Obstructive sleep apnea detection method, system and equipment and storage medium

Abstract

The invention provides an obstructive sleep apnea detection method, system and device and a storage medium, and relates to the technical field of signal processing, and the method comprises the steps: obtaining pressure signals detected by at least two bed-mounted pressure sensors in a human body sleep state, and the crowd type characteristics and the pre-sleep activity characteristics of a to-be-detected person; performing quality evaluation on each path of pressure signal, and determining a target pressure signal based on a quality evaluation result; preprocessing the target pressure signal, and performing time domain and frequency domain feature extraction on the preprocessed target pressure signal; and inputting the extracted multi-dimensional features, the crowd type features of the to-be-detected person and the pre-sleep activity features into a pre-trained machine learning model to obtain a classification result of whether the to-be-detected person suffers from obstructive sleep apnea. According to the method and the device, under the condition that PSG equipment is not used, accurate detection of the OSA is realized.

Description

Technical Field

[0001] This application relates to the field of signal processing technology, and in particular to a method, system, device and storage medium for detecting obstructive sleep apnea. Background Technology

[0002] Sleep-disordered breathing, especially obstructive sleep apnea (OSA), is a common and serious sleep disorder that significantly impacts health. During sleep, OSA patients experience repeated obstruction of the upper airway, leading to apnea or a significant reduction in ventilation, which in turn causes a series of problems such as hypoxemia and disrupted sleep structure. Long-term OSA increases the risk of developing various chronic diseases, including hypertension, coronary heart disease, arrhythmia, cerebrovascular disease, and diabetes, severely affecting patients' quality of life and lifespan.

[0003] Traditional OSA diagnosis primarily relies on polysomnography (PSG). While PSG can comprehensively and accurately monitor various physiological parameters during sleep, such as electroencephalography (EEG), electrooculography (EOG), electromyography (EMG), electrocardiography (ECG), respiratory airflow, chest and abdominal respiratory movements, and blood oxygen saturation, thus providing a reliable basis for OSA diagnosis and condition assessment, this method has many limitations. PSG requires patients to undergo the procedure in a hospital sleep monitoring room. The monitoring equipment is complex, and patients need to wear numerous electrodes and sensors. This not only causes significant inconvenience and discomfort for patients, affecting their sleep quality and potentially leading to biased monitoring results, but also results in high costs. Furthermore, it requires specialized technicians for operation and analysis, making it difficult to widely adopt PSG for routine screening and long-term home monitoring. Summary of the Invention

[0004] In view of this, this application proposes a method, system, device and storage medium for detecting obstructive sleep apnea.

[0005] In a first aspect, this application provides a method for detecting obstructive sleep apnea, comprising:

[0006] Acquire pressure signals detected by at least two bed-mounted pressure sensors during human sleep, as well as the population type characteristics and pre-sleep activity characteristics of the person being tested;

[0007] Each pressure signal is evaluated for quality, and the target pressure signal is determined based on the evaluation results.

[0008] The target pressure signal is preprocessed, and time-domain and frequency-domain features are extracted from the preprocessed target pressure signal.

[0009] The extracted multidimensional features, population type features of the individuals to be tested, and pre-sleep activity features are input into a pre-trained machine learning model to obtain classification results on whether or not the individual suffers from obstructive sleep apnea.

[0010] In one embodiment, the machine learning model is a Bayesian model; the step of inputting the extracted multi-dimensional features, the population type features of the person to be tested, and the pre-sleep activity features into the pre-trained machine learning model to obtain a classification result of whether or not the person suffers from obstructive sleep apnea includes:

[0011] The extracted multi-dimensional features, the population type features of the person to be detected, and the pre-sleep activity features are input into the Bayesian model so that the Bayesian model can call the population-specific feature parameter library according to the population type features of the person to be detected.

[0012] Obtain the population type weight coefficient and activity weight coefficient corresponding to the population type characteristics and pre-sleep activity characteristics of the person to be tested, respectively;

[0013] The weights of time-domain and frequency-domain features are adjusted based on the population type weight coefficient and the activity weight coefficient, and the posterior probabilities of whether the person to be tested has obstructive sleep apnea are determined by combining the population-specific parameters called.

[0014] The classification result of whether the person being tested has obstructive sleep apnea is determined based on the posterior probability of the two types of results.

[0015] In one embodiment, adjusting the weights of time-domain and frequency-domain features based on the population type weight coefficient and the activity weight coefficient, and combining this with the invoked population-specific parameters, to determine the posterior probability of whether the person to be tested suffers from obstructive sleep apnea includes:

[0016] The calibration weights for each time-domain feature and each frequency-domain feature are determined based on the population type weight coefficient and the activity weight coefficient.

[0017] Based on the population type of the person to be tested, retrieve the characteristic conditional probability distribution of the corresponding population from the corresponding population-specific characteristic parameter database;

[0018] The posterior probabilities of whether the person under test has obstructive sleep apnea are determined based on the extracted multi-dimensional features, the calibration weights of each time-domain feature and each frequency-domain feature, and the feature conditional probability distribution.

[0019] In one embodiment, the step of performing quality assessments on each pressure signal and determining the target pressure signal based on the quality assessment results includes:

[0020] The energy and standard deviation of each signal sub-signal in the target frequency band are calculated to obtain the quality assessment results in the target frequency band.

[0021] The fusion weights of each signal are determined based on the frequency band quality assessment results, and the sub-signals of each signal in the same frequency band are weighted and fused based on the fusion weights of each signal to obtain the target pressure signal.

[0022] In one embodiment, the step of performing quality assessments on each pressure signal and determining the target pressure signal based on the quality assessment results includes:

[0023] Perform a sliding window analysis for each signal with a preset time, and calculate the signal energy and standard deviation within the window;

[0024] The corresponding energy fluctuation coefficient is determined based on the signal energy within the window of each signal.

[0025] The optimal signal is determined based on the energy fluctuation coefficient and standard deviation of each signal, and the optimal signal is used as the target pressure signal.

