Sleep monitoring method and system based on combination of electroencephalogram and millimeter wave radar

Abstract

The invention relates to the technical field of sleep monitoring, in particular to a sleep monitoring method and system based on combination of electroencephalogram and millimeter wave radar, and aims to solve the problem that OSA judgment accuracy is insufficient due to complication. The method comprises the following steps: acquiring an electroencephalogram signal of a user based on a wireless EEG sensor, and acquiring a physiological signal of the user by using a millimeter wave radar; performing sleep staging analysis on the electroencephalogram signal, and detecting a micro-awakening event in combination with body movement data; the physiological signals are processed through a physiological signal processing model, physiological signal features of the user are extracted, and an apnea event, a hypopnea event, an arrhythmia event, a blood pressure fluctuation abnormal event and a periodic leg movement event of the user are monitored; calculating each OSA judgment index according to the event; and calculating OSA personalized scores according to the OSA judgment indexes, performing OSA severity grading, and generating an OSA severity grading report for assisting medical staff in diagnosis and treatment scheme formulation.

Description

Technical Field

[0001] The present invention relates to the technical field of sleep monitoring, and particularly to a sleep monitoring method and system based on the combination of electroencephalogram and millimeter-wave radar. Background Art

[0002] Obstructive sleep apnea (OSA) is a common sleep breathing disorder, which refers to the partial or complete obstruction of the upper airway during sleep, resulting in apnea or hypopnea, usually accompanied by a decrease in oxygenation. The occurrence of OSA not only affects sleep quality, but may also lead to a series of serious health problems, including hypertension, cardiovascular diseases, diabetes, stroke, and daytime sleepiness. OSA patients often have physical and psychological burdens due to frequent awakenings and decreased oxygen saturation, increasing the risk of accidents. Although the harm of OSA causes huge social and economic losses, it has not been fully recognized and treated at present. Therefore, timely and accurate diagnosis of OSA is crucial for preventing its harm and effective treatment.

[0003] Currently, the "gold standard" for diagnosing OSA is polysomnography, which can accurately identify the occurrence of apnea events by comprehensively monitoring multiple physiological indicators such as electroencephalogram, respiratory flow, oxygen saturation, and electrocardiogram. However, this method has some limitations. The diagnosis of PSG often relies on the apnea-hypopnea index (AHI) and oxygenation fluctuations, but these indicators do not fully consider the impact of comorbidities (such as cardiovascular diseases, periodic limb movement disorder, etc.) on OSA. Therefore, there is an urgent need for a more personalized sleep monitoring technology and analysis method to assist medical staff in improving the differential diagnosis efficiency of OSA in patients with comorbidities and providing an objective basis for formulating individualized treatment plans.

[0004] Therefore, a sleep monitoring method and system based on the combination of electroencephalogram and millimeter-wave radar are proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a sleep monitoring method and system based on the combination of electroencephalogram and millimeter-wave radar to overcome the problem of insufficient accuracy in OSA judgment due to comorbidities. It includes: obtaining the electroencephalogram signal of the user during the sleep stage based on a wireless EEG sensor, performing sleep stage analysis on the electroencephalogram signal, and detecting microarousal events; collecting the physiological signals of the user during the sleep stage using a millimeter-wave radar; respectively processing the physiological signals through a physiological signal processing model, extracting the physiological signal characteristics of the user, and monitoring the user's apnea events, hypopnea events, arrhythmia events, abnormal blood pressure fluctuation events, and periodic limb movement events; calculating various OSA determination indicators based on the above events; calculating an OSA personalized score based on the OSA determination indicators, performing OSA severity grading, and generating an OSA severity grading report for assisting medical staff in diagnosis and treatment plan formulation.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A sleep monitoring method based on the combination of electroencephalogram (EEG) and millimeter-wave radar, comprising:

[0008] Acquiring the EEG signals of a user during the sleep stage based on a wireless EEG sensor, performing sleep stage analysis on the EEG signals, and detecting micro-arousal events;

[0009] Collecting the physiological signals of the user during the sleep stage by using a millimeter-wave radar; the physiological signals include: respiratory signals, heart rate signals, and body movement data; respectively processing the physiological signals through a physiological signal processing model, extracting the physiological signal features of the user, and monitoring the user's apnea events, hypopnea events, arrhythmia events, abnormal blood pressure fluctuation events, and periodic leg movement events according to the physiological signal features;

[0010] Calculating various OSA determination indexes based on the micro-arousal events, the apnea events, the hypopnea events, the arrhythmia events, the abnormal blood pressure fluctuation events, and the periodic leg movement events;

[0011] Calculating an OSA personalized score according to the OSA determination indexes, and performing OSA severity grading according to the OSA personalized score, generating an OSA severity grading report for assisting medical staff in diagnosis and treatment plan formulation.

[0012] Preferably, the wireless EEG sensor is designed at the positions of occipital lobes O1 and O2 on the upper rear part of the pillow; the millimeter-wave radar is placed near the head of the test subject's bed.

[0013] Preferably, the specific process of performing sleep stage analysis on the EEG signals and detecting micro-arousal events in combination with the body movement data is: decomposing the EEG signals through wavelet transform to obtain the time-domain features and frequency-domain features of the EEG signals, and using a machine learning model to stage the sleep stage based on the time-domain features and the frequency-domain features, classifying and identifying the wakefulness stage, light sleep stage, deep sleep stage, and rapid eye movement sleep stage; analyzing the suddenly accelerated body movements by using the mutation point detection method according to the sleep stage in combination with the body movement data to detect micro-arousal events.

