Method for screening paroxysmal atrial fibrillation based on heart rate variability characteristics

Abstract

The invention discloses a paroxysmal atrial fibrillation screening method based on heart rate variability characteristics, and relates to the technical field of medical diagnosis, and the paroxysmal atrial fibrillation screening method comprises the following steps: collecting an electrocardiosignal of a subject, decomposing the electrocardiosignal into heart rate variability sub-signals, and mapping the heart rate variability sub-signals to a characteristic space; calculating a critical moderation index and a dynamic stability parameter, generating an atrial fibrillation precursor state map, and determining a state transition critical point; constructing a pulse coupled oscillator network, and measuring system synchronization change and chaotic transition characteristics; measuring the desynchronization degree of the pulse coupled oscillator network and the atrial fibrillation precursor state map matching degree, extracting atrial fibrillation precursor signal features by using a stochastic resonance technology, and generating an atrial fibrillation occurrence probability value; and establishing an individualized calibration mechanism, carrying out individualized adjustment on the occurrence probability of the atrial fibrillation, and carrying out early warning. According to the method, a multi-level individualized calibration mechanism is established, the atrial fibrillation occurrence probability is dynamically adjusted in combination with the historical attack mode, time distribution and induction factors of the patient, and precise early warning is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical diagnosis, and in particular to a method for screening paroxysmal atrial fibrillation based on heart rate variability characteristics. Background Art

[0002] Paroxysmal atrial fibrillation is a common and highly prevalent arrhythmia disease, characterized by intermittent and irregular episodes. Traditional diagnostic methods rely on electrocardiogram recordings during symptom attacks. However, due to the unpredictability of the attacks, it is often difficult to capture them in a timely manner. Although long-term electrocardiogram monitoring can improve the detection rate, the device is inconvenient to wear and consumes a large amount of medical resources.

[0003] Existing atrial fibrillation prediction methods mainly rely on single linear analysis or simple statistical features, failing to fully consider the complex non-linear dynamic characteristics of cardiac electrophysiological activities, resulting in low prediction accuracy and insufficient individualization. In addition, the lack of an effective identification mechanism for the critical state transition before the occurrence of atrial fibrillation limits the practicality of the clinical warning system. Summary of the Invention

[0004] The present invention provides a method for screening paroxysmal atrial fibrillation based on heart rate variability characteristics, which is used to solve the technical problems of low accuracy of atrial fibrillation prediction, insufficient individualization, and difficulty in identifying critical state transitions in the prior art.

[0005] In view of this, the first aspect of the present invention provides a method for screening paroxysmal atrial fibrillation based on heart rate variability characteristics, including: collecting the electrocardiogram signals of the subject and decomposing them into heart rate variability sub-signals, mapping the heart rate variability sub-signals to the feature space through phase space reconstruction technology; calculating the critical slowing index and dynamic stability parameters in the feature space, generating an atrial fibrillation precursor state map and determining the state transition critical point; constructing a pulse-coupled oscillator network, using the characteristic parameters of the heart rate variability sub-signals as inputs, measuring the system synchronization change and chaotic transition characteristics; measuring the desynchronization degree of the pulse-coupled oscillator network and the matching degree with the atrial fibrillation precursor state map, using stochastic resonance technology to extract the atrial fibrillation precursor signal characteristics, generating an atrial fibrillation occurrence probability value; establishing an individualized calibration mechanism, making an individualized adjustment to the atrial fibrillation occurrence probability and giving a warning.

[0006] Optionally, collect the electrocardiogram (ECG) signal of the subject and decompose it into heart rate variability sub-signals. Mapping the heart rate variability sub-signals to the feature space through phase space reconstruction technology includes: collecting the ECG signal of the subject, and decomposing the ECG signal into high-frequency sub-signals, low-frequency sub-signals, and very low-frequency sub-signals; performing time-delay embedding processing on each frequency band sub-signal to determine the embedding dimension and time delay; constructing a phase space trajectory based on the processed sub-signals, projecting the phase space trajectory onto the two-dimensional plane corresponding to the main feature direction to obtain the phase space feature distribution; calculating the dynamic characteristic parameters of the phase space feature distribution to form a non-linear dynamic feature set.

[0007] Optionally, generate an atrial fibrillation precursor state map and determine the state transition critical point, including: based on the non-linear dynamic feature set, calculate the critical slowing-down index and the dynamic stability parameter; construct a two-dimensional state plane, with the critical slowing-down index as the abscissa and the dynamic stability parameter as the ordinate, map the ECG activity state points of the subject to generate an atrial fibrillation precursor state map; determine the state transition critical point through cluster analysis, and calculate the distance between the ECG activity state points of the subject and the state transition critical point to analyze the change trend.

[0008] Optionally, construct a pulse-coupled oscillator network, including: constructing a plurality of non-linear oscillator units, each oscillator unit corresponding to a frequency component of the heart rate variability sub-signal, and the internal parameters of the oscillator unit being determined by the non-linear dynamic characteristics of the corresponding sub-signal; defining the coupling topology between the oscillator units and establishing a weighted coupling matrix; setting a pulse trigger threshold, and generating a pulse when the state variable of the oscillator unit exceeds the pulse trigger threshold, and the pulse affects other connected oscillator units through the coupling topology.

[0009] Optionally, measure the system synchronization change and chaos transition characteristics, including: setting initial conditions and iteratively solving the dynamic equations of the pulse-coupled oscillator network to obtain the time series of each oscillator unit; calculating the phase synchronization index between each oscillator unit and constructing a time-varying synchronization matrix; quantifying the synchronization change characteristics by calculating the change rate of the maximum eigenvalue of the time-varying synchronization matrix over time; based on the phase space reconstruction of the time series, calculating the maximum Lyapunov exponent and the correlation entropy to quantify the chaos transition characteristics; combining the synchronization change characteristics and the chaos transition characteristics to form a dynamic state descriptor.

[0010] Optionally, determining the degree of desynchronization of the pulse-coupled oscillator network and the matching degree with the premonitory state map of atrial fibrillation includes: defining a desynchronization metric function, calculating the overall desynchronization degree of the pulse-coupled oscillator network; mapping the dynamic state descriptors onto the premonitory state map of atrial fibrillation to form a state trajectory; calculating the minimum distance between the state trajectory and the state transition critical points in the premonitory state map of atrial fibrillation to quantify the matching degree; constructing a bistable stochastic resonance model, using the desynchronization degree and the matching degree as inputs, and introducing a noise signal; extracting the statistical features of the output of the bistable stochastic resonance model, calculating a normalization index based on the statistical features, and obtaining the atrial fibrillation occurrence probability value.

