Cardiopulmonary coupling feature extraction method and system based on adaptive Fourier decomposition

Abstract

The present invention discloses a cardiopulmonary coupling feature extraction method and system based on adaptive Fourier decomposition. The method can take into account the dynamics of electrophysiological signals, extract the spectral features of electrophysiological signals and realize the adaptive decomposition of key rhythms. The specific steps are as follows: preprocessing the collected original single-lead electrocardiogram (ECG) signal and respiratory signal; extracting the heart beat interval R-R interval time series from the preprocessed ECG signal; using the adaptive bandwidth Fourier decomposition algorithm to process the respiratory signal time series and the R-R interval time series respectively to obtain the Fourier natural frequency band functions of the above two time series; obtaining the Fourier decomposition coefficients from the analytical form of the Fourier natural frequency band functions of the two time series at different frequencies, and calculating the CPC value based on the Fourier decomposition coefficients of the R-R and respiratory signal time series to construct a cardiopulmonary coupling map; extracting multiple types of cardiopulmonary coupling features based on the cardiopulmonary coupling map.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring and information technology, and in particular to a cardiopulmonary coupling feature extraction method and system based on adaptive Fourier decomposition. Background Art

[0002] The autonomic nervous system controls and regulates the activities and secretions of various organs, blood vessels, smooth muscles, and glands in the body, and is closely related to endocrine regulatory activities such as body temperature, sleep, and blood pressure. The integrity and regulatory ability of autonomic nervous system function play an important role in maintaining a stable physiological state. Recent clinical and electrophysiological studies have shown that the incidence of autonomic dysfunction is very high in patients receiving treatment in the intensive care unit (ICU). Autonomic dysfunction may lead to various unexpected consequences. For example, critically ill patients are at risk of neurocognitive disorders, and this risk may persist after discharge. In addition, existing studies have shown that premature infants have underactive autonomic nervous system activity and incomplete development of the parasympathetic nervous system. Therefore, evaluating the function of the autonomic nervous system is crucial for the detection and treatment of related diseases.

[0003] Traditional methods use heart rate variability analysis to assess autonomic nervous system function, non-invasively monitoring and evaluating autonomic nervous system function by analyzing the time and frequency domain characteristics of a time series consisting of heartbeat intervals. However, heart rate variability analysis methods are affected by age and physical condition, and there are large individual differences. Due to the influence of multiple factors such as exercise, blood pressure, and physiological regulation, the signals required for heart rate variability analysis are usually unstable, short in length, and contain a large amount of noise artifacts, resulting in obvious limitations of this method in physiological research and clinical applications. In addition, studies have shown that analyzing only a single physiological variable such as heart rate variability cannot accurately and comprehensively describe the state of autonomic nervous system regulation function.

[0004] In recent years, research on cardiopulmonary coupling based on ECG signals has received widespread attention. In 2005, a research team from Harvard Medical School first proposed the concept of cardiopulmonary coupling (CPC). By fusing the dual-modal information of heart rate and respiratory time series, they accurately measured the strength of cardiopulmonary coupling and achieved the assessment of cardiopulmonary function and sleep quality during sleep. In addition, the original cardiopulmonary coupling algorithm was based on the traditional Fourier analysis method. Traditional Fourier analysis essentially assumes that the input signal is a stationary signal, while ECG signals are usually non-stationary and nonlinear signals, which cannot meet this requirement. On the other hand, the use of traditional Fourier transform cannot simultaneously obtain high frequency resolution and high time resolution, which inevitably leads to the ambiguity of state changes in the spectral results. Therefore, the cardiopulmonary coupling method based on traditional Fourier analysis is difficult to accurately assess cardiopulmonary function and autonomic nervous function.

[0005] Currently, there is an adaptive bandwidth Fourier decomposition method (FDM). Compared to traditional Fourier analysis methods, FDM is one of the more novel time-frequency processing algorithms. It can provide an analytical representation of the signal from the perspective of instantaneous frequency and instantaneous amplitude, making it more suitable for processing nonlinear and nonstationary time series. In addition, compared with empirical mode decomposition (EMD), FDM overcomes its lack of mathematical theoretical basis. Furthermore, the FDM algorithm has important and widespread applications in biomedical signal processing, where it can be used to extract useful features from signals or quantify their dynamic behavior to facilitate subsequent statistical analysis.

[0006] Currently, adaptive bandwidth Fourier decomposition methods have been used in research such as ECG signal denoising and EEG feature extraction for sleep staging. However, adaptive bandwidth Fourier decomposition algorithms have not yet been applied to CPC analysis. Summary of the Invention

[0007] In light of this, the present invention provides a cardiopulmonary coupling feature extraction method and system based on adaptive Fourier decomposition. This method utilizes adaptive bandwidth Fourier decomposition to extract cardiopulmonary coupling features, taking into account the dynamic nature of electrophysiological signals, extracting electrophysiological signal spectral features and achieving adaptive decomposition of key rhythms. This method is particularly suitable for analyzing electrophysiological signals with nonlinear, non-stationary, and highly dynamic characteristics.

[0008] To achieve the above-mentioned object, the present invention provides a cardiopulmonary coupling feature extraction method based on adaptive Fourier decomposition, the technical solution of which includes the following steps:

[0009] Preprocess the collected original single-lead ECG signal and respiratory signal.

[0010] Extract the heart beat interval RR interval time series from the preprocessed ECG signal.

[0011] The respiratory signal time series and the RR interval time series are processed separately using the adaptive bandwidth Fourier decomposition algorithm to obtain the Fourier natural frequency band functions of the two time series.

[0012] The Fourier decomposition coefficients were obtained from the analytical forms of the Fourier natural band functions of the two time series at different frequencies. The cardiopulmonary coupling CPC values ​​were calculated based on the Fourier decomposition coefficients of the RR and respiratory signal time series to construct a cardiopulmonary coupling map.

