A method and device for analyzing electroencephalogram during sleep

Abstract

The present invention relates to the technical field of sleep electroencephalogram analysis, and particularly to a sleep electroencephalogram analysis method and device, which comprehensively evaluate the sleep state through double verification of frequency domain analysis and image analysis, and improve the reliability and credibility of the results. The method includes: obtaining the sleep electroencephalogram time-domain information of the user to be analyzed; the sleep electroencephalogram time-domain information includes at least one physiological signal such as electrooculogram, electroencephalogram, electrocardiogram, and electromyogram of the masseter muscle; inputting the sleep electroencephalogram time-domain information into an electroencephalogram time-frequency domain conversion model to obtain a sleep electroencephalogram frequency-domain feature set; using a pre-set sleep activity feature benchmark to classify and statistically analyze the sleep electroencephalogram frequency-domain feature set to obtain a sleep state staging frequency domain analysis result of the user to be analyzed; which includes sleep states in different time periods; inputting the sleep electroencephalogram time-domain information of the user to be analyzed into an electroencephalogram activity image recognition model.

Description

Technical Field

[0001] The present invention relates to the technical field of sleep electroencephalogram analysis, and particularly to a sleep electroencephalogram analysis method and device. Background Art

[0002] As a weak electrical signal spontaneously generated by the human body, electroencephalogram (EEG) is a direct reflection of the activities of brain neurons and is of great significance for understanding brain functions, diagnosing nervous system diseases, and evaluating sleep quality. By amplifying these weak electrical signals through an EEG amplifier and converting them into digital information through A / D conversion technology, we can observe the electroencephalogram (EEG) on a screen, which provides an intuitive analysis tool for doctors and researchers. In current medical practices, most hospitals have equipped with digital EEG examination equipment and regard sleep EEG as one of the routine examination items.

[0003] Most of the existing sleep EEG analysis methods rely on simple fluctuation recognition of the electroencephalogram, mainly focusing on the morphology and amplitude changes of waveforms, while ignoring deeper information such as the frequency domain components and graphical trends of EEG. Although the existing methods can provide some basic sleep information, such as a rough division of sleep stages, they have certain limitations in accuracy and depth.

[0004] Therefore, there is an urgent need for a sleep EEG analysis method to solve the above problems. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a sleep EEG analysis method and device that can more comprehensively evaluate the sleep state through double verification of frequency domain analysis and image analysis, improving the reliability and credibility of the results.

[0006] In a first aspect, the present invention provides a sleep EEG analysis method, which includes:

[0007] Obtain the time-domain information of the sleep EEG of the user to be analyzed;

[0008] Input the time-domain information of the sleep EEG into an EEG time-frequency domain conversion model to obtain a sleep EEG frequency domain feature set;

[0009] Classify and statistically analyze the sleep EEG frequency domain feature set using a pre-set sleep activity feature benchmark to obtain a frequency domain analysis result of the sleep state staging of the user to be analyzed;

[0010] Input the time-domain information of the sleep EEG of the user to be analyzed into an EEG activity image recognition model to obtain an image analysis result of the sleep state staging of the user to be analyzed;

[0011] Compare the frequency-domain analysis results of the sleep state staging with the image analysis results of the sleep state staging, extract the time periods with different analysis results, and send them for manual analysis;

[0012] Summarize the manual analysis results and the analysis results without objections to obtain the sleep state staging results of the user to be analyzed.

[0013] On the other hand, the present application also provides a sleep electroencephalogram analysis device, which includes:

[0014] A sleep electroencephalogram time-domain information acquisition module for acquiring the sleep electroencephalogram time-domain information of the user to be analyzed; the sleep electroencephalogram time-domain information includes at least one physiological signal such as electrooculogram, electroencephalogram, electrocardiogram, and electromyogram of the masseter muscle;

[0015] An electroencephalogram time-frequency domain conversion module for inputting the acquired sleep electroencephalogram time-domain information into an electroencephalogram time-frequency domain conversion model to obtain a sleep electroencephalogram frequency-domain feature set;

[0016] A sleep state frequency-domain analysis module for classifying and statistically analyzing the sleep electroencephalogram frequency-domain feature set by using a pre-set sleep activity feature benchmark to obtain the frequency-domain analysis results of the sleep state staging of the user to be analyzed;

[0017] An electroencephalogram activity image recognition module for inputting the sleep electroencephalogram time-domain information of the user to be analyzed into an electroencephalogram activity image recognition model to obtain the image analysis results of the sleep state staging;

[0018] An analysis result comparison and manual review module for comparing the frequency-domain analysis results of the sleep state staging with the image analysis results of the sleep state staging, extracting the time periods with different analysis results, and sending these time periods to the manual analysis process;

[0019] A result summary module for summarizing the manual analysis results and the analysis results without objections to obtain the final sleep state staging results of the user to be analyzed.