[0026] In one embodiment, the preprocessing of the target pressure signal and the extraction of time-domain and frequency-domain features from the preprocessed target pressure signal include:

[0027] Low-frequency noise below a predetermined frequency is removed by adaptive Kalman filtering, and multiple frequency band signals are obtained by multi-layer wavelet decomposition, wherein the multiple frequency band signals include respiratory main frequency band signals and cardiovascular main frequency band signals.

[0028] The respiratory main frequency band signal is detected by a peak detection algorithm to identify respiratory peaks. A respiratory threshold is determined based on the signal mean of the respiratory main frequency band signal. The respiratory frequency and respiratory cycle variation coefficient are determined according to the identified respiratory peaks and the respiratory threshold. The maximum amplitude, minimum amplitude and amplitude ratio of the respiratory signal are extracted from the respiratory main frequency band signal.

[0029] Wavelet denoising is performed on the cardiovascular main frequency band signal to extract the heartbeat signal, and the heart rate variability index is calculated based on the heartbeat signal and a preset time window;

[0030] Fast Fourier Transform is performed on time period signals selected from the respiratory dominant frequency band signal and the cardiovascular dominant frequency band signal to calculate the energy proportion of the respiratory dominant frequency band and the energy proportion of the cardiovascular dominant frequency band, respectively.

[0031] Secondly, this application also provides an obstructive sleep apnea detection system, including: a processing module and at least two bed-mounted pressure sensors;

[0032] Each bed pressure sensor is designed to be installed on the mattress at the position corresponding to the human chest.

[0033] The processing module is connected to each bed load pressure sensor and is used to perform the obstructive sleep apnea detection method as described in the first aspect.

[0034] In one embodiment, the processing module includes a signal conditioning module, a low-pass filter, a high-pass filter, a data acquisition module, and a data processing module;

[0035] The signal conditioning module is used to receive the pressure signals output by each bed pressure sensor and amplify each pressure signal to obtain an amplified pressure signal.

[0036] The low-pass filter is used to filter out high-frequency noise in the amplified pressure signal;

[0037] The high-pass filter is used to filter the signal output by the low-pass filter in order to suppress baseline drift;

[0038] The data acquisition module is used to perform analog-to-digital conversion on the signal output by the high-pass filter to generate a digital signal;

[0039] The data processing module is used to process the digital signals output by the data acquisition module and to execute the obstructive sleep apnea detection method as described in the first aspect.

[0040] Thirdly, this application also provides an electronic device, including a processor and a memory; the memory has a computer program stored thereon, wherein the computer program, when executed by the processor, implements the obstructive sleep apnea detection method as described in the first aspect.

[0041] Fourthly, this application also provides a computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the obstructive sleep apnea detection method as described in the first aspect.

[0042] The obstructive sleep apnea detection method of this application has the following advantages over related technologies:

[0043] 1. The obstructive sleep apnea detection method of this application acquires pressure signals detected by at least two bed-mounted pressure sensors during human sleep, and performs quality assessment on each pressure signal. Based on the quality assessment results, a target pressure signal is determined, thus obtaining a target pressure signal with higher quality. After acquiring the target pressure signal, it can be preprocessed, and then time-domain and frequency-domain features can be extracted from the preprocessed target pressure signal to accurately extract multi-dimensional features related to OSA. On this basis, considering the influence of the population type characteristics and pre-sleep activity characteristics of the person being tested, the extracted multi-dimensional features, the population type characteristics of the person being tested, and the pre-sleep activity characteristics are input into a pre-trained machine learning model to improve diagnostic accuracy through multi-dimensional physiological features. At the same time, relying on population type characteristics, the model can break through the limitations of single-population standards, optimize the judgment logic for high-risk groups and pre-sleep activities, enhance personalized adaptation capabilities, and further improve the classification accuracy of OSA.

[0044] 2. The obstructive sleep apnea detection method of this application does not rely on complex and expensive PSG equipment. At least two bed-mounted pressure sensors are deployed on the mattress, and signals are collected through the bed-mounted pressure sensors. It is suitable for primary healthcare and home screening scenarios and is highly convenient to use.

[0045] 3. Based on population type weighting coefficients and activity weighting coefficients, the weights of time-domain and frequency-domain features related to respiratory stability can be adjusted, with the weights of these features being increased. Then, pre-constructed population-specific parameters are called, and the adjusted weighted time-domain / frequency-domain features are compared with these parameters, using Bayes' theorem to calculate the posterior probability. Based on the prior probability of OSA in this population, the "likelihood" is calculated using the adjusted weighted physiological features, and the influence of population / activity weights is incorporated. Finally, the posterior probabilities of two outcomes—"having OSA" and "not having OSA"—are output. The classification result of whether the person being tested has obstructive sleep apnea is determined based on the posterior probabilities of these two outcomes. This classification process uses probability values ​​as the core basis, combined with population risk characteristics, to balance the risks of missed and false positives, ensuring the accuracy of the classification results. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1This is a schematic flowchart of an obstructive sleep apnea detection method in one embodiment of this application;

[0048] Figure 2 This is a flowchart illustrating step S104 in one embodiment of this application;

[0049] Figure 3 This is a flowchart illustrating step S103 in one embodiment of this application;

[0050] Figure 4 This is a schematic diagram of the obstructive sleep apnea detection system in one embodiment of this application. Detailed Implementation

[0051] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0052] In some embodiments, such as Figure 1 As shown, this application provides a method for detecting obstructive sleep apnea, including the following steps S101 to S104.

[0053] S101: Acquire pressure signals detected by at least two bed-mounted pressure sensors during human sleep, as well as the population type characteristics and pre-sleep activity characteristics of the person to be tested.

[0054] The bed-mounted pressure sensor can be deployed either built into the mattress or installed on the bed surface. It activates before the subject falls asleep and continuously collects signals throughout the entire sleep cycle, converting minute pressure changes during sleep into electrical signals. The subject's demographic characteristics focus on high-risk and normal populations, encompassing features such as age, weight, body type, and underlying medical conditions. For example, it can categorize individuals into the elderly, overweight, and normal-weight populations. Pre-sleep activity characteristics target non-pathological factors that may interfere with OSA diagnosis within a certain timeframe before sleep, including dietary and beverage intake and physical activity, such as alcohol consumption and vigorous exercise before bed.