[0014] Preferably, the physiological signal processing model includes: a respiratory signal processing unit, a heart rate signal processing unit, and a body movement data processing unit;

[0015] The respiratory signal processing unit includes: a respiratory feature extraction layer and an apnea and hypopnea event recognition layer;

[0016] The respiratory feature extraction layer extracts the change in respiratory flow in the respiratory signal through signal filtering and denoising processing to obtain the respiratory feature;

[0017] The apnea and hypopnea event recognition layer determines the apnea-hypopnea duration and apnea-hypopnea frequency of the apnea event and / or the hypopnea event according to the respiratory feature;

[0018] The heart rate signal processing unit includes: a heart rate feature extraction layer, an arrhythmia event recognition layer, and a blood pressure fluctuation abnormality event recognition layer;

[0019] The heart rate feature extraction layer performs multi-scale decomposition on the heart rate signal through the MODWT algorithm, extracts the heart rate of the heart rate signal, and obtains the heart rate feature;

[0020] The arrhythmia event recognition layer analyzes the heart rate variability of the user through the heart rate feature to identify the arrhythmia event;

[0021] The blood pressure fluctuation abnormality event recognition layer predicts the blood pressure of the user through machine learning according to the heart rate feature and the heart rate variability, and identifies the blood pressure fluctuation abnormality event;

[0022] The body movement data processing unit includes: a body movement feature extraction layer and a periodic leg movement event recognition layer;

[0023] The body movement feature extraction layer extracts the periodic leg movement feature by using the body movement data, and the feature includes: body movement frequency and body movement amplitude;

[0024] The periodic leg movement event recognition layer identifies the periodic leg movement event according to the periodic leg movement feature and records the leg movement times and leg movement duration.

[0025] Preferably, the OSA determination indexes include: microarousal index, apnea-hypopnea index, sympathetic nerve activation degree, blood pressure fluctuation condition, and periodic leg movement index;

[0026] The microarousal index is calculated by using the number of microarousal events and the total sleep duration;

[0027] The apnea-hypopnea index is the total number of the apnea event and the hypopnea event occurring per hour, and is calculated according to the apnea-hypopnea duration and apnea-hypopnea frequency;

[0028] The sympathetic nerve activation degree is quantified according to the parameter reflecting the heart rate variability in the heart rate signal;

[0029] The blood pressure fluctuation condition is calculated based on the statistical standard deviation of blood pressure;

[0030] The periodic leg movement index is calculated using the number of periodic leg movement events and the total sleep duration.

[0031] Preferably, the calculation formula for the OSA personalized score is:

[0032] OSA_Score = ω 1 ·AHI + ω 2 ·ArI + ω 3 ·SNA + ω 4 ·σ BP + ω 5 ·PLMI;

[0033] Wherein, OSA_Score is the OSA personalized score; ω 1 is the weight of the apnea hypopnea index; AHI is the apnea hypopnea index; ω 2 is the weight of the microarousal index; ArI is the microarousal index; ω 3 is the weight of the degree of sympathetic nerve activation; SNA is the degree of sympathetic nerve activation; ω 4 is the weight of the blood pressure fluctuation condition; σ BP is the blood pressure fluctuation condition; ω 5 is the weight of the periodic leg movement index; PLMI is the periodic leg movement index.

[0034] Preferably, a sleep monitoring system based on the combination of electroencephalogram and millimeter wave radar includes:

[0035] A signal acquisition module, configured to obtain the electroencephalogram signal of the user during the sleep stage based on a wireless EEG sensor, and collect the physiological signals of the user during the sleep stage by using a millimeter wave radar; the physiological signals include: respiratory signals, heart rate signals, and body movement data;

[0036] A microarousal event detection module, configured to perform sleep stage analysis on the electroencephalogram signal, and detect microarousal events in combination with the body movement data;

[0037] A physiological signal processing module, configured to respectively process the physiological signals through a physiological signal processing model, extract the physiological signal features of the user, and monitor the apnea event, hypopnea event, arrhythmia event, and periodic leg movement event of the user according to the physiological signal features;

[0038] An OSA determination index calculation module, configured to calculate each OSA determination index based on the microarousal event, the apnea event, the hypopnea event, the arrhythmia event, and the periodic leg movement event;

[0039] The OSA personalized scoring and severity grading module is used to calculate the OSA personalized score according to the OSA determination index, and perform OSA severity grading according to the OSA personalized score, generating an OSA severity grading report to assist medical staff in diagnosis and treatment plan formulation.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] 1. By integrating signals collected by a wireless EEG sensor and a millimeter-wave radar, the present invention can accurately detect sleep stages and micro-arousal events without disturbing the user's natural sleep. At the same time, the collected respiration, heart rate, and body movement data provide a basis for comprehensively evaluating the user's physiological state, significantly improving the accuracy and reliability of sleep monitoring, and providing reliable input data for subsequent improving the diagnostic accuracy of OSA by integrating multiple comorbidity indicators.

[0042] 2. The present invention proposes a physiological signal processing model to process multiple physiological signals such as respiration, heart rate, and body movement respectively, and extract key features, which can accurately monitor abnormal events that may occur during the user's sleep, such as apnea, hypopnea, arrhythmia, abnormal blood pressure fluctuations, and periodic leg movement events. This multi-parameter comprehensive monitoring not only improves the accuracy and sensitivity of abnormal event detection, but also provides objective data support for clinical practice, helps to detect sleep disorders early and formulate personalized treatment plans, thereby reducing the health risks caused by these comorbidities, and providing comorbidity data reference for subsequent personalized determination of OSA severity.