[0020] In a third aspect, the present application provides an electronic device, including a bus, a transceiver, a memory, a processor, and a computer program stored on the memory and executable on the processor. The transceiver, the memory, and the processor are connected through the bus. When the computer program is executed by the processor, the steps in any one of the above methods are implemented.

[0021] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in any one of the above methods are implemented.

[0022] The beneficial effects of the present invention compared with the prior art are as follows: By converting the time-domain signal into frequency-domain features, the changes of different frequency components can be observed more clearly, so as to identify different sleep stages, capture the subtle changes in brain electrical activities, and improve the accuracy of analysis; Using the brain electrical activity image recognition model to perform graphical analysis on the sleep state, providing more intuitive and detailed sleep state information, and further improving the accuracy of analysis; Through frequency-domain analysis, more in-depth information about brain electrical activities can be extracted, such as the change trends of fast frequency components and slow frequency components; Through image recognition technology, complex patterns and structures in the electroencephalogram can be captured, and these patterns and structures are often associated with specific physiological states or pathological conditions; Through frequency-domain analysis and image recognition technology, most of the analysis work can be automated, reducing the workload of manual visual inspection analysis and improving the analysis efficiency; For the time periods with significant differences in the automatic analysis results, manual verification is carried out for correction to ensure the accuracy of the final results; Through the pre-set sleep activity feature benchmarks and classification statistics methods, the standardization and consistency of the analysis process can be ensured, reducing the result inconsistency caused by differences in the technical level and sense of responsibility of analysts; Through the double verification of frequency-domain analysis and image analysis, the sleep state can be evaluated more comprehensively, and the reliability and credibility of the results can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are 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 efforts.

[0024] Figure 1 is a flowchart of a sleep electroencephalogram analysis method provided by an embodiment of the present invention;

[0025] Figure 2 is a hardware architecture diagram of an electronic device provided by an embodiment of the present invention;

[0026] Figure 3 is a structural diagram of a sleep electroencephalogram analysis device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0028] Please refer to Figure 1 , an embodiment of the present invention provides a method for analyzing sleep electroencephalogram, and the method includes:

[0029] Step S1, obtaining the sleep electroencephalogram time-domain information of the user to be analyzed; the sleep electroencephalogram time-domain information includes at least one physiological signal such as electrooculogram, electroencephalogram, electrocardiogram, and electromyogram of the masseter muscle;

[0030] Step S2, inputting the sleep electroencephalogram time-domain information into an electroencephalogram time-frequency domain conversion model to obtain a sleep electroencephalogram frequency-domain feature set;

[0031] Step S3, classifying and statistically analyzing the sleep electroencephalogram frequency-domain feature set by using a pre-set sleep activity feature benchmark to obtain a frequency-domain analysis result of the sleep state staging of the user to be analyzed; it includes the sleep state in different time periods;

[0032] Step S4, inputting the sleep electroencephalogram time-domain information of the user to be analyzed into an electroencephalogram activity image recognition model to obtain an image analysis result of the sleep state staging of the user to be analyzed; it includes the sleep state in different time periods;

[0033] Step S5, comparing the frequency-domain analysis result of the sleep state staging with the image analysis result of the sleep state staging, extracting the time periods with different analysis results, and sending them for manual analysis;

[0034] Step S6, summarizing the manual analysis result and the analysis result without objection to obtain the sleep state staging result of the user to be analyzed.

[0035] In this embodiment, by converting the time-domain signal into frequency-domain features, the changes of different frequency components can be observed more clearly, so as to identify different sleep stages, capture the subtle changes in brain electrical activities, and improve the accuracy of analysis. By using the brain electrical activity image recognition model, the sleep state is analyzed graphically, providing more intuitive and detailed sleep state information, and further improving the accuracy of analysis. Through frequency-domain analysis, more in-depth information about brain electrical activities can be extracted, such as the change trends of fast-frequency components and slow-frequency components. Through image recognition technology, complex patterns and structures in the electroencephalogram can be captured, and these patterns and structures are often associated with specific physiological states or pathological conditions. Through frequency-domain analysis and image recognition technology, most of the analysis work can be automated, reducing the workload of manual visual inspection analysis and improving the analysis efficiency. For the time periods with large differences in the automatic analysis results, manual verification is carried out for correction to ensure the accuracy of the final results. Through the pre-set sleep activity feature benchmarks and classification and statistical methods, the standardization and consistency of the analysis process can be ensured, reducing the result inconsistency caused by differences in the technical levels and responsibilities of analysts. Through the double verification of frequency-domain analysis and image analysis, the sleep state can be evaluated more comprehensively, and the reliability and credibility of the results can be improved.

[0036] The following describes Figure 1 the execution manners of the respective steps shown.