[0055] S102: Perform quality assessment on each pressure signal separately, and determine the target pressure signal based on the quality assessment results.

[0056] In applications, corresponding characteristic parameters can be calculated based on each pressure signal. Then, the quality of each pressure signal can be evaluated based on preset standards. The signal with the better quality can be selected as the target pressure signal based on the evaluation results, or the pressure signals can be fused together and the fused signal can be used as the target pressure signal. The quality evaluation standard can consider anti-interference capability, i.e., evaluation based on the fluctuation parameters of each pressure signal. Alternatively, the quality evaluation standard can consider signal discernibility, assessing signal quality by determining whether the signal can clearly distinguish normal physiological movements during sleep from interference signals.

[0057] S103: Preprocess the target pressure signal and extract features in the time and frequency domains from the preprocessed target pressure signal.

[0058] Preprocessing may include noise reduction. For example, wavelet filtering or mean filtering can be used to remove signal noise, and interpolation can be used to correct for any breaks caused by brief signal loss, ensuring the signal is continuous and complete.

[0059] It is understandable that time-domain features focus on the intuitive representation of a signal in the time dimension, while frequency-domain features require Fourier transform to convert the time-domain signal to the frequency domain. Correspondingly, the frequency characteristics of the signal can uncover hidden information that cannot be intuitively presented in the time-domain signal, providing supplementary evidence for sleep state assessment. Both types of features together cover the temporal patterns and frequency characteristics of the signal, providing quantitative data support for subsequent sleep stage analysis and abnormal behavior identification, thereby improving the accuracy of OSA classification.

[0060] S104: Input the extracted multi-dimensional features, population type features of the person to be tested, and pre-sleep activity features into the pre-trained machine learning model to obtain the classification result of whether or not the person suffers from obstructive sleep apnea.

[0061] The machine learning model in this application can be of the type of model, such as random forest model, support vector machine and Bayesian model.

[0062] In the application, sample data of OSA patients diagnosed by PSG and healthy individuals can be collected, and features can be extracted to construct a training set.

[0063] It is understandable that the time-domain and frequency-domain features extracted from the detected pressure signals can directly reflect physiological signals related to OSA, such as changes in respiratory airflow and body posture fluctuations during sleep. Furthermore, the population type characteristics and pre-sleep activity characteristics of the individuals being tested are closely related to the risk of OSA, and these characteristics can assist the model in adapting to individual predispositions. Therefore, after inputting these three types of features into a pre-trained machine learning model, the model can first perform feature association matching, then calculate classification probabilities through internal algorithms, and finally output a classification result of "having OSA" or "not having OSA," achieving non-invasive and convenient detection of OSA.

[0064] It should be noted that in the application, the model can also be updated based on newly collected data. The update is initiated after a preset number (e.g., 50) of new samples have been collected. The newly collected sample data can be filtered first to ensure signal integrity (effective segment ratio > 80%, segments with motion interference > 30% removed) and label reliability (new samples are verified by PSG, AHI error < 5).

[0065] Model updates can be triggered based on feature drift detection. Specifically, KS tests (significance levels) are performed on the multi-dimensional features (e.g., 12-dimensional features) of the new sample and the original training set. If ≥3 features pass the test, a full update is triggered; otherwise, an incremental update is performed.

[0066] Model update strategies can include full updates and incremental updates.

[0067] Full update: Merge new samples with the original dataset and retrain the model. Use adaptive bandwidth Gaussian kernel density estimation, dynamically adjusting the bandwidth by 1 / 5 of the feature variance, and optimize the Laplacian smoothing coefficient. .

[0068] Incremental update: Only new samples are used to update the model parameters, and continuous features are used to update the mean and variance. The corresponding formulas are as follows:

[0069]

[0070]

[0071] in, This represents the mean of the continuous features after the update. This represents the variance of continuous features before the update. The variance represents the continuous features in the new sample set; This represents the variance of the continuous features after the update. This represents the variance of continuous features before the update. It represents the variance of continuous features in the new sample set.

[0072] Discrete feature update conditional probability

[0073]

[0074] This represents the conditional probability that the discrete feature takes the value x when the category is C after the update. This indicates the number of samples of category C before the update; This represents the conditional probability that the discrete feature takes the value x when the class is C before the update. This represents the number of samples of category C in the new sample set; This represents the conditional probability that the discrete feature takes the value x when the category is C in the new sample set.

[0075] The obstructive sleep apnea detection method of this application acquires pressure signals detected by at least two bed-mounted pressure sensors during human sleep, and performs quality assessment on each pressure signal. Based on the quality assessment results, a target pressure signal is determined, resulting in a higher-quality target pressure signal. After acquiring the target pressure signal, it can be preprocessed, and then time-domain and frequency-domain features can be extracted from the preprocessed target pressure signal to accurately extract multi-dimensional features related to OSA. Furthermore, considering the influence of the individual's population type characteristics and pre-sleep activity characteristics, the extracted multi-dimensional features, the individual's population type characteristics, and pre-sleep activity characteristics are input into a pre-trained machine learning model. This improves diagnostic accuracy through multi-dimensional physiological features. Simultaneously, relying on population type characteristics allows the model to overcome the limitations of a single population standard, optimizing the judgment logic for high-risk groups and individuals' pre-sleep activities, enhancing personalized adaptation capabilities, and further improving the classification accuracy of OSA. Meanwhile, the obstructive sleep apnea detection method of this application does not rely on complex and expensive PSG equipment. It simply deploys at least two bed-mounted pressure sensors on the mattress and collects signals through the bed-mounted pressure sensors, making it suitable for primary healthcare and home screening scenarios and highly convenient to use.

