves. The main components of the device include a driving power supply, an oscillator, a power amplifier, and an ultrasonic transducer. Among them, the oscillator is the core component of the ultrasonic generator, which drives other components to work by generating high-frequency oscillation signals. The oscillator usually uses technologies such as resonant circuits or piezoelectric ceramic chips to achieve high-frequency oscillation; the power amplifier amplifies the signal generated by the oscillator so that it has enough energy to drive the ultrasonic transducer; the ultrasonic transducer is a device that converts electrical energy into mechanical vibrations, and also a device that converts mechanical vibrations into ultrasonic waves. It is usually composed of piezoelectric materials. When voltage is applied, the piezoelectric material will vibrate mechanically and generate ultrasonic waves. According to the configured ultrasonic parameters, the ultrasonic generator can generate ultrasonic signals of specific frequency and intensity. The collimation module is a collimator, which is filled with coupling fluid. The collimator can reduce the attenuation of ultrasound during propagation and is limited to the set spatial range to emit to the Shenmen acupoint of the experimental object. The collimation module is specifically a collimator. The ultrasonic transducer generates ultrasound, the center frequency of which is set to 3 MHz, and a Mantoux phased array transducer with an appearance size of 8×8×23 mm is selected, and the model is 3L16-0.31×5.

[0084] The signal acquisition module is connected to the experimental subject and is used to collect the EEG signals of the experimental subject during sleep after ultrasound stimulation in real time; the experimental subject is set as a college student, healthy, without central nervous system diseases, and aged between 18 and 23 years old. Specifically, the signal acquisition and processing module is an EEG electrode; the EEG electrode is connected to the head of the experimental subject using a unipolar lead method, and the electrodes A1 and A2 on both sides of the earlobes are connected together and grounded as unrelated electrodes. The recording electrode itself does not generate noise and drift, and the collected EEG signal data is usually a continuous time series.

[0085] The neural signal processing module can select the BioSemiActiveTwo neural signal acquisition and processing system, which is connected to the PC through a USB interface and connected to the EEG electrodes at the same time, and is configured to convert the sleep EEG signals into corresponding digital signals, and downsample the converted digital signals, perform 50Hz notch denoising and 0.1-100Hz filtering to obtain the preliminarily processed sleep EEG signals; the preprocessing module is connected to the neural signal processing module, and the preprocessing module is implemented by the MATLAB program in the PC, and is used to remove artifacts from the preliminarily processed sleep EEG signals using the EEMD+FastICA method to obtain the preprocessed sleep EEG signals.

[0086] The steps of removing artifacts using the EEMD+FastICA method in the preprocessing module are as follows:

[0087] Step 1.1: Add an additive white noise with a standard normal distribution to the EEG signal;

[0088] Step 1.2: Use the EMD algorithm to decompose the new single-channel sleep EEG signal into a series of intrinsic mode functions (IMFs);

[0089] Step 1.3: Repeat the first two steps several times to obtain multiple IMFs sets;

[0090] Step 1.4: Take the average of the entire IMSFs set to obtain the average IMFs set;

[0091] Step 1.5: Use the FastICA algorithm on the average IMFs set to obtain the corresponding confusion matrix and de-confusion matrix W and independent components;

[0092] Step 1.6: Use the confusion matrix M to reconstruct the EEG signal source into an IMFs set with the noise source removed;

[0093] Step 1.7: Sum the reconstructed IMFs set to reconstruct the signal of interest, i.e., the denoised single-channel sleep EEG signal;

[0094] Perform independent component decomposition on the collected sleep EEG signal x to obtain an unmixing matrix W, which linearly decomposes the independent sources mixed in the multi-channel EEG signal. The output matrix S = xW, each row of which is the time series of each independent source, the mixing matrix M is the inverse matrix of the unmixing matrix W, and the denoised data matrix x′ = MS′;

[0095] The EEMD algorithm is as follows:

[0096] Step 2.1: Set the overall average number m;

[0097] Step 2.2: Substitute a white noise n with a standard normal distribution i(t) is added to the sleep EEG signal x(t) to generate a new signal sequence, x i (t) = x(t) + n i (t)(i=1,2,…,m);

[0098] Step 2.3: Find each set of noisy sleep EEG signals x i (t); and fit the upper and lower envelopes e max (t) and e min (t);

[0099] Step 2.4: Calculate the mean m of the upper and lower envelopes i (t), and calculate the intermediate signal h i (t) = x i (t)-m i (t);

