Sleep apnea detection method and system, medical equipment and storage medium

Abstract

The invention relates to the field of biomedical signal processing and health monitoring, and discloses a sleep apnea detection method and system, medical equipment and a storage medium, and the method comprises the following steps: collecting a respiratory signal of a user, constructing a Fokker-Planck equation model describing respiratory signal probability evolution, and establishing a Fokker-Planck equation model; the method comprises the following steps: extracting characteristic parameters such as state drift rate and entropy change amplitude, judging whether an apnea event occurs or not by combining a threshold condition and a statistical rule in a time window, and feeding back a detection result in time in manners such as sound-light alarm, remote notification or data recording. The method has the advantages of being high in detection precision, high in robustness, good in real-time performance, wide in adaptability and the like, the multi-scene requirements of family monitoring and medical diagnosis can be met, and efficient and reliable technical support is provided for early warning and health management of sleep apnea.

Description

Technical Field

[0001] The present invention relates to the technical field of biomedical signal processing and health monitoring, and in particular to a sleep apnea detection method, system, medical equipment and storage medium. Background Art

[0002] Sleep apnea is a common sleep disorder, the main feature of which is repeated pauses or significant weakening of breathing during sleep. In severe cases, it may lead to health problems such as hypoxemia, hypertension, and cardiovascular disease. As people pay more attention to sleep quality, the detection and diagnosis technology for sleep apnea has also been widely studied and developed. At present, the detection methods on the market mainly rely on polysomnography (PSG) or home detection devices based on simple breathing signals. However, these existing technologies have obvious defects and limitations.

[0003] First, traditional polysomnography equipment is complex to operate and relies on the precise installation of multiple sensors and leads, which not only increases the burden on users, but also places high demands on the portability and comfort of the equipment, making it difficult to meet the needs of daily home monitoring. In addition, due to the high cost of equipment and testing, it is difficult for ordinary users to rely on such devices for long-term continuous monitoring of sleep quality.

[0004] Secondly, some detection methods based on single respiratory signal analysis, although they can improve portability, are usually limited to time domain or frequency domain analysis techniques and lack in-depth modeling of the dynamic characteristics of respiratory signals. This method is easily affected by noise interference or unclear signal characteristics, resulting in insufficient detection accuracy and robustness, especially in complex environments or when there are large physiological differences between different individuals, and its detection effect is difficult to be stable and reliable.

[0005] In addition, existing technologies are also insufficient in terms of real-time anomaly detection. Many methods rely on long-term data accumulation or post-processing analysis and cannot provide immediate feedback when apnea occurs, which is obviously not ideal for some high-risk groups who need timely intervention. At the same time, the alarm mechanism of existing equipment is usually relatively simple, and it is difficult to flexibly adapt to user needs according to actual application scenarios.

[0006] Therefore, the existing technology urgently needs a sleep apnea detection method with high accuracy, strong robustness, good real-time performance and applicable to multiple scenarios, so as to overcome the above technical defects and provide users with efficient and reliable health management tools. Summary of the invention

[0007] In view of the shortcomings of the prior art, the present invention provides a sleep apnea detection method, system, medical device and storage medium, which solve the technical problems in the prior art of low sleep apnea detection accuracy, poor robustness, insufficient real-time performance and lack of adaptation to multiple scenario requirements.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: a sleep apnea detection method, comprising the following steps:

[0009] Collect the user's breathing signal, record the breathing amplitude data changing over time, and perform preprocessing;

[0010] Based on the preprocessed respiratory signal, a multidimensional state variable describing the respiratory state is constructed;

[0011] Based on the state variables, a dynamic evolution model of the respiratory signal is established, wherein the dynamic evolution model describes the dynamic evolution process of the respiratory signal;

[0012] Solving the state probability distribution in the dynamic evolution model to obtain the time evolution characteristics of the respiratory signal;

[0013] Extract characteristic parameters of state drift rate and signal complexity change;

[0014] Based on the characteristic parameters, it is determined whether a sleep apnea event occurs, and if it is determined to have occurred, an alarm signal is triggered.

[0015] Preferably, the pretreatment comprises:

[0016] The respiratory signal is decomposed into multiple scales using wavelet transform method;

[0017] Interference signals outside the frequency range are filtered out, and the respiratory frequency components within the range of 0.1-1 Hz are retained.

[0018] Preferably, the multidimensional state variables include:

[0019] Respiration amplitude is used as the first state variable;

[0020] The change rate of the respiratory amplitude is used as the second state variable, wherein the change rate is the derivative of the respiratory signal amplitude with respect to time.

[0021] Preferably, the dynamic evolution model includes:

[0022] The state evolution equation describing the change of respiratory state over time;

[0023] The state evolution equation defines a normal breathing state and a balanced pause state by setting a bistable potential energy function, wherein the normal breathing state and the pause state are two stable points of the potential energy function respectively.