[0076] In some embodiments, the machine learning model is a Bayesian model. For example... Figure 2 As shown, in step S104, the extracted multi-dimensional features, the population type features of the person to be tested, and the pre-sleep activity features are input into the pre-trained machine learning model to obtain the classification result of whether or not the person suffers from obstructive sleep apnea, including the following steps S201 to S204.

[0077] S201: Input the extracted multi-dimensional features, the population type features of the person to be tested, and the pre-sleep activity features into the Bayesian model so that the Bayesian model can call the population-specific feature parameter library according to the population type features of the person to be tested.

[0078] Among them, the population-specific characteristic parameter library pre-stores OSA-related specific parameters for different population types.

[0079] Understandably, after inputting the extracted multi-dimensional features, the population type features of the individuals to be tested, and the pre-sleep activity features into the Bayesian model, the Bayesian model will analyze the population type features of the individuals to be tested. Based on these features, it will automatically call upon a pre-built population-specific feature parameter library to determine the prior probability of the corresponding population type of the individuals to be tested. Then, the Bayesian model can use the population-specific parameters to perform probabilistic inference on the multi-dimensional features and pre-sleep activity features, avoiding the neglect of physiological differences between different populations due to the use of general parameters, thereby improving the accuracy of subsequent OSA detection.

[0080] S202: Obtain the population type weight coefficient and activity weight coefficient corresponding to the population type characteristics and bedtime activity characteristics of the person to be tested.

[0081] In applications, algorithms such as logistic regression and random forest can be used to calculate the contribution of population type characteristics and pre-sleep activity characteristics to OSA based on historical sample data. For example, overweight (BMI ≥ 30) is much more associated with OSA than normal weight (BMI 18.5-23.9), and drinking alcohol before bed significantly increases the risk of OSA due to relaxation of airway muscles, with a much higher association than using electronic devices before bed. Therefore, the weight coefficients of corresponding features will be different, with higher contribution resulting in larger weight coefficients, ensuring that the impact priority of different population characteristics on OSA detection is accurately reflected.

[0082] For example, a weighting coefficient ω can be introduced when calculating the posterior probability. For instance, if an overweight patient has a BMI > 30, the initial weight of the "breathing amplitude ratio" feature is increased to 1.4 times; if alcohol is detected at dinner that day, the weight of the "nighttime blood oxygen fluctuation" feature is increased by 1.3 times in the nighttime detection.

[0083] S203: Adjust the weights of time-domain and frequency-domain features based on population type weight coefficients and activity weight coefficients, and combine them with the population-specific parameters to determine the posterior probability of whether the person being tested has obstructive sleep apnea.

[0084] The core logic of adjusting the weights of time-domain and frequency-domain features is to align the influence of physiological signal features with the individual attributes and short-term behaviors of the person being tested. Based on population type weight coefficients and activity weight coefficients, the weights of time-domain and frequency-domain features related to respiratory stability can be adjusted, increasing their weights. Then, pre-constructed population-specific parameters are called, and the adjusted weighted time-domain / frequency-domain features are compared with these parameters, using Bayes' theorem to calculate the posterior probability. Based on the prior probability of OSA for this population, the "likelihood" is calculated using the adjusted weighted physiological features, and then the influence of population / activity weights is incorporated to finally output the posterior probabilities for the two outcomes: "has OSA" and "does not have OSA."

[0085] S204: Determine whether the person being tested has obstructive sleep apnea based on the posterior probability of the two outcomes.

[0086] It is understandable that after determining the posterior probabilities of having and not having OSA, three scenarios can be considered: First, the posterior probability of having OSA is significantly higher than a preset threshold and much greater than the probability of not having OSA. For example, the probability of having OSA is 0.89, and the probability of not having it is 0.11. Since 0.89 > 0.5 (the standard threshold), it is directly determined that the individual has obstructive sleep apnea. Second, the posterior probability of not having OSA is higher than the threshold, and the value is dominant. For example, for a young woman of normal weight, the posterior probability of not having OSA is 0.72, and the posterior probability of having it is 0.28. Since 0.72 > 0.5, it can be determined that the individual does not have OSA. This classification process uses probability values ​​as the core basis, combined with population risk characteristics, to balance the risks of missed and false positives, ensuring the accuracy of the classification results.

[0087] In some embodiments, step S203 involves adjusting the weights of time-domain and frequency-domain features based on population type weight coefficients and activity weight coefficients, and combining this with the invoked population-specific parameters to determine the posterior probability of whether the person under test has obstructive sleep apnea. This includes: determining the calibration weights of each time-domain and frequency-domain feature based on the population type weight coefficients and activity weight coefficients; retrieving the feature conditional probability distribution of the corresponding population from the corresponding population-specific feature parameter library based on the population type of the person under test; and determining the posterior probability of whether the person under test has obstructive sleep apnea based on the extracted multi-dimensional features, the calibration weights of each time-domain feature and each frequency-domain feature, and the feature conditional probability distribution.

[0088] In the application, for each population group, the feature conditional probability distribution of its PSG samples can be stored separately; at the same time, the population type weight coefficient for that population group can also be stored. For example, in the parameter library for overweight patients, the conditional probability of abnormal respiratory amplitude ratio is stored, and the population type weight coefficient corresponding to this feature is determined. For example, the population type weight coefficient for overweight patients is 1.4; while the population type weight coefficient for normal adults is 1.0.

[0089] Let X be the feature set of the person to be tested (including multi-dimensional physiological characteristics, population type characteristics, and pre-sleep activity characteristics). Then, P(C1|X) represents the posterior probability that the person to be tested has OSA given the feature X, and P(C2|X) represents the posterior probability that the person to be tested does not have OSA given the feature X. These two probabilities are calculated by a pre-trained machine learning model, and their sum is 1.

[0090] Combining the posterior probability formula of Naive Bayes, the weight coefficient ω directly affects the feature conditional probability, adjusting its contribution. The modified formula is as follows (taking the determination of whether a sample is an OSA patient as an example):

[0091]

[0092] Representative population type This indicates that, given an OSA-positive result and belonging to the G population, the characteristics... The conditional probability of occurrence; Representation of features Weighting coefficients; This represents the product of (weights × conditional probabilities) of all features. The higher the weight of a feature, the greater its contribution to the final molecule, thus making the posterior probability more biased towards the true class.