[0013] Extract multiple types of cardiopulmonary coupling features based on the cardiopulmonary coupling map.

[0014] Furthermore, the collected original single-lead ECG signal and respiratory signal are preprocessed as follows:

[0015] Collect ECG signal and respiratory signal, the sampling frequency is f s1 and fs2 , stored in the computer.

[0016] The detailed process includes: applying a bandpass filter to the entire ECG signal, retaining the bandpass frequency band, removing other frequency bands, removing power frequency noise and baseline drift, and preventing irrelevant frequency components from affecting the results.

[0017] Furthermore, the heart beat interval RR interval time series is extracted from the preprocessed ECG signal, specifically:

[0018] Mark the heartbeats in the single-lead ECG signal, calculate the time interval between two adjacent heartbeats, and obtain the RR time series.

[0019] After extracting the normal RR interval time series from the ECG signal, a sliding average filter with a 21-data-point window is used to remove outliers due to erroneous heartbeat detection; when the center point in the window is outside the preset value, it is removed.

[0020] Finally, cubic spline interpolation is used to uniformly resample the RR and respiratory signal time series at a frequency of 4 Hz to obtain the resampled RR time series R[n] and respiratory signal E[n], respectively, where n is the sampling point of the discrete time series.

[0021] Furthermore, the respiratory signal time series and the RR interval time series are processed respectively using the adaptive bandwidth Fourier decomposition algorithm to obtain the Fourier natural frequency band functions of the two time series, which specifically includes the following steps:

[0022] In the first step, the RR interval R[n] of length N and the respiratory signal time series E[n] are Fourier transformed to obtain the Fourier transform forms R[k] and E[k] of the two time series, where k is the frequency variable of the Fourier transform form of the time series.

[0023] In the second step, the minimum number N of AFIBFs obtained by adaptive bandwidth Fourier decomposition of RR and respiratory signals is obtained according to the above HTL-FS method. i , and the obtained AFIBFs meet the following conditions:

[0024]

[0025] where φ i Expressed as the instantaneous phase of the i-th AFIBFs, φ i [n+1] is the n+1th data point, φ i [n-1] is the n-1th data point, ω i Expressed as the instantaneous frequency of the i-th AFIBFs, ω i[n] is the nth data point, and thus we get:

[0026]

[0027] where N i-1 is the minimum number of the i-1th AFIBFs, X[k] is the Fourier transform form R[k] of the RR interval sequence or the Fourier transform form E[k] of the respiratory signal time series, a i [n] The instantaneous amplitude sequence of the i-th AFIBFs, and the corresponding residual component

[0028]

[0029] Finally, the RR interval R[n] and respiratory signal E[n] time series are processed by the adaptive bandwidth Fourier decomposition method to obtain the Fourier frequency band intrinsic function R of each time series. i [n],E i [n] and the remaining component r[n].

[0030]

[0031] And the analytical form of the intrinsic function of each Fourier band

[0032]

[0033] in, are the Fourier band intrinsic functions R i [n],E i Hilbert transform of [n], a R,i [n] is the instantaneous amplitude of the i-th AFIBFs in the RR interval time series, a E,i [n] is the instantaneous amplitude of the i-th AFIBFs in the respiratory signal time series, φ R,i [n],φ E,i [n] are the RR interval and the instantaneous phase of the respiratory signal time series respectively.

[0034] Furthermore, the Fourier decomposition coefficients are obtained from the analytical forms of the Fourier natural band functions of the two time series at different frequencies. The cardiopulmonary coupling CPC values ​​are calculated based on the Fourier decomposition coefficients of the RR and respiratory signal time series to construct the cardiopulmonary coupling map. The specific steps are as follows:

[0035] First, calculate the cross power spectrum of the RR interval time series and the respiratory signal time series, denoted as Γ ER .

[0036] Next, calculate the coherence between the two signals, denoted as Λ ER .

[0037] Finally, the cardiopulmonary coupling index at a given frequency is calculated. This index quantifies the coupling strength between the RR and respiratory signal time series by integrating the cross-spectral power and coherence. The cardiopulmonary coupling index is defined as follows:

[0038] CPC fn =<Γ ER > 2 ·Λ ER (15)

[0039] Based on the above-mentioned cardiopulmonary coupling CPC value, a cardiopulmonary coupling map is drawn.

[0040] Furthermore, multiple types of cardiopulmonary coupling features are extracted based on the cardiopulmonary coupling map, specifically:

[0041] After obtaining the complete cardiopulmonary coupling spectrum, the frequency bands were divided as follows: 0.15-0.40 Hz as the high frequency band, 0.04-0.15 Hz as the low frequency band, and 0.003-0.04 Hz as the ultra-low frequency band.

[0042] The extracted multi-category cardiopulmonary coupling features include the following:

[0043] The cardiopulmonary coupling power of each frequency band was calculated by summing up all the cardiopulmonary coupling values ​​in the divided high-frequency band, low-frequency band, and very low-frequency band, respectively, and denoted as HF, LF, and VLF. The ratio of the low-frequency band cardiopulmonary coupling value to the high-frequency band cardiopulmonary coupling value, LF / HF, was also calculated.

[0044] TP is the sum of all cardiopulmonary coupling values ​​within the specified frequency band, that is, the total power.

[0045] LFnorm and HFnorm represent the normalized low-frequency and high-frequency powers, respectively.

[0046] LF / HF is defined as the ratio of low frequency to high frequency power.

[0047] The cardiopulmonary coupling feature extraction system based on adaptive Fourier decomposition includes a signal acquisition module, a preprocessing module, a time series extraction module, an adaptive bandwidth Fourier decomposition module, a cardiopulmonary coupling map drawing module and a cardiopulmonary coupling feature extraction module.

[0048] The signal acquisition module is used to collect original single-lead ECG signals and respiratory signals.

[0049] The preprocessing module is used to preprocess the collected original single-lead ECG signals and respiratory signals.