[0011] Optionally, performing individualized adjustment on the atrial fibrillation occurrence probability and giving an early warning includes: based on the historical electrocardiogram monitoring data and atrial fibrillation attack records of the subject, extracting the atrial fibrillation attack pattern features, where the atrial fibrillation attack pattern features include the time distribution of atrial fibrillation attacks and the atrial fibrillation inducing factors; constructing an individualized feature analysis framework, where the individualized feature analysis framework includes a time feature layer, a physiological state layer, and an environmental factor layer; establishing an individualized parameter adjustment mechanism, converting the atrial fibrillation attack pattern features into calibration coefficients, multiplying them with the atrial fibrillation occurrence probability value to form an individualized probability assessment; establishing a time probability distribution map based on the time distribution of atrial fibrillation attacks; generating a graded early warning signal based on the individualized probability assessment and the time probability distribution map.

[0012] The second aspect of the present invention provides a paroxysmal atrial fibrillation screening system based on heart rate variability features, including: a signal acquisition and mapping module, configured to acquire the electrocardiogram signal of the subject and decompose it into heart rate variability sub-signals, and map the heart rate variability sub-signals to the feature space through phase space reconstruction technology; a premonitory state analysis module, configured to calculate the critical slowing index and the dynamic stability parameter in the feature space, generate a premonitory state map of atrial fibrillation and determine the state transition critical points; a network dynamics simulation module, configured to construct a pulse-coupled oscillator network, use the characteristic parameters of the heart rate variability sub-signals as inputs, and measure the system synchronization change and the chaotic transition characteristics; a signal feature extraction module, configured to determine the degree of desynchronization of the pulse-coupled oscillator network and the matching degree with the premonitory state map of atrial fibrillation, extract the premonitory atrial fibrillation signal features using stochastic resonance technology, and generate the atrial fibrillation occurrence probability value; an early warning module, configured to establish an individualized prediction model, and identify and give an early warning of the atrial fibrillation occurrence pattern of the subject through parameter update.

[0013] The beneficial effects of the present invention are as follows: By using phase space reconstruction technology and critical slowing down indicators, the present invention constructs an atrial fibrillation precursor state map, realizing the early identification of the state before atrial fibrillation attack and the precise positioning of the transition critical point. By introducing a pulse-coupled oscillator network and using the non-linear dynamic characteristics of heart rate variability sub-signals as input parameters, the ability to characterize complex cardiac dynamic systems is improved by measuring the changes in system synchrony and chaotic transition characteristics. Stochastic resonance technology is used to extract the weak signal characteristics of atrial fibrillation precursors from the noise background, and the degree of desynchronization is combined with the matching degree of the state map to enhance the detection sensitivity of atrial fibrillation precursor signals. A multi-level individual calibration mechanism is established, and by combining the patient's historical attack patterns, time distributions, and inducing factors, the probability of atrial fibrillation occurrence is dynamically adjusted to achieve precise early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following-described drawings are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0015] Figure 1 FIG. is the overall flowchart of a screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics.

[0016] Figure 2 FIG. is the signal processing flowchart of a screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics.

[0017] Figure 3 FIG. is the synchronization feature measurement flowchart of a screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the objectives, features, and advantages of the present invention more obvious and understandable, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the following-described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0019] Example 1, referring to Figures 1 to 3 , is the first embodiment of the present invention. This embodiment provides a screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics. The overall flowchart is as shown in Figure 1 and includes: S1: Collect the electrocardiogram (ECG) signals of the subject and decompose them into heart rate variability sub-signals, and map the heart rate variability sub-signals to the feature space through phase space reconstruction technology.

[0020] In a preferred embodiment, the signal processing flow chart is as Figure 2 shown, including the following steps: S1.1: Collect the ECG signals of the subject, and decompose the ECG signals into high-frequency sub-signals, low-frequency sub-signals, and very low-frequency sub-signals through frequency domain decomposition method.

[0021] First, use a standard ECG monitoring device (such as a Holter monitor) to collect the ECG signals of the subject for no less than 24 hours to ensure that sufficient ECG data is captured, including data during possible paroxysmal atrial fibrillation attack periods and non-attack periods. The sampling frequency is set to 200Hz - 1000Hz to ensure the validity and integrity of the signals.

[0022] Subsequently, preprocess the collected original ECG signals, including baseline drift correction, high-frequency noise filtering, and R-wave detection. Among them, wavelet transform method is used for baseline drift correction, low-pass filter (cutoff frequency 45Hz) is used for high-frequency noise filtering, and Pan-Tompkins algorithm is used for R-wave detection. A clear R-wave sequence is obtained through preprocessing, and the intervals between adjacent R-waves are calculated to obtain the RR interval sequence.

[0023] Then, interpolate and resample the RR interval sequence (sampling frequency 4Hz) to obtain an equally spaced sampling point sequence for subsequent spectral analysis. Then, use the autoregressive power spectral density estimation method to perform spectral analysis on the resampled RR interval sequence, and decompose the heart rate variability sub-signals into high-frequency sub-signals (0.15 - 0.4Hz, mainly reflecting parasympathetic nerve activity), low-frequency sub-signals (0.04 - 0.15Hz, reflecting the comprehensive regulation of sympathetic and parasympathetic nerves), and very low-frequency sub-signals (0.0033 - 0.04Hz, reflecting thermoregulation, vasomotor, and renin-angiotensin system activities). To obtain the time-domain representation of each frequency band sub-signal, use a 4th-order Butterworth band-pass filter to filter the resampled RR interval sequence, and extract the time-domain signals corresponding to the respective frequency bands for subsequent analysis.

[0024] S1.2: Perform time-delay embedding processing on each frequency band sub-signal, determine the embedding dimension through the nearest neighbor analysis method, and determine the time delay through information entropy analysis.

[0025] According to the Takens embedding theorem, perform phase space reconstruction on each frequency band sub-signal, and map the one-dimensional time series to a multi-dimensional phase space.

[0026] To determine the optimal parameters, the false nearest neighbor algorithm is used to determine the optimal embedding dimension. This algorithm is based on the following principle: when the embedding dimension is insufficient, points that should not be adjacent in the phase space will appear "pseudo" adjacent due to dimensional projection; as the embedding dimension increases, the proportion of such "pseudo" adjacent points will decrease. The specific steps are as follows: by calculating the distance change between the nearest neighbor point pairs under different embedding dimensions, determine the dimension when the ratio of distance change is first lower than the false nearest neighbor tolerance threshold as the optimal embedding dimension. At the same time, select the time delay corresponding to the first local minimum of the mutual information as the optimal delay.