[0100] Step 2.5: Determine h i (t) Whether it is IMF, that is, h i (t) Whether the two conditions of the eigenmode function are met;

[0101] Step 2.6: If h i (t) is not IMF, then use h i (t) replaces x i (t), repeat steps 2.2-2.5 until h i (t) satisfying the criteria;

[0102] Step 2.7: If yes, then h i (t) is the selected IMF1, denoted by c i (t), calculate the difference signal r i (t) = x i (t)-c i (t);

[0103] Step 2.8: i (t) replaces x i (t) Repeat the above process m times to decompose m IMF components and obtain the form of their respective IMF sums, that is,

[0104] Step 2.9: Repeat the above steps m times, adding white noise signals with different amplitudes each time to obtain the set of IMFs: c 1,j (t),c 2,j (t),…c M,j (t),j=1,2,…J;

[0105] Step 2.10: Using the principle that the statistical mean of unrelated sequences is zero, the corresponding IMF is subjected to the FastICA algorithm as follows:

[0106] Step 3.1: Normalize the sleep EEG signal x, that is, subtract its mean m = E{x} to make it have a zero mean;

[0107] Step 3.2: Whiten the sleep EEG signal, x′ = ED -1 / 2 E T x, the whitening process removes the correlation between signals and obtains a sleep EEG signal sequence that is uncorrelated with each other;

[0108] Step 3.3: Set the number m of independent components to be estimated and the number of iterations p;

[0109] Step 3.4: Initialize the weight vector matrix W p ;

[0110] Step 3.5: According to the maximum negative entropy principle, W p =E{x′g(W p T x′)}-E{g′(W p T x′)}W p ;

[0111] Step 3.6: make

[0112] Step 3.7: Determine W p Whether it converges, if W p If it does not converge, return to step 3.5;

[0113] Step 3.8: Let p = p + 1, repeat the iteration, if p ≤ m, return to step 3.4.

[0114] The final IMF after EEMD decomposition is obtained by ensemble average operation, namely:

[0115] The feature extraction module is implemented by MATLAB and Python programs on the PC. The standard sleep staging algorithm is used to perform sleep staging, which includes the wakefulness stage, the non-rapid eye movement NREM stage, and the rapid eye movement REM stage. Each time point is marked as the corresponding sleep stage; deep learning is used to detect the NREM sleep period; for the NREM sleep period, the fast Fourier transform is used to convert the signal from the time domain to the frequency domain to obtain the spectrum of the NREM sleep period; the 4Hz-8Hz theta wave is selected to obtain the theta wave frequency band; the theta wave power is integrated on the theta wave frequency band to obtain the theta rhythm EEG signal and the sleep theta wave EEG characteristics of the NREM stage. The specific steps are as follows:

[0116] Step 4.1: Let the preprocessed continuous-time signal be X pre (t);

[0117] Step 4.2: Use the standard sleep staging algorithm to perform sleep staging and obtain the continuous time signals of different sleep stages stage(t), including the continuous time signal of wakefulness stage Wakefulness(t), the continuous time signal of rapid eye movement stage REM(t), and the continuous time signal of non-rapid eye movement stage NREM(t);

[0118] Step 4.3: Extract the time domain features of each stage of stage(t), apply the trained classifier to the sleep data, and determine which sleep stage each time point is in based on the features. Define a binary sequence nrem(t), when nrem(t) = 1, it means that the time point is in the NREM sleep stage, when nrem(t) = 0, it means that the time point is in other sleep periods;

[0119] Step 4.4: Let the EEG signal of NREM sleep stage be X θ (t), X θ (t) = X pre (t)*nrem(t), perform fast Fourier transform X θ (f) = FFT(X θ (t)), then the spectrum is P θ (f) = |X θ (f)| 2 ;

[0120] Step 4.5: The frequency range of theta waves is [4Hz, 8Hz], i.e. f min =4Hz,f max =8Hz. Then the θ wave power is

[0121] Step 4.6: In real time, the θ wave power of each NREM sleep stage is integrated into the form of a feature vector, which is θ = [P1, P2, P3, …, P i ,…,P n ], where P i is the theta wave power of the i-th NREM sleep stage. In the present invention, the sleep theta wave EEG feature is a feature vector integrating the theta wave power of each NREM sleep stage.

[0122] The data visualization module is connected to the PC port, specifically a BrainMaster Discovery electroencephalograph.

[0123] The data receiving module is connected to the PC port, and the data is stored in the storage hard disk of the PC. The module is configured to receive the theta rhythm EEG signals in the monitored NREM stage and the sleep theta wave EEG characteristics.