[0043] 3. The present invention proposes a method for determining the severity of OSA based on multiple indicators such as the micro-arousal index, apnea-hypopnea index, degree of sympathetic nerve activation, blood pressure fluctuation, and periodic leg movement index, and dynamically adjusts the weights of these indicators for different users to achieve precise grading of the severity of OSA in different populations. By integrating multiple comorbidity indicators and dynamically adjusting these indicators, this method overcomes the problem of insufficient accuracy in OSA judgment due to comorbidities, improves the accuracy of OSA diagnosis, provides an objective basis for doctors' subsequent clinical diagnosis, and helps to formulate targeted diagnostic and treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flowchart of a sleep monitoring method based on the combination of electroencephalogram and millimeter-wave radar provided by an embodiment of the present invention;

[0045] Figure 2 It is a structural diagram of a sleep monitoring system based on the combination of electroencephalogram and millimeter-wave radar provided by an embodiment of the present invention;

[0046] Figure 3 This is the working principle diagram of the physiological signal processing model provided by the embodiments of the present invention;

[0047] Figure 4 This is the structural diagram of the OSA personalized scoring provided by the embodiments of the present invention. Detailed implementation manners

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0049] Obstructive sleep apnea (OSA) is a common sleep breathing disorder, which refers to the partial or complete obstruction of the upper airway during sleep, resulting in apnea or hypopnea, usually accompanied by a decrease in oxygenation. The occurrence of OSA not only affects sleep quality but may also lead to a series of serious health problems. Currently, the "gold standard" for diagnosing OSA is polysomnography technology. However, the diagnosis of PSG often relies on the apnea-hypopnea index (AHI) and oxygenation fluctuations, but these indicators do not fully consider the impact of comorbidities (such as cardiovascular diseases, periodic limb movement disorder, etc.) on OSA.

[0050] The present invention proposes a sleep monitoring method and system based on the combination of electroencephalogram and millimeter-wave radar, which realizes the customization of personalized OSA scores according to the physiological conditions of different users, thereby improving the judgment accuracy of OSA. In order to illustrate that the method of the present invention can play a role in customizing personalized OSA scores according to the physiological conditions of different users and thus improving the judgment accuracy of OSA, the effectiveness of the present invention will be described from two embodiments below.

[0051] Embodiment 1

[0052] In the embodiments of the present application, the process of customizing personalized OSA scores according to the physiological conditions of different users and thus improving the judgment accuracy of OSA by using the method proposed by the present invention is described in detail. The embodiments of the present application are directed to the OSA diagnosis of 150 patients of different ages in the Class III Grade A Hospital A. The following will be based on Figure 1 the content to describe the OSA diagnosis process of the patients in the Class III Grade A Hospital A in detail; among them, Figure 1The specific flowchart of the method proposed by the present invention includes: acquiring the electroencephalogram (EEG) signals of the user during the sleep stage based on a wireless EEG sensor, and collecting the physiological signals of the user during the sleep stage using a millimeter-wave radar; performing sleep staging analysis on the EEG signals, and detecting micro-arousal events in combination with body movement data; collecting the physiological signals of the user during the sleep stage using a millimeter-wave radar; respectively processing the physiological signals through a physiological signal processing model, extracting the physiological signal features of the user, and monitoring the apnea event, hypopnea event, arrhythmia event, abnormal blood pressure fluctuation event, and periodic leg movement event of the user based on the physiological signal features; calculating various OSA determination indexes based on the micro-arousal event, apnea event, hypopnea event, arrhythmia event, abnormal blood pressure fluctuation event, and periodic leg movement event; calculating the OSA personalized score according to the OSA determination indexes, and performing OSA severity grading according to the OSA personalized score, generating an OSA severity grading report for assisting medical staff in diagnosis and treatment plan formulation. Figure 2 The structural diagram of the system proposed by the present invention. In combination with Figure 1 the content in

[0053] A sleep monitoring method based on the combination of EEG and millimeter-wave radar includes:

[0054] acquiring the EEG signals of the user during the sleep stage based on a wireless EEG sensor, and collecting the physiological signals of the user during the sleep stage using a millimeter-wave radar; the physiological signals include: respiratory signals, heart rate signals, and body movement data;

[0055] The wireless EEG sensor is designed at the positions of the occipital lobes O1 and O2 in the upper rear part of the pillow. When the headrest is placed on the pillow, it can directly contact these two positions, and sheet electrodes are used; the millimeter-wave radar is placed near the head of the subject.

[0056] Specifically, a 2-channel EEG sensor is adopted to mainly monitor the EEG signals in the frontal area and the central area.

[0057] Before performing sleep staging analysis on the EEG signals, preprocessing of the EEG signals is also included; the preprocessing includes: using 0.5 Hz high-pass filtering and 50 Hz notch filtering to eliminate power frequency interference; transforming the EEG signals into the frequency domain through wavelet transform for decomposition, and separating the myoelectric artifact and eye movement interference signals in the EEG signals.

[0058] A 60 Hz millimeter-wave radar is used to perform non-contact physiological signal detection through frequency-modulated continuous wave.

[0059] The millimeter-wave radar obtains the respiratory signals by detecting the minute periodic movements of the chest and abdomen.

[0060] Extract the heart rate signal by detecting the weak chest wall movement caused by heart beats;

[0061] Record the number of turnovers and leg movement events by detecting the whole body movement situation to obtain the body movement data.

[0062] In the embodiment of the present application, a wireless EEG sensor is integrated into a pillow, and a millimeter wave radar is installed near the head of the subject's bed to achieve unconstrained monitoring, avoid the discomfort caused by wearing electrodes, improve user compliance, and is suitable for long-term sleep monitoring; a 4-channel EEG sensor is adopted to monitor the EEG signals in the frontal area and the central area, which helps to improve the accuracy of subsequent sleep staging; the EEG signals are preprocessed by filtering and wavelet transform, which can remove signals such as power frequency interference, myoelectric artifacts and eye movement interference, helps to improve the purity of the EEG signals, and at the same time improves the accuracy of subsequent sleep staging and micro-arousal detection.