[0024] Preferably, the state probability distribution of the dynamic evolution model is obtained by a numerical solution method, and the state probability distribution is the joint probability density of the breathing amplitude and the breathing change rate.

[0025] Preferably, the state drift rate is the probability change rate of the respiratory signal from a normal state to a pause state; the signal complexity change is obtained by calculating the amplitude of the entropy value change of the signal probability distribution, wherein the entropy value is used to quantify the degree of disorder of the signal.

[0026] Preferably, the rules for determining the apnea event include:

[0027] When the state drift rate exceeds a preset threshold, it is determined that the respiratory signal has a tendency to transfer to a pause state;

[0028] When the entropy value change amplitude of the signal complexity change exceeds a preset threshold, it is determined that the respiratory signal is in an abnormal state;

[0029] When the number of times the above two conditions are met within the continuous detection window exceeds a preset value, it is determined as a apnea event.

[0030] The present invention also provides a sleep apnea detection system, comprising:

[0031] A signal acquisition module, used to collect the user's breathing signal;

[0032] A signal processing module, used for performing denoising and filtering on the respiratory signal to extract an effective respiratory frequency component;

[0033] A state modeling module, used to construct state variables based on the processed respiratory signal and establish a dynamic evolution model of the respiratory signal;

[0034] A feature extraction module is used to extract characteristic parameters of state drift rate and signal complexity change from the dynamic evolution model;

[0035] A determination module, used for determining whether a apnea event occurs based on the characteristic parameters;

[0036] The alarm module is used to trigger an alarm signal when a respiratory arrest event is determined to have occurred.

[0037] The present invention also provides a medical device, comprising:

[0038] At least one sensor, used to collect a breathing signal of a user;

[0039] at least one processor for executing the sleep apnea detection method;

[0040] At least one alarm device is used to send an alarm signal to the user or medical institution when a apnea event is detected.

[0041] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the sleep apnea detection method as described above is implemented.

[0042] The invention provides a sleep apnea detection method, system, medical equipment and storage medium.

[0043] It has the following beneficial effects:

[0044] 1. The present invention constructs a dynamic evolution model based on the Fokker-Planck equation, comprehensively models the probability distribution and dynamic characteristics of the respiratory signal, and combines multi-dimensional characteristic parameters such as state drift rate and entropy change amplitude to improve the detection accuracy of sleep apnea events. Compared with traditional time domain or frequency domain analysis methods, the present invention can more accurately capture abnormal changes in respiratory signals and effectively reduce the missed detection rate.

[0045] 2. The present invention adopts a statistical analysis method within a continuous time window in the judgment of pause events, and reduces the interference of noise signals and short-term abnormalities on the detection results by combining the threshold conditions and accumulation rules of characteristic parameters. Especially in the case of large individual differences between different users and environmental interference, the robustness of the present invention is particularly outstanding, and it can run stably and output accurate results.

[0046] 3. The present invention ensures the high efficiency of the detection process by optimizing the feature parameter extraction and judgment algorithm, and can complete the processing of respiratory signals and the judgment of apnea events in a short time, thereby realizing real-time detection of sleep apnea events. The real-time performance makes the present invention suitable for dynamic monitoring scenarios, providing users with timely health warnings and intervention possibilities.

[0047] 4. The determination parameters and detection window length in the present invention can be dynamically adjusted according to the user's physiological characteristics and application scenarios, so as to adapt to the needs of users of different ages, body shapes and health conditions. Whether in home scenarios or medical scenarios, the present invention can be flexibly deployed and has strong applicability.

[0048] 5. After detecting a sleep apnea event, the present invention can promptly remind the user or guardian through a variety of alarm methods. This diversified alarm mechanism not only improves the user's convenience, but also provides data support for subsequent health data analysis and diagnosis.

[0049] 6. The technical solution of the present invention has a clear structure and concise algorithm implementation logic, and is suitable for implementation through the combination of software and hardware. Through the integration of respiratory signal acquisition equipment, processing modules with embedded algorithms, and alarm terminals, the present invention can be deployed in a variety of monitoring devices at low cost and high efficiency, reducing the difficulty of technical implementation and promotion costs.

[0050] 7. The present invention helps users detect sleep apnea problems early through accurate detection and alarm mechanisms, and take timely intervention measures to reduce long-term health risks. This active health management capability provides technical support for improving users' sleep quality and preventing and controlling related chronic diseases, which has important social and economic significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0052] Figure 2 Schematic diagram of the system architecture of the present invention.

[0053] Among them, 100, signal acquisition module; 200, signal processing module; 300, state modeling module; 400, feature extraction module; 500, determination module; 600, alarm module. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Please see attached Figure 1 The present invention provides a sleep apnea detection method based on a dynamic evolution model, which accurately identifies sleep apnea events by collecting the user's breathing signal and combining signal processing, modeling, feature extraction and judgment technology. The specific implementation of the present invention is described in detail below in combination with each step.