[0050] The time series extraction module is used to extract the heart beat interval RR interval time series from the preprocessed electrocardiogram signal.

[0051] The adaptive bandwidth Fourier decomposition module uses the adaptive bandwidth Fourier decomposition algorithm to process the respiratory signal time series and the RR interval time series respectively to obtain the Fourier natural frequency band functions of the above two time series.

[0052] The cardiopulmonary coupling map drawing module obtains the Fourier decomposition coefficients from the analytical form of the Fourier natural frequency band functions of the two time series at different frequencies, and calculates the cardiopulmonary coupling CPC value based on the Fourier decomposition coefficients of the RR and respiratory signal time series to construct the cardiopulmonary coupling map.

[0053] The cardiopulmonary coupling feature extraction module extracts multiple types of cardiopulmonary coupling features based on the cardiopulmonary coupling map.

[0054] Furthermore, the signal acquisition module is specifically configured to collect ECG signals and respiratory signals, with sampling frequencies of f and s1 and f s2 , stored in the computer.

[0055] The preprocessing module specifically applies a bandpass filter to the entire ECG signal, retains the bandpass frequency band, removes other frequency bands, removes power frequency noise and baseline drift, and avoids the influence of irrelevant frequency components on the results.

[0056] Furthermore, the time series extraction module specifically performs the following steps:

[0057] Mark the heartbeats in the single-lead ECG signal, calculate the time interval between two adjacent heartbeats, and obtain the RR time series.

[0058] After extracting the normal RR interval time series from the ECG signal, a sliding average filter with a 21-data-point window is used to remove outliers due to erroneous heartbeat detection; when the center point in the window is outside the preset value, it is removed.

[0059] Finally, cubic spline interpolation is used to uniformly resample the RR and respiratory signal time series at a frequency of 4 Hz to obtain the resampled RR time series R[n] and respiratory signal E[n], respectively, where n is the sampling point of the discrete time series.

[0060] The adaptive bandwidth Fourier decomposition module specifically performs the following steps:

[0061] In the first step, the RR interval R[n] of length N and the respiratory signal time series E[n] are Fourier transformed to obtain the Fourier transform forms R[k] and E[k] of the two time series, where k is the frequency variable of the Fourier transform form of the time series.

[0062] In the second step, the minimum number N of AFIBFs obtained by adaptive bandwidth Fourier decomposition of RR and respiratory signals is obtained according to the above HTL-FS method. i , and the obtained AFIBFs meet the following conditions:

[0063]

[0064] where φ i Expressed as the instantaneous phase of the i-th AFIBFs, φ i [n+1] is the n+1th data point, φ i [n-1] is the n-1th data point, ω i Expressed as the instantaneous frequency of the i-th AFIBFs, ω i [n] is the nth data point, and thus we get:

[0065]

[0066] where N i-1 is the minimum number of the i-1th AFIBFs, X[k] is the Fourier transform form R[k] of the RR interval sequence or the Fourier transform form E[k] of the respiratory signal time series, a i [n] The instantaneous amplitude sequence of the i-th AFIBFs, and the corresponding residual component

[0067]

[0068] Finally, the RR interval R[n] and respiratory signal E[n] time series are processed by the adaptive bandwidth Fourier decomposition method to obtain the Fourier frequency band intrinsic function R of each time series. i [n],E i [n] and the remaining component r[n].

[0069]

[0070] And the analytical form of the intrinsic function of each Fourier band

[0071]

[0072] in, are the Fourier band intrinsic functions R i [n],E i Hilbert transform of [n], a R,i [n] is the instantaneous amplitude of the i-th AFIBFs in the RR interval time series, a E,i [n] is the instantaneous amplitude of the i-th AFIBFs in the respiratory signal time series, φ R,i [n],φE,i [n] are the RR interval and the instantaneous phase of the respiratory signal time series respectively.

[0073] Furthermore, the cardiopulmonary coupling map drawing module specifically performs the following steps:

[0074] First, calculate the cross power spectrum of the RR interval time series and the respiratory signal time series, denoted as Γ ER .

[0075] Next, calculate the coherence between the two signals, denoted as Λ ER .

[0076] Finally, the cardiopulmonary coupling index at a given frequency is calculated. This index quantifies the coupling strength between the RR and respiratory signal time series by integrating the cross-spectral power and coherence. The cardiopulmonary coupling index is defined as follows:

[0077]

[0078] Based on the above-mentioned cardiopulmonary coupling CPC value, a cardiopulmonary coupling map is drawn.

[0079] The cardiopulmonary coupling feature extraction module specifically performs the following steps:

[0080] After obtaining the complete cardiopulmonary coupling spectrum, the frequency bands were divided as follows: 0.15-0.40 Hz as the high frequency band, 0.04-0.15 Hz as the low frequency band, and 0.003-0.04 Hz as the ultra-low frequency band.

[0081] The extracted multi-category cardiopulmonary coupling features include the following:

[0082] The cardiopulmonary coupling power of each frequency band was calculated by summing up all the cardiopulmonary coupling values ​​in the divided high-frequency band, low-frequency band, and very low-frequency band, respectively, and denoted as HF, LF, and VLF. The ratio of the low-frequency band cardiopulmonary coupling value to the high-frequency band cardiopulmonary coupling value, LF / HF, was also calculated.

[0083] TP is the sum of all cardiopulmonary coupling values ​​within the specified frequency band, that is, the total power.

[0084] LFnorm and HFnorm represent the normalized low-frequency and high-frequency powers, respectively.

[0085] LF / HF is defined as the ratio of low frequency to high frequency power.

