System and device for identifying obstructive sleep apnea disease based on lead electroencephalogram signal

Abstract

The invention provides an obstructive sleep apnea disease recognition system and device based on lead electroencephalogram signals. The system and device can be applied to the technical field of biomedicine and the technical field of artificial intelligence. The recognition system comprises a signal acquisition device used for acquiring a first electric signal in a first target area of the left brain of a user and acquiring a second electric signal in a second target area of the right brain of the user to obtain a lead electroencephalogram signal set; the processor is used for extracting a plurality of initial brain function connection features from the lead electroencephalogram signal set, and updating the plurality of initial brain function connection features according to feature reference values of the initial brain function connection features to obtain a plurality of target brain function connection features; inputting the multiple target brain function connection features into a disease recognition model to obtain a recognition result; and under the condition that the identification result represents the obstructive sleep apnea disease, identifying the disease degree of the obstructive sleep apnea disease according to the Pearson's correlation coefficient and the phase locking value.

Description

Technical Field

[0001] This disclosure relates to the fields of biomedical technology and artificial intelligence technology, and more specifically to a system and device for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals. Background Technology

[0002] Obstructive sleep apnea (OSA) is a common sleep-disordered breathing condition characterized by recurrent upper airway collapse during sleep, leading to apnea, decreased blood oxygen saturation, and disruption of sleep structure. The relevant technology relies on polysomnography (PSG), which includes signals from multiple channels such as electroencephalography (EEG), electrooculography (EOG), electromyography (EMG), electrocardiography (ECG), respiration, and blood oxygenation. However, the equipment is expensive, the procedure is complex, the operation is cumbersome, and it requires specialized personnel, limiting its application in primary healthcare institutions and home screening. Summary of the Invention

[0003] In view of the above problems, this disclosure provides a system and device for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals.

[0004] According to a first aspect of this disclosure, a system for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals is provided, comprising: a signal acquisition device for acquiring a first electrical signal in a first target region of the user's left brain and a second electrical signal in a second target region of the user's right brain, thereby obtaining a set of leaded EEG signals for the user during sleep; and a processor for performing the following operations: extracting multiple initial brain functional connectivity features from the set of leaded EEG signals, wherein the multiple initial brain functional connectivity features characterize the spatiotemporal synchronicity between the brain activity states of the user's left brain and the brain activity states of the user's right brain; updating the multiple initial brain functional connectivity features based on their respective feature reference values ​​to obtain multiple target brain functional connectivity features; inputting the multiple target brain functional connectivity features into a disease identification model to obtain an identification result; and, if the identification result indicates obstructive sleep apnea, identifying the severity of obstructive sleep apnea based on at least one of the Pearson correlation coefficient and phase lock value included in the multiple initial brain functional connectivity features.

[0005] According to embodiments of this disclosure, the feature benchmark values ​​include a normal upper limit and a normal lower limit; based on the feature benchmark values ​​of each of the multiple initial brain functional connectivity features, the multiple initial brain functional connectivity features are updated to obtain multiple target brain functional connectivity features, including: identifying connectivity sub-features in the initial brain functional connectivity features that are greater than the normal upper limit or less than the normal lower limit as outliers; updating the outliers based on their types to obtain intermediate brain functional connectivity features; and performing dimensionality reduction on the standardized intermediate brain functional connectivity features using a principal component analysis algorithm to obtain the target brain functional connectivity features.

[0006] According to embodiments of this disclosure, the upper and lower normal limits are determined as follows: the connective features in the initial brain functional connectivity features are sorted in ascending order of value to obtain a numerical sequence; the interquartile range is determined based on the difference between the first sequence sub-feature located at the first position and the second sequence sub-feature located at the second position in the numerical sequence, where the interquartile range characterizes the degree of dispersion between the first and second sequence sub-features; the lower normal limit is determined based on the first sequence sub-feature and the interquartile range; and the upper normal limit is determined based on the second sequence sub-feature and the interquartile range.

[0007] According to embodiments of this disclosure, updating outliers based on their type to obtain intermediate brain functional connectivity features includes: when the outlier type is a missing value, updating the missing value to a median connectivity sub-feature of the brain functional connectivity features to obtain intermediate brain functional connectivity features; wherein, the median connectivity sub-feature is the median of multiple connectivity sub-features arranged according to time information in the initial brain functional connectivity features.

[0008] According to embodiments of this disclosure, the various initial brain functional connectivity features also include at least one of the following: cross-correlation coefficient, wavelet coherence, spectral coherence, phase and amplitude coupling index, synchronization likelihood index, and mutual information.

[0009] According to embodiments of this disclosure, the disease identification model is a gradient boosting tree model, which is trained in the following manner: obtaining a training set of leaded EEG signals, the training set of leaded EEG signals containing multiple samples, the samples including various brain functional connectivity training features and label values, the label values ​​being the apnea-hypopnea index corresponding to the various brain functional connectivity training features; training an initial model based on the training set of leaded EEG signals to obtain the gradient boosting tree model.

[0010] According to embodiments of this disclosure, training an initial model based on a lead-based EEG signal training set to obtain a gradient boosting tree model includes: for each sample, inputting multiple brain function connectivity training features into the trained initial classification model to obtain multiple initial predicted values; determining an initial negative gradient based on the initial predicted values ​​and the loss function value between the initial predicted values ​​and the true values, and using the initial negative gradient as a new label; training a first classification model based on the lead-based EEG signal training set and the new label to obtain a decision tree, the decision tree including multiple leaf nodes, each leaf node corresponding to a subset of samples in the lead-based EEG signal training set; using a preset learning rate as the weight of the decision tree, and performing a weighted summation of the decision prediction values ​​of the decision tree and the initial prediction values ​​of the initial classification model based on the weights to obtain an updated prediction value of the updated model; using the updated model as the initial classification model, repeating the above steps until the target number of iterations is reached to obtain the gradient boosting tree model.

[0011] According to embodiments of this disclosure, the severity of obstructive sleep apnea is identified based on at least one of a variety of initial brain functional connectivity features, including Pearson correlation coefficient and phase-locked value, comprising: determining a target apnea-hypopnea index corresponding to the Pearson correlation coefficient based on a preset mapping relationship between a preset Pearson correlation coefficient and a preset apnea-hypopnea index; and identifying the severity of obstructive sleep apnea based on the target apnea-hypopnea index.

[0012] According to embodiments of this disclosure, identifying the severity of obstructive sleep apnea based on a target apnea-hypopnea index includes: determining the severity as mild when the target apnea-hypopnea index is in a first numerical range; determining the severity as moderate when the target apnea-hypopnea index is in a second numerical range, wherein the value in the first numerical range is less than the value in the second numerical range; and determining the severity as severe when the target apnea-hypopnea index is in the second numerical range, wherein the value in the second numerical range is less than the value in a third numerical range.