[0037] For step S1:

[0038] In step S1, the time-domain information of the sleep electroencephalogram of the user to be analyzed is obtained; the time-domain information of the sleep electroencephalogram refers to the characteristics of the signal changing with time. For sleep electroencephalogram analysis, the time-domain information of the sleep electroencephalogram includes, but is not limited to, physiological signals such as electrooculogram (EOG), electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram of the masseter muscle (EMG), etc. The specific signals are as follows:

[0039] Electrooculogram (EOG): The EOG records the potential changes generated during eye movement; eye movement is closely related to the sleep stage, especially the rapid eye movement (REM) sleep stage; the REM and non-REM sleep stages can be identified through the EOG.

[0040] Electroencephalogram (EEG): The EEG is the most important sleep electroencephalogram signal, reflecting the electrical activities of the cerebral cortex; the EEG signal shows different characteristic waveforms in different sleep stages, such as δ waves (deep sleep), θ waves (light sleep), and α waves (awake state); the acquisition of the EEG signal usually requires the use of multi-channel electrodes to ensure coverage of different regions of the brain.

[0041] Electrocardiogram (ECG): The ECG records the electrical activity of the heart, which is manifested as a series of heartbeat waveforms; there is a certain correlation between heart activity and sleep quality, and abnormal heart rate variability may indicate sleep disorders; the ECG signal helps to exclude sleep disturbances caused by heart problems.

[0042] Electromyogram of the mandible (EMG): The EMG records the electrical activity of the mandibular muscles, reflecting muscle tension and movement; during sleep, the EMG signal can help distinguish different sleep stages, especially in the REM sleep stage, where muscle tension is significantly reduced; abnormal EMG activity may be related to sleep apnea or other sleep-related diseases.

[0043] The specific data acquisition operations are as follows:

[0044] Step S11: The acquisition device includes an electroencephalogram amplifier, an A / D converter, and a multi-channel recorder; the electroencephalogram amplifier is used to amplify weak electroencephalogram signals to an amplitude suitable for A / D conversion; the A / D converter is used to convert the amplified analog signal into a digital signal for computer processing; the multi-channel recorder is used to synchronously record multiple physiological signals, such as EOG, EEG, ECG, and EMG.

[0045] Step S12: Electroencephalogram (EEG) electrodes are placed at specific positions on the scalp according to the international 10-20 system standard to record the electrical activity of different regions of the brain; electrooculogram (EOG) electrodes are placed above and below the eyes to record eye movements; electrocardiogram (ECG) electrodes are placed on the chest or limbs to record heart electrical activity; electromyogram of the mandible (EMG) electrodes are placed on the mandibular muscles to record muscle activity.

[0046] Step S13: Ensure that all electrodes start recording at the same time point to ensure data synchronization; select an appropriate sampling frequency, and the sampling frequency of the electroencephalogram is set to 256 Hz or higher to capture high-frequency components; store the acquired physiological signals in the computer for subsequent processing and analysis.

[0047] Step S14: Filter the acquired signals to remove noise and interference. Common filtering methods include band-pass filtering, low-pass filtering, and high-pass filtering; perform baseline correction on the signals to eliminate baseline drift; identify and remove artifacts caused by external interference or poor electrode contact.

[0048] Regarding step S2:

[0049] Step S2 is to convert the sleep EEG time-domain information obtained in Step S1 into a sleep EEG frequency-domain feature set; through the EEG time-frequency domain conversion model, the time-domain signal is converted into a frequency-domain signal, so as to extract the EEG signal features of different frequency components for identifying different sleep stages and abnormal events; the specific implementation is as follows:

[0050] Step S21: The sleep EEG time-domain information obtained from Step S1 includes electrooculogram (EOG), electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram of the masseter muscle (EMG); ensure that the input data is a digital signal and is time-series data, and each time point corresponds to a sampling value;

[0051] Step S22: Use the fast Fourier transform (FFT) as the core architecture of the EEG time-frequency domain conversion model. The input of the EEG time-frequency domain conversion model is the sleep EEG time-domain information, and the output is the sleep EEG frequency-domain feature set; the specific conversion is as follows:

[0052] Step S221: Segment the time-domain signal according to a certain time window. For example, a 30-second time window can be selected;

[0053] Step S222: Apply the fast Fourier transform to the data within each time window. The conversion process is as follows:

[0054] “import numpy as np

[0055] from scipy.fft import fft

[0056] # Assume data is a time-series data and fs is the sampling frequency

[0057] fs = 256 # Sampling frequency

[0058] window_length = 30 * fs # 30 seconds of data

[0059] num_windows = len(data) / / window_length

[0060] freq_features = []

[0061] for i in range(num_windows):

[0062] window_data = data[i * window_length:(i + 1) * window_length]

[0063] freq_spectrum = fft(window_data)

[0064] freq_features.append(np.abs(freq_spectrum))

[0065] Step S223: Divide the frequency spectrum into different frequency ranges as needed. For example, define the frequencies above 12 Hz as fast frequency components and the frequencies below 8 Hz as slow frequency components.