[0086] Beneficial effects:

[0087] The present invention provides a method and system for extracting cardiopulmonary coupling features based on adaptive Fourier decomposition. Building on the advantages of adaptive bandwidth Fourier decomposition algorithms in analyzing nonlinear and nonstationary signals and constructing visual high-resolution atlases, this paper proposes a new cardiopulmonary coupling feature extraction algorithm based on adaptive bandwidth Fourier decomposition for the first time. This algorithm can take into account the dynamic nature of electrophysiological signals, extract electrophysiological signal spectral features, and achieve adaptive decomposition of key rhythms. Compared to traditional Fourier transform-based cardiopulmonary coupling algorithms, the present invention is more suitable for analyzing electrophysiological signals with nonlinear, nonstationary, and highly dynamic characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Figure 1 This is a technical flow chart of the dynamic cardiopulmonary coupling algorithm based on adaptive bandwidth Fourier decomposition. DETAILED DESCRIPTION

[0089] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0090] Example 1:

[0091] To resolve the contradiction between the precise depiction of rhythm frequency and the high dynamic requirements of components in traditional cardiopulmonary coupling maps, the present invention proposes a high dynamic cardiopulmonary coupling algorithm based on the adaptive bandwidth Fourier decomposition method, and realizes the evaluation of cardiopulmonary function and autonomic nervous system function. The technical flow chart of the present invention is as follows Figure 1 The detailed process is as follows:

[0092] Step 1) Collect the subject's ECG signal and respiratory signal, the sampling frequencies are f s1 and f s2 , stored in the computer.

[0093] Step 2) Preprocess the original ECG signal. The detailed process includes: applying a bandpass filter to the entire ECG signal, retaining the bandpass frequency band (e.g., 0.5-30 Hz), removing other frequency bands, removing power frequency noise and baseline drift, and preventing irrelevant frequency components from affecting the results.

[0094] Step 3) Extract the heartbeat interval (RR) time series from the preprocessed ECG signal. Specifically, mark the heartbeats in the single-lead ECG signal, calculate the time interval between two adjacent heartbeats, and obtain the RR time series; after extracting the normal RR interval time series from the ECG signal, use a sliding average filter with a 21-data point window to remove outliers caused by erroneous heartbeat detection. When the center point in the window is outside the preset (average value 30%), it will be removed. Finally, use cubic spline interpolation to uniformly resample the RR and respiratory signal time series at a frequency of 4Hz to obtain the resampled RR time series R[n] and respiratory signal E[n], respectively, where n is the sampling point of the discrete time series.

[0095] Step 4) Perform adaptive bandwidth Fourier decomposition on the preprocessed RR and EDR time series to obtain Fourier intrinsic band functions (FIBFs). The adaptive bandwidth Fourier decomposition algorithm is based on the condition that the instantaneous frequency is only meaningful for a single component signal. It adaptively divides the frequency band in the Fourier domain and decomposes the original signal x(t) into a set of M physically meaningful single component signals y through zero-phase filtering. i (t), that is, the Fourier natural frequency band function, where the single component signal is infinitely differentiable on the interval [a, b], that is, {y i (t):y i (t)∈C ∞ [a, b], i = 1, ..., M}. The analytical form of Fourier intrinsic frequency band functions (AFIBFs) obtained by Fourier domain adaptive spectral decomposition can represent the different frequency components in the original signal from the perspective of instantaneous frequency and instantaneous amplitude, further improving the time-frequency resolution of the analysis results. This type of representation method is more suitable for nonlinear, non-stationary, and highly dynamic signal analysis. To ensure the accuracy of the decomposition results, FIBFs must meet the following conditions:

[0096] 1) FIBFs is a zero-mean function, that is

[0097] 2) FIBFs are mutually orthogonal functions, that is

[0098] 3) The analytical function of FIBFs can be expressed as: Among them, a i (t) is the instantaneous amplitude, φ i (t) is the instantaneous phase, and the instantaneous amplitude a i (t) and instantaneous frequency ω i (t) is non-negative, that is, j is the symbol of the imaginary part, is the Hilbert transform of FIBFs, that is τ is the integration variable.

[0099] The specific principle of adaptive bandwidth Fourier decomposition is as follows: for any finite length signal x(t) that satisfies the Dirichlet condition and is in the interval [t1, t1+T0], where t1 is the start time of the signal and T0 is the duration of the signal x(t), the corresponding periodic extension signal can be constructed. k is the summation variable, and the periodic extension signal The Fourier transform of is:

[0100]

[0101] in

[0102]

[0103] a0、a k and b k is the Fourier coefficient, ω k is the angular frequency

[0104] In order to obtain the analytical form of the Fourier natural frequency band function of the signal, equation (1) can be rewritten as

[0105]

[0106] Where j is the imaginary unit, * symbol indicates the complex conjugate, c k =(a k -jb k ), From formula (2), we can see that and analytical functions The complex conjugate of definition for In order to better represent the nonlinear, non-stationary and highly dynamic characteristics of AFIBFs, Rewritten in terms of instantaneous phase and instantaneous amplitude:

[0107]

[0108] M is the number of decompositions, a i (t) is the instantaneous frequency of the i-th component, φ i (t) is the instantaneous phase of the i-th component

[0109] Combining equations (3) and (4), and using the high-to-low frequency scanning method to sort out the frequency components after Fourier decomposition (High-to-Low Frequency Scanning, HTL-FS), we can obtain:

[0110]

[0111] Therefore, the following representation relationship can be obtained by the HTL-FS method:

[0112]

[0113] Where N0=M,N M = 1. In order to obtain the minimum number of AFIBFs (to ensure that the frequency components contained in each AFIBFs are reasonable and the time-frequency resolution is as high as possible), we start from N i-1 Start with -1, decrease and select the smallest N i value, such that 1≤N i ≤N i-1 -1 and when i=1,...,M, it satisfies:

[0114]

[0115] where a i (t) and ω i (t)=2πf i (t) are the instantaneous amplitude and instantaneous frequency of the i-th FIBF respectively. i If the value is satisfied and the condition of formula (7) is met, the original signal x(t) can be decomposed and the Fourier natural frequency band function and its analytical form can be obtained.