[0013] The second aspect of this disclosure provides a device for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals, comprising: an extraction module for extracting multiple initial brain functional connectivity features from a set of leaded EEG signals for a user during sleep, the set of leaded EEG signals including a first electrical signal acquired in a first target region of the user's left brain and a second electrical signal acquired in a second target region of the user's right brain, the multiple brain functional connectivity features representing the spatiotemporal synchronicity between the brain activity states of the user's left brain and the brain activity states of the user's right brain; a processing module for updating the multiple initial brain functional connectivity features based on their respective feature reference values ​​to obtain multiple target brain functional connectivity features; an input module for inputting the multiple target brain functional connectivity features into a disease identification model to obtain an identification result; and an identification module for identifying the severity of obstructive sleep apnea based on at least one of the Pearson correlation coefficient and phase lock value included in the multiple initial brain functional connectivity features when the identification result indicates obstructive sleep apnea.

[0014] According to embodiments of this disclosure, a first electrical signal is acquired in a first target region of the user's left brain and a second electrical signal is acquired in a second target region of the user's right brain via a signal acquisition device, resulting in a set of leaded EEG signals for the user during sleep. Acquiring only the first and second electrical signals simplifies the data acquisition process, improves the system's computational performance and convenience, and can be applied in primary healthcare institutions and home screening. Simultaneously, the processor of the obstructive sleep apnea disease identification system based on leaded EEG signals is also used to extract multiple initial brain functional connectivity features from the leaded EEG signal set. Based on the feature benchmark values ​​of each of the multiple initial brain functional connectivity features, the features are updated to obtain multiple target brain functional connectivity features, thereby improving the quality of the feature data input to the disease identification model. Furthermore, since the multiple initial brain functional connectivity features characterize the spatiotemporal synchronization between the brain activity states of the user's left and right brains, using the target brain functional connectivity features for disease identification can improve the accuracy of the identification results. When the identification result is characterized as obstructive sleep apnea disease, the severity of obstructive sleep apnea disease can be identified based on at least one of multiple initial brain functional connectivity features, including Pearson correlation coefficient and phase lock value, thereby improving the accuracy of obstructive sleep apnea disease severity identification. Attached Figure Description

[0015] The foregoing contents, as well as other objects, features, and advantages of this disclosure, will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0016] Figure 1A schematic diagram illustrates the structure of a system for identifying obstructive sleep apnea based on leaded electroencephalogram signals according to an embodiment of the present disclosure;

[0017] Figure 2 The illustration shows a schematic diagram of the identification technology process for obstructive sleep apnea based on leaded electroencephalogram signals according to an embodiment of the present disclosure;

[0018] Figure 3 The diagram illustrates the classification performance of a system for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals according to an embodiment of the present disclosure.

[0019] Figure 4 The diagram illustrates a binary confusion matrix of a system for identifying obstructive sleep apnea based on lead-based electroencephalogram signals according to an embodiment of the present disclosure.

[0020] Figure 5 The diagram illustrates a four-class confusion matrix of a system for identifying obstructive sleep apnea based on lead-based electroencephalogram signals according to an embodiment of the present disclosure.

[0021] Figure 6 A flowchart illustrating a method for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals according to an embodiment of the present disclosure is shown schematically.

[0022] Figure 7 A schematic diagram illustrates the structure of a device for identifying obstructive sleep apnea based on leaded electroencephalogram signals according to an embodiment of the present disclosure;

[0023] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing a method for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals, according to an embodiment of the present disclosure. Detailed Implementation

[0024] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0025] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0026] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0027] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0028] In the technical solution disclosed herein, the user information (including but not limited to user personal information, user image information, user device information, such as EEG signals) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation entry points are provided for users to choose to authorize or refuse.

[0029] In scenarios involving automated decision-making using personal information, the methods, devices, and systems provided in this disclosure all offer users corresponding entry points for choosing to agree to or reject the automated decision-making results. If the user chooses to reject, the process proceeds to the expert decision-making stage. Here, "automated decision-making" refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests, or economic, health, and credit status through computer programs, and then making a decision. Here, "expert decision-making" refers to the activity of making decisions by personnel who specialize in a particular field, possess specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0030] OSA (Obstructive Sleep Apnea) is a common sleep-disordered breathing phenomenon characterized by recurrent upper airway collapse during sleep, leading to apnea, decreased blood oxygen saturation, and disruption of sleep structure. The relevant technology relies on polysomnography, which includes signals from multiple channels such as electroencephalography (EEG), electrooculography (EOG), electromyography (EMG), electrocardiography (ECG), respiration, and blood oxygenation. This method is expensive, complex, cumbersome, and requires specialized personnel, limiting its application in primary healthcare institutions and home screening.

[0031] Electroencephalogram (EEG) signals can reflect the perturbation characteristics of neural activity caused by obstructive sleep apnea (OSA), including changes in frequency structure, nonlinear complexity, and functional connectivity. If OSA could be identified using only EEG leads, it would greatly simplify the screening system and improve ease of use.

[0032] In view of this, embodiments of the present disclosure provide a system for identifying obstructive sleep apnea based on leaded EEG signals, comprising: a signal acquisition device for acquiring a first electrical signal in a first target region of the user's left brain and a second electrical signal in a second target region of the user's right brain, thereby obtaining a set of leaded EEG signals for the user during sleep; and a processor for performing the following operations: extracting multiple initial brain functional connectivity features from the set of leaded EEG signals, wherein the multiple brain functional connectivity features characterize the spatiotemporal synchronization between the brain activity states of the user's left brain and the brain activity states of the user's right brain; updating the multiple initial brain functional connectivity features based on their respective feature reference values ​​to obtain multiple target brain functional connectivity features; inputting the multiple target brain functional connectivity features into a disease identification model to obtain an identification result; and, if the identification result indicates obstructive sleep apnea, identifying the severity of obstructive sleep apnea based on at least one of the Pearson correlation coefficient and phase lock value included in the multiple initial brain functional connectivity features.

[0033] Figure 1 A schematic block diagram of a system for identifying obstructive sleep apnea based on leaded electroencephalogram signals according to an embodiment of the present disclosure is shown.

[0034] like Figure 1 As shown, the obstructive sleep apnea identification system 100 based on leaded EEG signals in this embodiment includes a signal acquisition device 110 and a processor 120.

[0035] The signal acquisition device 110 is used to acquire a first electrical signal in a first target area of ​​the user's left brain and a second electrical signal in a second target area of ​​the user's right brain, thereby obtaining a set of leaded electroencephalogram (EEG) signals for the user during sleep.

[0036] The processor 120 is configured to perform the following operations: extract multiple initial brain functional connectivity features from the set of leaded EEG signals, wherein the multiple initial brain functional connectivity features characterize the spatiotemporal synchronicity between the brain activity state of the user's left brain and the brain activity state of the user's right brain; update the multiple initial brain functional connectivity features based on the feature benchmark values ​​of each of the multiple initial brain functional connectivity features to obtain multiple target brain functional connectivity features; input the multiple target brain functional connectivity features into a disease identification model to obtain identification results; and, if the identification results indicate obstructive sleep apnea disease, identify the severity of obstructive sleep apnea disease based on at least one of the Pearson correlation coefficient and phase lock value included in the multiple initial brain functional connectivity features.