[0093] For example, suppose an overweight patient (BMI=32) drank alcohol at dinner that day, and the feature vector at the time of detection... Call the overweight patient parameter library to obtain... , Determine the weighting coefficients: The inherent weight of the population is ω1=1.4. The dynamic weighting of alcohol consumption is ω2=1.3.

[0094] Calculation of molecules (core part):

[0095]

[0096] If no weights are applied, the numerator is only... It is evident that after the weight adjustment, the molecular values ​​of the more critical features for overweight drinkers were significantly increased, making the posterior probability more reflective of the true OSA risk.

[0097] In this embodiment, the prior probability of OSA in the population type corresponding to the person to be tested is first used as a basis. Then, the extracted multi-dimensional features are matched with the corresponding feature conditional probability distribution. By calibrating the weights, higher computational weights are given to highly correlated features. Finally, the posterior probabilities of having OSA and not having OSA are calculated, providing a quantitative basis for subsequent classification, so as to ultimately determine whether the person to be tested has OSA.

[0098] In some embodiments, step S102 involves performing quality assessments on each pressure signal and determining the target pressure signal based on the quality assessment results. This includes: calculating the energy and standard deviation of the sub-signals of each signal in the target frequency band to obtain the quality assessment results in the target frequency band; determining the fusion weight of each signal based on the frequency band quality assessment results; and performing weighted fusion of the sub-signals of each signal in the same frequency band based on the fusion weight of each signal to obtain the target pressure signal.

[0099] The target frequency band needs to be determined based on the sleep monitoring requirements, and could be, for example, the dominant respiratory frequency band (0.1-0.5Hz) and the dominant cardiovascular frequency band (1-5Hz). For each signal sub-signal within each target frequency band, the energy and standard deviation are calculated. Energy reflects the signal strength of the sub-signal in that frequency band; higher energy indicates clearer physiological signals in that band. For example, high energy in the respiratory frequency band indicates more pronounced pressure fluctuations caused by respiration. Standard deviation reflects the stability of the sub-signal within that frequency band; a smaller standard deviation indicates less signal interference. These two indicators provide the sub-band quality assessment results for each signal within each target frequency band.

[0100] Then, based on the frequency band quality assessment results, the fusion weights of each signal are determined. The core principle is to ensure that the higher-quality frequency bands have a greater weight in the fusion process. For example, in the respiratory frequency band, if channel A scores 85 points and channel B scores 60 points, then the fusion weight of channel A in this frequency band is set to 0.6 (85 / (85+60)) and that of channel B is set to 0.4. In the heart rate frequency band, if channel B scores 80 points and channel A scores 65 points, then the weight of channel B is set to 0.55 and that of channel A is set to 0.45, ensuring that the advantageous frequency bands of each signal can play a full role. Finally, after determining the fusion weights corresponding to each signal, the sub-signals of each signal in this frequency band are multiplied by their corresponding fusion weights and summed to obtain the fused sub-signals for that frequency band. Then, the fused sub-signals of all target frequency bands are integrated to finally form a target pressure signal that covers key physiological information, has a high signal-to-noise ratio, and is highly stable, which serves as the final target pressure signal.

[0101] In some other embodiments, step S102 involves performing quality assessments on each pressure signal and determining the target pressure signal based on the quality assessment results. This includes: performing a sliding window analysis of each signal for a preset time, calculating the signal energy and standard deviation within the window; determining the corresponding energy fluctuation coefficient based on the signal energy within the window for each signal; determining the optimal signal based on the energy fluctuation coefficient and standard deviation of each signal, and using the optimal signal as the target pressure signal.

[0102] It's understandable that performing a sliding window analysis on each signal over a preset time is to continuously capture the temporal changes of the signals. The preset time can be set to 20-40 seconds (e.g., 30 seconds), and the sliding step size can be set to 2-4 seconds. After dividing each pressure signal into multiple consecutive overlapping windows, the signal energy and standard deviation can be calculated for the signal data within each window. Then, based on the formula for calculating the energy fluctuation coefficient, the energy fluctuation coefficient corresponding to each signal is determined. The formula for calculating the energy fluctuation coefficient is: Energy fluctuation coefficient = Maximum energy difference within the window / Average energy within the window. Finally, based on the preset quality standards, the energy fluctuation coefficients and standard deviations of each signal, it is determined whether each signal meets the quality requirements. Signals that meet the quality standards are selected first. If multiple signals meet the quality requirements, the signal with the smaller energy fluctuation coefficient is selected from the signals that meet the quality standards.

[0103] For example, a 30-second sliding window analysis is performed on two chest signals to calculate the signal energy within the window:

[0104]

[0105] Calculate the standard deviation of the signal within the window:

[0106]

[0107] When the energy fluctuation coefficient of a signal and When the value is less than 0.2 (i.e., the quality standard example mentioned above), it is determined to be a high-quality signal; if both channels meet the requirement, the one with the smaller energy fluctuation coefficient is selected; if only one channel meets the requirement, that channel is selected; if neither meets the requirement, an early warning is issued and the best quality buffer segment within the previous 30 seconds is activated.

[0108] In some embodiments, such as Figure 3 As shown, in step S103, the target pressure signal is preprocessed, and the time domain and frequency domain features of the preprocessed target pressure signal are extracted, including the following steps S301 to S304.

[0109] S301: Low-frequency noise below a predetermined frequency is removed by adaptive Kalman filtering, and multiple frequency band signals are obtained by multi-layer wavelet decomposition. Among them, the multiple frequency band signals include respiratory main frequency band signals and cardiovascular main frequency band signals.

[0110] In applications, low-frequency noise below 0.1Hz, such as bed vibration, can be removed using adaptive Kalman filtering. Then, after five layers of wavelet decomposition, five frequency bands are obtained (0.1-0.5Hz, 0.5-1Hz, 1-2Hz, 2-5Hz, 5-10Hz), where the 0.1-0.5Hz band represents the dominant respiratory frequency band, and the 1-2Hz band represents the dominant cardiovascular frequency band.