[0056] like Figure 1 As shown, the sleep apnea detection method may include the following steps:

[0057] S1, respiratory signal acquisition and preprocessing;

[0058] S2, constructing multidimensional state variables based on respiratory signals;

[0059] S3. Based on the state variables, a dynamic evolution model of the respiratory signal is established;

[0060] S4, solving the state probability distribution in the dynamic evolution model to obtain the time evolution characteristics of the respiratory signal;

[0061] S5, extracting characteristic parameters of state drift rate and signal complexity change;

[0062] S6. Determine whether a sleep apnea event occurs based on the characteristic parameters.

[0063] Each step of the method of the present invention is described in detail below.

[0064] For step S1, in one possible implementation, the user wears a portable respiratory signal acquisition device, such as a nasal airflow sensor, a chest displacement sensor, or a smart mattress sensor, to acquire the user's respiratory signal in real time. The acquisition device records the signal x(t) of the respiratory amplitude changing over time in a non-invasive manner. The signal sampling frequency f s It can be set according to specific needs, such as 100 Hz, to ensure that the dynamic changes of the respiratory signal are fully captured.

[0065] As an option, the acquisition device can combine multiple signal sources, such as recording nasal airflow and chest displacement signals simultaneously, to improve the reliability and accuracy of signal acquisition. These signals can be recorded independently or a comprehensive respiratory signal can be generated through a multi-sensor fusion algorithm.

[0066] It should be noted that the breathing signal may be affected by environmental noise, equipment interference or user movement. In order to ensure the signal quality, the collected raw data needs to be preprocessed.

[0067] Specifically, the preprocessing includes the following steps:

[0068] 1. Denoising

[0069] In an exemplary implementation, wavelet transform is used to perform multi-scale decomposition of the respiratory signal to effectively filter out environmental noise and interference. The wavelet transform uses the Daubechies wavelet family for decomposition, the decomposition scale is set to 4 levels, and the third and fourth level components are selected for signal reconstruction, thereby retaining the main frequency band of the respiratory signal.

[0070] In some embodiments, the denoising process may also use a low-pass filtering method, setting the cutoff frequency range to 0.1 Hz to 1 Hz to filter out high-frequency noise and low-frequency trends other than the respiratory signal. The specific filter type may be a FIR filter or an IIR filter, and the filter order may be adjusted according to the actual application scenario.

[0071] 2. Normalization

[0072] To facilitate the analysis in subsequent steps, this embodiment performs normalization processing on the denoised signal to normalize the signal amplitude to the range of [0, 1].

[0073] It can be understood that normalization processing can not only improve the numerical stability of the signal, but also enable the signals of different users to be analyzed on the same standardized scale, which facilitates the universal design of the algorithm.

[0074] 3. Interpolation completion

[0075] In a possible implementation, if the collected respiratory signal has missing or discontinuous samples, an interpolation method is used to complete the signal. Specifically, a linear interpolation method or a spline interpolation method can be used to estimate the missing points.

[0076] It should be noted that interpolation completion can ensure the temporal continuity of the signal and provide complete data support for subsequent dynamic modeling.

[0077] In this embodiment, after completing signal acquisition and preprocessing, the signal obtained is the processed breathing amplitude data x(t), which is characterized in that the signal frequency component conforms to the physiological breathing range, the noise interference has been significantly reduced, and the signal amplitude is normalized and has time continuity.

[0078] It is understandable that high-quality preprocessing of the signal is crucial for the subsequent steps of kinetic modeling and feature extraction. Through the above method, the preprocessed signal can effectively improve the robustness and accuracy of apnea detection, while ensuring the reliability and efficiency of the subsequent algorithm operation.

[0079] For step S2, in this embodiment, step S2 describes the dynamic evolution characteristics of the respiratory signal by constructing a multidimensional state variable. The multidimensional state variable can fully reflect the amplitude of the respiratory signal and its dynamic changes, and provide support for subsequent dynamic modeling. The specific implementation method is as follows:

[0080] In a possible implementation, the multidimensional state variable includes the amplitude information and dynamic change information of the respiratory signal. The amplitude information describes the state value of the respiratory signal at a certain point in time, and the dynamic change information represents the rate of change of the respiratory signal over time. This dual variable design can take into account both the static and dynamic characteristics of the signal.

[0081] Specifically, the first state variable is the amplitude x(t) of the respiratory signal, which is derived from the respiratory signal that has been filtered, denoised and normalized in step S1. The amplitude x(t) is defined as follows:

[0082] x(t)=x norm (t)

[0083] Among them, x norm(t) is the normalized signal.