[0063] Preferably, perform sleep staging analysis on the EEG signals, and detect micro-arousal events in combination with the body movement data; the specific process is as follows:

[0064] Decompose the EEG signals by wavelet transform to obtain the time domain features and frequency domain features of the EEG signals, and use a machine learning model to stage the sleep stages based on the time domain features and the frequency domain features, and classify and identify the wakefulness period, light sleep period, deep sleep period and rapid eye movement sleep period;

[0065] Analyze the suddenly accelerated body movement by using the mutation point detection method according to the sleep staging in combination with the body movement data to detect micro-arousal events.

[0066] Specifically, the time domain features include: mean value of EEG signals, standard deviation, peak-to-peak value and zero crossing rate; the frequency domain features include: delta wave (0.5 - 4Hz), theta wave (4 - 8Hz), alpha wave (8 - 13Hz), beta wave (14 - 30Hz);

[0067] Train a CNN-LSTM model to classify the sleep stages according to the time domain features and the frequency domain features;

[0068] Based on the phenomenon that the energy of beta wave is enhanced in a short time while the theta wave decreases, initially identify the time window in which the micro-arousal event may occur;

[0069] Identify the mutation points of the suddenly accelerated body movement by using a sliding window mutation detection algorithm, and further determine whether the micro-arousal is accompanied by significant body movement; if a body movement mutation point is detected within ±2 seconds of the occurrence time of the micro-arousal signal, it is determined as the micro-arousal event, otherwise it is marked as an isolated micro-arousal; the isolated micro-arousal is: only brain arousal without obvious body movement.

[0070] In the embodiments of the present application, wavelet transform is used to extract the time-domain and frequency-domain features of electroencephalogram (EEG) signals, which can effectively separate the time-frequency domain features of different sleep stages, and an automatic classification is performed through a CNN-LSTM deep learning model, improving the accuracy of sleep staging. This method not only analyzes micro-arousal events using EEG signals, but also combines the body movement data of millimeter-wave radar for verification, enhancing the reliability of micro-arousal detection. Preliminary screening is carried out using the phenomena of short-term enhancement of β-wave energy and reduction of θ-wave, and then the sliding window mutation detection method is used to analyze the changes in body movement, making the identification of micro-arousal events more accurate and reducing the misjudgment that may be caused by relying solely on EEG signals. By matching the body movement mutation points within a time window of 2 seconds before and after the occurrence time of micro-arousal signals, it is possible to effectively distinguish micro-arousals accompanied by body movement from isolated micro-arousals, providing more targeted parameter support for the subsequent assessment of the severity of obstructive sleep apnea (OSA). This method provides a more accurate, non-invasive and efficient sleep monitoring solution, which helps in the early screening and long-term management of OSA patients.

[0071] Preferably, the physiological signal processing model processes the physiological signals respectively to extract the physiological signal features of the user, and monitors the user's apnea events, hypopnea events, arrhythmia events, abnormal blood pressure fluctuations events and periodic leg movement events according to the physiological signal features; reference Figure 3 ;

[0072] The physiological signal processing model includes: a respiratory signal processing unit, a heart rate signal processing unit and a body movement data processing unit;

[0073] The respiratory signal processing unit includes: a respiratory feature extraction layer and an apnea and hypopnea event identification layer;

[0074] The respiratory feature extraction layer extracts the changes in respiratory flow in the respiratory signal through signal filtering and denoising processing to obtain the respiratory features;

[0075] The apnea and hypopnea event identification layer determines the apnea-hypopnea duration and apnea-hypopnea frequency of the apnea event and / or the hypopnea event according to the respiratory features;

[0076] Specifically, the respiratory feature extraction layer uses a Butterworth low-pass filter to remove high-frequency noise, and uses an adaptive Kalman filter to reduce baseline drift; the envelope curve of the respiratory signal is calculated through Hilbert transform, and the respiratory features are extracted, including respiratory frequency, respiratory amplitude, inspiratory time, expiratory time and respiratory cycle stability;

[0077] The apnea and hypopnea event recognition layer detects apnea (no airflow for ≥10 seconds) and hypopnea events (airflow reduction by more than 30% and duration ≥10 seconds) based on the respiratory characteristics by setting apnea threshold and hypopnea threshold.

[0078] The heart rate signal processing unit includes: a heart rate feature extraction layer, an arrhythmia event recognition layer, and a blood pressure fluctuation abnormality event recognition layer;

[0079] The heart rate feature extraction layer performs multi-scale decomposition on the heart rate signal through the MODWT algorithm, extracts the heart rate of the heart rate signal, and obtains heart rate features;

[0080] The arrhythmia event recognition layer analyzes the heart rate variability of the user through the heart rate features and recognizes the arrhythmia events;

[0081] The blood pressure fluctuation abnormality event recognition layer predicts the user's blood pressure through machine learning based on the heart rate features and the heart rate variability, and recognizes the blood pressure fluctuation abnormality events;

[0082] Specifically, after performing multi-scale decomposition on the heart rate signal through the MODWT algorithm, the heart rate feature extraction layer uses the peak detection algorithm to extract the heart rate intervals of the heart rate signal and obtains the heart rate features;

[0083] The arrhythmia event recognition layer calculates heart rate variability parameters based on the heart rate features, including: standard deviation, low-frequency power, and high-frequency power;

[0084] The blood pressure fluctuation abnormality event recognition layer combines the heart rate features and heart rate variability parameters, uses a regression model based on XGBoost to predict the user's nocturnal blood pressure changes, and recognizes blood pressure fluctuation abnormality events, such as nocturnal hypertension and / or hypotension.