[0027] In this embodiment, the false nearest neighbor tolerance threshold is set to 0.01. The optimal embedding dimension of different frequency band sub-signals is usually between 3 and 10, and the value range of the time delay is usually 5 to 30 sampling points. The specific values are determined according to the signal characteristics and sampling frequency to adapt to the characteristic differences of signals from different subjects.

[0028] S1.3: Construct a phase space trajectory based on the processed sub-signal, and project the phase space trajectory onto the two-dimensional plane corresponding to the main eigenvector through cross-section mapping to obtain the phase space feature distribution.

[0029] Based on the time-delay embedded processed sub-signal, construct a phase space trajectory for each frequency band sub-signal. For the convenience of analysis and visualization, determine the main eigenvector direction of the phase space trajectory through principal component analysis (PCA), and project the trajectory onto the two-dimensional plane composed of the first two principal components.

[0030] This process adopts a standardized processing flow, which specifically includes: first, perform standardized processing; then calculate the covariance matrix of the phase space trajectory, perform eigenvalue decomposition on the covariance matrix, and select the eigenvectors corresponding to the two largest eigenvalues as the main eigenvector directions. These two eigenvectors can usually retain 70%-90% of the variance information of the original data; finally, project the phase space trajectory onto the two-dimensional plane composed of these two main eigenvector directions to obtain the phase space feature distribution on the two-dimensional plane.

[0031] S1.4: Calculate the dynamic characteristic parameters of the phase space feature distribution to form a non-linear dynamic feature set representing heart rate variability.

[0032] Based on the phase space feature distribution constructed in the foregoing steps, calculate the dynamic characteristic parameters, including dimension features, stability indices, and complexity parameters. Among them, the dimension features include correlation dimension, box dimension, and information dimension; the stability indices include the maximum Lyapunov exponent (reflecting the degree of sensitivity of the system to initial conditions) and the Poincaré plot analysis index (reflecting the short-term and long-term change characteristics of heart rate variability); the complexity parameters include approximate entropy, sample entropy, and multi-scale entropy.

[0033] The specific calculation methods are as follows: The correlation dimension is calculated by the Grassberger-Procaccia algorithm, which reflects the complexity and degrees of freedom of the system; the box dimension is calculated by the grid counting method, which characterizes the filling degree of the phase space trajectory; the information dimension is calculated by information entropy, which is used to quantify the information complexity of the system; the calculation of the largest Lyapunov exponent adopts the Rosenstein algorithm; the Poincaré plot analysis indicators include SD1, SD2 and their ratio SD1 / SD2, where SD1 represents short-term variability and SD2 represents long-term variability; the calculation of approximate entropy and sample entropy adopts the template matching method, and the multi-scale entropy is realized by calculating the sample entropy after multi-scale coarse graining.

[0034] By integrating the above-obtained features, a non-linear dynamics feature set characterizing heart rate variability is formed. These features together reflect the non-linear dynamics characteristics of the heart rate variability sub-signals and can be used for subsequent screening and classification of paroxysmal atrial fibrillation.

[0035] Through the above steps, the present invention realizes a comprehensive analysis of the heart rate variability sub-signals, overcoming the limitation that traditional linear analysis methods cannot capture the complex dynamic characteristics of heart rate variability. Compared with the prior art, through phase space reconstruction and non-linear dynamics feature extraction, the present invention can more accurately describe the complex characteristics of the cardiac autonomic regulation mechanism, improve the screening sensitivity and specificity of paroxysmal atrial fibrillation, and provide a more reliable diagnostic basis for clinical practice.

[0036] S2: Calculate the critical slowing-down index and dynamic stability parameter in the feature space, generate the atrial fibrillation precursor state map and determine the state transition critical point; In a preferred embodiment, the following steps are included: S2.1: Based on the non-linear dynamics feature set, calculate the critical slowing-down index and dynamic stability parameter.

[0037] First of all, the critical slowing-down index in the present invention refers to the dynamic characteristic change parameter shown before the system approaches the critical transition, including the autoregressive coefficient, the variance change rate and the autocorrelation function decay rate. Among them, the autoregressive coefficient is obtained by performing a first-order autoregressive modeling on the heart rate variability time series, the variance change rate is obtained by calculating the variance change of the heart rate variability features within a continuous time window, and the autocorrelation function decay rate is obtained by calculating the decay rate of the autocorrelation function of the heart rate variability feature time series with time delay.

[0038] Secondly, the dynamic stability parameter in the present invention refers to an index of the system's response ability to external disturbances, including the disturbance recovery time, the steady-state deviation degree, and the system elasticity index. Among them, the disturbance recovery time is obtained by analyzing the time required for the heart rate variability characteristics to return to a stable state after experiencing significant fluctuations, the steady-state deviation degree is obtained by calculating the deviation degree of the heart rate variability characteristics relative to their historical steady state, and the system elasticity index is obtained by analyzing the response curve of the heart rate variability characteristics to minute disturbances.

[0039] S2.2: Construct a two-dimensional state plane, with the critical slowing index as the abscissa and the dynamic stability parameter as the ordinate, map the electrocardiogram activity state points of the subject, and generate a pre-atrial fibrillation state map.

[0040] After calculating the critical slowing index and the dynamic stability parameter, the present invention constructs a two-dimensional state plane using these two parameters, where the critical slowing index is used as the abscissa and the dynamic stability parameter is used as the ordinate; subsequently, the critical slowing index and the dynamic stability parameter values of the subject at different time points are used as coordinate points and mapped onto this two-dimensional state plane. By connecting adjacent state points in the time series, a state trajectory is generated, thereby intuitively reflecting the time-evolution characteristics of the subject's electrocardiogram activity state. Then, based on historical case data, different risk-level regions are marked on the two-dimensional state plane to form a pre-atrial fibrillation state map, providing an intuitive reference for clinical judgment.

[0041] S2.3: Identify the region where the state point density changes through cluster analysis, determine the state transition critical point, and calculate the distance between the electrocardiogram activity state point of the subject and the state transition critical point to analyze the change trend of the electrocardiogram activity state point of the subject.