[0084] As an option, the second state variable is the rate of change v(t) of the respiratory signal, which is calculated by taking the derivative of the signal amplitude with respect to time. The rate of change is calculated as follows:

[0085]

[0086] In actual operation, considering the discreteness of digital signals, the rate of change is usually approximated using the finite difference method. Its expression is:

[0087]

[0088] Among them, t i and t i-1 are the time between two adjacent sampling points, Δt=t i -t i-1 is the sampling interval.

[0089] It should be noted that the sampling frequency f s It will affect the sampling time interval Δt. In one possible implementation, when f s When =100Hz, Δt is 0.01s.

[0090] In some embodiments, the definition of state variables can be further extended. In addition to the above-mentioned amplitude x(t) and change rate v(t), higher-order change information can also be introduced, such as acceleration (second-order derivative of the signal):

[0091]

[0092] The acceleration variable can provide acceleration or deceleration information of the signal change, providing richer features for describing the dynamic behavior of the signal. However, in this embodiment, only amplitude and rate are used as state variables for explanation, and the use of extended variables needs to be determined according to specific scene requirements.

[0093] Exemplarily, the result of constructing the multidimensional state variable can be expressed as a two-dimensional state space, where each state point is composed of [x(t), v(t)]. The state space describes the amplitude value and change rate of the respiratory signal at a certain moment. It can be understood that this state space expression can fully reflect the temporal dynamic characteristics of the respiratory signal, which is convenient for the establishment of the subsequent dynamic model.

[0094] It should be noted that the construction of multidimensional state variables is not limited to a single sampling point, but can also batch process continuous sampling points. For example, the signal is divided into time windows of fixed length (such as 10 seconds), and a series of state points [x(t), v(t)] are calculated in each window to form a state trajectory. The state trajectory can more intuitively reflect the changing trend of the respiratory signal over time.

[0095] In a possible implementation, in order to enhance the robustness of the state variable, the change rate v(t) can be smoothed, for example, by using a sliding average method to reduce the impact of instantaneous fluctuations. The formula is:

[0096]

[0097] Where N is the length of the sliding window.

[0098] It is understandable that the core of constructing multidimensional state variables is to combine the static amplitude characteristics of the signal with the dynamic change behavior, so as to provide sufficient input information for subsequent steps. The state variable construction method described in this embodiment is universal, can be adapted to different types of respiratory signal acquisition equipment, and provide a unified mathematical expression for subsequent dynamic evolution modeling.

[0099] For step S3, in this embodiment, step S3 establishes a dynamic evolution model describing the dynamic behavior of the respiratory signal based on the multidimensional state variables [x(t), v(t)] constructed in step S2. The model is used to describe the evolution process of the respiratory signal from the normal state to the pause state, and provides a mathematical basis for the subsequent state probability distribution solution and feature extraction.

[0100] In one possible implementation, the dynamic behavior of the respiratory signal is modeled as a nonlinear damping system. The system describes the dynamic evolution relationship of the amplitude and change rate of the respiratory signal over time through the state evolution equation. The dynamic evolution equation is as follows:

[0101]

[0102] in:

[0103] x: the amplitude of the signal, describing the static characteristics of the breathing state;

[0104] v: the rate of change of the signal, describing the dynamic changes of the respiratory signal;

[0105] γ: damping coefficient, reflecting the energy attenuation of the respiratory signal over time;

[0106] U(x): potential energy function, defining the normal breathing state and the balanced pause state;

[0107] ξ(t): Gaussian white noise, used to model external random interference, satisfying <ξ(t)>=0 and <ξ(t)ξ(t')>=2Dδ(t-t'), where D is the noise intensity.

[0108] It should be noted that the various parameters in the above dynamic equations can be adjusted according to the actual application scenario. For example, the value of the damping coefficient γ is usually between 0.1 and 0.5 to adapt to the breathing signal characteristics of different users.

[0109] Specifically, the design of the potential energy function U(x) is the core of the dynamic evolution model. As an option, a bistable potential energy function is used in this embodiment to describe the two main equilibrium states of the respiratory signal, namely the normal breathing state and the pause state. The expression of the potential energy function is:

[0110] U(x)=a(x)-x 1 ) 2 (xx 2 ) 2

[0111] in:

[0112] x 1 : The balance point of normal breathing state;

[0113] x 2 : The balance point of the pause state;

[0114] a: Parameter that controls the transition speed.

[0115] In an exemplary implementation, x can be set 1 = 0.6 and x 2 =0.2, representing the typical amplitude values ​​of the normal state and the paused state, respectively. The shape of the potential energy function is optimized by adjusting the value of a, and the common value range is 0.1 to 1.0. This function can ensure that the normal state and the paused state form two stable points in the dynamic system and provide the necessary energy barrier when transitioning between states.