[0111] S302: The respiratory main frequency band signal is detected by a peak detection algorithm to identify respiratory peaks. The respiratory threshold is determined based on the signal mean of the respiratory main frequency band signal. The respiratory frequency and respiratory cycle variation coefficient are determined based on the identified respiratory peaks and respiratory thresholds. The maximum amplitude, minimum amplitude and amplitude ratio of the respiratory signal are extracted from the respiratory main frequency band signal.

[0112] In application, the dominant respiratory frequency band signal can be smoothed first to reduce residual noise. Then, a local maximum criterion can be set, requiring a data point to be greater than 3-5 adjacent data points before and after it. Simultaneously, spurious peaks are eliminated by combining the overall signal strength, ultimately locating the highest point of pressure fluctuation in each respiratory cycle, i.e., the respiratory peak. Next, the respiratory threshold is determined based on the signal mean of the dominant respiratory frequency band signal. The signal mean over the entire analysis period can be calculated first; this mean reflects the baseline level of respiratory pressure fluctuations. Then, the respiratory threshold is set based on the signal mean, for example, it can be set to 1.5 times the signal mean to avoid misjudgment or missed judgment of peaks due to baseline drift.

[0113] Subsequently, respiratory rate and the coefficient of variation of the respiratory cycle are determined based on the identified respiratory peaks and thresholds. Respiratory rate calculation involves first statistically analyzing the time interval between two adjacent effective respiratory peaks, then converting the number of respiratory cycles within one minute into respiratory rates per minute. The coefficient of variation of the respiratory cycle is obtained by calculating the standard deviation of all respiratory cycles and dividing by the mean of the cycles; this coefficient reflects the stability of the respiratory rhythm. Finally, the maximum amplitude, minimum amplitude, and amplitude ratio of the respiratory signal are extracted from the dominant respiratory frequency signal. Amplitude is the pressure difference between the peak and trough within a single respiratory cycle; maximum amplitude is the maximum value among all amplitudes within the analysis period; minimum amplitude is the minimum value; and amplitude ratio is the ratio of the maximum amplitude to the minimum amplitude. These indicators reflect the fluctuation of respiratory intensity. It should be noted that during an OSA attack, the amplitude ratio is usually <0.3.

[0114] The formula for calculating the coefficient of variation of the respiratory cycle is as follows:

[0115] CV R =std(TR ) / mean(T R )

[0116] Among them, T R For a respiratory cycle, std(T) R The standard deviation of the respiratory cycle, mean(T) R () indicates the periodic mean.

[0117] S303: Perform wavelet denoising on the cardiovascular main frequency band signal to extract the heartbeat signal, and calculate the heart rate variability index based on the heartbeat signal and a preset time window.

[0118] In the cardiovascular feature extraction stage of obstructive sleep apnea (OSA) detection, wavelet denoising is first performed on the cardiovascular main frequency band signal obtained through multi-layer wavelet decomposition. The signal is first decomposed into high-frequency detail coefficients and low-frequency approximation coefficients at different scales. Soft thresholding is applied to the high-frequency detail coefficients, setting coefficients with absolute values ​​less than the threshold to zero and shrinking coefficients greater than the threshold, retaining effective coefficients related to heartbeat. The signal is then reconstructed using wavelet denoising, and finally, the periodic heartbeat signal is extracted from the denoised signal. Subsequently, heart rate variability (HRV) is calculated using a preset time window. The preset time window needs to balance the timeliness and stability of HRV analysis; in sleep monitoring, it is typically set to 5-10 minutes (e.g., 5 branches). The time interval between two adjacent heartbeat peaks, i.e., the RR interval, is first identified from the heartbeat signal, and all RR interval data within the preset time window are statistically analyzed. Then, time-domain indices such as SDNN are calculated based on these data. SDNN is the standard deviation of the RR interval, reflecting the overall variability of heart rate; OSA patients often show a decreased SDNN due to autonomic nervous system dysfunction.

[0119] S304: Select time period signals from the respiratory dominant frequency band signal and the cardiovascular dominant frequency band signal, perform fast Fourier transform, and calculate the energy proportion of the respiratory dominant frequency band and the energy proportion of the cardiovascular dominant frequency band respectively.

[0120] In applications, a continuous signal segment with no significant interference within 10-40 seconds can be selected; for example, a 30-second segment can be chosen. After selecting the signal segment, a Fast Fourier Transform (FFT) is performed on each segment. The FFT converts the pressure signal in the time domain into a frequency-energy spectrum in the frequency domain, visually presenting the signal energy distribution corresponding to different frequency components, where the horizontal axis represents frequency and the vertical axis represents energy value. FFT can accurately capture the energy contribution of each sub-frequency within the respiratory and cardiovascular main frequency bands. For the respiratory main frequency band, first locate the frequency range of 0.1-0.5Hz in the frequency domain spectrum, calculate the sum of the energy of all frequency points within this range, then calculate the total energy of the entire frequency band, and finally obtain the energy percentage of the respiratory main frequency band by "total energy of the respiratory main frequency band ÷ total energy of the entire frequency band × 100%". The calculation logic for the energy percentage of the cardiovascular main frequency band is the same; simply replace the frequency range with 1-2Hz and calculate the ratio of the sum of the energy in this range to the total energy of the entire frequency band.

[0121] For example, the formula for calculating the energy proportion of the dominant respiratory frequency band (0.1-0.5Hz) after FFT transformation of a 30-second signal (8192 points) is as follows:

[0122] P R =E 0.1-0.5HZ / E total

[0123] Among them, P R E represents the energy percentage of the dominant respiratory frequency band. 0.1-0.5HZ E represents the total energy of the dominant respiratory frequency band. total This represents the total energy across the entire frequency band. The calculation method for the energy proportion of the cardiovascular main frequency band, PH, is similar. Among them, under OSA conditions, PR decreases by >30%, and PH fluctuation amplitude increases by >20%.