[0042] Based on the generated pre-atrial fibrillation state map, the present invention applies clustering methods such as the density peak clustering algorithm or DBSCAN to perform cluster analysis on the historical state points on the state plane, thereby identifying the regions where the state point density changes significantly. By calculating the state point density gradient of each cluster boundary region, the present invention determines the state transition critical point, which represents the critical position where the system transitions from a stable state to an unstable state and is a key indicator for early warning.

[0043] Next, the present invention calculates the Euclidean distance between the current electrocardiogram activity state point of the subject and the critical point and analyzes the change trend of this distance over time. Among them, the time interval parameter is preferably set to 1 / 10 of the monitoring period. When the distance difference is negative, it indicates that the state point is approaching the critical point and the risk increases; when the distance difference is positive, it indicates that the state point is moving away from the critical point and the risk decreases.

[0044] In addition, the present invention also calculates the rate at which the state point approaches the critical point. When this rate is positive and continuously increasing, it indicates that the subject's electrocardiogram activity state is accelerating towards the critical point, and vigilance should be increased. When the rate reaches or exceeds a preset threshold (determined based on clinical experience), the system will trigger an early warning mechanism.

[0045] In summary, by calculating the critical slowing-down index and the dynamic stability parameter, combining with the atrial fibrillation precursor state map and the analysis of the state transition critical point, the present invention realizes the early warning of paroxysmal atrial fibrillation, and has higher prediction accuracy and lead time compared with traditional methods. This method can capture the atrial fibrillation precursor features that are difficult to detect by conventional electrocardiogram analysis, effectively identify patients in a high-risk state of atrial fibrillation, provide a time window for clinical intervention, and improve the effect of atrial fibrillation prevention and treatment.

[0046] S3: Construct a network of pulse-coupled oscillators, use the characteristic parameters of the heart rate variability sub-signals as inputs, and measure the changes in system synchrony and the characteristics of chaotic transitions.

[0047] In a preferred embodiment, the flow chart for measuring the synchronization characteristics is as Figure 3 shown, and includes the following steps: S3.1: Construct multiple non-linear oscillator units, each oscillator unit corresponding to a frequency component of the heart rate variability sub-signal, and the internal parameters of the oscillator unit are determined by the non-linear dynamic characteristics of the corresponding sub-signal.

[0048] Specifically, when constructing the oscillator unit, a modified Van der Pol oscillator model is adopted, and its general form can be expressed as:

[0049] where x is the oscillator state variable, representing the current position of the oscillator; is the damping coefficient, controlling the convergence of the oscillation; is the natural angular frequency, determining the oscillation period; is the non-linear term, exciting the oscillation at small amplitudes and suppressing it at large amplitudes; is the first derivative of the state variable, that is, the velocity of a single oscillator unit; the second derivative of the state variable, that is, the acceleration of a single oscillator unit.

[0050] Furthermore, for each frequency band sub-signal of the heart rate variability (high-frequency sub-signal, low-frequency sub-signal, and very low-frequency sub-signal), a corresponding oscillator unit is constructed respectively, thereby forming a network structure of at least three basic oscillator units.

[0051] Furthermore, a mapping relationship is established between the internal parameters of each oscillator unit and the non-linear dynamic characteristics of the corresponding sub-signal. Specifically, the natural frequency of the oscillator Obtain the main frequency value of the corresponding sub-signal; damping coefficient Is proportional to the dimensionality feature of the sub-signal, and its value range is from 0.1 to 5.0; the non-linear term coefficient is related to the complexity parameter of the sub-signal and is mapped to the range of 0.5 to 3.0 through normalization processing.

[0052] S3.2: Define the coupling topology between oscillator units and establish a weighted coupling matrix according to physiological relevance.

[0053] Specifically, according to physiological relevance, define the coupling topology between oscillator units. High-frequency sub-signals mainly reflect parasympathetic nerve activities, low-frequency sub-signals reflect the comprehensive regulation of sympathetic and parasympathetic nerves, and extremely low-frequency sub-signals are related to physiological processes with long time scales such as body temperature regulation. Based on these physiological relationships, establish a weighted coupling matrix C, where the element C ij Represents the coupling strength of the i-th oscillator unit to the j-th oscillator unit. The dimension of the weighted coupling matrix C is N×N, where N is the total number of oscillator units (at least 3).

[0054] Among them, the coupling strength is proportional to the correlation between heart rate variability sub-signals. For any two sub-signals S i and S j , calculate their Pearson correlation coefficient r ij , and map it to the coupling strength C ij :

[0055] Among them, Is the coupling strength adjustment factor, and its value range is from 0.01 to 1.0; Is the absolute value of the correlation coefficient, representing the correlation strength; Is the sign function of the correlation coefficient, and takes the value of 1 when , and takes the value of 0 when , which is used to determine the inhibitory or promoting nature of the coupling.

[0056] S3.3: Set the pulse trigger threshold. When the state variable of the oscillator unit exceeds the pulse trigger threshold, a pulse is generated, and the pulse affects other connected oscillator units through the coupling topology.

[0057] Specifically, set a pulse trigger threshold for each oscillator unit. When the state variable of the oscillator unit exceeds this threshold, a normalized pulse is generated. The pulse value is 1 at the moment when the state variable crosses the threshold, and 0 at other times, and the pulse is generated only at the moment when the state variable first crosses the threshold to avoid generating multiple pulses due to continuous crossing of the threshold.

[0058] In addition, the selection of the pulse trigger threshold is based on the statistical analysis of the corresponding sub-signal amplitude, and is usually set to 1.5 to 2.5 times the standard deviation of the sub-signal to ensure an appropriate frequency of pulse generation.

[0059] S3.4: Map the nonlinear dynamic characteristics to the control parameters of the pulse-coupled oscillator network.

[0060] Among them, the control parameters include the natural frequency, damping coefficient, and nonlinear term coefficient. The natural frequency is determined by the main frequency of the sub-signal; the damping coefficient is obtained by adding an adjustment value proportional to the dimensionality feature to the basic damping coefficient; the nonlinear term coefficient is determined by adding an adjustment value proportional to the complexity parameter to the basic nonlinear coefficient; the stability index is mapped by adjusting the recovery time parameter of the pulse-coupled oscillator, and the lower the stability, the longer the recovery time.

[0061] S3.5: Set the initial conditions and iteratively solve the dynamic equations of the pulse-coupled oscillator network to obtain the time series of each oscillator unit.

[0062] Based on the above parameter settings, a complete dynamic equation of the pulse-coupled oscillator network is constructed. This equation describes the state changes of each oscillator unit and the interaction relationship between units, as follows:

[0063] where i = 1, 2,..., N, and N is the total number of oscillator units. is the natural frequency. is the damping coefficient. is the nonlinear term coefficient. The left side of the equation describes the free motion of the i-th oscillator, and the right side describes the pulse-coupled influence from other oscillators.