[0116] It should be noted that the specific parameters of the potential energy function can be fitted and optimized based on the characteristics of the collected breathing signal. By statistically analyzing the historical signal data of different users, the optimal parameter combination can be selected to enhance the adaptability and robustness of the model.

[0117] In a possible implementation, the external random disturbance is modeled by Gaussian white noise ξ(t). The intensity of the noise D determines the sensitivity of the system to random disturbances.

[0118] It can be understood that the dynamic evolution model can simplify the complex dynamic changes of the respiratory signal into a mathematical state evolution process, thereby providing a unified modeling basis for subsequent steps. The input of the model is the state variable [x(t), v(t)] calculated in step S2, and the output is the state evolution behavior of the signal at each moment.

[0119] In some embodiments, in order to improve the computational efficiency of the model, a numerical method can be used to solve the dynamic equations. For example, the explicit Euler method or the fourth-order Runge-Kutta method is used to iteratively update the state evolution, and the discrete form is as follows:

[0120] xn +1 =x n +Δt·v n

[0121]

[0122] Among them, Δt is the time step, and the typical value is 0.01s.

[0123] It should be further explained that the model can adapt to the input characteristics of various respiratory signal acquisition devices and can adapt to the physiological characteristics of different users through parameter optimization. Through the establishment of the above dynamic evolution model, the evolution characteristics of the respiratory signal from the normal state to the pause state can be effectively captured, laying a solid theoretical foundation for solving the subsequent state probability distribution and determining the pause event.

[0124] For step S4, in this embodiment, step S4 is based on the dynamic evolution model established in step S3, and obtains the time evolution characteristics of the respiratory signal by solving the state probability distribution. The state probability distribution can describe the distribution of the amplitude and rate of the respiratory signal, and provide a basis for subsequent feature extraction and pause event determination.

[0125] In one possible implementation, the solution to the state probability distribution is based on the Fokker-Planck equation. The Fokker-Planck equation is used to describe the evolution of probability density in a random dynamic system, and its form is as follows:

[0126]

[0127] in:

[0128] P(x,v,t) is the joint probability density at time t;

[0129] It is the deterministic driving force of the system;

[0130] D is the noise intensity, which represents the impact of random disturbances.

[0131] It should be noted that this equation introduces the diffusion term The random fluctuations in the respiratory signal are simulated, and the deterministic term f(x,v) is combined to simulate the nonlinear dynamic behavior of the signal.

[0132] Specifically, the goal of solving the Fokker-Planck equation is to obtain the joint probability density P(x, v, t). In one possible implementation, a numerical method can be used for discretization. For example, the time and space of the equation are discretized using the finite difference method. The specific steps are as follows:

[0133] 1. Discretize the state space [x, v] into a finite grid of points, with grid steps of Δx and Δv respectively, and use P as the probability density on the grid points i,j (t) indicates.

[0134] 2. The time step is set to Δt, and time iteration is performed by explicit finite difference method or implicit finite difference method.

[0135] 3. Calculate the probability density change at each grid point, and the update formula is:

[0136]

[0137] in, Represents the probability density at time nΔt.

[0138] In some embodiments, to ensure the stability of the numerical solution, the implicit difference method or the Crank-Nicholson method may be preferably used. In addition, to reduce the influence of boundary effects, reasonable boundary conditions may be set for the state space, such as an absorbing boundary or a reflecting boundary.

[0139] As an option, to simplify the computational complexity, the marginal probability density P(x, t) can be extracted from the joint probability density P(x, v, t), that is:

[0140]

[0141] The marginal probability density describes the distribution of the respiratory signal in the amplitude dimension without considering the influence of the rate dimension.

[0142] Specifically, the marginal probability density can be calculated by numerical integration methods, such as trapezoidal integration method or Gaussian integration method. Its discrete form is:

[0143]

[0144] Where M is the number of grid points in the rate dimension, and Δv is the rate grid step size.

[0145] It should be noted that the marginal probability density is more suitable for the subsequent feature extraction step because it directly reflects the amplitude distribution characteristics of the respiratory signal.

[0146] It is understandable that the core of solving the Fokker-Planck equation is to balance the calculation accuracy and efficiency. In practical applications, different solution methods and parameter settings can be selected according to specific scenarios. For example, in real-time detection scenarios, low-resolution grids and larger time steps can be used to speed up the calculation; while in offline analysis scenarios, high-resolution grids can be selected to improve accuracy.

[0147] Exemplarily, after a user's breathing signal is solved by the above steps, its state probability distribution presents the following characteristics:

[0148] In normal breathing state (amplitude x ≈ x 1 ), the probability density value is high, indicating that the system is in a stable state.

[0149] In the pause state (amplitude x ≈ x 2 ), the probability density value is low, but with the effect of random disturbance, the probability distribution gradually drifts toward the pause state.