[0116] In the present invention, the specific steps of processing the RR interval signal and the respiratory signal time series by the adaptive bandwidth Fourier decomposition method are as follows:

[0117] The first step is to perform Fourier transform on the RR interval R[n] of length N and the respiratory signal time series E[n] to obtain the Fourier transform forms R[k] and E[k] of the two time series, where k is the frequency variable of the Fourier transform form of the time series. k frequency point

[0118] In the second step, the minimum number N of AFIBFs obtained by adaptive bandwidth Fourier decomposition of RR and respiratory signals is obtained according to the above HTL-FS method. i , and the obtained AFIBFs meet the following conditions:

[0119]

[0120] where φ iExpressed as the instantaneous phase of the i-th AFIBFs, φ i [n+1] is the n+1th data point, ω i Expressed as the instantaneous frequency of the i-th AFIBFs, ω i [n] is the nth data point, and from this we can get:

[0121]

[0122] where N i-1 is the minimum number of the i-1th AFIBFs, X[k] is the Fourier transform form R[k] of the RR interval sequence or the Fourier transform form E[k] of the respiratory signal time series, a i [n] The instantaneous amplitude sequence of the i-th AFIBFs, and the corresponding residual component

[0123]

[0124] Finally, the RR interval R[n] and respiratory signal E[n] time series are processed by the adaptive bandwidth Fourier decomposition method to obtain the Fourier frequency band intrinsic function R of each time series. i [n],E i [n] and the remaining component r[n]

[0125]

[0126] And the analytical form of the intrinsic function of each Fourier band

[0127]

[0128] in, are the Fourier band intrinsic functions R i [n],E i Hilbert transform of [n], a R,i [n] is the instantaneous amplitude of the i-th AFIBFs in the RR interval time series, a E,i [n] is the instantaneous amplitude of the i-th AFIBFs in the respiratory signal time series, φ R,i [n],φ E,i [n] are the RR interval and the instantaneous phase of the respiratory signal time series respectively.

[0129] Step 5) Calculate the cardiopulmonary coupling (CPC) value and draw the corresponding cardiopulmonary coupling map. First, calculate the cross power spectrum of the RR interval time series and the respiratory signal time series, denoted as Γ ERIn the specific calculation, a sliding window with a window length of 8 seconds is used to average the instantaneous frequency of the RR interval and the respiratory signal, and use it as the frequency of the current sub-window. From the RR interval and respiratory signal time series, find the i-th and i'th windows respectively, so that the frequencies of the two windows are closest, which is the frequency f n , and then calculate the cross power spectrum Γ of the two time series ER , the specific expression is as follows:

[0130]

[0131] in, and They are the RR interval and respiratory signal time series at a given frequency f n Adaptive bandwidth Fourier decomposition coefficients under , * indicates complex conjugate.

[0132] Next, calculate the coherence between the two signals, denoted as Λ ER The coherence is defined as the product of the square of the average cross spectrum divided by the average spectral power of the individual signals, that is:

[0133]

[0134] Where <> indicates averaging frequencies in the original spectrum or averaging multiple measurements at a given frequency. This is because coherence is a statistical measure, so statistical averaging is necessary. '||' represents the absolute value or modulus of an imaginary number. In the original literature, the average spectral power is obtained by dividing each observation window into subwindows and averaging the subwindows.

[0135] Finally, the cardiopulmonary coupling index at a given frequency is calculated. This index quantifies the coupling strength between the RR and respiratory signal time series by integrating the cross-spectral power and coherence. The cardiopulmonary coupling index is defined as follows:

[0136]

[0137] Based on the aforementioned cardiopulmonary coupling CPC values, a cardiopulmonary coupling map is drawn. By observing the coupling patterns in the cardiopulmonary coupling map, it is possible to identify whether the subject's cardiopulmonary function is abnormal.

[0138] Step 6) Extract multiple cardiopulmonary coupling features based on the cardiopulmonary coupling spectrum. In cardiopulmonary coupling analysis, after obtaining a complete cardiopulmonary coupling spectrum, all cardiopulmonary coupling values ​​within the high-frequency band (0.15-0.40 Hz), low-frequency band (0.04-0.15 Hz), and ultra-low-frequency band (0.003-0.04 Hz) are summed as the cardiopulmonary coupling power of each frequency band, denoted as HF, LF, and VLF. The ratio of the low-frequency band cardiopulmonary coupling value to the high-frequency band cardiopulmonary coupling value, LF / HF, is then calculated to provide a reference for subsequent cardiopulmonary function assessment. In addition, new indices obtained by standardizing the above indicators are also widely used in research related to autonomic nervous system assessment. Table 1 summarizes common cardiopulmonary coupling indices and their corresponding frequency band ranges. As shown in Table 1, TP is defined as the sum of all cardiopulmonary coupling values ​​within a specified frequency band, that is, the total power; HF, LF, and VLF represent the power in the high-frequency, low-frequency, and very low-frequency ranges, respectively; LFnorm and HFnorm represent the standardized low-frequency and high-frequency powers, respectively; and LF / HF is defined as the ratio of low-frequency to high-frequency power.

[0139] Table 1 Cardiopulmonary coupling feature extraction

[0140] index Frequency range (Hz) explain TP ≤0.4 Total power, the sum of all cardiopulmonary coupling values ​​within the specified frequency band HF 0.1–0.4 High-frequency power, the sum of the cardiopulmonary coupling values ​​in the high-frequency range HFnorm Normalized high frequency power, HF / TP LF 0.01–0.1 Low-frequency power, the sum of the cardiopulmonary coupling values ​​in the low-frequency range LFnorm Normalized low frequency power, LF / TP LF / HF Low-frequency to high-frequency power ratio VLF ≤0.01 Power in the ultra-low frequency range

[0141] The present invention has been verified on the data sets of healthy people, diseased patients and premature infants in the Physionet database. By drawing the cardiopulmonary coupling maps of different types of subjects, it was found that the dynamic cardiopulmonary coupling map based on adaptive bandwidth Fourier decomposition can visualize the cardiopulmonary coupling status of different types of subjects. In addition, there are significant statistical differences between the new cardiopulmonary coupling features obtained by people with different cardiopulmonary functions, which provides support for subsequent cardiopulmonary function assessment. In summary, the cardiopulmonary coupling algorithm based on adaptive bandwidth Fourier decomposition can be used as a supplementary tool for routine clinical diagnosis, which helps to improve the ability to assess cardiopulmonary function and has certain potential value and application prospects in portable wearable health monitoring.

