CN117860266A

CN117860266A - Intelligent insomnia treatment method and device based on real-time EEG monitoring technology

Info

Publication number: CN117860266A
Application number: CN202410214435.XA
Authority: CN
Inventors: 缪国栋
Original assignee: Xiamen Dnake Intelligent Technology Co ltd
Current assignee: Xiamen Dnake Intelligent Technology Co ltd
Priority date: 2024-02-27
Filing date: 2024-02-27
Publication date: 2024-04-12

Abstract

The present invention discloses an intelligent insomnia treatment method and device based on real-time EEG monitoring technology, and the intelligent insomnia treatment method includes the following steps: S1: using a computer tomography device to place electrodes at specific positions on the patient's head, collect electrical signals from different brain regions of the patient, and conduct a comprehensive assessment of the patient's sleep condition; S2: performing preprocessing operations such as filtering, artifact removal, and signal correction on the data, and extracting useful features from the preprocessed EEG signals. The present invention can formulate a personalized treatment plan that meets the patient's needs by collecting and analyzing the patient's brain region signals, and use intelligent algorithms to perform personalized treatment interventions, thereby improving the treatment effect, and being able to feed back the treatment results to the patient, so that the patient can clearly understand his or her own situation.

Description

Intelligent insomnia treatment method and device based on real-time electroencephalogram monitoring technology

Technical Field

The invention relates to the technical field of insomnia treatment, in particular to an intelligent insomnia treatment method and device based on a real-time electroencephalogram monitoring technology.

Background

Normal sleep generally comprises a process from awake to deep sleep, in which brain waves show obvious regularity, when sleep disorder occurs, brain electric activity in a certain stage of the sleep process and even in the whole sleep process is abnormal, the first appearance of the sleep disorder is that the patient is difficult to fall asleep (namely insomnia), the period from awake to falling asleep is long, the trend of gradually lowering the frequency of the brain waves is slow and insignificant, and unstable frequency jump occurs along with the slow trend, the second appearance of the sleep disorder is that the sleep quality is poor, the patient is easy to wake and dream, the patient feels like sleeping like non-sleep, like waking, dizziness, trance, the brain electric signals at the moment are mostly in a theta wave frequency band, the second stage is mainly light sleep me, the third stage of deep sleep is deep sleep, and the proportion of the W stage is very small;

in order to improve sleep quality, a patient usually adopts a magnetic field sleep promoting instrument to promote sleep, the existing magnetic field sleep promoting instrument usually uses a certain waveform with frequency close to brain waves, or the brain waves stored in a memory in advance generate a time-varying magnetic field with corresponding waveform, and the brain waves of the patient can be influenced by the magnetic field energy in a given way, so that sleep promotion can be realized according to ideal sleep promoting steps; however, the sleep-promoting mode in the prior art has the following defects when in use: because the brain electrical mode and the response of each person are unique, the individual treatment scheme can not be proposed to realize targeted treatment according to the condition of a patient, and the same treatment scheme leads to unsatisfactory treatment effect and can not meet the use requirement, and the conditions are summarized, so that the intelligent insomnia treatment method and device based on the real-time brain electrical monitoring technology are provided.

Disclosure of Invention

Based on the technical problems in the background technology, the invention provides an intelligent insomnia treatment method and device based on a real-time electroencephalogram monitoring technology.

The invention provides an intelligent insomnia treatment method based on a real-time electroencephalogram monitoring technology, which comprises the following steps of:

s1: placing electrodes at specific positions of the head of a patient by using a computer graph, collecting electrical signals of different brain areas of the patient, and comprehensively evaluating the sleeping condition of the patient;

s2: performing preprocessing operations such as filtering, artifact removal, signal correction and the like on the data, and extracting useful features from the preprocessed electroencephalogram signals;

s3: according to the characteristics of the computer signals, using a professional sleep analysis algorithm to enable the sleep atmosphere of the patient to be different stages, such as a waking period, an N1 period, an N2 period, an N3 period, a REM period and the like;

s4: identifying whether the patient has insomnia according to the sleep characteristics and evaluation criteria of the patient, formulating a customized treatment scheme according to personal differences, and adopting an intelligent algorithm to perform personalized treatment intervention;

s5: during treatment, the sleeping condition of the patient can be timely fed back and prompted by means of mobile application degree, intelligent equipment or e-mail, and feedback content can comprise sleeping quality scores, sleeping period analysis, measures for suggesting improvement and the like.