[0064] Furthermore, set the initial conditions. For each oscillator unit, the initial position and initial velocity are determined based on the initial segment data of the corresponding sub-signal and are normalized; use the fourth-order Runge-Kutta numerical integration method to solve the above dynamic equations, set the time step to 0.01 seconds, and the total simulation time is the same as the acquisition time of the original electrocardiogram signal. This process generates the time series of each oscillator unit, and these time series will be used for subsequent synchronization and chaos characteristic analysis.

[0065] S3.6: Calculate the phase synchronization index between each oscillator unit, construct a time-varying synchronization matrix, and characterize the cooperative relationship between oscillator units.

[0066] For the time series of each oscillator unit, its instantaneous phase is calculated through Hilbert transform; based on the instantaneous phase, the phase synchronization index between any two oscillator units i and j is calculated. Among them, the phase synchronization index is obtained by the complex exponential average of the phase differences between the two units, and the value range of the phase synchronization index is from 0 to 1. The closer the value is to 1, the higher the degree of synchronization, and the closer the value is to 0, the lower the degree of synchronization.

[0067] S3.7: Quantify the change characteristics of the synchronization of the pulse-coupled oscillator network by calculating the rate of change of the maximum eigenvalue of the time-varying synchronization matrix over time.

[0068] For each time point t, calculate the eigenvalues of the time-varying synchronization matrix and determine the maximum eigenvalue; calculate the rate of change of the maximum eigenvalue over time, which is approximately calculated by the central difference method for discrete time series. Among them, the absolute value of the rate of change characterizes the rate of change of the synchronization of the pulse-coupled oscillator network, and its change trend is closely related to the stability of cardiac electrical activity. Under normal sinus rhythm, the fluctuation of the rate of change is small; while in the pre-fibrillation state, the fluctuation of the rate of change increases significantly.

[0069] S3.8: Based on the phase space reconstruction of the time series, calculate the largest Lyapunov exponent and the correlation entropy of the pulse-coupled oscillator network to quantify the chaotic transition characteristics.

[0070] For the time series of each oscillator unit, the time-delay embedding method is used for phase space reconstruction to construct a multi-dimensional state vector. The selection method and steps of the embedding dimension and time delay are the same as those in step S1.2. Calculate the largest Lyapunov exponent in the reconstructed phase space to characterize the chaotic degree of the system. The largest Lyapunov exponent is obtained by analyzing the divergence rate of initially close trajectories in the phase space. The largest Lyapunov exponent greater than 0 indicates that the system has chaotic characteristics, and the larger the largest Lyapunov exponent, the higher the degree of chaos. Calculate the correlation entropy to quantify the unpredictability of the system. The correlation entropy is obtained by analyzing the conditional probability of the system state under different embedding dimensions. The larger the correlation entropy, the stronger the unpredictability of the system.

[0071] S3.9: Combine the change characteristics of synchronization and the chaotic transition characteristics to form a dynamic state descriptor of the pulse-coupled oscillator network.

[0072] It should be noted that the dynamic state descriptor of the pulse-coupled oscillator network provides a high-level feature representation of the heart rate variability sub-signal and can effectively capture the subtle changes in the pre-fibrillation state. Compared with traditional linear or single non-linear indicators, this feature set comprehensively considers the system's synergy and chaotic dynamics characteristics, improving the sensitivity and specificity of paroxysmal atrial fibrillation prediction.

[0073] Preferably, through the above steps, the present invention establishes a mapping mechanism from the heart rate variability sub-signal to the pulse-coupled oscillator network and extracts the key indicators characterizing the dynamic characteristics of the system. This method based on complex system theory can effectively capture the subtle changes in cardiac electrical activity and provide a new technical means for the early warning of paroxysmal atrial fibrillation.

[0074] S4: Determine the degree of desynchronization of the pulse-coupled oscillator network and the matching degree with the atrial fibrillation precursor state map, and use the stochastic resonance technology to extract the characteristics of the atrial fibrillation precursor signal and generate the atrial fibrillation occurrence probability value.

[0075] In a preferred embodiment, the following steps are included: S4.1: Define the desynchronization metric function and calculate the overall desynchronization degree of the pulse-coupled oscillator network.

[0076] Specifically, the desynchronization metric function D is defined as the statistical measure of the phase difference between oscillator units in the pulse-coupled oscillator network, and the specific expression is:

[0077] where N is the total number of oscillator units, are the phases of the i-th and j-th oscillator units at time t, respectively.

[0078] Furthermore, the following processing is required to calculate the overall desynchronization degree of the pulse-coupled oscillator network: perform Hilbert transform on the time series of each oscillator unit obtained in S3.5 to extract the instantaneous phase information; calculate the absolute value of the phase difference between any two oscillator units; normalize the phase differences of all oscillator unit pairs so that the value range of the desynchronization metric function D is [0, 1]; take the average value of the desynchronization metric function within a certain time window (preferably 10 seconds) as the overall desynchronization degree index.

[0079] It should be understood that when the D value is close to 0, it indicates that the network is in a highly synchronous state; when the D value is close to 1, it indicates that the network is in a completely desynchronized state. Under normal heart rhythm conditions, the D value is usually less than 0.3, while the D value increases significantly in the pre-onset stage of atrial fibrillation and usually exceeds 0.6.

[0080] S4.2: Map the dynamic state descriptor to the atrial fibrillation precursor state map to form a state trajectory.

[0081] Specifically, map the synchronization change characteristics and chaos transition characteristics in the dynamic state descriptor to the abscissa increment and ordinate increment of the state plane respectively; based on the two-dimensional state plane coordinate system constructed in S2.2, starting from the current electrocardiogram activity state point, superimpose the above coordinate increments to obtain a new state point; continuously record the position change of the state point over time to form a state trajectory.

[0082] It should be noted that the morphological characteristics of the state trajectory can reflect the dynamic characteristics of cardiac electrical activity. Under normal heart rhythm, the state trajectory usually shows stable small-range fluctuations; while in the pre-onset stage of atrial fibrillation, the state trajectory will show obvious directional drift characteristics and gradually approach the state transition critical point.

[0083] S4.3: Calculate the minimum distance between the state trajectory and the state transition critical point in the atrial fibrillation precursor state atlas to quantify the matching degree.