[0085] The body movement data processing unit includes: a body movement feature extraction layer and a periodic leg movement event recognition layer;

[0086] The body movement feature extraction layer extracts the periodic leg movement features from the body movement data, and the features include: body movement frequency and body movement amplitude;

[0087] The periodic leg movement event recognition layer recognizes periodic leg movement events based on the periodic leg movement features and records the leg movement times and leg movement durations.

[0088] Specifically, the body movement feature extraction layer analyzes the body movement spectrum using the short-time Fourier transform and extracts the leg movement features;

[0089] The periodic leg movement event recognition layer uses a threshold detection method and a pattern matching algorithm to recognize periodic leg movement events, that is, leg movements occurring every 5 - 90 seconds, and records the number of leg movements and the duration of leg movements.

[0090] In the embodiment of the present application, this method comprehensively analyzes the respiratory signal, heart rate signal, and body movement data of the user during the sleep stage by constructing a physiological signal processing model, so as to achieve accurate monitoring of various sleep-related physiological abnormal events. Through the Butterworth low-pass filter, adaptive Kalman filter, and Hilbert transform, this method can accurately extract the characteristics of the respiratory signal, such as respiratory frequency, respiratory amplitude, inspiratory time, and expiratory time, significantly improving the recognition accuracy of apnea and hypopnea events. By using MODWT multi-scale decomposition and peak detection algorithm, the heart rate intervals are extracted from the heart rate signal, and heart rate variability parameters such as standard deviation, low-frequency power, and high-frequency power are calculated, making the results of arrhythmia detection more accurate. At the same time, through the regression model based on XGBoost combined with heart rate variability characteristics, accurate prediction of abnormal nocturnal blood pressure fluctuations is realized, which is beneficial to evaluating the impact of OSA on the cardiovascular system. In terms of body movement data processing, the body movement spectrum is analyzed by short-time Fourier transform to make the extraction of leg movement characteristics more accurate. Combining the threshold detection method and the pattern matching algorithm, periodic leg movement events are effectively identified, improving the detection ability of periodic leg movement syndrome. Further combining body movement data with sleep stage analysis can distinguish isolated leg movements from leg movement events related to micro-awakenings, providing richer information for sleep quality assessment. Overall, this method improves the accuracy and automation level of sleep monitoring based on multi-modal physiological signal fusion analysis and intelligent algorithms, providing strong support for the early detection, severity assessment, and personalized intervention of subsequent OSA and related sleep disorders.

[0091] Preferably, based on the micro-awakening events, the apnea events, the hypopnea events, the arrhythmia events, the abnormal blood pressure fluctuation events, and the periodic leg movement events, various OSA determination indexes are calculated; the OSA determination indexes include: micro-awakening index, apnea-hypopnea index, degree of sympathetic nerve activation, blood pressure fluctuation situation, and periodic leg movement index; refer to Figure 4 ;

[0092] The micro-awakening index is calculated by using the number of micro-awakening events and the total sleep duration; the calculation formula is:

[0093]

[0094] where ArI is the micro-awakening index; N arousal is the number of micro-awakening events; T sleep is the total sleep duration;

[0095] The apnea-hypopnea index is the total number of the apnea events and the hypopnea events occurring per hour, and is calculated according to the apnea-hypopnea duration and the apnea-hypopnea frequency; the calculation formula is:

[0096]

[0097] where AHI is the microarousal index; N apnea is the number of apnea events; N hypopnea is the number of hypopnea events;

[0098] The degree of sympathetic nerve activation is quantified according to the parameters of heart rate variability reflected in the heart rate signal; the calculation formula is:

[0099]

[0100] where SNA is the degree of sympathetic nerve activation; LF is the power of the low-frequency component of heart rate variability; HF is the power of the high-frequency component of heart rate variability;

[0101] The blood pressure fluctuation situation is calculated based on the statistical standard deviation of blood pressure; the calculation formula is:

[0102]

[0103] where σ BP is the blood pressure fluctuation situation; BP i is the i-th blood pressure value; is the average blood pressure value; N is the total number of blood pressure values; the periodic leg movement index is calculated using the number of periodic leg movement events and the total sleep duration; the calculation formula is:

[0104]

[0105] where PLMI is the periodic leg movement index; N leg is the number of periodic leg movement events.

[0106] By comprehensively analyzing micro-arousal events, apnea events, hypopnea events, arrhythmia events, abnormal blood pressure fluctuation events, and periodic limb movement events, the embodiments of the present application can comprehensively reflect the sleep quality and physiological state of the user, which helps to provide more accurate and detailed indicators for the diagnosis and evaluation of OSA. The micro-arousal index and the periodic limb movement index provide a basis for identifying possible sleep interruptions during sleep by measuring the frequencies of micro-arousal events and leg movements; the apnea-hypopnea index can accurately reflect the occurrence frequency of apnea and hypopnea events, and further evaluate the severity of OSA; the degree of sympathetic nerve activation reflects the state of the autonomic nerve function of the user by analyzing heart rate variability, which helps to reveal the excessive activation of the sympathetic nerve and indicates possible cardiovascular problems; the blood pressure fluctuation situation quantifies abnormal blood pressure fluctuations by statistically calculating the standard deviation of blood pressure, helps to predict the rules and risks of blood pressure changes, and further evaluates the cardiovascular health status of the patient. Combining these indicators helps to provide a personalized OSA score for the patient and a personalized sleep intervention plan in the follow-up, and provides a scientific basis for the formulation of subsequent diagnosis and treatment plans, so as to achieve early detection and precise management of OSA and related diseases, and improve the overall health level of the patient.