[0124] The parameter adjustment module is a MALTAB program on the PC side, in which preset expectations can be manually input. The preset expectations are the average power of theta waves in the NREM stage set based on previous studies and literature reports, which are vector data. The sleep theta wave EEG characteristics received from the data receiving module, that is, the characteristic vector of theta wave power, are calculated with the preset expectations to obtain the difference, and then the difference is analyzed and calculated through the control strategy to obtain the ultrasonic parameter adjustment amount. The control strategy is to adjust the ultrasonic parameters according to the difference between the preset expectations and the actually observed sleep theta wave EEG characteristics, and adjust the frequency offset according to the positive and negative direction and size of the difference. If the difference is positive, it means that the actually observed sleep theta wave EEG characteristics are lower than the preset expectations, and the ultrasonic frequency is increased; if the difference is negative, it means that the actually observed sleep theta wave EEG characteristics are higher than the preset expectations, and the ultrasonic frequency is reduced; for the ultrasonic intensity and stimulation duration, the intensity or duration is adjusted according to the size of the difference. If the difference is large, the ultrasonic intensity or duration is increased; if the difference is small, the ultrasonic intensity or duration is reduced. The PID module is a MALTAB program on the PC side. It uses the PID algorithm to comprehensively consider the size, change trend and accumulation of errors through proportional terms, integral terms and differential terms, and accurately adjusts and optimizes the ultrasonic stimulation parameter adjustment amount calculated by the parameter adjustment module to obtain the final ultrasonic stimulation parameter adjustment amount, and sends the parameter adjustment amount to the ultrasonic parameter configuration module. Based on the initial ultrasonic parameters, the ultrasonic parameter configuration module calculates the subsequent ultrasonic parameters in real time by calculating the real-time ultrasonic parameter adjustment amount obtained by the feedback control module, and then continues to generate corresponding ultrasonic stimulation for the subsequent ultrasonic parameters to stimulate the Shenmen acupoint.

[0125] A real-time closed-loop sleep rhythm control method based on ultrasonic stimulation of Shenmen acupoint, such as Figure 2 As shown,

[0126] S1: Initial ultrasound parameters were manually set based on experience;

[0127] S12: Limit the range of ultrasound stimulation parameters to ensure safety for the human body;

[0128] S13: generating a corresponding ultrasonic stimulation signal according to the ultrasonic parameters, and stimulating the Shenmen acupoint of the experimental subject;

[0129] S2: Use EEG electrode patches to collect EEG signals of the experimental subjects during sleep after the Shenmen acupoint is stimulated in real time;

[0130] S3: Processing and feature extraction of sleep EEG signals collected in real time;

[0131] S4: The average power of theta waves in the NREM stage was set based on previous studies and literature reports, and the expectation was manually preset;

[0132] S42: calculating the ultrasonic parameter adjustment amount according to the difference between the preset expectation and the actually observed sleep theta wave EEG characteristics and the control strategy;

[0133] S43: adjusting the ultrasonic parameter according to the ultrasonic parameter adjustment amount;

[0134] S5: Repeat the above steps and adjust the ultrasound parameters continuously to make the actually observed sleep theta wave EEG characteristics gradually approach and maintain within the preset expected range, so as to achieve real-time regulation of NREM stage EEG signals.