[0084] Specifically, extract the position coordinates of the nearest K state points on the state trajectory. In the present invention, K is 100; for each critical point in the set of state transition critical points, calculate the Euclidean distance respectively; take the minimum value of all distances as the minimum distance from the state trajectory to the critical point. Based on the minimum distance d min , define the matching degree M as:

[0085] where is the scaling coefficient, determined according to the scale of the state plane, and the preferred value range is 0.5 - 2.0. This design makes the value range of the matching degree M be (0, 1). The larger the value, the closer the state trajectory is to the critical point, and the higher the risk of atrial fibrillation.

[0086] S4.4: Construct a bistable stochastic resonance model with the degree of desynchronization and the matching degree as inputs.

[0087] Specifically, construct a bistable stochastic resonance model based on the overdamped Langevin equation, and its mathematical expression is:

[0088] where z is the system state variable, a and b are system parameters ( , ), is the input signal, is the noise term. This model has a double-well structure and can effectively amplify weak signals.

[0089] In addition, the input signal is constructed based on the degree of desynchronization D and the matching degree M, and its expression is:

[0090] where is the signal strength coefficient, w1 and w2 are characteristic frequencies, which are set to 0.1 Hz and 0.05 Hz respectively. Through this design, the information of the degree of desynchronization and the matching degree can be effectively coupled into the stochastic resonance system.

[0091] It should be noted that the selection of the parameters a and b of the bistable stochastic resonance model is crucial for the system performance. In the present invention, the value range of the parameter a is 1.0 - 2.0, and the value range of the parameter b is 0.5 - 1.5. Specifically, it can be appropriately adjusted according to the physiological characteristics of the subject. For example, for older subjects, smaller a values and larger b values can be selected to adapt to the dynamic characteristics of their cardiovascular system.

[0092] S4.5: Introduce a noise signal into the bistable stochastic resonance model, and optimize the system parameters to maximize the gain of the weak pre - atrial fibrillation signal.

[0093] It should be noted that the noise signal introduced into the bistable stochastic resonance model selects Gaussian white noise, and its variance is a key parameter. The noise intensity needs to satisfy:

[0094] to ensure the generation of the stochastic resonance effect.

[0095] Furthermore, to maximize the gain of the weak pre - atrial fibrillation signal, the system parameters need to be optimized. The specific steps are as follows: fix the values of a and b, and within the allowable range of the noise intensity scan in small steps (such as 0.05a); for each value, calculate the signal - to - noise ratio SNR of the system output; select the noise intensity that maximizes the SNR as the optimization result.

[0096] The bistable stochastic resonance system with optimized parameters can amplify the weak pre - atrial fibrillation signal by 5 - 10 times while keeping the time - frequency characteristics of the signal unchanged. This characteristic is crucial for identifying the early features of atrial fibrillation.

[0097] S4.6: Extract the statistical features of the output of the bistable stochastic resonance model.

[0098] Among them, the statistical features include the transition rate, the dwell - time distribution, and the spectral peak power. The transition rate is defined as the average number of times the system state crosses zero per unit time, which reflects the instability of the system. Under normal conditions, the R value is stable within the range of 0.5 - 2.0 times per second; in the pre - atrial fibrillation stage, the R value increases significantly and shows obvious fluctuation characteristics, providing an early indication of pre - atrial fibrillation. The dwell time refers to the length of time that the system state continuously stays at a steady - state position. Specifically, by statistically analyzing the dwell - time sequences of the system in the positive and negative steady - states respectively and calculating their probability distributions, under normal conditions, the dwell - time distribution usually shows an exponential distribution; in the pre - atrial fibrillation stage, the distribution pattern will change significantly, showing a power - law distribution characteristic, which can be used as an important sign of pre - atrial fibrillation.

[0099] In addition, power spectral density analysis is performed on the output signal of the bistable stochastic resonance model to extract the frequencies and powers of the main spectral peaks. Under normal conditions, the main spectral peak frequency is stable within the range of 0.04 - 0.15 Hz; while in the pre-atrial fibrillation stage, the main spectral peak frequency will shift to a higher frequency band (0.15 - 0.4 Hz), and the spectral peak power increases significantly. This change in spectral characteristics is consistent with the frequency domain characteristics of cardiac electrical activities before the onset of atrial fibrillation.

[0100] S4.7: Calculate the normalized index based on statistical features and map it to the atrial fibrillation occurrence probability value through a non-linear conversion function.

[0101] Specifically, the Sigmoid function is used as the non-linear conversion function to map the normalized index to the atrial fibrillation occurrence probability value.

[0102] Preferably, through the above steps, the present invention can detect significant changes 10 - 30 minutes before the onset of atrial fibrillation, providing a critical time window for early clinical intervention. In addition, the stochastic resonance technology is used to effectively amplify weak atrial fibrillation precursor signals, improving the detection sensitivity, and the accuracy rate can reach more than 85% in clinical verification, solving the technical problems of short atrial fibrillation early warning time and low accuracy rate in the prior art.

[0103] S5: Establish an individualized calibration mechanism to individually adjust the atrial fibrillation occurrence probability and give an early warning.

[0104] In a preferred embodiment, it includes the following steps: S5.1: Based on the subject's historical electrocardiogram monitoring data and atrial fibrillation attack records, extract the atrial fibrillation attack mode characteristics. Among them, the atrial fibrillation attack mode characteristics include the time distribution of atrial fibrillation attacks and atrial fibrillation inducing factors. Specifically, collect the electrocardiogram monitoring data of the subject for several consecutive days, mark the atrial fibrillation attack time points, record the start time, duration, and end time of each attack to form a time series; then, analyze the time distribution characteristics of atrial fibrillation attacks, including intra-day distribution, inter-day distribution, and attack interval distribution; next, correlate the subject's activity log and physiological parameter records to identify potential atrial fibrillation inducing factors, mainly including physical activity, emotional state, dietary factors, medication conditions, sleep quality, etc.; finally, conduct a statistical analysis of various inducing factors within 4 hours before each atrial fibrillation attack, and calculate the association strength between each factor and atrial fibrillation attacks. The greater the association strength, the stronger the correlation between the factor and atrial fibrillation attacks, and it may be an important inducing factor.

[0105] S5.2: Construct an individualized feature analysis framework, which includes a time feature layer, a physiological state layer, and an environmental factor layer, for analyzing atrial fibrillation attack mode characteristics.