[0037] According to embodiments of this disclosure, the signal acquisition device 110 may be a head-mounted portable EEG (Electroencephalography) device, which acquires and records the activity of brain neuron populations by placing electrodes on the scalp to obtain brain electrical signals.

[0038] According to embodiments of this disclosure, the first target region of the user's left brain may be the central region C3 of the left hemisphere. The first electrical signal may be an EEG signal in lead C3.

[0039] According to embodiments of this disclosure, the second target region of the user's right brain may be the central region C4 of the right brain. The second electrical signal may be an EEG signal in lead C4.

[0040] According to embodiments of this disclosure, the lead EEG signal set includes multiple lead EEG signal subsets ordered by time, and each lead EEG signal subset includes a first electrical signal and a second electrical signal at that time.

[0041] In embodiments of this disclosure, the user's consent or authorization may be obtained before acquiring the set of leaded EEG signals during sleep.

[0042] According to an embodiment of this disclosure, the processor 120 is communicatively connected to the signal acquisition device 110 to acquire a set of leaded electroencephalogram (EEG) signals from the signal acquisition device 110.

[0043] According to embodiments of this disclosure, feature benchmark values ​​are used to identify outliers in initial brain functional connectivity features.

[0044] For example, anomaly detection can be performed on the initial brain functional connectivity features using feature benchmark values ​​to obtain abnormal sub-features in the initial brain functional connectivity features. These abnormal sub-features can then be processed by removing or filling them to obtain the target brain functional connectivity features.

[0045] According to embodiments of this disclosure, the disease identification model can be a machine learning algorithm, such as a neural network model, a random forest algorithm, a support vector machine algorithm, etc.

[0046] According to embodiments of this disclosure, the identification result can be a binary classification result or a multi-class classification result. For example, a binary classification result can include two types: "obstructive sleep apnea" and "healthy". A multi-class classification result can include multiple types such as "obstructive sleep apnea", "normal", "central sleep disorder" and "sleep-related motor disorder".

[0047] According to embodiments of this disclosure, the Pearson correlation coefficient (PCC) is used to quantify the linear synchronicity of the first and second electrical signals in the time domain. The magnitude of the Pearson correlation coefficient shows a significant negative correlation with the severity of obstructive sleep apnea; that is, the higher the severity of obstructive sleep apnea, the smaller the Pearson correlation coefficient between the first and second electrical signals.

[0048] According to embodiments of this disclosure, the phase locking value (PLV) is an indicator that quantifies the degree of phase synchronization between a first electrical signal and a second electrical signal within a specific frequency band. The magnitude of the phase locking value shows a significant negative correlation with the severity of obstructive sleep apnea; that is, the higher the severity of obstructive sleep apnea, the smaller the phase locking value between the first and second electrical signals.

[0049] According to embodiments of this disclosure, a first electrical signal is acquired in a first target region of the user's left brain and a second electrical signal is acquired in a second target region of the user's right brain via a signal acquisition device, resulting in a set of leaded EEG signals for the user during sleep. Acquiring only the first and second electrical signals simplifies the data acquisition process, improves the system's computational performance and convenience, and can be applied in primary healthcare institutions and home screening. Simultaneously, the processor of the obstructive sleep apnea disease identification system based on leaded EEG signals is also used to extract multiple initial brain functional connectivity features from the leaded EEG signal set. Based on the feature benchmark values ​​of each of the multiple initial brain functional connectivity features, the features are updated to obtain multiple target brain functional connectivity features, thereby improving the quality of the feature data input to the disease identification model. Furthermore, since the multiple initial brain functional connectivity features characterize the spatiotemporal synchronization between the brain activity states of the user's left and right brains, using target brain functional connectivity features for disease identification can improve the accuracy of the initial identification results. When the identification result is characterized as obstructive sleep apnea disease, the severity of obstructive sleep apnea disease can be identified based on at least one of multiple initial brain functional connectivity features, including Pearson correlation coefficient and phase lock value, thereby improving the accuracy of obstructive sleep apnea disease severity identification.

[0050] According to embodiments of this disclosure, the various initial brain functional connectivity features also include at least one of the following: cross-correlation coefficient, wavelet coherence, spectral coherence, phase and amplitude coupling index, synchronization likelihood index, and mutual information.

[0051] According to embodiments of this disclosure, the cross-correlation function (XCOR) characterizes the degree of similarity between the first electrical signal and the second electrical signal under different time delays.

[0052] According to embodiments of this disclosure, wavelet coherence (WCOH) is used to characterize the linear correlation strength between a first electrical signal and a second electrical signal at a specific time point and a specific frequency band.

[0053] According to embodiments of this disclosure, Spectral Coherence (SCOH) characterizes whether the amplitude and phase changes of the first electrical signal and the second electrical signal are synchronized at different frequencies.

[0054] According to embodiments of this disclosure, the Phase-Amplitude Coupling Index (PAC), also known as the RHO index, is used to characterize the degree of correlation between the phase of a first electrical signal and the amplitude of a second electrical signal.

[0055] According to embodiments of this disclosure, the synchronization likelihood index characterizes the degree of synchronization between the first electrical signal and the second electrical signal in the state space.

[0056] According to embodiments of this disclosure, mutual information represents the amount of information contained in the first electrical signal regarding the second electrical signal.

[0057] According to embodiments of this disclosure, various initial brain functional connectivity features may include eight functional connectivity features: Pearson correlation coefficient, phase lock value, cross-correlation coefficient, wavelet coherence, spectral coherence, phase and amplitude coupling index, synchronization likelihood index, and mutual information. These functional connectivity features can reflect the spatiotemporal synchronicity and coupling degree between different brain regions and can also serve as important neuroelectrophysiological information on OSA brain functional impairment. Therefore, the accuracy of disease identification can be improved based on these eight functional connectivity features.

[0058] According to embodiments of this disclosure, the feature benchmark values ​​include a normal upper limit and a normal lower limit; based on the feature benchmark values ​​of each of the multiple initial brain functional connectivity features, the multiple initial brain functional connectivity features are updated to obtain multiple target brain functional connectivity features, including: identifying connectivity sub-features in the initial brain functional connectivity features that are greater than the normal upper limit or less than the normal lower limit as outliers; updating the outliers based on their types to obtain intermediate brain functional connectivity features; and performing dimensionality reduction on the standardized intermediate brain functional connectivity features using a principal component analysis algorithm to obtain the target brain functional connectivity features.

[0059] According to embodiments of this disclosure, connectivity features in the initial brain functional connectivity features that are greater than the lower limit of normal and less than the upper limit of normal are determined to be normal values; connectivity features in the initial brain functional connectivity features that are greater than the upper limit of normal or less than the lower limit of normal are determined to be abnormal values.

[0060] According to embodiments of this disclosure, the types of outliers may include missing values, noise values, etc.