[0135] 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 real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint, characterized in that: include: Ultrasonic parameter configuration module: used to configure the parameters of ultrasonic stimulation, including frequency, intensity and duration; Ultrasonic stimulation generation module: generates corresponding ultrasonic stimulation signals according to ultrasonic parameters, and performs ultrasonic stimulation on the Shenmen acupoint of the experimental subject; Signal acquisition module: set on the head of the sleeping experimental subject, collecting the sleeping EEG signal of the experimental subject after the Shenmen acupoint is stimulated by ultrasound; Data processing and feature extraction module: used to pre-process and extract features of the collected EEG signal data, including connecting the following modules in sequence: Neural signal processing module: receiving sleep EEG signals and performing preliminary processing. The preliminary processing steps are to downsample the sleep EEG signals, perform 50Hz notch denoising, and perform 0.1Hz-100Hz filtering to obtain preliminary processed sleep EEG signals; Preprocessing module: The preliminarily processed sleep EEG signals are subjected to EEMD+FastICA method to remove artifacts and obtain preprocessed sleep EEG signals; Feature extraction module: receives the pre-processed sleep EEG signals and performs sleep stage classification and spectrum analysis, divides different sleep cycles and extracts spectrum features to obtain the θ rhythm EEG signals and sleep θ wave EEG features of the NREM stage; The feedback control module is used to compare the sleep θ wave EEG characteristics with the preset expectations, and to adjust the frequency, intensity and duration of ultrasound stimulation to achieve real-time adjustment of ultrasound stimulation parameters, including the following modules connected in sequence: Data receiving module: receiving the θ rhythm EEG signals and sleep θ wave EEG characteristics in the monitored NREM stage; Parameter adjustment module: manually preset expectations, perform difference calculation on the currently obtained θ rhythm EEG signal and sleep θ wave EEG characteristics and the preset expectations, and calculate the ultrasound parameter adjustment amount according to the difference result and control strategy; the control strategy in the parameter adjustment module is to adjust the ultrasound parameters according to the difference between the preset expectation and the actually observed sleep θ wave EEG characteristics. For the ultrasound frequency, the frequency offset is adjusted according to the positive and negative direction and size of the difference. If the difference is positive, it means that the actually observed sleep θ wave EEG characteristics are lower than the preset expectations, and the ultrasound frequency is increased; if the difference is negative, it means that the actually observed sleep θ wave EEG characteristics are higher than the preset expectations, and the ultrasound frequency is reduced; for the ultrasound intensity and stimulation duration, the intensity or duration is adjusted according to the size of the difference. If the difference is large, the ultrasound intensity or duration is increased; if the difference is small, the ultrasound intensity or duration is reduced; PID module: Using the PID control algorithm, the error size, change trend and accumulation are comprehensively considered through proportional terms, integral terms and differential terms to achieve precise regulation and optimization of the adjustment amount of ultrasound stimulation parameters.

2. A real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint according to claim 1, characterized in that: The ultrasonic parameter configuration module initially configures ultrasonic parameters that are empirically selected or based on effective solutions in previous studies. The real-time ultrasonic parameter adjustment amount obtained by the feedback control module is calculated in real time based on the initial ultrasonic parameters to obtain subsequent ultrasonic parameters. To ensure human safety, the ultrasonic frequency is selected in the low frequency range of 20KHz to 100KHz; the ultrasonic intensity is set at 0.1mW / cm 2 Up to 3mW / cm 2 between.

3. A real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint according to claim 1, characterized in that: The steps of removing artifacts using the EEMD+FastICA method in the preprocessing module are as follows: Step 1.1: Add an additive white noise with a standard normal distribution to the EEG signal; Step 1.2: Use the EMD algorithm to decompose the new single-channel sleep EEG signal into a series of intrinsic mode functions (IMFs); Step 1.3: Repeat the first two steps several times to obtain multiple IMFs sets; Step 1.4: Take the average of the entire IMSFs set to obtain the average IMFs set; Step 1.5: Use the FastICA algorithm on the average IMFs set to obtain the corresponding confusion matrix and de-confusion matrix W and independent components; Step 1.6: Use the confusion matrix M to reconstruct the EEG signal source into an IMFs set with the noise source removed; Step 1.7: Sum the reconstructed IMFs set to reconstruct the signal of interest, i.e., the denoised single-channel sleep EEG signal; The collected sleep EEG signal x is decomposed into independent components to obtain an unmixing matrix W, which linearly decomposes the independent sources mixed in the multi-channel EEG signals; the output matrix S = xW, each row of which is the time series of each independent source, the mixing matrix M is the inverse matrix of the unmixing matrix W, and the denoised data matrix x′ = MS′.

4. A real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint according to claim 3, characterized in that: The EEMD algorithm is as follows: Step 2.1: Set the overall average number m; Step 2.2: Substitute a white noise n with a standard normal distribution i (t) is added to the sleep EEG signal x(t) to generate a new signal sequence, x i (t) = x(t) + n i (t)(i=1,2,…,m); Step 2.3: Find each set of noisy sleep EEG signals x i (t); and fit the upper and lower envelopes e max (t) and e min (t); Step 2.4: Calculate the mean m of the upper and lower envelopes i (t), and calculate the intermediate signal h i (t) = x i (t)-m i (t); Step 2.5: Determine h i (t) Whether it is IMF, that is, h i (t) Whether the two conditions of the eigenmode function are met; Step 2.6: If h i (t) is not IMF, then use h i (t) replaces x i (t), repeat steps 2.2-2.5 until h i (t) satisfy the criteria; Step 2.7: If yes, then h i (t) is the selected IMF1, denoted by c i (t), calculate the difference signal r i (t) = x i (t)-c i (t); Step 2.8: i (t) replaces x i (t) Repeat the above process m times to decompose m IMF components and obtain the form of their respective IMF sums, that is, Step 2.9: Repeat the above steps m times, adding white noise signals with different amplitudes each time to obtain the set of IMFs: c 1,j (t),c 2,j (t),…c M,j (t),j=1,2,…J; Step 2.10: Based on the principle that the statistical mean of unrelated sequences is zero, the corresponding IMFs are averaged to obtain the final IMF after EEMD decomposition, namely:

5. A real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint according to claim 3, characterized in that: The FastICA algorithm is as follows: Step 3.1: Normalize the sleep EEG signal x, that is, subtract its mean m = E{x} to make it have a zero mean; Step 3.2: Whiten the sleep EEG signal, x′ = ED -1 / 2 E T x, the whitening process removes the correlation between signals and obtains a sleep EEG signal sequence that is uncorrelated with each other; Step 3.3: Set the number m of independent components to be estimated and the number of iterations p; Step 3.4: Initialize the weight vector matrix W p ; Step 3.5: According to the maximum negative entropy principle, W p =E{x′g(W p T x′)}-E{g′(W p T x′)}W p ; Step 3.6: make Step 3.7: Determine W p Whether it converges, if W p If it does not converge, return to step 3.5; Step 3.8: Let p = p + 1, repeat the iteration, if p ≤ m, return to step 3.

4.

6. A real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint according to claim 1, characterized in that: The feature extraction module specifically calculates as follows: using a standard sleep staging algorithm to perform sleep staging, sleep staging includes wakefulness stage, non-rapid eye movement (NREM) stage and rapid eye movement (REM) stage, marking each time point as a corresponding sleep stage; using deep learning to detect the NREM sleep period; for the NREM sleep period, using fast Fourier transform to convert the signal from the time domain to the frequency domain, and obtaining the spectrum of the NREM sleep period; selecting the 4Hz-8Hz theta wave to obtain the theta wave frequency band; integrating the theta wave power on the theta wave frequency band to obtain the theta rhythm EEG signal of the NREM stage and the sleep theta wave EEG feature, the specific steps are as follows: Step 4.1: Let the preprocessed continuous-time signal be X pre (t); Step 4.2: Use the standard sleep staging algorithm to perform sleep staging and obtain the continuous time signal stage(t) of different sleep stages; Including the continuous time signal of wakefulness (t), the continuous time signal of rapid eye movement (REM) (t), and the continuous time signal of non-rapid eye movement (NREM) (t); Step 4.3: Extract the time domain features of each stage of stage(t), apply the trained classifier to the sleep data, and determine which sleep stage each time point is in based on the features; define a binary sequence nrem(t), when nrem(t) = 1, it means that the time point is in the NREM sleep stage, when nrem(t) = 0, it means that the time point is in other sleep periods; Step 4.4: Let the EEG signal of NREM sleep stage be X θ (t), X θ (t) = X pre (t)*nrem(t), perform fast Fourier transform X θ (f) = FFT(X θ (t)), then the spectrum is P θ (f) = |X θ (f)| 2 ; Step 4.5: The frequency range of theta waves is [4Hz, 8Hz], i.e. f min =4Hz,f max =8Hz; then the θ wave power is Step 4.6: In real time, the θ wave power of each NREM sleep stage is integrated into the form of a feature vector, which is θ = [P1, P2, P3, …, P i ,…,P n ], where P i is the theta wave power of the i-th NREM sleep stage.

7. A real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint according to claim 1, characterized in that: The data processing and feature extraction module is connected to a data visualization module for presenting the processed data and features in a visual manner, and helps users intuitively understand and analyze NREM sleep stage data by drawing θ waveform graphs and spectrum graphs.

8. The real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint according to claim 1, characterized in that: In the parameter adjustment module, the preset expectation is the average power of the theta waves in the NREM stage set based on previous studies and literature reports, which is vector data.

9. The real-time sleep rhythm closed-loop control system based on ultrasonic stimulation of Shenmen acupoint according to claim 1, characterized in that: The ultrasonic stimulation generation module includes the following modules connected in sequence: Signal generation module: generates corresponding ultrasonic stimulation signal output according to the configured ultrasonic parameters; Collimation module: limits the output ultrasonic stimulation signal within a set spatial range and transmits it to the Shenmen acupoint of the experimental subject.

Images