[0142] The present invention proposes for the first time a new algorithm for extracting dynamic cardiopulmonary coupling features based on adaptive bandwidth Fourier decomposition, which is suitable for accurate assessment of cardiopulmonary function, is efficient, reliable, and easy to software.

[0143] Example 2:

[0144] Embodiment 2 of the present invention provides a cardiopulmonary coupling feature extraction system based on adaptive Fourier decomposition, comprising a signal acquisition module, a preprocessing module, a time series extraction module, an adaptive bandwidth Fourier decomposition module, a cardiopulmonary coupling mapping module, and a cardiopulmonary coupling feature extraction module. The system is configured to execute the cardiopulmonary coupling feature extraction method based on adaptive Fourier decomposition provided in embodiment 1.

[0145] The signal acquisition module is used to collect the original single-lead ECG signal and respiratory signal; in the embodiment of the present invention, the signal acquisition module specifically performs the following steps: collecting the ECG signal and respiratory signal, the sampling frequencies are f s1 and f s2 , stored in the computer.

[0146] The preprocessing module is used to preprocess the collected original single-lead ECG signal and respiratory signal. In the embodiment of the present invention, the preprocessing module specifically performs the following steps: using a bandpass filter on the entire ECG signal, retaining the bandpass frequency band, removing other frequency bands, removing power frequency noise and baseline drift, and preventing irrelevant frequency components from affecting the results.

[0147] The time series extraction module is used to extract the heart beat interval RR interval time series from the preprocessed ECG signal. In the embodiment of the present invention, the time series extraction module specifically performs the following steps:

[0148] Mark the heartbeats in the single-lead ECG signal, calculate the time interval between two adjacent heartbeats, and obtain the RR time series.

[0149] After extracting the normal RR interval time series from the ECG signal, a sliding average filter with a 21-data-point window is used to remove outliers due to erroneous heartbeat detection; when the center point in the window is outside the preset value, it is removed.

[0150] Finally, cubic spline interpolation is used to uniformly resample the RR and respiratory signal time series at a frequency of 4 Hz to obtain the resampled RR time series R[n] and respiratory signal E[n], respectively, where n is the sampling point of the discrete time series.

[0151] The adaptive bandwidth Fourier decomposition module uses the adaptive bandwidth Fourier decomposition algorithm to process the respiratory signal time series and the RR interval time series respectively to obtain the Fourier natural frequency band functions of the above two time series. In the embodiment of the present invention, the adaptive bandwidth Fourier decomposition module specifically performs the following steps:

[0152] In the first step, the RR interval R[n] of length N and the respiratory signal time series E[n] are Fourier transformed to obtain the Fourier transform forms R[k] and E[k] of the two time series, where k is the frequency variable of the Fourier transform form of the time series.

[0153] In the second step, the minimum number N of AFIBFs obtained by adaptive bandwidth Fourier decomposition of RR and respiratory signals is obtained according to the above HTL-FS method. i , and the obtained AFIBFs meet the following conditions:

[0154]

[0155] where φ i Expressed as the instantaneous phase of the i-th AFIBFs, φ i [n+1] is the n+1th data point, φ i [n-1] is the n-1th data point, ω i Expressed as the instantaneous frequency of the i-th AFIBFs, ω i [n] is the nth data point, and thus we get:

[0156]

[0157] where N i-1 is the minimum number of the i-1th AFIBFs, X[k] is the Fourier transform form R[k] of the RR interval sequence or the Fourier transform form E[k] of the respiratory signal time series, a i [n] The instantaneous amplitude sequence of the i-th AFIBFs, and the corresponding residual component

[0158]

[0159] Finally, the RR interval R[n] and respiratory signal E[n] time series are processed by the adaptive bandwidth Fourier decomposition method to obtain the Fourier frequency band intrinsic function R of each time series. i [n],E i [n] and the remaining component r[n];

[0160]

[0161] And the analytical form of the intrinsic function of each Fourier band

[0162]

[0163] in, are the Fourier band intrinsic functions R i [n],E i Hilbert transform of [n], a R,i [n] is the instantaneous amplitude of the i-th AFIBFs in the RR interval time series, a E,i [n] is the instantaneous amplitude of the i-th AFIBFs in the respiratory signal time series, φ R,i [n],φ E,i [n] are the RR interval and the instantaneous phase of the respiratory signal time series respectively.

[0164] The cardiopulmonary coupling map drawing module obtains the Fourier decomposition coefficients from the analytical form of the Fourier natural frequency band functions of the two time series at different frequencies, calculates the cardiopulmonary coupling CPC values ​​based on the Fourier decomposition coefficients of the RR and respiratory signal time series, and constructs a cardiopulmonary coupling map. In this embodiment of the present invention, the cardiopulmonary coupling map drawing module specifically performs the following steps:

[0165] First, calculate the cross power spectrum of the RR interval time series and the respiratory signal time series, denoted as Γ ER .

[0166] Next, calculate the coherence between the two signals, denoted as Λ ER .

[0167] Finally, the cardiopulmonary coupling index at a given frequency is calculated. This index quantifies the coupling strength between the RR and respiratory signal time series by integrating the cross-spectral power and coherence. The cardiopulmonary coupling index is defined as follows:

[0168]

[0169] Based on the above-mentioned cardiopulmonary coupling CPC value, a cardiopulmonary coupling map is drawn.