[0150] It should be further explained that the evolution of the state probability distribution can not only reflect the current characteristics of the respiratory signal, but also capture the dynamic change trend of the signal from the normal state to the pause state, providing comprehensive and reliable data support for subsequent feature extraction and pause event judgment.

[0151] The solution of the state probability distribution directly affects the accuracy of the pause event determination. Therefore, the discrete grid resolution, time step and boundary condition settings used in this embodiment are experimentally optimized to ensure a reasonable balance between calculation accuracy and efficiency. Through the above method, accurate modeling and analysis of the complex dynamic behavior of the respiratory signal can be achieved, laying a solid foundation for subsequent steps.

[0152] As for step S5, in this embodiment, step S5 is intended to extract key characteristic parameters from the state probability distribution obtained in step S4, including state drift rate and signal complexity change. These characteristic parameters are used to quantify the dynamic evolution behavior of the respiratory signal and provide a basis for subsequent pause event determination.

[0153] Specifically, the state drift rate is used to describe the probability change rate of the respiratory signal from the normal state to the pause state. As an option, the calculation formula of the state drift rate Δx(t) is:

[0154]

[0155] in:

[0156] P(x,t) is the marginal probability density function of the respiratory signal amplitude;

[0157] It represents the partial derivative of the probability density function with respect to time, reflecting the change of probability distribution over time;

[0158] x 1 and x 2 They are the amplitude values ​​corresponding to normal breathing state and pause state respectively.

[0159] It should be noted that the integral interval [x 1 ,x 2 ]The amplitude range from normal state to pause state is selected to ensure that the calculated drift rate accurately reflects the trend of state transfer.

[0160] In one possible implementation, the calculation When , the difference method can be used for approximation, that is:

[0161]

[0162] Wherein, Δt is the time step, which is usually consistent with the time step when solving the probability density function in step S4.

[0163] It should be noted that, in order to improve the calculation accuracy, P(x, t) can be smoothed or a higher-order differential approximation method can be used.

[0164] The signal complexity change is quantified by calculating the entropy value of the probability distribution. The entropy value S(t)S(t)S(t) is calculated as:

[0165] S(t)=-∫P(x,t)lnP(x,t)dx

[0166] Here, ln represents the natural logarithm.

[0167] Alternatively, the entropy change S(t) is defined as:

[0168] ΔS(t)=|S(t)-S(t-Δt)|

[0169] This value reflects the degree to which the complexity of the signal changes over time.

[0170] It should be noted that the calculation of entropy value requires numerical integration of probability density function P(x, t). Commonly used numerical integration methods include trapezoidal method and Simpson method. The specific choice can be determined according to the calculation accuracy and efficiency requirements.

[0171] In a possible implementation, the probability density function can be discretized to improve computational efficiency. The amplitude range is divided into N equally spaced small intervals, and the value of the probability density function P(x, t) in each small interval is represented by P i (t) represents, the discrete form of entropy value is:

[0172]

[0173] Wherein, Δx is the step size of the amplitude interval.

[0174] It should be noted that the calculation of the state drift rate and the entropy change amplitude requires a high accuracy of the probability density function. Therefore, when solving the probability density function in step S4, it should be ensured that the grid division is sufficiently detailed and the time step is reasonably selected to avoid the influence of numerical errors on the calculation of characteristic parameters.

[0175] It can be understood that the state drift rate Δx(t) and the entropy change amplitude ΔS(t) together constitute the key features of the dynamic behavior of the respiratory signal. Among them, the state drift rate reflects the rate at which the signal transitions from the normal state to the pause state, while the entropy change amplitude quantifies the change in signal complexity.

[0176] In some embodiments, other characteristic parameters may be further combined, such as the position change of the probability density peak, the variance change, etc., to enhance the accuracy of the pause event determination. However, this embodiment mainly focuses on the extraction of the state drift rate and the entropy change amplitude.

[0177] It should be emphasized that the extraction process of characteristic parameters should be real-time and efficient to meet the real-time monitoring needs in practical applications. To this end, optimized numerical calculation methods can be used in the algorithm implementation, and technologies such as parallel computing can be used to accelerate the calculation process.

[0178] By executing the above step S5, this embodiment extracts key characteristic parameters from the state probability distribution. These parameters provide necessary basis for the subsequent pause event determination and alarm steps, and realize accurate detection of sleep apnea.

[0179] As for step S6, in this embodiment, step S6 determines whether the user has a sleep apnea event by combining characteristic parameters such as the state drift rate and signal complexity change (entropy change amplitude) extracted in step S5. Once a sleep apnea event is detected, the system will trigger an alarm signal to achieve timely reminder or intervention.