[0066] Step S23: Extract the frequency components of interest preset from the frequency spectrum. For example, extract the fast frequency components above 12 Hz and the slow frequency components below 8 Hz; perform statistical analysis on the extracted frequency components to extract useful feature values, such as maximum value, average value, variance, etc.

[0067] Step S24: Save the extracted frequency features as a sleep EEG frequency domain feature set. Each sleep EEG frequency domain feature set contains multiple feature values corresponding to different time windows; ensure that the format of the sleep EEG frequency domain feature set is suitable for subsequent classification statistics and analysis, usually a two-dimensional array, where each row corresponds to a time window and each column corresponds to a feature value.

[0068] By converting the sleep EEG time domain information into a frequency domain feature set in Step S2, the features of different frequency components are extracted, providing rich information for subsequent sleep state staging; it can not only capture the subtle changes in EEG activities, but also improve the accuracy and depth of analysis, providing strong support for clinical diagnosis and research.

[0069] Regarding Step S3:

[0070] Step S3 uses a pre-set sleep activity feature benchmark to perform classification statistics on the sleep EEG frequency domain feature set obtained in Step S2, thereby obtaining the frequency domain analysis result of the sleep state staging of the user to be analyzed; the frequency domain analysis result of the sleep state staging includes the sleep states in different time periods, such as wakefulness period, light sleep period, deep sleep period, and REM period; the method for setting the sleep activity feature benchmark includes:

[0071] Step S31: Obtain the personal basic information of the user to be analyzed and the basic configuration information of the monitoring device; considering the influence on the sleep pattern and electroencephalogram, collect personal basic information including age, gender, weight, height, health status, drug use situation, etc.; the basic configuration information of the monitoring device involves the model, sampling rate, filter settings, etc. of the EEG monitoring device used; the device configuration directly affects the quality of the data and the accuracy of the analysis result.

[0072] Step S32: Extract the sleep impact feature information set from the personal basic information. Analyze the user's personal basic information to extract features that may affect sleep, including changes in sleep requirements related to age, different sleep patterns caused by gender differences, and the direct impact of health status on sleep quality, etc. Through feature extraction, a set containing multiple sleep impact features is obtained, namely the sleep impact feature information set.

[0073] Step S33: Input the sleep impact feature information set into the sleep impact comprehensive analysis model to obtain the sleep impact comprehensive index. The sleep impact comprehensive analysis model uses machine learning or deep learning algorithms, which can process multi-source data and output the sleep impact comprehensive index. The sleep impact comprehensive index reflects the overall impact of the user's personal characteristics on their sleep pattern.

[0074] Step S34: Extract the data acquisition quality impact features from the basic configuration information of the monitoring device to obtain the heterogeneous vector of monitoring device impact features. The heterogeneous vector of monitoring device impact features includes the following:

[0075] Device model feature: Different models of devices may have different performances and precisions.

[0076] Electrode position feature: The impact of the position and number of electrodes on the signal quality.

[0077] Sampling frequency feature: The impact of the sampling frequency on the signal resolution.

[0078] Data transmission mode feature: The stability difference between wired transmission and wireless transmission.

[0079] Calibration information feature: The accuracy and timeliness of device calibration.

[0080] Step S35: Conduct error-related analysis based on the heterogeneous vector of monitoring device impact features to obtain the device impact error factor. Through error-related analysis, quantify the impact of the monitoring device on data acquisition quality to obtain the device impact error factor. Specifically, establish an error model to analyze the impact of different features on data acquisition errors. Use statistical methods (such as regression analysis) to calculate the error factor of each feature, which is used to correct the device errors in data acquisition.

[0081] Step S36: Correct the sleep impact comprehensive index based on the device impact error factor to obtain the sleep impact correction index. Use the device impact error factor to adjust the sleep impact comprehensive index to compensate for the impact of device errors on the sleep analysis results. The corrected index can more accurately reflect the user's sleep characteristics and reduce the interference of device errors on the analysis results.

[0082] Step S37: Using the sleep impact correction index as an indexing condition, traverse the pre-established sleep activity benchmark database to extract the sleep activity feature benchmark corresponding to the sleep impact correction index; the sleep activity feature benchmark includes the EEG frequency domain benchmarks corresponding to sleep activities in different stages; the sleep activity benchmark database contains a large number of sleep activity feature benchmarks of different people under different conditions (i.e., sleep impact correction index). By matching the sleep impact correction index, the sleep activity feature benchmark closest to the user's current situation can be found, providing an accurate reference for the subsequent frequency domain analysis of sleep state staging.