Preferably, in the step S1, the computer graphics instrument is used to collect the electrical signals of different brain regions of the patient, and the acquisition logic is as follows:

positioning electrode position: according to the international 10-20 system, the corresponding position of the electrode placed on the scalp is determined, the scalp is divided into different areas, and the specific position of the electrode placed can be determined by measuring the position of the head circumference and using the mark points;

preparing an electrode and an electrode cap: attaching an electrode to the electrode cap and ensuring that the electrode is in close proximity to the patient's scalp, wherein the electrode may be a metal disk or needle electrode which is attached to the scalp by a conductive adhesive;

cleaning the scalp: the degreasing shampoo or alcohol is smeared on the scalp, so that the scalp grease can be removed to ensure good contact quality;

connecting the electrodes to an amplifier: connecting the electrodes to an amplifier or recording device via wires, the amplifier being to receive and amplify the weak electrical signals captured by the electrodes for subsequent processing and analysis;

collecting an electric signal: when a patient performs a specific task (such as eye closure, relaxation, etc.) or performs a specific stimulus (such as light flickering, auditory stimulus, etc.), the acquired electric signals are started to be acquired, and the acquired electric signals are transmitted to an amplifier through an electrode and recorded;

data analysis and interpretation: the acquired electrical signals may be processed and analyzed to obtain various electroencephalographic features and to perform event-related potential analysis.

Preferably, in S2, the specific steps of preprocessing the data and extracting the features are as follows:

(1) And (3) data preprocessing:

and (3) data filtering: filtering the acquired brain electrical signals by using a digital filter to remove high-frequency noise and low-frequency noise;

denoising signals: various noise components in the electroencephalogram signals, such as power line interference, myoelectric interference and the like, are reduced by adopting a denoising algorithm (such as wavelet denoising, independent component analysis and the like);

artifact removal: removing artifacts by using a filter or other methods according to the sampling frequency of the electroencephalogram signals and the possible occurrence of artifact interference (such as ocular artifacts caused by blinking);

removing pseudo-body motion: for electroencephalogram signal fluctuation caused by head movement, false body movement removal can be realized through a signal processing technology (such as correction algorithm or average removal and the like);

(2) And (3) extracting features:

extracting time domain features: extracting a series of time domain features such as average energy, variance, peak value and the like from the original time sequence brain electrical signals;

extracting frequency domain features: converting the brain electrical signal into the frequency domain by performing a Fourier transform spectrum analysis method and extracting a series of frequency domain features such as amplitude spectral density, band power ratio, etc., which are used in the analysisThe mathematical expression of (2) is:where X (f) is the frequency spectrum in the frequency domain, X (t) is the signal in the time domain, f is the frequency, and e is the base of the natural logarithm e;

time-frequency domain feature extraction: using a time-frequency analysis method wavelet transformation to obtain a time-frequency image of the electroencephalogram signal and extracting time-frequency domain features such as energy distribution, phase distribution and the like from the time-frequency image, wherein the wavelet transformation uses the expression: CWT (a, b) = ζ x (t) = { (t-b)/a } dt, where x (t) represents a continuous time signal, ψ (a, b) represents a wavelet function, a represents a scale parameter, b represents a shift parameter, ψ represents a convolution operation, and ψ represents the conjugate of the wavelet function;

nonlinear feature extraction: by introducing a nonlinear dynamics model-image reconstruction method, nonlinear characteristics of the electroencephalogram signals, such as fitting degree, bifurcation analysis and the like, are extracted, and the mathematical expressions used by the method are as follows: y=f (Ax), where f represents a nonlinear function, a represents the measurement matrix of the system, and x represents the image to be reconstructed.