[0107] Preferably, calculate the personalized OSA score according to the OSA determination index, and perform the grading of the severity of OSA according to the personalized OSA score to generate a report on the grading of the severity of OSA, which is used to assist the diagnosis and treatment plan formulation of medical staff;

[0108] The calculation formula of the personalized OSA score is:

[0109] OSA_Score = ω 1 ·AHI + ω 2 ·ArI + ω 3 ·SNA + ω 4 ·σ BP + ω 5 ·PLMI;

[0110] wherein, OSA_Score is the personalized OSA score; ω 1 is the weight of the apnea-hypopnea index; AHI is the apnea-hypopnea index; ω 2 is the weight of the micro-arousal index; ArI is the micro-arousal index; ω 3 is the weight of the degree of sympathetic nerve activation; SNA is the degree of sympathetic nerve activation; ω 4 is the weight of the blood pressure fluctuation situation; σ BP is the blood pressure fluctuation situation; ω 5 is the weight of the periodic limb movement index; PLMI is the periodic limb movement index; ω 1 、ω 2 、ω3 , ω 4 and ω 5 Be dynamically adjusted according to the comorbidities of different populations.

[0111] Specifically, for different populations, such as patients with comorbidities like hypertension and diabetes, the OSA personalized score may be adjusted due to the presence of these comorbidities. For example, patients with hypertension may need to have their scores appropriately increased according to the weight of blood pressure fluctuations, while patients with diabetes may be affected by microarousal events or the degree of sympathetic nerve activation, resulting in an increased score;

[0112] The OSA severity grading includes:

[0113] Mild OSA: The OSA personalized score is lower than the mild OSA threshold, indicating fewer apnea events, relatively stable physiological signals, and milder symptoms in the patient.

[0114] Moderate OSA: The OSA personalized score is within the moderate OSA threshold range, indicating that the patient may experience moderate-frequency apnea events and more significant problems such as microarousals or heart rate fluctuations.

[0115] Severe OSA: The OSA personalized score is higher than the severe OSA threshold, usually indicating that the patient has frequent apnea events, severe heart rate irregularities, blood pressure fluctuations, etc., and the impact of comorbidities may also be more obvious;

[0116] Among them, the mild OSA threshold, the moderate OSA threshold, and the severe OSA threshold are customized by experts based on the differences in individual comorbidities.

[0117] Table 1 gives an example of the weights for dynamic adjustment according to different comorbidities.

[0118] Table 1 Example of weights for dynamic adjustment according to different comorbidities

[0119]

[0120] Based on the calculation formula of OSA personalized score, the embodiments of this application can effectively provide a customized OSA assessment for each patient by performing weighted calculations on multiple OSA determination indicators and dynamically adjusting in combination with the comorbidities of different populations. This method can fully reflect the specific physiological conditions and health risks of individuals, contributing to the accurate assessment of the severity of OSA. By considering the weights of various indicators and dynamic adjustments, it can ensure that for different patient groups, such as patients with hypertension, diabetes, cardiovascular diseases, etc., the generated scores are more individualized, enhancing the clinical practicality and accuracy of the scores. This not only provides clinicians with more dimensions of assessment basis but also can provide a more accurate reference basis for subsequent diagnosis and treatment plan formulation, thereby improving the treatment effect and the implementation of personalized treatment. This method can also support the long-term management and tracking of patients' conditions, helping to detect potential health risks in a timely manner, promoting early intervention, and improving the quality of life of patients.

[0121] The embodiments of this application provide a sleep monitoring method based on the combination of electroencephalogram and millimeter-wave radar. By integrating the monitoring and analysis of multiple physiological signals, it can comprehensively and accurately evaluate the sleep quality and health status of users. By simultaneously acquiring electroencephalogram signals and physiological signals through wireless EEG sensors and millimeter-wave radar, it can not only effectively detect the sleep stages of users but also accurately identify microarousal events in combination with body movement data, providing a reliable basis for in-depth analysis of sleep disorders. By using a physiological signal processing model to monitor important events such as apnea, hypopnea, arrhythmia, abnormal blood pressure fluctuations, and periodic limb movements in real time, it helps to subsequently determine whether the user has sleep-related diseases such as OSA. By accurately calculating OSA determination indicators and personalized scores, the system can automatically classify the severity of OSA, providing more detailed health data support for clinicians and helping to formulate more personalized treatment plans for patients in a timely manner. This method not only improves the accuracy of sleep monitoring but also can provide personalized and accurate OSA judgment criteria and treatment suggestions for patients, helping to improve the treatment effect, improve the quality of life of patients, and provide a scientific basis for long-term health management and risk assessment.

[0122] Embodiment 2

[0123] In Embodiment 1, the method proposed by the present invention successfully achieved the customization of personalized OSA scores according to the physiological conditions of different users, thereby improving the judgment accuracy of OSA. To further verify the effectiveness of the present invention, in the embodiments of this application, another 150 patients in the third-level class-A hospital B were also subjected to OSA judgment.

[0124] A sleep monitoring system based on the combination of electroencephalogram and millimeter-wave radar, comprising:

[0125] The signal acquisition module is used to obtain the electroencephalogram (EEG) signals of the user during the sleep stage based on a wireless EEG sensor, and collect the physiological signals of the user during the sleep stage by using a millimeter-wave radar; the physiological signals include: respiratory signals, heart rate signals, and body movement data.

[0126] The wireless EEG sensor is designed at the positions of the occipital lobes O1 and O2 at the upper rear part of the pillow. The headrest can directly contact these two positions when placed on the pillow, and the electrodes are sheet electrodes; the millimeter-wave radar is placed near the head of the test subject's bed.