[0083] In this step, by collecting the personal basic information of the user to be analyzed (such as age, gender, health status, etc.), the influence of individual differences on sleep patterns can be considered, realizing personalized sleep state analysis; Steps S34 to S36 detail how to analyze the basic configuration information of the monitoring device, extract the features affecting data acquisition quality, and obtain the device influence error factor through error correlation analysis; ensuring that relatively accurate data can be obtained even when using devices of different models and configurations, improving the accuracy of analysis; by inputting the personal basic information and monitoring device information into the sleep impact comprehensive analysis model, a comprehensive index can be output to reflect the overall impact of the user's personal characteristics on their sleep pattern. This comprehensive evaluation method can more comprehensively reflect the user's sleep status than a single indicator; using the pre-established sleep activity benchmark database for traversal to extract the sleep activity feature benchmark corresponding to the sleep impact correction index; the database-based matching method can quickly find the sleep activity feature benchmark closest to the user's current situation, providing an accurate reference for subsequent analysis; in summary, by comprehensively considering personal characteristics, device errors, and sleep activity feature benchmarks, the uncertainty and errors in the analysis process can be reduced, improving the reliability and accuracy of the analysis; it can adapt to the sleep analysis needs of different users, different devices, and different conditions, with high flexibility and versatility.

[0084] Regarding step S4:

[0085] Step S4 is to input the sleep EEG time domain information obtained in step S1 into the EEG activity image recognition model, and analyze the EEG activity through image recognition technology to obtain the sleep state staging image analysis result of the user to be analyzed; the sleep state staging image analysis result includes the sleep states in different time periods, such as the wake period, light sleep period, deep sleep period, and REM period; the specific implementation is as follows:

[0086] Step S41: Use the time-domain information of the sleep electroencephalogram of the user to be analyzed as input data. The time-domain information of the sleep electroencephalogram includes one or more physiological signals such as electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), and electromyogram of the masseter muscle (EMG). The time-domain signal of the sleep electroencephalogram will present specific patterns and changes during sleep.

[0087] Step S42: Input the time-domain information of the sleep electroencephalogram into a pre-trained electroencephalogram activity image recognition model. The electroencephalogram activity image recognition model is constructed based on deep learning or machine learning techniques and can automatically extract and identify key features in the sleep electroencephalogram signal. The key features include the morphology of the waveform, amplitude changes, and dynamic features in the time series, etc.

[0088] Step S43: Inside the electroencephalogram activity image recognition model, the above-mentioned extracted key features will be further processed and transformed into image forms, including but not limited to spectrograms, waveform diagrams, trend diagrams, etc., which can intuitively display the changes in the sleep electroencephalogram signal at different time periods.

[0089] Step S44: Based on the above-transformed images, the electroencephalogram activity image recognition model will further perform stage analysis of the sleep state. According to the preset sleep state classification criteria (such as the light sleep stage, deep sleep stage of non-rapid eye movement period NREM, and rapid eye movement period REM, etc.), the sleep state in each time period will be classified, and the corresponding stage analysis results of the sleep state image will be generated. The stage analysis results of the sleep state image include sleep state labels and corresponding confidence levels or probability values in different time periods.

[0090] In this step, using the electroencephalogram activity image recognition model to deeply analyze the time-domain information of the sleep electroencephalogram can automatically extract key features and transform them into intuitive image forms, improving the accuracy and reliability of sleep state staging. Through the preset sleep state classification criteria, the model can classify the sleep states in different time periods and generate image analysis results including confidence levels or probability values, which not only reduces the dependence on manual analysis but also improves the analysis efficiency and accuracy, providing more detailed and intuitive sleep state staging information for doctors and researchers.

[0091] Specifically, the electroencephalogram activity image recognition model is a model constructed based on deep learning or machine learning techniques, used to analyze the time-domain information of the sleep electroencephalogram, analyze the electroencephalogram activity through image recognition technology, so as to obtain the stage analysis results of the sleep state image of the user to be analyzed. The model can automatically extract and identify key features in the sleep electroencephalogram signal, transform them into image forms, and further perform stage analysis of the sleep state. The model architecture is as follows:

[0092] Data Input Layer: Time-domain information of sleep electroencephalogram, including physiological signals such as electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), and electromyogram of the masseter muscle (EMG); normalize, denoise, and adjust the size of the input data to ensure that the data format meets the model requirements;

[0093] Feature Extraction Layer: Extract key features of time-domain signals, such as the morphology of waveforms, amplitude changes, and dynamic features in the time series; convert the extracted features into image forms, such as waveform diagrams, trend charts, etc., which can intuitively display the changes in sleep electroencephalogram signals;

[0094] Deep Learning Model: Use a convolutional neural network model to automatically extract high-level features in the image through multiple layers of convolution, pooling, and activation functions; use pre-trained convolutional neural network models (such as ResNet, VGG, etc.) and fine-tune them on this basis to adapt to specific electroencephalogram image recognition tasks; map the extracted features to classification labels and output the sleep state classification results for each time period;

[0095] Classification Layer: According to the preset sleep state classification criteria (such as light sleep, deep sleep, and rapid eye movement sleep REM in non-rapid eye movement NREM), classify the sleep states for each time period; generate the analysis results of sleep state staging images, including sleep state labels and corresponding confidence levels or probability values for different time periods.