Preferably, the specific step of S3 is as follows:

selecting a proper deep learning model, wherein the model is one of a Convolutional Neural Network (CNN), a cyclic neural network (RNN) or an attention mechanism, and constructing a corresponding model architecture;

dividing the data set: dividing the data set into a training set, a verification set and a test set, wherein the training set is used for model training, the verification set is used for adjusting model super parameters and monitoring model performance, and the test set is used for evaluating final model performance;

model training: the training set is used for carrying out iterative training on the model, the model parameters are adjusted by optimizing the loss function so as to improve the accuracy of sleep segmentation, and the expression used is as follows: θ=argminl (y (x), y (x)), where θ represents an optimal parameter value obtained in the optimization process, argmin represents a parameter value corresponding to the minimum value, L represents a loss function, and is used to measure a difference or an error between a predicted result y (x) of the model and a true value y (x), y (x) represents a true value or a label corresponding to the input sample x, and y (x) represents a predicted value of the model to the input sample x, which is an output of the model adjusted by the model parameter θ;

model verification and tuning: evaluating the performance of the model by using the verification set, and performing model tuning according to the need, such as super parameter adjustment, regularization addition, model structure adjustment and the like;

model test and evaluation: evaluating the finally trained model by using a test set to obtain performance indexes of the model on new data, such as accuracy, recall, F1 score and the like;

model application and deployment: the trained model is applied to new sleep data, segment prediction is performed, and subsequent analysis and processing are performed as needed, so that the whole night sleep is divided into different stages, such as a awake period, a sleep period (including light sleep and deep sleep), and REM sleep.

Preferably, in the step S4, it is identified whether the patient has insomnia, and the acquiring logic is as follows:

extracting features contributing to insomnia recognition from the sleep stage results, wherein the features can comprise a sleep stage proportion, a sleep depth index, a sleep fragmentation index, a sleep quality assessment index and the like;

selecting the features with the most distinguishing and predicting capabilities for insomnia identification by a feature selection algorithm and combining a statistical analysis method;

using a machine learning or artificial intelligence method to establish a classification model-support vector machine, taking the extracted characteristics as input, dividing the patient into insomnia groups and non-insomnia groups, wherein the mathematical expression used is as follows: f (x) =sign (w·x+b), where w is the normal vector of the hyperplane, b is the offset of the hyperplane, for the input sample x, its distance from the hyperplane is calculated, if the result is greater than 0, the sample belongs to the positive class (y=1), and if the result is less than 0, the sample belongs to the negative class (y= -1).

Preferably, in the step S4, a personalized therapeutic intervention is performed by using an intelligent algorithm, which specifically includes the following steps:

algorithm model selection: according to the feature extraction result, selecting a proper intelligent algorithm model for personalized analysis and prediction, wherein the intelligent algorithm model can be one of a machine learning algorithm (such as a support vector machine, a random forest, a neural network and the like) and a deep learning algorithm (such as a convolutional neural network, a cyclic neural network and the like);

establishing a personalized model: establishing a personalized treatment model according to the electroencephalogram characteristics and other related information of the individual by using the selected algorithm model, wherein the model can be used for predicting according to real-time data of a patient and evaluating sleep quality and insomnia symptoms of the patient;

making a treatment scheme: based on the prediction result of the personalized model, a targeted treatment scheme is formulated, and according to the requirements and characteristics of a patient, proper therapeutic intervention means such as sound, light, respiratory training, cognitive behavior therapy and the like are adopted to improve the sleeping quality of the patient and relieve insomnia symptoms;

real-time monitoring and adjustment: after implementing the therapeutic intervention scheme, continuously monitoring the electroencephalogram signal and the therapeutic effect of the patient, and timely adjusting the therapeutic scheme according to the monitoring result so as to achieve the optimal effect.

Preferably, the support vector machine is a supervised learning model, the support vector machine is supported by constructing one or more hyperplanes (parting lines or curved surfaces) and dividing samples of different classes, the input of the support vector machine is a set of feature vectors, the feature vectors can be used for describing various attributes of the samples, the output of the support vector machine is a classification result of a new sample and indicates which class of training data the sample belongs to, and the main function of the support vector machine is to perform two-class and multi-class tasks, which can be suitable for linear separable, linear inseparable and nonlinear classification problems;

the random forest is an integrated learning method, classification or regression tasks are completed by establishing a plurality of sets of decision trees, the input of the random forest is a group of feature vectors, the feature vectors are used for constructing nodes of the decision trees, the output of the random forest is a result of classifying or carrying out regression prediction on new samples and can be a discrete class label or a continuous numerical value, the random forest can be used for solving the problems of classification, regression, feature selection and the like, and the random forest has the advantages of overfitting resistance, high-dimensional data processing capability and the like;