[0170] The cardiopulmonary coupling feature extraction module extracts multiple types of cardiopulmonary coupling features based on the cardiopulmonary coupling map. In the embodiment of the present invention, the cardiopulmonary coupling feature extraction module specifically performs the following steps:

[0171] After obtaining the complete cardiopulmonary coupling spectrum, the frequency bands were divided as follows: 0.15-0.40 Hz as the high frequency band, 0.04-0.15 Hz as the low frequency band, and 0.003-0.04 Hz as the ultra-low frequency band.

[0172] The extracted multi-category cardiopulmonary coupling features include the following:

[0173] The cardiopulmonary coupling power of each frequency band was calculated by summing up all the cardiopulmonary coupling values ​​in the divided high-frequency band, low-frequency band, and very low-frequency band, respectively, and denoted as HF, LF, and VLF. The ratio of the low-frequency band cardiopulmonary coupling value to the high-frequency band cardiopulmonary coupling value, LF / HF, was also calculated.

[0174] TP is the sum of all cardiopulmonary coupling values ​​within the specified frequency band, that is, the total power.

[0175] LFnorm and HFnorm represent the normalized low-frequency and high-frequency powers, respectively.

[0176] LF / HF is defined as the ratio of low frequency to high frequency power.

[0177] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A cardiopulmonary coupling feature extraction method based on adaptive Fourier decomposition, characterized in that: include: Preprocess the collected original single-lead ECG signal and respiratory signal; Extracting heart beat intervals from preprocessed ECG signals R - R Interval time series; specifically: Mark the heartbeat in the single-lead ECG signal and calculate the time interval between two adjacent heartbeats to obtain R - R Time series; Extracting normal R - R After the interval time series, a sliding average filter with a window of 21 data points is used to remove outliers due to false heartbeat detection; when the center point in the window is outside the preset value, it is removed; Finally, use cubic spline interpolation with 4 Hz Frequency is evenly spaced R - R The two signals of the respiratory signal time series are resampled to obtain the resampled R - R Time Series and respiratory signals ,in is the sampling point of the discrete time series; The adaptive bandwidth Fourier decomposition algorithm is used to analyze the respiratory signal time series and R - R The interval time series is processed to obtain the above respiratory signal time series and R - R The Fourier natural frequency band function of the interval time series specifically includes the following steps: The first step is to of R - R Interphase and respiratory signal time series Perform Fourier transform to obtain respiratory signal time series and R - R Fourier transform form of interval time series and ,in is the frequency variable in the Fourier transform form of the time series; The second step is to use the high to low frequency scanning method HTL - FS The methods are respectively obtained R - R The adaptive bandwidth Fourier decomposition of the respiratory signal can be obtained indivual AFIBFs The minimum number of , and make the obtained AFIBFs The following conditions are met: （8） in Expressed as indivual AFIBFs The instantaneous phase of For the first data points, is the n-1th data point, Expressed as indivual AFIBFs The instantaneous frequency, For the first data points, and thus obtain: （9） in For the indivual AFIBFs The minimum number of for R - R Fourier transform of the interval series Or the Fourier transform of the respiratory signal time series , No. indivual AFIBFs The instantaneous amplitude sequence of , and the corresponding residual component （10） Finally, the adaptive bandwidth Fourier decomposition method is used to R - R Interphase and respiratory signals The time series is processed to obtain the Fourier band intrinsic function of each time series and remaining quantity ; （11） And the analytical form of the intrinsic function of each Fourier band （12） in, Fourier band intrinsic functions The Hilbert transform of for R - R Interval time series indivual AFIBFs The instantaneous amplitude, is the respiratory signal time series indivual AFIBFs The instantaneous amplitude, They are R - R Intervals and instantaneous phase of respiratory signal time series; From the respiratory signal time series and R - R The Fourier decomposition coefficients are obtained from the analytical form of the Fourier natural frequency band function of the interval time series at different frequencies, and the Fourier decomposition coefficients are obtained according to the R - R Calculate cardiopulmonary coupling using the Fourier decomposition coefficients of the respiratory signal time series CPC The cardiopulmonary coupling map is constructed by the following steps: First, calculate the cross power spectrum of the RR interval time series and the respiratory signal time series, denoted as ; Next, we calculate the coherence between these two signals, which is expressed as ; Finally, the cardiopulmonary coupling index at a given frequency is calculated. This index quantifies the coupling strength between the RR and respiratory signal time series by integrating the cross-spectral power and coherence. The cardiopulmonary coupling index is defined as follows: （15） Based on the above-mentioned cardiopulmonary coupling CPC value, a cardiopulmonary coupling map is drawn; Based on the cardiopulmonary coupling map, multiple types of cardiopulmonary coupling features are extracted, specifically: After obtaining the complete cardiopulmonary coupling spectrum, the frequency bands were divided as follows: 0.15–0.40 Hz as the high-frequency band, 0.04–0.15 Hz as the low-frequency band, and 0.003–0.04 Hz as the ultra-low-frequency band; The extracted multi-category cardiopulmonary coupling features include the following: The cardiopulmonary coupling values ​​in the divided high-frequency band, low-frequency band, and very low-frequency band are summed up as the cardiopulmonary coupling power of each frequency band, respectively recorded as HF, LF, and VLF, and the ratio of the low-frequency band cardiopulmonary coupling value to the high-frequency band cardiopulmonary coupling value, LF / HF, is calculated; TP is the sum of all cardiopulmonary coupling values ​​within the specified frequency band, i.e., the total power; LFnorm and HFnorm represent the normalized low-frequency and high-frequency powers, respectively; LF / HF is defined as the ratio of low frequency to high frequency power.