[0096] The electroencephalogram activity image recognition model can automatically extract and recognize key features in sleep electroencephalogram signals by combining deep learning technology and image recognition technology, generate intuitive image forms, and perform staging analysis of sleep states; this model not only improves the accuracy and depth of analysis, but also provides strong support for clinical diagnosis and research.

[0097] Regarding step S5:

[0098] Step S5 compares the sleep state staging results based on frequency-domain analysis (obtained in step S3) with the sleep state staging results based on image comprehensive processing (obtained in step S4) to identify the time periods with differences between the two and perform further manual analysis on these time periods; the aim is to improve the accuracy and reliability of the analysis and ensure that the final sleep state staging results are more accurate; the specific operations are as follows:

[0099] Step S51: Compare the frequency-domain analysis results obtained in step S3 with the image analysis results obtained in step S4, and compare the sleep states (such as awake, light sleep, deep sleep, REM sleep, etc.) of each time period in the two results one by one;

[0100] Step S52: Each time period has a clear start time and end time, as well as corresponding sleep state labels and confidence levels; align the time periods of the frequency-domain analysis results and the image analysis results to ensure that corresponding records can be found for each time period in both results; compare the sleep state labels (such as awake, light sleep, deep sleep, REM sleep, etc.) within each time period to check if they are consistent; during the comparison process, if it is found that the sleep state labels of a certain time period are inconsistent in the two analysis results, mark that time period as a "discrepancy time period".

[0101] Step S53: For the identified discrepancy time periods, they need to be sent to professionals for manual analysis; the manual analysis includes a detailed review of the original sleep EEG time-domain information and a comprehensive consideration of relevant physiological signals (such as electrooculogram, electrocardiogram, submandibular electromyogram, etc.); professionals will make an accurate judgment on the sleep state within the discrepancy time periods based on their professional knowledge and experience and give the final analysis results.

[0102] Step S5 effectively improves the accuracy and reliability of sleep EEG analysis by introducing a manual review mechanism; ensures the accuracy of the automated analysis results and reduces misjudgments caused by model limitations or errors; at the same time, it also provides an opportunity for professionals to directly participate in the analysis process and enables them to verify and correct the automated analysis results using their professional knowledge and experience.

[0103] Regarding Step S6:

[0104] Step S6 is to summarize the results of the manual analysis in Step S5 and the non-disputed automated analysis results to generate the final sleep state staging results; the specific operations are as follows:

[0105] Step S61: Initially integrate the sleep state staging results obtained through frequency-domain analysis in Step S3 and the sleep state staging results obtained through image recognition technology in Step S4; at the same time, collect the results of the manual analysis in Step S5, especially the analysis results of the time periods with differences in the automated analysis stage.

[0106] Step S62: For the time periods without disputes in the automated analysis stage, directly confirm the accuracy of their analysis results; for the time periods with differences, verify and adjust them with reference to the results of the manual analysis to ensure the reliability of the final results; when necessary, combine the professional knowledge and clinical experience of doctors to conduct further comprehensive judgment and analysis on the aggregated data, which helps to solve possible complex situations or controversial points and improve the overall analysis accuracy.

[0107] Step S63: Integrate all the verified and adjusted sleep stage classification results into a complete report. The report should detail the sleep state and its confidence or probability value for each time period, and mark any key points with manual intervention.

[0108] Step S64: Feed back the generated final report to relevant doctors or researchers as an important basis for diagnosis and treatment decisions, and it can also be used for subsequent research and improving existing algorithms and models.

[0109] Step S6 combines automated analysis with manual analysis to fully utilize the advantages of both, improving the accuracy and reliability of sleep EEG analysis. At the same time, Step S6 can also provide clear and coherent sleep stage classification results, providing valuable information for doctors and researchers, helping to better understand the patient's sleep pattern and quality, and thus making more accurate diagnosis and treatment decisions.

[0110] As Figure 2 、 Figure 3 shown, an embodiment of the present invention provides a sleep EEG analysis device. The device embodiment can be implemented by software, or by hardware or a combination of software and hardware. At the hardware level, as Figure 2 shown, it is a hardware architecture diagram of an electronic device where the sleep EEG analysis device provided by the embodiment of the present invention is located. In addition to Figure 2 the shown processor, memory, network interface, and non-volatile memory, the electronic device where the device is located in the embodiment usually may also include other hardware, such as a forwarding chip responsible for processing packets, etc. Taking software implementation as an example, as Figure 3 shown, as a logically meaningful device, it is formed by the CPU of its corresponding electronic device reading the computer program in the non-volatile memory into the memory and running.