the neural network is a network structure composed of a plurality of neurons (or called nodes), wherein the neurons are subjected to information transmission and processing through connection weights, the input of the neural network is a group of feature vectors, the feature vectors are used as nodes of an input layer and used for representing input features of samples, the output of the neural network is a prediction result of classifying, regressing or other tasks on a new sample, the prediction result can be discrete class labels or continuous numerical values, and the neural network can be used for tasks such as pattern recognition, classification, regressing and clustering and has strong fitting capacity and nonlinear modeling capacity;

the convolutional neural network consists of a convolutional layer (Convolitional layer), a pooling layer (PoolingLayer) and a full-connection layer (FullyConnectedLayer), the input of the convolutional neural network is a two-dimensional image, and the convolutional neural network can be a gray image (single channel) or a color image (multiple channels), and the convolutional neural network has the following functions: the method has excellent performance in the field of image processing, and the local characteristics and the spatial structure of the image can be extracted through the convolution layer and the pooling layer;

the cyclic neural network mainly comprises one or more cyclic layers, wherein an activation function takes an output of a previous moment as an input of a current moment, the input of the cyclic neural network is suitable for processing sequence data, such as natural language, time sequence data and the like, and the functions of the cyclic neural network are as follows: the memory unit of the loop layer can capture and utilize the previous information in the sequence data, so as to predict, classify or generate future output.

The invention also provides an intelligent insomnia treatment device based on the real-time electroencephalogram monitoring technology, which comprises a data acquisition module, a data preprocessing module, a feature extraction module, a sleep stage module, an insomnia identification module, a therapeutic intervention module, a feedback prompt module, a data storage and management module and a user interface module;

the medical staff and the patient can operate the system through the user interface module;

the data acquisition module is used for acquiring electric signals of different brain areas of a patient and carrying out storage management on data through the data storage and management module;

the data preprocessing module is used for preprocessing the acquired data such as filtering, artifact removal, signal correction and the like, and extracting useful features from the preprocessed electroencephalogram signals through the feature extraction module;

the sleep stage module uses a professional sleep analysis algorithm to stage different sleep atmospheres of a patient according to the characteristics of the computer signals;

the insomnia recognition module is used for recognizing whether the patient has insomnia;

the therapeutic intervention module adopts an intelligent algorithm to perform personalized therapeutic intervention on a patient;

the feedback prompt module can timely feed back and prompt the sleeping condition of the patient in a mode of mobile application degree, intelligent equipment or electronic mail and the like.

Compared with the prior art, the invention has the beneficial effects that:

1. the electrical signals of the brain region of the patient are collected, analyzed and evaluated to prepare a personalized treatment scheme which accords with the patient, and an intelligent algorithm is adopted to perform personalized treatment intervention, so that the treatment effect can be improved;

2. the sleeping conditions of the patient are fed back and prompted in time in a mode of moving application degree, intelligent equipment or electronic mail and the like, so that the patient can clearly know own conditions;

according to the invention, through collecting and analyzing brain area signals of a patient, a personalized treatment scheme which accords with the patient can be prepared, and personalized treatment intervention is performed by adopting an intelligent algorithm, so that the treatment effect can be improved, the treatment result can be fed back to the patient, and the patient can clearly know the situation of the patient conveniently.

Drawings

FIG. 1 is a flow chart of an intelligent insomnia treatment method based on a real-time electroencephalogram monitoring technology;

fig. 2 is a block diagram of an intelligent insomnia treatment device based on a real-time electroencephalogram monitoring technology.

Detailed Description

The invention is further illustrated below in connection with specific embodiments.

Examples

Referring to fig. 1, the embodiment provides an intelligent insomnia treatment method based on a real-time electroencephalogram monitoring technology, which comprises the following steps:

s1: placing electrodes at specific positions of the head of a patient by using a computer graph, collecting electrical signals of different brain areas of the patient, and comprehensively evaluating the sleeping condition of the patient; wherein, the computer graph instrument is used for collecting the electric signals of different brain areas of a patient, and the acquisition logic is as follows:

data analysis and interpretation: the acquired electrical signals can be processed and analyzed to obtain various electroencephalogram characteristics and perform event-related potential analysis;

s2: performing preprocessing operations such as filtering, artifact removal, signal correction and the like on the data, and extracting useful features from the preprocessed electroencephalogram signals; the specific steps of preprocessing the data and extracting the characteristics are as follows:

(1) And (3) data preprocessing:

(2) And (3) extracting features:

extracting frequency domain features: by performing a fourier transform spectrum analysis method, an electroencephalogram signal is converted into a frequency domain, and a series of frequency domain features such as amplitude spectral density, frequency band power ratio and the like are extracted, and mathematical expressions used in analysis are as follows:where X (f) is the frequency spectrum in the frequency domain, X (t) is the signal in the time domain, f is the frequency, and e is the base of the natural logarithm e;

nonlinear feature extraction: by introducing a nonlinear dynamics model-image reconstruction method, nonlinear characteristics of the electroencephalogram signals, such as fitting degree, bifurcation analysis and the like, are extracted, and the mathematical expressions used by the method are as follows: y=f (Ax), where f represents a nonlinear function, a represents the measurement matrix of the system, and x represents the image to be reconstructed;

s3: according to the characteristics of computer signals, a professional sleep analysis algorithm is used for carrying out different stages of sleep atmosphere of a patient, such as a waking stage, an N1 stage, an N2 stage, an N3 stage, a REM stage and the like, and the specific steps are as follows:

model application and deployment: applying the trained model to new sleep data, carrying out subsection prediction, and carrying out subsequent analysis and processing according to the requirement, so as to divide the whole night sleep into different stages, such as a waking period, a sleeping period (including light sleep and deep sleep) and REM sleep;

wherein, whether the patient has insomnia problem is identified, the acquisition logic is as follows:

using a machine learning or artificial intelligence method to establish a classification model-support vector machine, taking the extracted characteristics as input, dividing the patient into insomnia groups and non-insomnia groups, wherein the mathematical expression used is as follows: f (x) =sign (w·x+b), where w is the normal vector of the hyperplane, b is the offset term of the hyperplane, for the input sample x, calculating its distance from the hyperplane, if the result is greater than 0, the sample belongs to the positive class (y=1), if the result is less than 0, the sample belongs to the negative class (y= -1);

the intelligent algorithm is adopted for personalized therapeutic intervention, and the specific steps are as follows:

algorithm model selection: according to the feature extraction result, selecting a proper intelligent algorithm model for personalized analysis and prediction, wherein the intelligent algorithm model can be one of a machine learning algorithm (such as a support vector machine, a random forest, a neural network and the like) and a deep learning algorithm (such as a convolutional neural network, a cyclic neural network and the like); wherein the support vector machine is a supervised learning model, the support vector machine is supported by constructing one or more hyperplanes (parting lines or curved surfaces) and dividing samples of different classes, the input of the support vector machine is a set of eigenvectors which can be used to describe various properties of the samples, the output of the support vector machine is a classification result of a new sample, which indicates which class of training data the sample belongs to, the main function of the support vector machine is to perform two-classification and multi-classification tasks, which can be applied to the problems of linear separable, linear inseparable and nonlinear classification, the random forest in this step is an integrated learning method, the classification or regression task is accomplished by constructing a set of decision trees, the input of the random forest is a set of eigenvectors which are used to construct nodes of the decision tree, the output of the random forest is the result of classifying or predicting the new sample, which can be discrete class label or continuous value, the random forest can be used for solving the problems of classification, regression, feature selection and the like, has the advantages of anti-overfitting capability, high-dimensional data processing capability and the like, the neural network in the step is a network structure composed of a plurality of neurons (or called nodes), wherein the neurons are in information transmission and processing through connection weights, the input of the neural network is a group of feature vectors which are used as nodes of an input layer and are used for representing the input features of the sample, the output of the neural network is the predicting result of classifying, regressing or other tasks on the new sample, can be discrete class label or continuous value, the neural network can be used for tasks such as pattern recognition, classification, regression, clustering and the like, the method has strong fitting capability and nonlinear modeling capability, the convolutional neural network in the step consists of a convolutional layer (Convolitional layer), a pooling layer (PoolingLayer) and a full-connection layer (FullyConnectcedLayer), the input of the convolutional neural network is a two-dimensional image, which can be a gray level image (single channel) or a color image (multiple channels), and the convolutional neural network has the following functions: the method has excellent performance in the field of image processing, local features and spatial structures of images can be extracted through a convolution layer and a pooling layer, the cyclic neural network in the step mainly consists of one or more cyclic layers, wherein an activation function takes output at the previous moment as input at the current moment, the input of the cyclic neural network is suitable for processing sequence data, such as natural language, time sequence data and the like, and the functions of the cyclic neural network are as follows: the memory unit of the circulating layer can capture and utilize the previous information in the sequence data, so that future output is predicted, classified or generated;