[0061] According to embodiments of this disclosure, z-score standardization is performed on the mesobrain functional connectivity features to unify the dimensions of different features. Principal component analysis (PCA) can be used to reduce the dimensionality of the standardized mesobrain functional connectivity features, retaining the information of the main contributors to reduce redundancy and improve the computational efficiency of the disease identification model.

[0062] Since different features may have different dimensions and value ranges, directly inputting the raw feature dataset into a machine learning model may cause the model to become overly sensitive to individual features, thus affecting classification performance. Therefore, the mesobrain functional connectivity features are standardized, scaling them to a distribution with a mean of 0 and a standard deviation of 1. Standardization not only improves the model's convergence speed but is also particularly important for some distance-based algorithms (such as support vector machines), as these algorithms are sensitive to feature scale. The calculation formula is as follows:

[0063] (1)

[0064] Among them, X ij It is the mesobrain functional connectivity sub-feature in the i-th row and j-th column of the mesobrain functional connectivity features. and These are the mean and standard deviation of the feature in column x, respectively. These are standardized features of the mesobrain functional connectives.

[0065] Although the mesobrain connectivity features were standardized, the feature matrix may still suffer from excessive dimensionality. Principal Component Analysis (PCA) was used to reduce the dimensionality of the standardized mesobrain functional connectivity features. PCA maps the original variables to a set of orthogonal principal components through a linear transformation, thereby retaining most of the data information while reducing the redundancy caused by dimensionality. The retention threshold for PCA was set to cover 95% of the total variance information of the original data, i.e.:

[0066] (2)

[0067] in, Let be the j-th eigenvalue of the covariance matrix of the standardized mesobrain functional connectivity features, and m be the total number of eigenvalues ​​in the covariance matrix of the standardized mesobrain functional connectivity features. The k-th eigenvalue is the covariance matrix of the standardized mesobrain functional connectivity features, arranged in descending order. p represents the number of principal components selected (the first p largest eigenvalues ​​are chosen for principal component analysis). After dimensionality reduction, the number of features is significantly reduced while retaining 95% of the data information. This method effectively reduces computational complexity and avoids the risk of overfitting due to excessive dimensionality, making it well-suited for processing high-dimensional feature data.

[0068] According to embodiments of this disclosure, the upper and lower normal limits are determined as follows: the connective features in the initial brain functional connectivity features are sorted in ascending order of value to obtain a numerical sequence; the interquartile range is determined based on the difference between the first sequence sub-feature located at the first position and the second sequence sub-feature located at the second position in the numerical sequence, where the interquartile range characterizes the degree of dispersion between the first and second sequence sub-features; the lower normal limit is determined based on the first sequence sub-feature and the interquartile range; and the upper normal limit is determined based on the second sequence sub-feature and the interquartile range.

[0069] According to embodiments of this disclosure, the first position may be located at one-quarter (25%) of the length of the numerical sequence, and the second position may be located at three-quarters (75%) of the length of the numerical sequence.

[0070] The connective features in the initial brain functional connectivity features are arranged in ascending order of value. The first sequence feature Q1, located in the first quarter of the numerical sequence, is subtracted from the second sequence feature Q3, located in the first three-quarters of the numerical sequence, to obtain the interquartile range (IQR). The lower limit of the normal value is Q1 - 1.5 × IQR, and the upper limit of the normal value is Q3 + 1.5 × IQR.

[0071] Connectivity sub-features with values ​​below Q1 - 1.5 × IQR and above Q3 + 1.5 × IQR in the initial brain functional connectivity features were identified as missing values. The interquartile range (IQR) method was used to detect and process outliers, which is suitable for non-normally distributed data, effectively identifying outliers and avoiding reliance on assumptions about data distribution.

[0072] According to embodiments of this disclosure, updating outliers based on their type to obtain intermediate brain functional connectivity features includes: when the outlier type is a missing value, updating the missing value to a median connectivity sub-feature of the brain functional connectivity features to obtain intermediate brain functional connectivity features; wherein, the median connectivity sub-feature is the median of multiple connectivity sub-features arranged according to time information in the initial brain functional connectivity features.

[0073] According to embodiments of this disclosure, missing values ​​may originate from the data acquisition process and the outlier handling process. To ensure the integrity of the feature matrix, all missing values ​​are processed using a median imputation strategy. Compared to mean imputation, the median is more robust to outliers and skewed distributions, and can provide a more stable estimate, making it well-suited for processing EEG signal data.

[0074] According to embodiments of this disclosure, the severity of obstructive sleep apnea is identified based on at least one of a variety of initial brain functional connectivity features, including Pearson correlation coefficient and phase-locked value, comprising: determining a target apnea-hypopnea index corresponding to the Pearson correlation coefficient based on a preset mapping relationship between a preset Pearson correlation coefficient and a preset apnea-hypopnea index; and identifying the severity of obstructive sleep apnea based on the target apnea-hypopnea index.

[0075] According to embodiments of this disclosure, the preset mapping relationship between the preset Pearson correlation coefficient and the preset apnea-hypopnea index is negatively correlated. That is, the higher the degree of obstructive sleep apnea, the smaller the Pearson correlation coefficient between the first electrical signal and the second electrical signal.

[0076] According to embodiments of this disclosure, the closer the Pearson correlation coefficient is to 1 (the higher the value), the lower the target apnea-hypopnea index, and the lower the likelihood of having obstructive sleep apnea. Conversely, the closer the Pearson correlation coefficient is to 0 (the lower the value), the higher the target apnea-hypopnea index, and the higher the likelihood (or severity) of having obstructive sleep apnea.

[0077] A higher phase lock-in value corresponds to a lower target apnea-hypopnea index and a lower likelihood of having obstructive sleep apnea. Conversely, a lower phase lock-in value corresponds to a higher target apnea-hypopnea index and a higher likelihood (or more severe) of obstructive sleep apnea.

[0078] According to embodiments of this disclosure, a target apnea-hypopnea index corresponding to the Pearson correlation coefficient is determined based on a preset mapping relationship between a preset Pearson correlation coefficient and a preset apnea-hypopnea index, thereby obtaining a simple and quick quantitative indicator of obstructive sleep apnea disease. By identifying the severity of obstructive sleep apnea disease through the target apnea-hypopnea index, rapid and accurate identification of the severity of the disease can be achieved, improving the clinical performance of the system.

[0079] According to embodiments of this disclosure, identifying the severity of obstructive sleep apnea based on a target apnea-hypopnea index includes: determining the severity as mild when the target apnea-hypopnea index is in a first numerical range; determining the severity as moderate when the target apnea-hypopnea index is in a second numerical range, wherein the value in the first numerical range is less than the value in the second numerical range; and determining the severity as severe when the target apnea-hypopnea index is in the second numerical range, wherein the value in the second numerical range is less than the value in a third numerical range.