[0127] Preferably, the micro-arousal event detection module is used to perform sleep stage analysis on the EEG signals and detect micro-arousal events in combination with the body movement data; the specific process is as follows: the time-domain features and frequency-domain features of the EEG signals are obtained through wavelet transform decomposition, and a machine learning model is used to classify the sleep stages based on the time-domain features and frequency-domain features, identifying the wakefulness stage, light sleep stage, deep sleep stage, and rapid eye movement (REM) sleep stage; based on the sleep stage classification and the body movement data, the sudden acceleration of body movement is analyzed by using the mutation point detection method to detect micro-arousal events.

[0128] Preferably, the physiological signal processing module is used to process the physiological signals respectively through a physiological signal processing model, extract the physiological signal features of the user, and monitor the user's apnea events, hypopnea events, arrhythmia events, and periodic leg movement events according to the physiological signal features.

[0129] The physiological signal processing model includes: a respiratory signal processing unit, a heart rate signal processing unit, and a body movement data processing unit.

[0130] The respiratory signal processing unit includes: a respiratory feature extraction layer and an apnea and hypopnea event recognition layer.

[0131] The respiratory feature extraction layer extracts the change in respiratory flow in the respiratory signal through signal filtering and denoising processing to obtain the respiratory features.

[0132] The apnea and hypopnea event recognition layer determines the apnea / hypopnea duration and apnea / hypopnea frequency of the apnea event and / or the hypopnea event according to the respiratory features.

[0133] The heart rate signal processing unit includes: a heart rate feature extraction layer, an arrhythmia event recognition layer, and a blood pressure fluctuation abnormality event recognition layer.

[0134] The heart rate feature extraction layer performs multi-scale decomposition on the heart rate signal through the MODWT algorithm, extracts the heart rate of the heart rate signal, and obtains the heart rate features.

[0135] The arrhythmia event recognition layer analyzes the user's heart rate variability through the heart rate characteristics to identify the arrhythmia events;

[0136] The blood pressure fluctuation abnormal event recognition layer predicts the user's blood pressure through machine learning according to the heart rate characteristics and the heart rate variability, and identifies the blood pressure fluctuation abnormal events;

[0137] The body movement data processing unit includes: a body movement feature extraction layer and a periodic leg movement event recognition layer;

[0138] The body movement feature extraction layer extracts the periodic leg movement features by using the body movement data, and the features include: body movement frequency and body movement amplitude;

[0139] The periodic leg movement event recognition layer identifies the periodic leg movement events according to the periodic leg movement features and records the leg movement times and leg movement durations.

[0140] Preferably, the OSA determination index calculation module is used to calculate various OSA determination indexes based on the microarousal events, the apnea events, the hypopnea events, the arrhythmia events and the periodic leg movement events; the OSA determination indexes include: microarousal index, apnea hypopnea index, sympathetic nerve activation degree, blood pressure fluctuation condition and periodic leg movement index;

[0141] The microarousal index is calculated by using the number of microarousal events and the total sleep duration;

[0142] The apnea hypopnea index is the total number of apnea events and hypopnea events occurring per hour, and is calculated according to the apnea hypopnea duration and the apnea hypopnea frequency;

[0143] The sympathetic nerve activation degree is quantified according to the parameters reflecting the heart rate variability in the heart rate signal;

[0144] The blood pressure fluctuation condition is calculated based on the statistical standard deviation of blood pressure;

[0145] The periodic leg movement index is calculated by using the number of periodic leg movement events and the total sleep duration.

[0146] Preferably, the OSA personalized scoring and severity grading module is used to calculate the OSA personalized score according to the OSA determination indexes, and perform OSA severity grading according to the OSA personalized score, and generate an OSA severity grading report for assisting medical staff in diagnosis and treatment plan formulation;

[0147] The calculation formula of the OSA personalized score is:

[0148] OSA_Score = ω 1 ·AHI + ω 2 ·ArI + ω 3 ·SNA + ω 4 ·σ BP + ω 5 ·PLMI;

[0149] Among them, OSA_Score is the personalized score of OSA; ω 1 is the weight of the apnea-hypopnea index; AHI is the apnea-hypopnea index; ω 2 is the weight of the arousal index; ArI is the arousal index; ω 3 is the weight of the degree of sympathetic nerve activation; SNA is the degree of sympathetic nerve activation; ω 4 is the weight of the blood pressure fluctuation; σ BP is the blood pressure fluctuation; ω 5 is the weight of the periodic leg movement index; PLMI is the periodic leg movement index; ω 1 、ω 2 、ω 3 、ω 4 and ω 5 are dynamically adjusted according to the comorbidities of different populations.

[0150] Table 2 gives the comparison table of the accuracy of the OSA personalized scoring method for diagnosing OSA.

[0151] Table 2 Comparison of the accuracy of the OSA personalized scoring method for diagnosing OSA

[0152]

[0153] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A sleep monitoring method based on the combination of EEG and millimeter wave radar, characterized in that: include: The wireless EEG sensor is used to obtain the user's EEG signals during sleep, and the millimeter-wave radar is used to collect the user's physiological signals during sleep. The physiological signals include: breathing signals, heart rate signals and body movement data; Performing sleep stage analysis on the EEG signal, and detecting micro-arousal events in combination with the body movement data; The physiological signals are processed respectively by a physiological signal processing model to extract the physiological signal characteristics of the user, and the user's apnea events, hypopnea events, arrhythmia events, abnormal blood pressure fluctuation events and periodic leg movement events are monitored according to the physiological signal characteristics; Calculating various OSA determination indicators based on the micro-arousal events, the apnea events, the hypopnea events, the arrhythmia events, the abnormal blood pressure fluctuation events and the periodic leg movement events; An OSA personalized score is calculated based on the OSA determination index, and OSA severity is graded based on the OSA personalized score to generate an OSA severity grading report for assisting medical staff in diagnosis and formulation of treatment plans.