real-time monitoring and adjustment: after implementing the therapeutic intervention scheme, continuously monitoring the electroencephalogram signal and the therapeutic effect of the patient, and timely adjusting the therapeutic scheme according to the monitoring result so as to achieve the optimal effect;

According to the embodiment, through collecting and analyzing brain area signals of a patient, a personalized treatment scheme which accords with the patient can be made, and personalized treatment intervention is performed by adopting an intelligent algorithm, so that the treatment effect can be improved, the treatment result can be fed back to the patient, and the patient can clearly know the situation of the patient conveniently.

Example 2

Referring to fig. 2, the embodiment provides an intelligent insomnia treatment device based on a real-time electroencephalogram monitoring technology, which comprises a data acquisition module, a data preprocessing module, a feature extraction module, a sleep stage module, an insomnia identification module, a treatment intervention module, a feedback prompt module, a data storage and management module and a user interface module;

The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical scheme of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.

Claims

1. An intelligent insomnia treatment method based on real-time EEG monitoring technology, characterized in that it includes the following steps:

S1: Use a computer to place electrodes at specific locations on the patient's head to collect electrical signals from different brain regions and conduct a comprehensive assessment of the patient's sleep status;

S2: Perform preprocessing operations such as filtering, artifact removal and signal correction on the data to extract useful features from the preprocessed EEG signals;

S3: Based on the characteristics of computer signals, a professional sleep analysis algorithm is used to classify the patient's sleep atmosphere into different stages, such as wakefulness, N1, N2, N3 and REM;

S4: Identify whether the patient has insomnia based on their sleep characteristics and assessment criteria, develop customized treatment plans based on individual differences, and use intelligent algorithms for personalized treatment interventions;

S5: During treatment, timely feedback and reminders on the patient's sleep status can be provided through mobile applications, smart devices or emails. The feedback may include sleep quality scores, sleep cycle analysis and suggested improvement measures.

2. The intelligent insomnia treatment method based on real-time EEG monitoring technology according to claim 1 is characterized in that, in S1, a computer imager is used to collect electrical signals from different brain regions of the patient, and the acquisition logic is as follows:

Locating the electrode position: According to the international 10-20 system, determine the corresponding position of the electrode on the scalp and divide the scalp into different areas. By measuring the head circumference and using the position of the landmark points, the specific location of the electrode can be determined;

Prepare electrodes and electrode cap: Connect the electrodes to the electrode cap and make sure that the electrodes fit tightly against the patient's scalp. The electrodes can be metal disks or needle electrodes, which are attached to the scalp with conductive glue.

Clean the scalp: Apply degreasing shampoo or alcohol on the scalp to remove scalp oil to ensure good contact quality;

Connecting the electrodes to an amplifier: The electrodes are connected to an amplifier or recording device via wires. The amplifier will receive and amplify the weak electrical signals captured by the electrodes for subsequent processing and analysis.

Collecting electrical signals: When the patient performs a specific task (such as closing the eyes, relaxing, etc.) or receives a specific stimulus (such as light flashing, auditory stimulation, etc.), the electrical signal is collected and transmitted to the amplifier through the electrodes and recorded;

Data Analysis and Interpretation: The acquired electrical signals can be processed and analyzed to obtain various EEG features and perform event-related potential analysis.

3. The intelligent insomnia treatment method based on real-time EEG monitoring technology according to claim 1 is characterized in that, in S2, the specific steps of preprocessing the data and extracting features are as follows:

(1) Data preprocessing steps:

Data filtering: Use digital filters to filter the collected EEG signals to remove high-frequency and low-frequency noise;

Signal denoising: Denoising algorithms are used to reduce various noise components in EEG signals, such as power line interference, myoelectric interference, etc.