2. The cardiopulmonary coupling feature extraction method based on adaptive Fourier decomposition according to claim 1, characterized in that: The preprocessing of the collected original single-lead ECG signal and respiratory signal is specifically as follows: Collect ECG signals and respiratory signals, the sampling frequencies are fs1 and fs2 , stored in a computer; The detailed process includes: applying a bandpass filter to the entire ECG signal, retaining the bandpass frequency band, removing other frequency bands, removing power frequency noise and baseline drift, and preventing irrelevant frequency components from affecting the results.

3. Cardiopulmonary coupling feature extraction system based on adaptive Fourier decomposition, characterized by: It includes signal acquisition module, preprocessing module, time series extraction module, adaptive bandwidth Fourier decomposition module, cardiopulmonary coupling map drawing module and cardiopulmonary coupling feature extraction module; The signal acquisition module is used to collect original single-lead ECG signals and respiratory signals; The preprocessing module is used to preprocess the collected original single-lead ECG signal and respiratory signal; The time series extraction module is used to extract the heart beat interval from the preprocessed ECG signal R - R Interval time series; The adaptive bandwidth Fourier decomposition module uses the adaptive bandwidth Fourier decomposition algorithm to respectively decompose the respiratory signal time series and R - R The interval time series is processed to obtain the Fourier natural frequency band function of the above respiratory signal time series and RR interval time series; The cardiopulmonary coupling map drawing module obtains the Fourier decomposition coefficients from the analytical form of the Fourier inherent frequency band function of the respiratory signal time series and the RR interval time series, and R - R Calculate cardiopulmonary coupling using the Fourier decomposition coefficients of the respiratory signal time series CPC Value, construct cardiopulmonary coupling map; The cardiopulmonary coupling feature extraction module extracts multiple types of cardiopulmonary coupling features based on the cardiopulmonary coupling map; The time series extraction module specifically performs the following steps: Mark the heartbeat in the single-lead ECG signal and calculate the time interval between two adjacent heartbeats to obtain R - R Time series; Extracting normal R - R After the interval time series, a sliding average filter with a window of 21 data points is used to remove outliers due to false heartbeat detection; when the center point in the window is outside the preset value, it is removed; Finally, use cubic spline interpolation with 4 Hz Frequency is evenly spaced R - R The two signals of the respiratory signal time series are resampled to obtain the resampled R - R Time Series and respiratory signals ,in is the sampling point of the discrete time series; The adaptive bandwidth Fourier decomposition module specifically performs the following steps: The first step is to of R - R Interphase and respiratory signal time series Perform Fourier transform to obtain the Fourier transform form of respiratory signal time series and RR interval time series and ,in is the frequency variable in the Fourier transform form of the time series; The second step is to use the high to low frequency scanning method HTL - FS The methods are respectively obtained R - R The adaptive bandwidth Fourier decomposition of the respiratory signal can be obtained indivual AFIBFs The minimum number of , and make the obtained AFIBFs The following conditions are met: （8） in Expressed as indivual AFIBFs The instantaneous phase of For the first data points, is the n-1th data point, Expressed as indivual AFIBFs The instantaneous frequency, For the first data points, and thus obtain: （9） in For the indivual AFIBFs The minimum number of for R - R Fourier transform of the interval series Or the Fourier transform of the respiratory signal time series , No. indivual AFIBFs The instantaneous amplitude sequence of , and the corresponding residual component （10） Finally, the adaptive bandwidth Fourier decomposition method is used to R - R Interphase and respiratory signals The time series is processed to obtain the Fourier band intrinsic function of each time series and remaining quantity ; （11） And the analytical form of the intrinsic function of each Fourier band （12） in, Fourier band intrinsic functions The Hilbert transform of for R - R Interval time series indivual AFIBFs The instantaneous amplitude, is the respiratory signal time series indivual AFIBFs The instantaneous amplitude, They are R - R Intervals and instantaneous phase of respiratory signal time series; The cardiopulmonary coupling map drawing module specifically performs the following steps: First, calculate the cross power spectrum of the RR interval time series and the respiratory signal time series, denoted as ; Next, we calculate the coherence between these two signals, which is expressed as ; Finally, the cardiopulmonary coupling index at a given frequency is calculated. This index quantifies the coupling strength between the RR and respiratory signal time series by integrating the cross-spectral power and coherence. The cardiopulmonary coupling index is defined as follows: （15） Based on the above-mentioned cardiopulmonary coupling CPC value, a cardiopulmonary coupling map is drawn; The cardiopulmonary coupling feature extraction module specifically performs the following steps: After obtaining the complete cardiopulmonary coupling spectrum, the frequency bands were divided as follows: 0.15–0.40 Hz as the high-frequency band, 0.04–0.15 Hz as the low-frequency band, and 0.003–0.04 Hz as the ultra-low-frequency band; The extracted multi-category cardiopulmonary coupling features include the following: The cardiopulmonary coupling values ​​in the divided high-frequency band, low-frequency band, and very low-frequency band are summed up as the cardiopulmonary coupling power of each frequency band, respectively recorded as HF, LF, and VLF, and the ratio of the low-frequency band cardiopulmonary coupling value to the high-frequency band cardiopulmonary coupling value, LF / HF, is calculated; TP is the sum of all cardiopulmonary coupling values ​​within the specified frequency band, i.e., the total power; LFnorm and HFnorm represent the normalized low-frequency and high-frequency powers, respectively; LF / HF is defined as the ratio of low frequency to high frequency power.

4. The cardiopulmonary coupling feature extraction system based on adaptive Fourier decomposition according to claim 3, characterized in that: The signal acquisition module specifically performs the following steps: collecting ECG signals and respiratory signals, with sampling frequencies of fs1 and fs2 , stored in a computer; The preprocessing module specifically performs the following steps: applying a bandpass filter to the entire ECG signal, retaining the bandpass frequency band, removing other frequency bands, removing power frequency noise and baseline drift, and avoiding the influence of irrelevant frequency components on the results.

Images