2. The sleep monitoring method based on the combination of EEG and millimeter wave radar according to claim 1, characterized in that: The wireless EEG sensor is designed at the positions of O1 and O2 of the occipital lobe at the upper back of the pillow; the millimeter wave radar is placed near the head of the subject's bed.

3. The sleep monitoring method based on the combination of EEG and millimeter wave radar according to claim 1, characterized in that: The specific process of performing sleep staging analysis on the EEG signal and detecting micro-arousal events in combination with the body movement data is as follows: obtaining the time domain characteristics and frequency domain characteristics of the EEG signal through wavelet transform decomposition, and using a machine learning model to stage the sleep stages based on the time domain characteristics and the frequency domain characteristics, and classify and identify the wakefulness period, light sleep period, deep sleep period and rapid eye movement sleep period; using a mutation point detection method to analyze sudden accelerated body movements according to the sleep staging combined with the body movement data, and detect micro-arousal events.

4. The sleep monitoring method based on the combination of EEG and millimeter wave radar according to claim 1, characterized in that: The physiological signal processing model comprises: a respiratory signal processing unit, a heart rate signal processing unit and a body motion data processing unit; the respiratory signal processing unit comprises: a respiratory feature extraction layer and a apnea and hypopnea event recognition layer; the respiratory feature extraction layer extracts the change of respiratory flow in the respiratory signal through signal filtering and denoising processing to obtain the respiratory feature; the apnea and hypopnea event recognition layer determines the apnea-hypopnea duration and apnea-hypopnea frequency of the apnea event and / or the hypopnea event according to the respiratory feature; the heart rate signal processing unit comprises: a heart rate feature extraction layer, an arrhythmia event recognition layer and an abnormal blood pressure fluctuation event recognition layer; the heart rate feature extraction layer uses the MODWT algorithm to identify the heart rate signal Multi-scale decomposition is performed to extract the heart rate frequency of the heart rate signal and obtain the heart rate characteristics; the arrhythmia event recognition layer analyzes the user's heart rate variability through the heart rate characteristics and identifies the arrhythmia events; the abnormal blood pressure fluctuation event recognition layer predicts the user's blood pressure through machine learning based on the heart rate characteristics and the heart rate variability, and identifies the abnormal blood pressure fluctuation events; the body movement data processing unit includes: a body movement feature extraction layer and a periodic leg movement event recognition layer; the body movement feature extraction layer uses the body movement data to extract the periodic leg movement features, and the periodic leg movement features include: body movement frequency and body movement amplitude; the periodic leg movement event recognition layer identifies periodic leg movement events based on the periodic leg movement features and records the number of leg movements and the duration of leg movements.

5. The sleep monitoring method based on the combination of EEG and millimeter wave radar according to claim 1, characterized in that: The OSA determination index includes: micro-arousal index, apnea-hypopnea index, sympathetic nerve activation level, blood pressure fluctuation and periodic leg movement index; the micro-arousal index is calculated using the number of micro-arousal events and the total sleep duration; the apnea-hypopnea index is the total number of apnea events and hypopnea events occurring per hour, calculated based on the apnea-hypopnea duration and apnea-hypopnea frequency; the sympathetic nerve activation level is quantified based on the parameters of the heart rate variability reflected in the heart rate signal; the blood pressure fluctuation is calculated based on the statistical standard deviation of the blood pressure; the periodic leg movement index is calculated using the number of periodic leg movement events and the total sleep duration.

6. The sleep monitoring method based on the combination of EEG and millimeter wave radar according to claim 1, characterized in that: The calculation formula of the OSA personalized score is: OSA_Score=ω1·AHI+ω2·ArI+ω3·SNA+ω4·σ BP +ω5·PLMI; Among them, OSA_Score is the personalized score of OSA; ω1 is the weight of the apnea-hypopnea index; AHI is the apnea-hypopnea index; ω2 is the weight of the micro-arousal index; ArI is the micro-arousal index; ω3 is the weight of the sympathetic nerve activation degree; SNA is the sympathetic nerve activation degree; ω4 is the weight of the blood pressure fluctuation; σ BP is the blood pressure fluctuation; ω5 is the weight of the periodic leg movement index; PLMI is the periodic leg movement index.

7. A sleep monitoring system based on the combination of EEG and millimeter wave radar, characterized in that: include: The signal acquisition module is used to obtain the user's EEG signals during sleep based on the wireless EEG sensor and to collect the user's physiological signals during sleep using the millimeter wave radar; The physiological signals include: breathing signals, heart rate signals and body movement data; A micro-arousal event detection module, used to perform sleep staging analysis on the EEG signal and detect micro-arousal events in combination with the body movement data; A physiological signal processing module, used to process the physiological signals respectively through a physiological signal processing model, extract the physiological signal characteristics of the user, and monitor the user's apnea events, hypopnea events, arrhythmia events and periodic leg movement events according to the physiological signal characteristics; An OSA determination index calculation module, used for calculating various OSA determination indexes based on the micro-arousal event, the apnea event, the hypopnea event, the arrhythmia event and the periodic leg movement event; The OSA personalized scoring and severity grading module is used to calculate the OSA personalized score according to the OSA determination index, and to grade the OSA severity according to the OSA personalized score, and to generate an OSA severity grading report to assist medical staff in diagnosis and formulation of treatment plans.

Images