[0106] Specifically, a three - layer nested individualized feature analysis framework is constructed, including: a time feature layer, a physiological state layer, and an environmental factor layer. The time feature layer processes the atrial fibrillation attack pattern related to time and serves as the basic layer of the framework; the physiological state layer analyzes the relationship between the physiological state of the subject and the atrial fibrillation risk and serves as the middle layer of the framework; the environmental factor layer evaluates the impact of external environmental factors on the atrial fibrillation attack and serves as the outer layer of the framework.

[0107] In the time feature layer, a time weight function is defined, comprehensively considering the hour distribution weight, date distribution weight, and interval distribution weight; in the physiological state layer, a physiological state scoring function is defined to perform weighted evaluation on each physiological parameter; in the environmental factor layer, an environmental impact function is defined to perform weighted evaluation on each environmental factor. Integrate the three - layer features to construct a comprehensive evaluation function, comprehensively considering the impacts of time features, physiological states, and environmental factors to form a comprehensive risk assessment mechanism.

[0108] S5.3: Establish an individualized parameter adjustment mechanism to convert the atrial fibrillation attack pattern features into calibration coefficients, which are multiplied by the atrial fibrillation occurrence probability value to form an individualized probability assessment for the subject.

[0109] Specifically, calculate the individualized calibration coefficient, which combines the comprehensive evaluation function and the basic calibration coefficient. As data accumulates, the influence degree of the individualized features gradually increases, making the calibration mechanism more in line with the actual situation of the subject. To improve the prediction accuracy, establish an adaptive adjustment mechanism to dynamically update the individualized weights according to the difference between the actual attack and the expected attack. This mechanism can continuously optimize the calibration coefficient according to the prediction results and improve the accuracy of individualized early warning. Then, multiply the individualized calibration coefficient by the atrial fibrillation occurrence probability value to obtain the individualized atrial fibrillation occurrence probability, which can better reflect the actual atrial fibrillation risk status of the subject compared with the original probability value.

[0110] S5.4: Based on the time distribution of atrial fibrillation attacks, establish a time probability distribution map to characterize the atrial fibrillation attack risk distribution in different time periods.

[0111] Furthermore, construct a two - dimensional time probability distribution map with the hours of a day and the dates of a week as the coordinate axes. This distribution map is calculated based on the time distribution features extracted in the previous steps and can visually display the atrial fibrillation attack risks in different time periods. To reduce the data sparsity problem, apply a smoothing processing method to optimize the time probability distribution map, making the risk assessment more continuous and reliable. The smoothed distribution map can more accurately reflect the relationship between time and atrial fibrillation risk. Then, combined with the current moment, extract the time risk factor from the time probability distribution map, and this risk factor reflects the atrial fibrillation attack probability based on historical statistics at the current moment.

[0112] S5.5: Generate a hierarchical early warning signal based on the individualized probability assessment and the time probability distribution map.

[0113] Furthermore, first, the individual atrial fibrillation occurrence probability and time risk factors are comprehensively considered to calculate the final risk score. Subsequently, multiple levels of warning thresholds are set, and the risk score is mapped to graded warning signals, including green warning (low-risk state), yellow warning (medium-low risk state), orange warning (medium-high risk state), and red warning (high-risk state).

[0114] S5.6: When the graded warning signal indicates a high-risk state of atrial fibrillation, preventive intervention suggestions are generated in combination with the inducing factors of atrial fibrillation.

[0115] It should be noted that the present invention has established a mapping relationship library between the inducing factors of atrial fibrillation and intervention measures, including specific intervention suggestions for inducing factors such as exercise, emotion, caffeine, alcohol, medication, and sleep. The intervention priorities are determined according to the correlation strength of the inducing factors, and the higher the correlation strength, the higher the intervention priority. When the warning level reaches orange or red, personalized intervention suggestions are generated, including risk assessment, immediate intervention measures, and medical contact reminders. An intervention effect feedback mechanism is established to evaluate the effectiveness of the intervention measures and update the intervention priority accordingly.

[0116] Preferably, through the above steps, the present invention has established a complete individualized atrial fibrillation warning and calibration mechanism. This mechanism fully considers the individual differences of the subjects and the characteristics of the atrial fibrillation attack pattern, integrates the effects of time, physiological, and environmental factors through a multi-level analysis framework, and realizes the precise calibration of the atrial fibrillation occurrence probability. Compared with the traditional fixed-threshold warning method, the individualized calibration mechanism of the present invention significantly improves the warning accuracy rate, reduces the false positive rate from 30%-40% of the traditional method to 15%-25%, and maintains a sensitivity of more than 85% at the same time. In addition, the graded warning signal and personalized intervention suggestions provide patients with timely and effective self-management tools, which helps to reduce the number of atrial fibrillation attacks and the risk of related complications.

[0117] Furthermore, this embodiment also provides a paroxysmal atrial fibrillation screening system based on heart rate variability characteristics, including: a signal acquisition and mapping module, configured to acquire the electrocardiogram signal of a subject and decompose it into heart rate variability sub-signals, and map the heart rate variability sub-signals to a feature space through phase space reconstruction technology; a precursor state analysis module, configured to calculate the critical slowing index and dynamic stability parameters in the feature space, generate an atrial fibrillation precursor state map, and determine the state transition critical point; a network dynamics simulation module, configured to construct a pulse-coupled oscillator network, use the characteristic parameters of the heart rate variability sub-signals as inputs, and measure the system synchronization change and chaos transition characteristics; a signal feature extraction module, configured to determine the desynchronization degree of the pulse-coupled oscillator network and the matching degree with the atrial fibrillation precursor state map, extract the atrial fibrillation precursor signal features using stochastic resonance technology, and generate an atrial fibrillation occurrence probability value; an early warning module, configured to establish an individualized prediction model, and identify and early warn the atrial fibrillation occurrence pattern of the subject through parameter update.

[0118] In summary, the present invention constructs an atrial fibrillation precursor state map through phase space reconstruction technology and critical slowing index, realizes the early identification of the state before atrial fibrillation attack and the precise positioning of the transition critical point. By introducing a pulse-coupled oscillator network and using the non-linear dynamic characteristics of heart rate variability sub-signals as input parameters, the ability to depict complex cardiac dynamic systems is improved by measuring the system synchronization change and chaos transition characteristics. Stochastic resonance technology is used to extract the weak signal features of atrial fibrillation precursors from the noise background, and the desynchronization degree is combined with the matching degree of the state map to enhance the detection sensitivity of atrial fibrillation precursor signals. A multi-level individualized calibration mechanism is established, and the atrial fibrillation occurrence probability is dynamically adjusted by combining the patient's historical attack pattern, time distribution, and inducing factors to achieve precise early warning.