[0080] According to embodiments of this disclosure, the identification result can include two types: "obstructive sleep apnea disease" and "healthy person". When the identification result indicates obstructive sleep apnea disease, the severity of the obstructive sleep apnea disease is identified based on the target apnea-hypopnea index (AHI), resulting in a severity level. The severity level can be categorized as mild, moderate, and severe. Therefore, a four-category task (mild, moderate, and severe OSA patients) can be achieved.

[0081] For example, the first numerical range can be AHI 5~14 times / hour, the second numerical range can be AHI 15~29 times / hour, and the third numerical range can be AHI ≥30 times / hour.

[0082] According to embodiments of this disclosure, since the apnea-hypopnea index is a quantitative indicator of obstructive sleep apnea disease, the severity of obstructive sleep apnea disease can be quickly determined based on the numerical range of the target apnea-hypopnea index.

[0083] According to embodiments of this disclosure, the disease identification model is a gradient boosting tree model, which is trained in the following manner: obtaining a training set of leaded EEG signals, the training set of leaded EEG signals containing multiple samples, the samples including various brain functional connectivity training features and label values, the label values ​​being the apnea-hypopnea index corresponding to the various brain functional connectivity training features; training an initial model based on the training set of leaded EEG signals to obtain the gradient boosting tree model.

[0084] According to embodiments of this disclosure, subjects wear a portable EEG acquisition device, select leads C3, C4, or a combination thereof, and record their complete nighttime sleep process. The portable EEG acquisition device can be a head-mounted portable EEG device.

[0085] The original sampling frequency can be higher than 125 Hz. After acquisition, the signal frequency is unified to 125 Hz through downsampling to ensure the efficiency and consistency of subsequent processing. Then, the EEG signal is preprocessed to obtain the initial electrical signal training set.

[0086] For example, a 0.3–50 Hz bandpass filter is used to remove low-frequency interference (such as baseline drift and breathing artifacts) and high-frequency noise (such as environmental electromagnetic noise and electromyographic interference) to ensure the accuracy of frequency domain and connectivity features.

[0087] For example, continuous EEG signals can be segmented into independent analysis units based on pre-labeled sleep structure tags or according to fixed time windows (e.g., every 30 seconds).

[0088] For example, by using threshold judgment or manual review, segments containing large abnormal artifacts (such as large electromyographic activity or sudden electrode detachment) can be removed to ensure the effectiveness and repeatability of subsequent feature extraction.

[0089] Electroencephalography (EEG) functional connectivity analyzes the synchronicity and phase relationships between different brain regions using EEG signals. This reveals the information interaction and synergistic effects of different brain regions during task execution, capturing dynamic changes in neural networks and providing a powerful tool for understanding cognitive processes, neural regulation mechanisms, and identifying brain dysfunction. Whether it's linear indicators like the Pearson correlation coefficient and cross-correlation function, or information transfer indicators like the synchronization likelihood index and mutual information, as well as spectral coherence, wavelet correlation, phase lock value, and p-index—all eight indicators—OSA patients show significantly lower values ​​than healthy individuals, playing a crucial role in distinguishing between patients and healthy individuals.

[0090] Therefore, eight indicators are extracted from the initial electrical signal training set: linear indicators such as Pearson correlation coefficient and cross-correlation function, information transmission indicators such as synchronization likelihood index and mutual information, as well as spectral coherence, wavelet correlation, phase lock value, and ρ index. These eight indicators are used to obtain the lead EEG signal training set.

[0091] According to embodiments of this disclosure, the initial model can be a weak classifier. Gradient Boosting Decision Tree (GBDT) is an ensemble method based on an iterative strategy. Its basic principle is to sequentially stack multiple weak classifiers with limited performance (usually shallow decision trees) to form a composite model with stronger predictive power. The gradient boosting tree model can be represented as an additive model of decision trees:

[0092] (3)

[0093] in, Let f represent the m-th decision tree, where M is the number of trees, and f M (x) represents the updated predicted value.

[0094] GBDT's strength lies in its ability to fit the residuals from the previous training iteration in each round of model training, thereby gradually narrowing the gap between the updated predicted values ​​and the true labels. GBDT uses gradient descent to optimize the loss function, capturing complex nonlinear relationships between features and exhibiting high predictive power. The GBDT model is implemented using the Scikit-learn GradientBoostingClassifier function. An initial setup of 100 trees is chosen to balance performance and computational cost. A learning rate of 0.1 is selected to control the contribution of each tree and prevent overfitting due to rapid convergence. A maximum depth of 3 is chosen to appropriately limit the complexity of each tree and avoid overfitting.

[0095] The EEG signal training set was divided into training and test sets. To more reliably evaluate the model's generalization performance and reduce the possibility of overfitting, a five-fold stratified cross-validation strategy was used. The StratifiedKFold function was used for stratified sampling, and `shuffle=True` was set to ensure random data shuffling, while `random_state=42` ensured repeatability. This method divides the original dataset into five similarly sized and independent subsets. In each iteration, four subsets are selected for gradient boosting tree model training, and the remaining subset is used as the validation set. This process is repeated in turn, ensuring that each subset is used as test data once, thus achieving a more robust performance evaluation.

[0096] To comprehensively measure the performance of machine learning models in the OSA classification task, a variety of evaluation metrics were selected, and the average value of the results after each fold cross-validation was calculated. The evaluation metrics included: confusion matrix, accuracy, precision, recall, F1 score, Kappa statistic, and AUC (Area Under ROC Curve). Among them, ROC (Receiver Operating Characteristic Curve) is the core tool for evaluating the performance of classification models.

[0097] The classification tasks are divided into a binary classification task (distinguishing OSA patients from healthy individuals) and a four-class classification task (further distinguishing mild, moderate, and severe OSA patients from healthy individuals), both using the AHI index as the label value. For the validation set or newly input unknown samples, their functional connectivity feature vectors are input, and the trained GBDT model makes predictions, outputting the identification result for each sample, along with the confidence score corresponding to each identification result (e.g., based on Softmax or leaf output probability), facilitating doctors or users to evaluate the reliability of the identification results.

[0098] According to embodiments of this disclosure, training an initial model based on a lead-based EEG signal training set to obtain a gradient boosting tree model includes: for each sample, inputting multiple brain function connectivity training features into the trained initial classification model to obtain multiple initial predicted values; determining an initial negative gradient based on the initial predicted values ​​and the loss function value between the initial predicted values ​​and the true values, and using the initial negative gradient as a new label; training a first classification model based on the lead-based EEG signal training set and the new label to obtain a decision tree, the decision tree including multiple leaf nodes, each leaf node corresponding to a subset of samples in the lead-based EEG signal training set; using a preset learning rate as the weight of the decision tree, and performing a weighted summation of the decision prediction values ​​of the decision tree and the initial prediction values ​​of the initial classification model based on the weights to obtain an updated prediction value of the updated model; using the updated model as the initial classification model, repeating the above steps until the target number of iterations is reached to obtain the gradient boosting tree model.