[0124] Based on the same inventive concept, such as Figure 4 As shown, this application also provides an obstructive sleep apnea detection system 400, including: a processing module 402 and at least two bedside pressure sensors 401.

[0125] Each bed-mounted pressure sensor 401 is adapted to be installed on the mattress at the position corresponding to the human chest. The processing module 402 is connected to each bed-mounted pressure sensor 401 and is used to execute any of the above-described obstructive sleep apnea detection methods.

[0126] The obstructive sleep apnea detection system 400 described above, after installing the bed-mounted pressure sensors 401 on the mattress at the position corresponding to the human chest, can detect pressure signals detected by at least two of the bed-mounted pressure sensors 401 during human sleep. After receiving the pressure signals, the processing module 402 can perform quality assessments on each pressure signal and determine the target pressure signal based on the quality assessment results, thereby obtaining a target pressure signal of higher quality. After the processing module 402 acquires the target pressure signal, it can preprocess the signal and then extract features in the time and frequency domains to accurately extract multi-dimensional features related to OSA. Based on this, considering the influence of the individual's population type and pre-sleep activity characteristics, the processing module 402 inputs the extracted multi-dimensional features, the individual's population type characteristics, and pre-sleep activity characteristics into a pre-trained machine learning model. This improves diagnostic accuracy through multi-dimensional physiological features. Simultaneously, it leverages population type characteristics to allow the model to overcome the limitations of a single population standard, optimizing the judgment logic for high-risk groups and individuals' pre-sleep activities, enhancing personalized adaptation capabilities, and further improving the classification accuracy of OSA. Moreover, the obstructive sleep apnea detection system 400 of this application does not rely on complex and expensive PSG equipment. It deploys at least two bed-mounted pressure sensors 401 on the mattress, collecting signals through these sensors, making it suitable for primary healthcare and home screening scenarios, and highly convenient to use.

[0127] In some embodiments, the processing module 402 includes a signal conditioning module, a low-pass filter, a high-pass filter, a data acquisition module, and a data processing module;

[0128] The signal conditioning module receives pressure signals from each bed-mounted pressure sensor 401 and amplifies them to obtain an amplified pressure signal. A low-pass filter removes high-frequency noise from the amplified pressure signal. A high-pass filter filters the signal output from the low-pass filter to suppress baseline drift. The data acquisition module performs analog-to-digital conversion on the signal output from the high-pass filter to generate a digital signal. The data processing module processes the digital signal output from the data acquisition module and executes any of the above-described obstructive sleep apnea detection methods.

[0129] The signal conditioning module uses a gain-adjustable instrumentation amplifier to amplify the two signals to 2-5V. A 5th-order Butterworth low-pass filter can be used. The data acquisition module can perform A / D conversion at a 500Hz sampling rate to generate a 16-bit digital signal.

[0130] It is understandable that the pressure signal directly output by the 401 bed-mounted pressure sensor is typically in the millivolt range due to limitations in its detection principle. This signal is easily masked by environmental noise during transmission. Therefore, the instrumentation amplifier built into the signal conditioning module is needed to precisely amplify each raw signal while suppressing the amplifier's own noise interference. This ensures that the amplified pressure signal clearly retains the subtle pressure fluctuations caused by respiration and heartbeat, laying the foundation for subsequent filtering. Next, a low-pass filter removes high-frequency noise mixed into the amplified pressure signal. Its cutoff frequency can be set to 30Hz, allowing only useful physiological signals below this frequency to pass through, blocking high-frequency noise from interfering with the signal baseline. Subsequently, a high-pass filter focuses on solving the "baseline drift" problem. The signal after low-pass filtering may experience low-frequency drift due to slow changes in sensor temperature, prolonged slight deformation of the bed, or slow turning of the patient. The module filters out these low-frequency drift components by setting a cutoff frequency of approximately 0.05Hz, stabilizing the signal baseline and ensuring that the subsequently extracted respiratory and heartbeat characteristics are not affected by baseline fluctuations. Next, the data acquisition module samples and quantizes the high-pass filtered analog signal using its built-in analog-to-digital converter (ADC), ultimately generating a transmittable and computable digital signal. Finally, the data processing module, as the core of signal processing, receives the digital signal output from the data acquisition module and executes the key operations from the previous process, realizing the transformation from the raw stress signal to the sleep health assessment result.

[0131] It should be noted that the obstructive sleep apnea detection system 400 provided in this application embodiment and the obstructive sleep apnea detection method provided in this application embodiment are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned obstructive sleep apnea detection method, and the repeated parts will not be described again.

[0132] In some embodiments, an electronic device provided in this application includes a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the above-described obstructive sleep apnea detection method.

[0133] Specifically, the processor may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor may also include onboard memory for caching purposes. The processor may be a single processing unit or multiple processing units for performing different actions of the method flow according to embodiments of this application.

[0134] Memory can be any medium capable of containing, storing, transmitting, propagating, or transmitting instructions. For example, memory can include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, instruments, or propagation media. Specific examples of memory include: magnetic storage devices such as magnetic tape or hard disk drives (HDDs); optical storage devices such as optical discs (CD-ROMs); and also random access memory (RAM) or flash memory; and / or wired / wireless communication links.

[0135] This application also provides a computer-readable medium storing a computer program thereon, which, when executed by a processor, implements the obstructive sleep apnea detection method described above. This computer-readable medium may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into that device / apparatus / system. The aforementioned computer-readable medium carries one or more programs, which, when executed, implement the method as described in the embodiments of this application.

[0136] According to embodiments of this application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wired, optical fiber, radio frequency signals, etc., or any suitable combination thereof.