Artifact removal: According to the sampling frequency of the EEG signal and possible artifact interference, use filters or other methods to remove artifacts;

Pseudo-motion removal: For EEG signal fluctuations caused by head movement, pseudo-motion removal can be achieved through signal processing technology;

(2) Steps of feature extraction:

Time domain feature extraction: Extract a series of time domain features from the original time series EEG signals, such as average energy, variance, peak value, etc.;

Frequency domain feature extraction: The EEG signal is converted to the frequency domain by performing Fourier transform spectrum analysis, and a series of frequency domain features are extracted, such as amplitude spectrum density, frequency band power ratio, etc. The mathematical expression used in the analysis is: Where X(f) is the spectrum in the frequency domain, x(t) is the signal in the time domain, f is the frequency, and e is the base of the natural logarithm e;

Time-frequency domain feature extraction: Use the time-frequency analysis method wavelet transform to obtain the time-frequency image of the EEG signal, and extract the time-frequency domain features, such as energy distribution, phase distribution, etc. The expression used by the wavelet transform is: CWT(a,b)＝∫x(t)*ψ*{(t-b)/a}dt, where x(t) represents the continuous time signal, ψ(a,b) represents the wavelet function, a represents the scale parameter, b represents the translation parameter, * represents the convolution operation, and ψ* represents the conjugate of the wavelet function;

Nonlinear feature extraction: By introducing the nonlinear dynamic model-image reconstruction method, the nonlinear characteristics of EEG signals are extracted, such as fitting degree, bifurcation analysis, etc. The mathematical expression used is: y=f(Ax), where f represents the nonlinear function, A represents the measurement matrix of the system, and x represents the image to be reconstructed.

4. The intelligent insomnia treatment method based on real-time EEG monitoring technology according to claim 1, characterized in that the specific steps of S3 are as follows:

Select a suitable deep learning model, which is one of the convolutional neural network (CNN), recurrent neural network (RNN) or attention mechanism, and build the corresponding model architecture;

Divide the dataset: Divide the dataset into training set, validation set and test set. The training set is used for model training, the validation set is used to adjust model hyperparameters and monitor model performance, and the test set is used to evaluate the final model performance.

Model training: Use the training set to iteratively train the model, and adjust the model parameters by optimizing the loss function to improve the accuracy of sleep segmentation. The expression used is: θ* = argminL(y(x), y(x)), where θ*: represents the optimal parameter value obtained during the optimization process, argmin: represents the parameter value corresponding to the minimum value, L: represents the loss function, which is used to measure the difference or error between the model's predicted result y(x) and the true value y(x), y(x): represents the true value or label corresponding to the input sample x, y(x): represents the model's predicted value for the input sample x, which is the output of the model adjusted by the model parameter θ;

Model validation and tuning: Use the validation set to evaluate the performance of the model and perform model tuning as needed, such as adjusting hyperparameters, adding regularization, adjusting the model structure, etc.

Model testing and evaluation: Use the test set to evaluate the final trained model and obtain the performance indicators of the model on new data, such as accuracy, recall, F1 score, etc.

Model application and deployment: Apply the trained model to new sleep data for segmented predictions and subsequent analysis and processing as needed to divide a night’s sleep into different stages, such as wakefulness, sleep (including light and deep sleep), and REM sleep.

5. The intelligent insomnia treatment method based on real-time EEG monitoring technology according to claim 1 is characterized in that in S4, whether the patient has insomnia is identified, and the acquisition logic is as follows:

Extract features that are helpful for insomnia identification from the results of sleep staging. These features may include the proportion of sleep stages, sleep depth index, sleep fragmentation index, sleep quality assessment index, etc.

Through feature selection algorithms, combined with statistical analysis methods, the most discriminative and predictive features for insomnia identification are selected;

A classification model, support vector machine, is established using machine learning or artificial intelligence methods. The extracted features are used as input to divide patients into insomnia group and non-insomnia group. The mathematical expression used is: f(x) = sign(w·x+b), where w is the normal vector of the hyperplane, and b is the bias term of the hyperplane. For the input sample x, its distance from the hyperplane is calculated. If the result is greater than 0, the sample belongs to the positive class (y=1). If the result is less than 0, the sample belongs to the negative class (y=-1).