[0099] According to embodiments of this disclosure, for each sample, an initial negative gradient of the loss function with respect to the initial predicted value is calculated, and this initial negative gradient is used as the residual to be corrected. The residual to be corrected is used as a new label, and a new decision tree (first classification model) is trained using the lead EEG signal training set and the new label, resulting in the leaf node partitions of the decision tree. For each leaf node of the decision tree, a gradient descent step size is calculated to minimize the loss function.

[0100] A preset learning rate is used as the weights of the decision trees. Based on these weights, the predicted values ​​of the decision trees and the initial predicted values ​​of the initial classification model are weighted and summed to obtain the updated predicted values ​​of the updated model. The preset learning rate is used to control the contribution of each tree and prevent overfitting.

[0101] Figure 2 The illustration shows a schematic diagram of the identification technology process for obstructive sleep apnea based on leaded electroencephalogram (EEG) signals according to an embodiment of the present disclosure.

[0102] The technical process for identifying obstructive sleep apnea based on leaded EEG signals mainly includes an information acquisition and preprocessing module, a feature extraction and processing module, and a classification model calculation module.

[0103] In the information acquisition and preprocessing module, a few leads of EEG signals (such as C3 and C4 leads) are acquired using a portable EEG device to record the complete nighttime sleep process. The sampling frequency is recommended to be 125 Hz or higher. The raw EEG signals are then preprocessed, including downsampling, bandpass filtering, segmentation, and artifact removal.

[0104] Downsampling can reduce the signal frequency to 125Hz.

[0105] Bandpass filtering can be achieved by using a 0.3–50 Hz bandpass filter to remove low-frequency interference (such as baseline drift and breathing artifacts) and high-frequency noise (such as electromagnetic interference and electromyography artifacts).

[0106] Segmentation can be done by sleep structure tags or by dividing the data into segments of 30 seconds each.

[0107] Artifact removal can be used to remove abnormal segments (such as large electromyographic interference) to improve the accuracy of analysis.

[0108] The feature extraction and processing module extracts features from the preprocessed raw EEG signals. For example, it extracts eight types of brain functional connectivity features, including Pearson correlation coefficient and cross-correlation function, as initial brain functional connectivity features. Based on the IQR method, connectivity features in the initial brain functional connectivity features that are greater than the upper limit of normal or less than the lower limit of normal are identified as outliers. If the outlier is a missing value, it is updated with the median connectivity feature (median imputation) to obtain intermediate brain functional connectivity features. Principal component analysis (PCA) is used to reduce the dimensionality of the z-score-normalized intermediate brain functional connectivity features to obtain the target brain functional connectivity features.

[0109] The classification model calculation module uses the target brain functional connectivity features as the dataset. The GBDT model is selected as the training and evaluation model to output the recognition results and their confidence scores. Finally, the binary classification (OSA patients and healthy individuals) and four-class classification (mild, moderate, and severe patients and healthy individuals) of the lead EEG signal dataset are used for clinical analysis and feedback on the OSA risk level of patients.

[0110] The embodiments disclosed herein innovatively use portable EEG devices to collect signals from leads C3 and C4, and mainly use the electrophysiological characteristics of brain functional connectivity to construct efficient binary and quadri-classification recognition models, which significantly simplifies the PSG process.

[0111] Figure 3 The diagram illustrates the classification performance of a system for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals according to an embodiment of the present disclosure.

[0112] like Figure 3 As shown, the obstructive sleep apnea disease identification system based on lead-based EEG signals achieved an accuracy of 100% and an AUC of 1 for the optimal binary classification (OSA patients and healthy individuals). The accuracy of the optimal quadratic classification (healthy individuals and mild, moderate, and severe OSA patients) reached 79.40% and the AUC was 0.9021.

[0113] Figure 4 The diagram illustrates a binary confusion matrix of a system for identifying obstructive sleep apnea based on lead-based electroencephalogram signals according to an embodiment of the present disclosure.

[0114] like Figure 4 As shown, the horizontal axis represents the predicted label, with OSA and Health set sequentially. The vertical axis represents the actual label, with Health and OSA set sequentially. From the binary classification confusion matrix, it can be seen that the accuracy is 100% when both the predicted and actual labels are "OSA". The accuracy is also 100% when both the predicted and actual labels are "Health".

[0115] Figure 5 The diagram illustrates a four-class confusion matrix of a system for identifying obstructive sleep apnea based on lead-based electroencephalogram signals according to an embodiment of the present disclosure.

[0116] like Figure 5 As shown, the horizontal axis represents the predicted labels, set sequentially as Health, Mild OSA, Moderate OSA, and Severe OSA. The vertical axis represents the actual labels, set sequentially as Severe OSA, Moderate OSA, Mild OSA, and Health. From the four-class confusion matrix, it can be seen that when both the predicted and actual labels are "Health," the accuracy is 100%. When both the predicted and actual labels are "Mild OSA," the accuracy is also 60%, which is higher than the accuracy when the predicted and actual labels are "Moderate OSA" and "Severe OSA," respectively. The accuracy when the predicted and actual labels are the same is higher than the accuracy when the predicted and actual labels are different, achieving accurate four-class classification.

[0117] The obstructive sleep apnea (OSA) identification system based on lead-based EEG signals fully leverages the high temporal resolution and ease of acquisition of EEG signals, reducing the drawbacks of cumbersome operation, high cost, and reliance on manual analysis in polysomnography. It is particularly suitable for home screening and primary healthcare settings, possessing significant practicality and promotional value, and powerfully driving the development and application of sleep medicine and intelligent EEG analysis technology. Simultaneously, the classification model exhibits excellent performance and high recognition accuracy, along with good scalability and real-time integration capabilities. It overcomes the limitations of current OSA screening methods in terms of portability and intelligence, and can improve low-cost, portable intelligent OSA screening systems, helping to increase early diagnosis rates and intervention efficiency, resulting in considerable social and economic benefits.

[0118] Figure 6 A flowchart illustrating a method for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals according to an embodiment of the present disclosure is shown.

[0119] like Figure 6 As shown, the method for identifying obstructive sleep apnea based on leaded EEG signals in this embodiment includes operations S610~S640.

[0120] During operation of S610, multiple initial brain function connectivity features are extracted from the set of leaded EEG signals for the user during sleep. The set of leaded EEG signals includes a first electrical signal collected from a first target region in the user's left brain and a second electrical signal collected from a second target region in the user's right brain. These multiple initial brain function connectivity features characterize the spatiotemporal synchronization between the brain activity states of the user's left brain and the brain activity states of the user's right brain.

[0121] In operation S620, multiple initial brain functional connectivity features are updated based on their respective feature benchmark values ​​to obtain multiple target brain functional connectivity features.

[0122] When operating the S630, multiple target brain function connectivity features are input into the disease recognition model to obtain recognition results.

[0123] In operation S640, when the identification result is characterized as obstructive sleep apnea disease, the severity of obstructive sleep apnea disease is identified based on at least one of a variety of initial brain functional connectivity features, including Pearson correlation coefficient and phase lock value.

[0124] Figure 7 The diagram schematically illustrates a structural block diagram of an obstructive sleep apnea identification device based on leaded electroencephalogram (EEG) signals according to an embodiment of the present disclosure.