[0137] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application. Therefore, the scope of this application should not be limited to the above embodiments. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method of detecting obstructive sleep apnea, comprising: The method comprises the following steps: obtaining pressure signals detected by at least two bed-mounted pressure sensors in a sleep state of a human body, and obtaining a crowd type feature and a pre-sleep activity feature of a person to be detected; respectively performing quality assessment on each pressure signal, and determining a target pressure signal based on the quality assessment result; performing preprocessing on the target pressure signal, and performing feature extraction on the preprocessed target pressure signal in a time domain and a frequency domain; inputting the extracted multi-dimensional features, the crowd type feature and the pre-sleep activity feature of the person to be detected into a pre-trained machine learning model to obtain a classification result of whether the person to be detected has obstructive sleep apnea.

2. The obstructive sleep apnea detection method of claim 1, wherein, The machine learning model is a Bayesian model; the inputting of the extracted multi-dimensional features, the crowd type feature and the pre-sleep activity feature of the person to be detected into the pre-trained machine learning model to obtain the classification result of whether the person to be detected has obstructive sleep apnea comprises: inputting the extracted multi-dimensional features, the crowd type feature and the pre-sleep activity feature of the person to be detected into the Bayesian model, so that the Bayesian model calls a crowd-specific feature parameter library according to the crowd type feature of the person to be detected; obtaining a crowd type weight coefficient and an activity weight coefficient corresponding to the crowd type feature and the pre-sleep activity feature of the person to be detected respectively; adjusting the weights of the time domain features and the frequency domain features based on the crowd type weight coefficient and the activity weight coefficient, and determining the posterior probability of two results of whether the person to be detected has obstructive sleep apnea by combining the called crowd-specific parameters; determining the classification result of whether the person to be detected has obstructive sleep apnea according to the posterior probability of the two results.

3. The obstructive sleep apnea detection method of claim 2, wherein, The adjusting of the weights of the time domain features and the frequency domain features based on the crowd type weight coefficient and the activity weight coefficient, and the determination of the posterior probability of the two results of whether the person to be detected has obstructive sleep apnea by combining the called crowd-specific parameters comprise: determining the calibration weights of each time domain feature and each frequency domain feature according to the crowd type weight coefficient and the activity weight coefficient; calling the feature conditional probability distribution of the corresponding crowd from the corresponding crowd-specific feature parameter library according to the crowd type of the person to be detected; determining the posterior probability of the two results of whether the person to be detected has obstructive sleep apnea according to the extracted multi-dimensional features, the calibration weights of each time domain feature and each frequency domain feature, and the feature conditional probability distribution.

4. The obstructive sleep apnea detection method of claim 1, wherein, The quality assessment on each pressure signal and the determination of the target pressure signal based on the quality assessment result comprise: calculating the energy and the standard deviation of a sub-signal in a target frequency band of each signal to obtain a quality assessment result in the target frequency band; determining a fusion weight of each signal according to the quality assessment result in the frequency band, and performing weighted fusion on a same frequency band sub-signal of each signal based on the fusion weight of each signal to obtain the target pressure signal.

5. The obstructive sleep apnea detection method of claim 1, wherein, The quality assessment on each pressure signal and the determination of the target pressure signal based on the quality assessment result comprise: respectively performing sliding window analysis on each signal in a preset time to calculate the signal energy and the standard deviation in the window; determining an energy fluctuation coefficient corresponding to each signal based on the signal energy in the window of each signal; Determine the optimal signal based on the energy fluctuation coefficient and standard deviation of each signal, and take the optimal signal as the target pressure signal.

6. The obstructive sleep apnea detection method of claim 1, wherein, The target pressure signal is preprocessed, and the preprocessed target pressure signal is subjected to feature extraction in time domain and frequency domain, comprising: Remove low-frequency noise below a predetermined frequency through adaptive Kalman filtering, and obtain a plurality of frequency band signals through multi-layer wavelet decomposition, wherein the plurality of frequency band signals include a respiratory main frequency band signal and a cardiovascular main frequency band signal; Detect the respiratory main frequency band signal through a peak detection algorithm to identify respiratory wave peaks, determine a respiratory threshold based on the signal mean of the respiratory main frequency band signal, determine a respiratory frequency and a respiratory cycle variation coefficient according to the identified respiratory wave peaks and the respiratory threshold, and extract a respiratory signal maximum amplitude, a minimum amplitude and an amplitude ratio from the respiratory main frequency band signal; Wavelet denoise the cardiovascular main frequency band signal to extract a heartbeat signal, and calculate a heart rate variability index according to the heartbeat signal and a preset time window; Perform fast Fourier transform on the time period signals selected from the respiratory main frequency band signal and the cardiovascular main frequency band signal, and calculate an energy proportion of the respiratory main frequency band and an energy proportion of the cardiovascular main frequency band, respectively.

7. A system for detecting obstructive sleep apnea, comprising: Comprise: a processing module and at least two bed-mounted pressure sensors; Each bed-mounted pressure sensor is adapted to be mounted at a position corresponding to the chest of a human body on a mattress; The processing module is connected with each bed-mounted pressure sensor, and is used to execute the obstructive sleep apnea detection method according to any one of claims 1 to 6.

8. The obstructive sleep apnea detection system of claim 7, wherein, The processing module comprises a signal conditioning module, a low-pass filter, a high-pass filter, a data acquisition module and a data processing module; The signal conditioning module is used to receive pressure signals output by each bed-mounted pressure sensor, and amplify each pressure signal to obtain an amplified pressure signal; The low-pass filter is used to filter high-frequency noise of the amplified pressure signal; The high-pass filter is used to filter the signal output by the low-pass filter to suppress baseline drift; The data acquisition module is used to perform analog-to-digital conversion on the signal output by the high-pass filter to generate a digital signal; The data processing module is used to process the digital signal output by the data acquisition module, and execute the obstructive sleep apnea detection method according to any one of claims 1 to 6.

9. An electronic device, comprising: Comprise a processor and a memory; the memory has a computer program stored therein, wherein the computer program, when executed by the processor, implements the obstructive sleep apnea detection method according to any one of claims 1 to 6.

10. A computer storage medium, characterized in that A computer program is stored thereon, wherein the computer program, when executed by a processor, implements the obstructive sleep apnea detection method according to any one of claims 1 to 6.

Images