[0111] As Figure 3 shown, a sleep EEG analysis device provided in this embodiment includes:

[0112] A sleep EEG time-domain information acquisition module, used to acquire the sleep EEG time-domain information of the user to be analyzed. The sleep EEG time-domain information includes various physiological signals such as electrooculogram, electroencephalogram, electrocardiogram, and electromyogram of the masseter muscle, providing basic data for subsequent analysis.

[0113] An EEG time-frequency domain conversion module, used to input the acquired sleep EEG time-domain information into an EEG time-frequency domain conversion model to obtain a sleep EEG frequency-domain feature set. This module converts the time-domain signal into frequency-domain features through time-frequency domain conversion technology, facilitating subsequent in-depth analysis of the sleep state.

[0114] The sleep state frequency domain analysis module is used to classify and statistically analyze the sleep EEG frequency domain feature set by using a pre-set benchmark of sleep activity characteristics, so as to obtain the frequency domain analysis result of the sleep state staging of the user to be analyzed; this module can identify and classify the sleep states in different time periods, such as non-rapid eye movement (NREM) and rapid eye movement (REM), etc.;

[0115] The EEG activity image recognition module is used to input the sleep EEG time domain information of the user to be analyzed into the EEG activity image recognition model to obtain the image analysis result of the sleep state staging; this module performs image processing on physiological signals such as electroencephalograms through image recognition technology and identifies the characteristic patterns therein to achieve the image analysis of the sleep state;

[0116] The analysis result comparison and manual review module is used to compare the frequency domain analysis result of the sleep state staging with the image analysis result of the sleep state staging, extract the time periods with different analysis results, and send these time periods to the manual analysis link; this module can automatically detect and identify the inconsistencies in the analysis results to ensure the accuracy and reliability of the analysis results; at the same time, for the controversial analysis results, this module can also provide means for manual review to ensure the accuracy of the final analysis results;

[0117] The result summary module is used to summarize the manual analysis results and the analysis results without objections to obtain the final sleep state staging result of the user to be analyzed; this module can integrate all the analysis results to generate a complete sleep state staging report, providing intuitive and comprehensive analysis data for doctors and researchers.

[0118] It can be understood that the structure schematically shown in the embodiments of the present invention does not constitute a specific limitation on a sleep EEG analysis device. In other embodiments of the present invention, a sleep EEG analysis device may include more or fewer components than shown in the figure, or combine certain components, or split certain components, or have different component arrangements. The components shown in the figure can be implemented in hardware, software, or a combination of software and hardware.

[0119] For the information interaction, execution process, etc. between the modules in the above device, since they are based on the same concept as the method embodiments of the present invention, the specific content can be referred to the description in the method embodiments of the present invention and will not be elaborated here.

[0120] The embodiments of the present invention also provide an electronic device, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, a sleep EEG analysis method in any embodiment of the present invention is implemented.

[0121] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the processor is caused to execute a sleep EEG analysis method according to any one of the embodiments of the present invention.

[0122] Specifically, a system or device equipped with a storage medium can be provided, on which software program code for implementing the functions of any one of the above embodiments is stored, and the computer (or CPU or MPU) of the system or device is caused to read and execute the program code stored in the storage medium.

[0123] In this case, the program code read from the storage medium itself can implement the functions of any one of the above embodiments, so the program code and the storage medium storing the program code constitute a part of the present invention.

[0124] Examples of storage media for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Optionally, the program code can be downloaded from a server computer via a communication network.

[0125] In addition, it should be clear that not only can the functions of any one of the above embodiments be realized by executing the program code read by the computer, but also by causing an operating system or the like operating on the computer based on the instructions of the program code to complete part or all of the actual operations.

[0126] In addition, it can be understood that the program code read from the storage medium is written into the memory provided in the expansion board inserted into the computer or into the memory provided in the expansion module connected to the computer, and then based on the instructions of the program code, the CPU or the like installed on the expansion board or expansion module is caused to execute part and all of the actual operations, thereby realizing the functions of any one of the above embodiments.