[0125] like Figure 7 As shown, the obstructive sleep apnea identification device 700 based on leaded EEG signals in this embodiment includes an extraction module 710, a processing module 720, an input module 730, and an identification module 740.

[0126] The extraction module 710 is used to extract various initial brain functional connectivity features from a set of leaded EEG signals collected from the user during sleep. The set of leaded EEG signals includes a first electrical signal acquired from a first target region in the user's left brain and a second electrical signal acquired from a second target region in the user's right brain. These various initial brain functional connectivity features characterize the spatiotemporal synchronization between the brain activity states of the user's left and right brains. In one embodiment, the extraction module 710 can be used to perform the operation S610 described above, which will not be repeated here.

[0127] The processing module 720 is used to update the various initial brain functional connectivity features based on their respective feature benchmark values ​​to obtain various target brain functional connectivity features. In one embodiment, the processing module 720 can be used to perform the operation S620 described above, which will not be repeated here.

[0128] The input module 730 is used to input multiple target brain function connectivity features into the disease recognition model to obtain recognition results. In one embodiment, the input module 730 can be used to perform the operation S630 described above, which will not be repeated here.

[0129] The identification module 740 is used to identify the severity of obstructive sleep apnea disease based on at least one of several initial brain functional connectivity features, including Pearson correlation coefficient and phase lock value, when the identification result characterizes obstructive sleep apnea disease. In one embodiment, the identification module 740 may be used to perform the operation S640 described above, which will not be repeated here.

[0130] According to embodiments of this disclosure, the feature baseline values ​​include a normal upper limit and a normal lower limit; the processing module 720 includes a first determination submodule, an update submodule, and a dimensionality reduction submodule. The first determination submodule is used to determine the connectivity sub-features in the initial brain functional connectivity features that are greater than the normal upper limit or less than the normal lower limit as outliers; the update submodule is used to update the outliers based on their type to obtain intermediate brain functional connectivity features; the dimensionality reduction submodule is used to reduce the dimensionality of the standardized intermediate brain functional connectivity features based on a principal component analysis algorithm to obtain the target brain functional connectivity features.

[0131] According to embodiments of this disclosure, the upper and lower normal limits are determined as follows: the connective features in the initial brain functional connectivity features are sorted in ascending order of value to obtain a numerical sequence; the interquartile range is determined based on the difference between the first sequence sub-feature located at the first position and the second sequence sub-feature located at the second position in the numerical sequence, where the interquartile range characterizes the degree of dispersion between the first and second sequence sub-features; the lower normal limit is determined based on the first sequence sub-feature and the interquartile range; and the upper normal limit is determined based on the second sequence sub-feature and the interquartile range.

[0132] According to embodiments of this disclosure, the update submodule includes an update unit. The update unit is used to update missing values ​​to median connectivity features of brain functional connectivity features when the outlier type is a missing value, thereby obtaining intermediate brain functional connectivity features; wherein the median connectivity feature is the median of multiple connectivity features arranged according to time information in the initial brain functional connectivity features.

[0133] According to embodiments of this disclosure, the various initial brain functional connectivity features also include at least one of the following: cross-correlation coefficient, wavelet coherence, spectral coherence, phase and amplitude coupling index, synchronization likelihood index, and mutual information.

[0134] According to embodiments of this disclosure, the disease identification model is a gradient boosting tree model, which is trained in the following manner: obtaining a training set of leaded EEG signals, the training set of leaded EEG signals containing multiple samples, the samples including various brain functional connectivity training features and label values, the label values ​​being the apnea-hypopnea index corresponding to the various brain functional connectivity training features; training an initial model based on the training set of leaded EEG signals to obtain the gradient boosting tree model.

[0135] According to embodiments of this disclosure, training an initial model based on a lead-based EEG signal training set to obtain a gradient boosting tree model includes: for each sample, inputting multiple brain function connectivity training features into the trained initial classification model to obtain multiple initial predicted values; determining an initial negative gradient based on the initial predicted values ​​and the loss function value between the initial predicted values ​​and the true values, and using the initial negative gradient as a new label; training a first classification model based on the lead-based EEG signal training set and the new label to obtain a decision tree, the decision tree including multiple leaf nodes, each leaf node corresponding to a subset of samples in the lead-based EEG signal training set; using a preset learning rate as the weight of the decision tree, and performing a weighted summation of the decision prediction values ​​of the decision tree and the initial prediction values ​​of the initial classification model based on the weights to obtain an updated prediction value of the updated model; using the updated model as the initial classification model, repeating the above steps until the target number of iterations is reached to obtain the gradient boosting tree model.

[0136] According to embodiments of this disclosure, the identification module 740 includes a second determining submodule and an identification submodule. The second determining submodule is used to determine a target apnea-hypopnea index corresponding to the Pearson correlation coefficient based on a preset mapping relationship between a preset Pearson correlation coefficient and a preset apnea-hypopnea index; the identification submodule is used to identify the severity of obstructive sleep apnea based on the target apnea-hypopnea index.

[0137] According to embodiments of this disclosure, the identification submodule includes a first determining unit, a second determining unit, and a third determining unit. The first determining unit is used to determine the severity of the illness as mild when the target apnea-hypopnea index is in a first numerical range; the second determining unit is used to determine the severity of the illness as moderate when the target apnea-hypopnea index is in a second numerical range, wherein the value in the first numerical range is less than the value in the second numerical range; and the third determining unit is used to determine the severity of the illness as severe when the target apnea-hypopnea index is in the second numerical range, wherein the value in the second numerical range is less than the value in the third numerical range.

[0138] According to embodiments of this disclosure, any plurality of modules among the extraction module 710, processing module 720, input module 730, and identification module 740 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules can be combined with at least part of the functionality of other modules and implemented in one module. According to embodiments of this disclosure, at least one of the extraction module 710, processing module 720, input module 730, and identification module 740 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the extraction module 710, processing module 720, input module 730, and identification module 740 can be at least partially implemented as a computer program module, which, when run, can perform corresponding functions.

[0139] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing a method for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals, according to an embodiment of the present disclosure.

[0140] like Figure 8 As shown, an electronic device 800 according to an embodiment of this disclosure includes a device processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage portion 808 into a random access memory (RAM) 803. The device processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), 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 device processor 801 may also include onboard memory for caching purposes. The device processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this disclosure.

[0141] RAM 803 stores various programs and data required for the operation of electronic device 800. Device processor 801, ROM 802, and RAM 803 are interconnected via bus 804. Device processor 801 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 802 and / or RAM 803. It should be noted that the programs may also be stored in one or more memories other than ROM 802 and RAM 803. Device processor 801 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in said one or more memories.

[0142] According to embodiments of this disclosure, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to a bus 804. The electronic device 800 may also include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0143] This disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of this disclosure.

[0144] According to embodiments of this disclosure, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this disclosure, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this disclosure, the computer-readable storage medium may include ROM 802 and / or RAM 803 and / or one or more memories other than ROM 802 and RAM 803 described above.