6. The intelligent insomnia treatment method based on real-time EEG monitoring technology according to claim 1 is characterized in that, in S4, an intelligent algorithm is used to perform personalized treatment intervention, and the specific steps are as follows:

Algorithm model selection: According to the feature extraction results, select a suitable intelligent algorithm model for personalized analysis and prediction. The intelligent algorithm model can be one of the machine learning algorithms (such as support vector machine, random forest, neural network, etc.) and deep learning algorithms (such as convolutional neural network, recurrent neural network, etc.);

Establish a personalized model: Use the selected algorithm model to establish a personalized treatment model based on the individual's EEG signal characteristics and other relevant information. The model can make predictions based on the patient's real-time data and evaluate the patient's sleep quality and insomnia symptoms;

Develop treatment plans: Based on the prediction results of the personalized model, develop targeted treatment plans and adopt appropriate treatment interventions such as sound, light, breathing training, cognitive behavioral therapy, etc. according to the needs and characteristics of the patients to improve the patients' sleep quality and relieve insomnia symptoms;

Real-time monitoring and adjustment: After implementing the treatment intervention plan, continuously monitor the patient's EEG signals and treatment effects, and adjust the treatment plan in a timely manner based on the monitoring results to achieve the best effect.

7. The intelligent insomnia treatment method based on real-time EEG monitoring technology according to claim 6 is characterized in that the support vector machine is a supervised learning model, which separates samples of different categories by constructing one or more hyperplanes. The input of the support vector machine is a set of feature vectors, which can be used to describe various attributes of the samples. The output of the support vector machine is the classification result of the new sample, indicating which category the sample belongs to in the training data. The main function of the support vector machine is to perform binary and multi-classification tasks, and it can be applied to linearly separable, linearly inseparable and nonlinear classification problems;

Random forest is an ensemble learning method that completes classification or regression tasks by building a collection of multiple decision trees. The input of random forest is a set of feature vectors, which are used to build nodes of decision trees. The output of random forest is the result of classification or regression prediction of new samples, which can be discrete category labels or continuous values. Random forest can be used to solve problems such as classification, regression and feature selection. It has the advantages of anti-overfitting ability and high-dimensional data processing ability.

A neural network is a network structure composed of multiple neurons, in which neurons transmit and process information through connection weights. The input of the neural network is a set of feature vectors, which are used as nodes of the input layer to represent the input features of the sample. The output of the neural network is the prediction result of classification, regression or other tasks of new samples, which can be discrete category labels or continuous values. Neural networks can be used for tasks such as pattern recognition, classification, regression, and clustering. It has powerful fitting and nonlinear modeling capabilities.

Convolutional neural network consists of convolutional layer, pooling layer and fully connected layer. The input of convolutional neural network is a two-dimensional image, which can be a grayscale image or a color image. The functions of convolutional neural network are: it has excellent performance in the field of image processing, and can extract local features and spatial structures of images through convolutional layer and pooling layer;

A recurrent neural network is mainly composed of one or more recurrent layers, in which the activation function uses the output of the previous moment as the input of the current moment. The input of the recurrent neural network is suitable for processing sequence data, such as natural language, time series data, etc. The function of the recurrent neural network is: through the memory units of the recurrent layer, it can capture and utilize previous information in the sequence data, so as to predict, classify or generate future outputs.

8. An intelligent insomnia treatment device based on real-time EEG monitoring technology, used to implement the method described in any one of claims 1 to 7, characterized in that it comprises a data acquisition module, a data preprocessing module, a feature extraction module, a sleep staging module, an insomnia identification module, a treatment intervention module, a feedback prompt module, a data storage and management module, and a user interface module;

Both medical staff and patients can operate the system through the user interface module;

The data acquisition module is used to collect electrical signals from different brain regions of the patient, and store and manage the data through the data storage and management module;

The data preprocessing module is used to perform preprocessing operations such as filtering, artifact removal and signal correction on the collected data, and extract useful features from the preprocessed EEG signals through the feature extraction module;

The sleep staging module uses a professional sleep analysis algorithm to classify the patient's sleep atmosphere into different stages according to the characteristics of the computer signal;

The insomnia identification module is used to identify whether the patient has insomnia problem;

The treatment intervention module uses intelligent algorithms to perform personalized treatment intervention on patients;

The feedback prompt module can provide timely feedback and prompts on the patient's sleep status through mobile applications, smart devices or emails.