[0119] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent substitution on some of the technical features; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics, characterized in that, Comprising: Collecting the electrocardiogram signals of a subject and decomposing them into heart rate variability sub-signals, and mapping the heart rate variability sub-signals to a feature space through phase space reconstruction technology; Calculating the critical slowing-down index and dynamic stability parameter in the feature space, generating an atrial fibrillation precursor state map and determining the state transition critical point; Constructing a pulse-coupled oscillator network, using the characteristic parameters of the heart rate variability sub-signals as inputs, and measuring the system synchronization change and chaotic transition characteristics; Determining the degree of desynchronization of the pulse-coupled oscillator network and the matching degree with the atrial fibrillation precursor state map, extracting the characteristics of the atrial fibrillation precursor signal by using stochastic resonance technology, and generating an atrial fibrillation occurrence probability value; Establishing an individualized calibration mechanism to individually adjust the atrial fibrillation occurrence probability and give an early warning.

2. The screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics according to claim 1, wherein The collecting the electrocardiogram signals of a subject and decomposing them into heart rate variability sub-signals, and mapping the heart rate variability sub-signals to a feature space through phase space reconstruction technology includes: Collecting the electrocardiogram signals of a subject, and decomposing the electrocardiogram signals into high-frequency sub-signals, low-frequency sub-signals and very low-frequency sub-signals; Performing time-delay embedding processing on each frequency band sub-signal to determine the embedding dimension and time delay; Constructing a phase space trajectory based on the processed sub-signals, and projecting the phase space trajectory onto a two-dimensional plane corresponding to the main feature direction to obtain a phase space feature distribution; Calculating the dynamic characteristic parameters of the phase space feature distribution to form a non-linear dynamic characteristic set.

3. The screening method for paroxysmal atrial fibrillation based on heart rate variability features according to claim 1, wherein The generating an atrial fibrillation precursor state map and determining the state transition critical point includes: Based on the non-linear dynamic characteristic set, calculating the critical slowing-down index and dynamic stability parameter; Constructing a two-dimensional state plane, with the critical slowing-down index as the abscissa and the dynamic stability parameter as the ordinate, mapping the electrocardiogram activity state points of the subject, and generating an atrial fibrillation precursor state map; Determining the state transition critical point through cluster analysis, and calculating the distance between the electrocardiogram activity state point of the subject and the state transition critical point to analyze the change trend.

4. The screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics according to claim 1, wherein, The constructing a pulse-coupled oscillator network includes: Constructing a plurality of non-linear oscillator units, each oscillator unit corresponding to a frequency component of the heart rate variability sub-signal, and the internal parameters of the oscillator unit being determined by the non-linear dynamic characteristics of the corresponding sub-signal; Defining the coupling topological structure between oscillator units and establishing a weighted coupling matrix; Setting a pulse trigger threshold, generating a pulse when the state variable of the oscillator unit exceeds the pulse trigger threshold, and the pulse affects other connected oscillator units through the coupling topological structure.

5. The screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics according to claim 1, characterized in that The measuring the system synchronization change and chaotic transition characteristics includes: Setting initial conditions and iteratively solving the dynamic equations of the pulse-coupled oscillator network to obtain the time series of each oscillator unit; Calculating the phase synchronization index between each oscillator unit and constructing a time-varying synchronization matrix; Quantifying the synchronization change characteristics by calculating the change rate of the maximum eigenvalue of the time-varying synchronization matrix with time; Based on the phase space reconstruction of the time series, calculating the largest Lyapunov exponent and correlation entropy to quantify the chaotic transition characteristics; Combining the synchronization change characteristics and the chaotic transition characteristics to form a dynamic state descriptor.

6. The screening method for paroxysmal atrial fibrillation based on heart rate variability characteristics according to claim 1, characterized in that The determination of the degree of desynchronization of the pulse-coupled oscillator network and the matching degree with the atrial fibrillation precursor state map includes: Defining a desynchronization metric function and calculating the overall desynchronization degree of the pulse-coupled oscillator network; Mapping the dynamic state descriptors onto the atrial fibrillation precursor state map to form a state trajectory; Calculating the minimum distance between the state trajectory and the state transition critical points in the atrial fibrillation precursor state map to quantify the matching degree; Constructing a bistable stochastic resonance model with the desynchronization degree and the matching degree as inputs and introducing a noise signal; Extracting the statistical features output by the bistable stochastic resonance model, calculating a normalization index based on the statistical features, and obtaining an atrial fibrillation occurrence probability value.

7. The screening method for paroxysmal atrial fibrillation based on heart rate variability features according to claim 1, wherein The individual adjustment of the atrial fibrillation occurrence probability and the early warning include: Based on the subject's historical electrocardiogram monitoring data and atrial fibrillation attack records, extracting atrial fibrillation attack pattern features, where the atrial fibrillation attack pattern features include the time distribution of atrial fibrillation attacks and atrial fibrillation inducing factors; Constructing an individual feature analysis framework, where the individual feature analysis framework includes a time feature layer, a physiological state layer, and an environmental factor layer; Establishing an individual parameter adjustment mechanism, converting the atrial fibrillation attack pattern features into calibration coefficients, multiplying them with the atrial fibrillation occurrence probability value to form an individual probability assessment; Establishing a time probability distribution map based on the time distribution of atrial fibrillation attacks; Generating a graded early warning signal based on the individual probability assessment and the time probability distribution map.

8. A paroxysmal atrial fibrillation screening system based on heart rate variability characteristics, characterized in that, Including: A signal acquisition and mapping module for acquiring the subject's electrocardiogram signal, decomposing it into heart rate variability sub-signals, and mapping the heart rate variability sub-signals to a feature space through phase space reconstruction technology; A precursor state analysis module for calculating the critical slowing index and dynamic stability parameters in the feature space, generating an atrial fibrillation precursor state map, and determining state transition critical points; A network dynamics simulation module for constructing a pulse-coupled oscillator network, using the characteristic parameters of the heart rate variability sub-signals as inputs, and measuring the system synchronization change and chaotic transition characteristics; A signal feature extraction module for determining the degree of desynchronization of the pulse-coupled oscillator network and the matching degree with the atrial fibrillation precursor state map, using stochastic resonance technology to extract atrial fibrillation precursor signal features, and generating an atrial fibrillation occurrence probability value; An early warning module for establishing an individual prediction model and identifying and warning the subject's atrial fibrillation occurrence pattern through parameter update.

Images