[0180] In a possible implementation, the determination of a pause event is based on the following two core characteristic parameters:

[0181] State drift rate Δx(t): reflects the probability change rate of the respiratory signal from the normal state to the pause state.

[0182] Entropy change amplitude ΔS(t): quantifies the degree to which the complexity of the respiratory signal changes over time.

[0183] It can be understood that these two parameters describe the dynamic change behavior of the respiratory signal from different angles and can effectively capture abnormal conditions.

[0184] Specifically, the decision rule can be defined by a threshold condition. As an option, the decision condition includes:

[0185] When the state drift rate Δx(t) exceeds the preset threshold ∈ 1 , it indicates that the signal probability density has significantly deviated from the normal state, and a pause event may have occurred.

[0186] When the entropy change amplitude ΔS(6) exceeds the preset threshold ∈ 2 When , it means that the signal complexity is significantly reduced, which further indicates that the respiratory state is abnormal.

[0187] Exemplarily, the determination formula for the pause event can be expressed as:

[0188] If Δx(t)>∈ 1 And ΔS(t)>ε 2 , it is considered a pause event.

[0189] It should be noted that the threshold ε 1 and ε 2 It can be optimized based on the user's physiological characteristics or sampled data. 1 The typical range is 0.1~0.5, ε 2 The typical range is 0.01 to 0.1.

[0190] In a possible implementation, in order to improve the reliability of detection and avoid false alarms, a decision rule within a continuous time window can be set. For example, the respiratory signal is divided into detection windows of fixed length (such as 10 seconds), and the number of characteristic parameters that meet the conditions is counted in each window. If the number of times the condition is met exceeds a preset threshold N, the decision rule is set. th , it is determined as an apnea event. The rule can be expressed as:

[0191] If (Δx(t)>∈ 1 And ΔS(t)>ε 2 ) > N th , it is considered a pause event.

[0192] It should be noted that the length of the detection window and the threshold N th The choice of N will affect the sensitivity and specificity of the system. Typically, the detection window length can be set to 10 seconds to 30 seconds, and the threshold Nth Can be adjusted dynamically according to experimental data.

[0193] As an option, the alarm mechanism in this embodiment will immediately trigger a corresponding alarm signal when a apnea event is detected. The alarm method may include but is not limited to:

[0194] Sound and light alarm: a warning sound is emitted through the speaker, or a flashing LED indicator is used to remind you.

[0195] Notification system: sends alarm information to the user's mobile device or guardian's terminal through a wireless communication module (such as Wi-Fi, Bluetooth or cellular network).

[0196] Data Logging: Record the time and frequency of pause events on a local storage device or in the cloud for subsequent analysis or diagnosis.

[0197] It should be noted that the implementation of the alarm mechanism can be adjusted according to specific scenarios. For example, in medical scenarios, the alarm signal can be directly transmitted to the monitoring device or the doctor's end; in home scenarios, the alarm signal can trigger a health reminder on the user's device.

[0198] In a possible implementation, the process of determining a pause event can be simplified into the following steps:

[0199] Get the characteristic parameters Δx(t) and ΔS(t) in the current time window;

[0200] Check whether the characteristic parameters meet the threshold conditions;

[0201] If the condition is met more than the preset number of times, a pause alarm signal is triggered.

[0202] It is understandable that, through the above method, this embodiment can accurately detect apnea events and issue an alarm in a timely manner. This detection and alarm mechanism has a high degree of real-time and robustness, and can adapt to the physiological characteristics and breathing patterns of different users.

[0203] It should be further explained that the determination rules and alarm mechanism in this embodiment can be optimized according to specific application scenarios. For example, in certain high-risk user groups, the threshold ∈ 1 and ∈ 2 , in order to improve the sensitivity of detection; for general users, the detection window length can be increased or N th , in order to reduce the occurrence of false alarms.

[0204] By executing step S6, this embodiment not only completes the accurate determination of apnea events, but also ensures the timely feedback of the detection results, thereby providing effective technical support for the health management of users.

[0205] In general, the present invention accurately captures the dynamic change characteristics of the respiratory signal from the normal state to the pause state through the systematic processing of the user's respiratory signal collection, preprocessing, state variable construction, dynamic modeling, characteristic parameter extraction and pause event determination. Specifically, the present invention uses the Fokker-Planck equation to describe the probabilistic evolution process of the respiratory signal, and realizes efficient detection of apnea events through key characteristic parameters such as state drift rate and entropy change amplitude, and cooperates with sound and light alarm or remote notification functions to provide users with a real-time and reliable health monitoring solution. The present invention has the characteristics of high precision, strong robustness and good adaptability, can meet the multi-scenario requirements of home monitoring and medical diagnosis, and provides effective technical support for early warning and intervention of sleep apnea.

[0206] The sleep apnea detection system described below and the sleep apnea detection method described above may refer to each other.