[0127] It should be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0128] Those of ordinary skill in the art will understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including those of the above method embodiments. The aforementioned storage medium includes various media that can store program codes, such as ROM, RAM, magnetic disks, or optical discs.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sleep EEG analysis method, characterized in that: The method comprises: Obtain the sleep EEG time domain information of the user to be analyzed; Inputting the sleep EEG time domain information into an EEG time-frequency domain conversion model to obtain a sleep EEG frequency domain feature set; Using the pre-set sleep activity feature benchmark, the sleep EEG frequency domain feature set is classified and counted to obtain the sleep state stage frequency domain analysis results of the user to be analyzed; Input the sleep EEG time domain information of the user to be analyzed into the EEG activity image recognition model to obtain the sleep state stage image analysis result of the user to be analyzed; Comparing the sleep state staging frequency domain analysis result with the sleep state staging image analysis result, extracting different time periods of the analysis results, and sending them to manual analysis; The manual analysis results and the undisputed analysis results are aggregated to obtain the sleep state staging results of the user to be analyzed; The sleep EEG time domain information includes at least one physiological signal including the electroencephalogram; The method for setting the sleep activity characteristic benchmark includes: Obtain basic personal information of the user to be analyzed and basic configuration information of the monitoring device; Extracting sleep impact features from the basic personal information to obtain a sleep impact feature information set; Inputting the sleep impact feature information set into a sleep impact comprehensive analysis model to obtain a sleep impact comprehensive index; Extracting features influencing data collection quality from the basic configuration information of the monitoring equipment to obtain heterogeneous vectors of monitoring equipment influencing features; Perform error correlation analysis based on the monitoring equipment influencing feature heterogeneous vectors to obtain equipment influencing error factors; Correcting the sleep impact comprehensive index based on the device impact error factor to obtain a sleep impact correction index; The sleep impact correction index is used as an index condition, and a pre-built sleep activity benchmark database is traversed to extract a sleep activity feature benchmark corresponding to the sleep impact correction index; the sleep activity feature benchmark includes an EEG frequency domain benchmark corresponding to sleep activities at different stages.

2. A sleep EEG analysis method as claimed in claim 1, characterized in that: The operation method of the EEG time-frequency domain conversion model is as follows: Segmenting the sleep EEG time domain signal according to a set time window; Applying fast Fourier transform to the sleep EEG time domain signal in each time window to convert the EEG signal from time domain representation to frequency domain representation; Based on the preset frequency classification requirements, the spectrum is divided into fast frequency components and slow frequency components; Extracting preset frequency components of interest from the spectrum, and performing statistical analysis on the extracted frequency components to extract characteristic values ​​of the frequency components; The extracted frequency features are saved as sleep EEG frequency domain feature sets, each of which contains multiple feature values ​​corresponding to different time windows.

3. A sleep EEG analysis method as claimed in claim 1, characterized in that: The sleep impact characteristic information set includes age characteristics, gender characteristics and health condition characteristics.

4. A sleep EEG analysis method as claimed in claim 3, characterized in that: The monitoring equipment influencing feature heterogeneous vector includes equipment model features, electrode position features, sampling frequency features, data transmission mode features and calibration information features.

5. A sleep EEG analysis method as claimed in claim 4, characterized in that: The structural framework of the EEG image recognition model includes a data input layer, a feature extraction layer, a deep learning model and a classification layer; Among them, the input data of the data input layer is the sleep EEG time domain information; The feature extraction layer is used to extract the key features of the sleep EEG time domain information, including the waveform shape, amplitude change and dynamic characteristics in the time series; and convert the extracted features into image form; The deep learning model uses a convolutional neural network model to automatically extract high-level features in the image form through multiple layers of convolution, pooling and activation functions; maps the extracted high-level features to classification labels, and outputs the sleep state classification results in each time period; The classification layer classifies the sleep state in each time period according to the preset sleep state classification standard and generates the sleep state stage image analysis results.

6. A sleep EEG analysis device, characterized in that: A sleep EEG analysis method according to any one of claims 1 to 4, wherein the device comprises: A sleep EEG time domain information acquisition module is used to obtain the sleep EEG time domain information of the user to be analyzed; the sleep EEG time domain information includes at least one physiological signal including an electroencephalogram; The EEG time-frequency domain conversion module is used to input the collected sleep EEG time-domain information into the EEG time-frequency domain conversion model to obtain a sleep EEG frequency domain feature set; The sleep state frequency domain analysis module is used to classify and count the sleep EEG frequency domain feature set using a preset sleep activity feature benchmark to obtain the sleep state stage frequency domain analysis result of the user to be analyzed; The EEG activity image recognition module is used to input the sleep EEG time domain information of the user to be analyzed into the EEG activity image recognition model to obtain the sleep state stage image analysis results; The analysis result comparison and manual review module is used to compare the sleep state staging frequency domain analysis results with the sleep state staging image analysis results, extract different time periods of the analysis results, and send these time periods to the manual analysis link; The result summary module is used to summarize the manual analysis results and the undisputed analysis results to obtain the final sleep state staging results of the user to be analyzed.

7. An electronic device for analyzing sleep EEG, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and characterized in that: When the computer program is executed by the processor, the steps in the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 5 are implemented.

Images