[0145] Embodiments of this disclosure also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code enables the computer system to implement the method for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals provided in embodiments of this disclosure.

[0146] When the computer program is executed by the device processor 801, it performs the functions defined in the system / apparatus of this disclosure embodiments. According to embodiments of this disclosure, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0147] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0148] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 809, and / or installed from the removable medium 811. When the computer program is executed by the device processor 801, it performs the functions defined in the system of this disclosure embodiment. According to embodiments of this disclosure, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0149] According to embodiments of this disclosure, program code for executing the computer programs provided in embodiments of this disclosure can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can execute entirely on a user's computing device, partially on a user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0150] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0151] Those skilled in the art will understand that the features described in the various embodiments of this disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments of this disclosure can be combined and / or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0152] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A system for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals, characterized in that, The system includes: A signal acquisition device is used to acquire a first electrical signal in a first target area of ​​the user's left brain and a second electrical signal in a second target area of ​​the user's right brain, thereby obtaining a set of leaded electroencephalogram (EEG) signals for the user during sleep. Processor, used to perform the following operations: Multiple initial brain function connectivity features are extracted from the EEG signals of the leads. These multiple initial brain function connectivity features characterize the spatiotemporal synchronization between the brain activity state of the user's left brain and the brain activity state of the user's right brain. Based on the feature benchmark values ​​of each of the initial brain functional connectivity features, the initial brain functional connectivity features are updated to obtain multiple target brain functional connectivity features. The various target brain function connectivity features are input into the disease identification model to obtain the identification results; In cases where the identification result is characterized as obstructive sleep apnea, the severity of the obstructive sleep apnea is identified based on at least one of the initial brain functional connectivity features, including the Pearson correlation coefficient and phase lock value.

2. The system according to claim 1, characterized in that, The characteristic benchmark values ​​include a normal upper limit and a normal lower limit; The method involves updating the initial brain functional connectivity features based on their respective feature benchmark values ​​to obtain multiple target brain functional connectivity features, including: Connectivity sub-features in the initial brain functional connectivity features that are greater than the upper limit of normal or less than the lower limit of normal are identified as outliers. The outliers are updated based on their types to obtain the interbrain functional connectivity features; The target brain functional connectivity features are obtained by dimensionality reduction of the standardized intermediate brain functional connectivity features based on the principal component analysis algorithm.

3. The system according to claim 2, characterized in that, The normal upper limit and the normal lower limit are determined based on the following method: The connector features in the initial brain functional connectivity features are sorted in ascending order of value to obtain a numerical sequence; The interquartile range is determined based on the difference between the first sequence sub-feature located at the first position and the second sequence sub-feature located at the second position in the numerical sequence. The interquartile range characterizes the degree of dispersion between the first sequence sub-feature and the second sequence sub-feature. Based on the first sequence sub-features and the interquartile range, the lower limit of the normal value is determined; Based on the second sequence sub-features and the interquartile range, the upper limit of the normal value is determined.

4. The system according to claim 2, characterized in that, The process of updating the outliers based on their types to obtain intermediate brain functional connectivity features includes: In the case where the outlier is a missing value, the missing value is updated to the median connectivity sub-feature of the brain functional connectivity feature to obtain the intermediate brain functional connectivity feature. The median connector feature is the median of multiple connector features arranged according to time information in the initial brain functional connectivity features.

5. The system according to any one of claims 1 to 4, characterized in that, Multiple initial brain functional connectivity features also include at least one of the following: cross-correlation coefficient, wavelet coherence, spectral coherence, phase and amplitude coupling index, synchronization likelihood index, and mutual information.

6. The system according to claim 1, characterized in that, The disease identification model is a gradient boosting tree model, which is trained using the following method: A training set of EEG signals in leads is obtained. The training set of EEG signals in leads contains multiple samples. The samples include multiple brain functional connectivity training features and label values. The label values ​​are apnea-hypopnea indices corresponding to the multiple brain functional connectivity training features. The initial model is trained based on the training set of the aforementioned leads in the EEG signal, and the gradient boosting tree model is obtained.

7. The system according to claim 6, characterized in that, The process of training an initial model based on the aforementioned lead EEG signal training set to obtain the gradient boosting tree model includes: For each sample, the multiple brain function connection training features are input into the trained initial classification model to obtain multiple initial prediction values; Based on the initial predicted value and the loss function value between the initial predicted value and the true value, the initial negative gradient is determined and used as the new label. Based on the training set of the leaded EEG signals and the newly added labels, a first classification model is trained to obtain a decision tree. The decision tree includes multiple leaf nodes, and each leaf node corresponds to a subset of samples in the training set of the leaded EEG signals. The preset learning rate is used as the weight of the decision tree, and the decision prediction value of the decision tree and the initial prediction value of the initial classification model are weighted and summed based on the weight to obtain the updated prediction value of the updated model. The updated model is used as the initial classification model, and the above steps are repeated until the target number of iterations is reached to obtain the gradient boosting tree model.

8. The system according to claim 6, characterized in that, The method of identifying the severity of obstructive sleep apnea based on at least one of the following initial brain functional connectivity features: Pearson correlation coefficient and phase lock value, includes: Based on the preset mapping relationship between the preset Pearson correlation coefficient and the preset apnea-hypopnea index, the target apnea-hypopnea index corresponding to the Pearson correlation coefficient is determined. The severity of obstructive sleep apnea is identified based on the target apnea-hypopnea index.

9. The system according to claim 8, characterized in that, The method of identifying the severity of obstructive sleep apnea based on the target apnea-hypopnea index includes: If the target apnea-hypopnea index is in the first numerical range, the severity of the disease is determined to be mild. When the target apnea-hypopnea index is in the second numerical range, the severity of the disease is determined to be moderate, and the value in the first numerical range is less than the value in the second numerical range. When the target apnea-hypopnea index is in the second numerical range, the severity of the disease is determined to be severe, and the value in the second numerical range is less than the value in the third numerical range.

10. A device for identifying obstructive sleep apnea based on leaded electroencephalogram (EEG) signals, characterized in that, The device includes: An extraction module is used to extract multiple initial brain functional connectivity features from a set of leaded EEG signals for a user during sleep. The set of leaded EEG signals includes a first electrical signal collected from a first target region in the user's left brain and a second electrical signal collected from a second target region in the user's right brain. The multiple initial brain functional connectivity features characterize the spatiotemporal synchronization between the brain activity state of the user's left brain and the brain activity state of the user's right brain. The processing module is used to update the various initial brain functional connectivity features based on the feature benchmark values ​​of each of the various initial brain functional connectivity features to obtain various target brain functional connectivity features; The input module is used to input various target brain function connectivity features into the disease recognition model to obtain recognition results; The identification module is used to identify the severity of obstructive sleep apnea disease based on at least one of the following initial brain functional connectivity features: Pearson correlation coefficient and phase lock value, when the identification result is characterized as obstructive sleep apnea disease.

Images