[0207] Please see attached Figure 2 The present invention also provides a sleep apnea detection system, comprising:

[0208] The signal acquisition module 100 is used to collect the user's breathing signal;

[0209] The signal processing module 200 is used to perform denoising and filtering on the respiratory signal to extract the effective respiratory frequency component;

[0210] A state modeling module 300, for constructing state variables based on the processed respiratory signal and establishing a dynamic evolution model of the respiratory signal;

[0211] A feature extraction module 400 is used to extract feature parameters of state drift rate and signal complexity change from the dynamic evolution model;

[0212] A determination module 500, configured to determine whether a apnea event occurs based on the characteristic parameters;

[0213] The alarm module 600 is used to trigger an alarm signal when it is determined that a apnea event occurs.

[0214] The system of this embodiment can be used to execute the above method embodiments, and its principles and technical effects are similar, which will not be repeated here.

[0215] The present invention also provides a medical device, comprising:

[0216] At least one sensor, used to collect a breathing signal of a user;

[0217] At least one processor, configured to execute the sleep apnea detection method as described above;

[0218] At least one alarm device is used to send an alarm signal to the user or medical institution when a apnea event is detected.

[0219] The present invention also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the sleep apnea detection method as described above is executed.

[0220] Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0221] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A sleep apnea detection method, characterized in that: The following steps are involved: Collect the user's breathing signal, record the breathing amplitude data changing over time, and perform preprocessing; Based on the preprocessed respiratory signal, a multidimensional state variable describing the respiratory state is constructed; Based on the state variables, a dynamic evolution model of the respiratory signal is established, wherein the dynamic evolution model describes the dynamic evolution process of the respiratory signal; Solving the state probability distribution in the dynamic evolution model to obtain the time evolution characteristics of the respiratory signal; Extract characteristic parameters of state drift rate and signal complexity change; Based on the characteristic parameters, it is determined whether a sleep apnea event occurs, and if it is determined to have occurred, an alarm signal is triggered.

2. The sleep apnea detection method according to claim 1, characterized in that: The pre-processing comprises: The respiratory signal is decomposed into multiple scales using wavelet transform method; Interference signals outside the frequency range are filtered out, and the respiratory frequency components within the range of 0.1-1 Hz are retained.

3. The sleep apnea detection method according to claim 1, characterized in that: The multidimensional state variables include: Respiration amplitude is used as the first state variable; The change rate of the respiratory amplitude is used as the second state variable, wherein the change rate is the derivative of the respiratory signal amplitude with respect to time.

4. The sleep apnea detection method according to claim 1, characterized in that: The dynamic evolution model includes: The state evolution equation describing the change of respiratory state over time; The state evolution equation defines a normal breathing state and a balanced pause state by setting a bistable potential energy function, wherein the normal breathing state and the pause state are two stable points of the potential energy function respectively.

5. The sleep apnea detection method according to claim 1, characterized in that: The state probability distribution of the dynamic evolution model is obtained by a numerical solution method, and the state probability distribution is the joint probability density of the breathing amplitude and the breathing change rate.

6. The sleep apnea detection method according to claim 1, characterized in that: The state drift rate is the probability change rate of the respiratory signal from the normal state to the pause state; the signal complexity change is obtained by calculating the change amplitude of the entropy value of the signal probability distribution, wherein the entropy value is used to quantify the disorder degree of the signal.

7. The sleep apnea detection method according to claim 1, characterized in that: The rules for determining the apnea event include: When the state drift rate exceeds a preset threshold, it is determined that the respiratory signal has a tendency to transfer to a pause state; When the entropy value change amplitude of the signal complexity change exceeds a preset threshold, it is determined that the respiratory signal is in an abnormal state; When the number of times the above two conditions are met within the continuous detection window exceeds a preset value, it is determined as a apnea event.

8. A sleep apnea detection system, used to perform the sleep apnea detection method according to any one of claims 1 to 7, characterized in that: include: A signal acquisition module, used to collect the user's breathing signal; A signal processing module, used for performing denoising and filtering on the respiratory signal to extract an effective respiratory frequency component; A state modeling module, used to construct state variables based on the processed respiratory signal and establish a dynamic evolution model of the respiratory signal; A feature extraction module is used to extract characteristic parameters of state drift rate and signal complexity change from the dynamic evolution model; A determination module, used for determining whether a apnea event occurs based on the characteristic parameters; The alarm module is used to trigger an alarm signal when a respiratory arrest event is determined to have occurred.

9. A medical device, characterized in that: include: At least one sensor, used to collect a breathing signal of a user; at least one processor, configured to execute the sleep apnea detection method according to any one of claims 1 to 7; At least one alarm device is used to send an alarm signal to the user or medical institution when a apnea event is detected.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the sleep apnea detection method according to any one of claims 1 to 7 is implemented.

Images