Implementation method of cardiac impedance signal detail separation reconstruction and feature extraction

Abstract

The present application relates to a kind of heart impedance signal detail separation reconstruction and feature extraction implementation method, belong to heart impedance signal processing technical field.The method includes: according to the main frequency relatively concentrated characteristics of human heart impedance signal ICG, using ICEEMDAN method to heart impedance signal is separated, obtains heart impedance main signal and detail signal, and then constructs PSO-VMD method to secondary decomposition reconstruction processing of detail signal;Using the physiological characteristics of heart impedance signal C point, using two sides difference to heart impedance signal, fixed C point, then segmented extraction other key feature points.The present application relates to the signal processing method of heart impedance signal detail separation has better robustness and stable denoising effect on the basis of retaining signal detail characteristics, on the basis of combining human physiology and feature point time sequence relationship, the feature point extracted with the contraction and relaxation relationship of cardiac function related has higher accuracy.

Description

Technical Field

[0001] This invention belongs to the field of cardiac impedance signal processing technology, and relates to the technique of preserving details of cardiac impedance signals, specifically to a method for separating, reconstructing and extracting details of cardiac impedance signals. Background Technology

[0002] The waveform, amplitude, and other characteristics of cardiac impedance signals can describe the systolic and diastolic functions of the heart, as well as parameters related to peripheral vascular compliance. Advances in related technologies have made the non-invasive acquisition of cardiac impedance signals from patients safe and convenient, suitable for diagnostic and medical monitoring purposes.

[0003] Uncertainty in physiological signal acquisition can introduce environmental electromagnetic interference, electrical equipment interference, or other bioelectrical signal interference into cardiac impedance signals, causing deviations in the characteristic localization of cardiac impedance signals. This, in turn, affects the accurate calculation of hemodynamic parameters, thus limiting its application in cardiovascular diagnosis and treatment.

[0004] The heart has a physiological mechanism of periodic contraction and relaxation, and cardiac impedance signals are obtained by measuring the impedance changes of blood flow in the heart and blood vessels through impedance mapping. These impedance changes also show periodic changes with the heartbeat. Therefore, the frequency of cardiac impedance signals is relatively concentrated, and the main frequency of cardiac impedance signals is concentrated below 20Hz.

[0005] The VMD algorithm can accurately capture the time-varying characteristics of a signal, decomposing it into a more refined set of mode functions containing information at different time scales. However, due to individual differences or variations in the sources and characteristics of noise in different scenarios, the number of decomposition levels (Kvmd) and the penalty factor α become limiting factors for VMD decomposition performance in practical applications. Specifically, K The number of layers, as a modal component, directly affects the precision and complexity of signal decomposition. On the other hand, the penalty factor... α In the VMD method, a larger penalty factor is used to balance the reconstruction error and bandwidth constraints of the signal. This larger penalty factor makes the decomposition result smoother, but may lead to the loss of signal details; a smaller penalty factor may increase noise components. Therefore, it is necessary to select appropriate VMD decomposition parameters according to the actual situation.

[0006] The generation of the C wave is related to cardiac contraction. By observing and analyzing the C wave in impedance signals, doctors and researchers can obtain important information about cardiac function and hemodynamic status, thus aiding in disease diagnosis and treatment. For example, stroke volume can be calculated based on impedance signals. The characteristics of the C wave must not be lost or distorted, and its location must be accurate. However, current signal processing methods have not clearly addressed the waveform distortion and amplitude changes caused by impedance signal denoising. Furthermore, existing methods for impedance signal processing rarely consider the influence of individual differences and the quality of the acquired impedance signals. Therefore, the final impedance signal processing effect and feature extraction accuracy are uncertain and lack robustness. Summary of the Invention

[0007] In view of this, the purpose of this invention is to provide a method for detail separation, reconstruction and feature extraction of cardiac impedance signals. This method adaptively matches different control parameters to different cardiac impedance signals collected from the front-end monitoring system for detail separation and reconstruction, in order to solve the problems of strong subjectivity, insufficient robustness, poor signal adaptability, feature loss and inaccurate feature point extraction in existing cardiac impedance signal processing technologies.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A method for detail separation, reconstruction, and feature extraction of cardiac impedance signals, specifically including the following steps: S1: Using the ICEEMDAN method, the cardiac impedance signal (ICG) is decomposed into NIMF components representing different frequency and amplitude components of the signal: (1) in, This represents the first decomposition by ICEEMDAN. k Layer NIMF eigenmode functions, k =1,..., K , K This indicates the number of decomposition layers of the ICG signal after ICEEMDAN decomposition. n The length of each NIMF component is equal to that of the input signal; ICEEMDAN represents the improved adaptive noise complete set empirical mode decomposition; and NIMF represents the mode components obtained after ICEEMDAN decomposition of the ICG signal. S2: Select layer K1 to divide the NIMF component into detailed components related to cardiac impedance signal and reconstruct them into ICG1, and reconstruct the remaining components that still contain some cardiac impedance signal information into ICG2: (2) (3) S3: Select the number of decomposition layers KvmdVMD decomposition of ICG2 with penalty factor α yields... Kvmd Group VIMF components: (4) Wherein, VMD represents variational mode decomposition, and VIMF represents the modal components obtained after VMD decomposition of the ICG signal; S4: Adaptive thresholding of VIMF components is performed using the following formula: (5) in, Indicates the first i Layer VIMF components, Indicates the number of thresholds after threshold processing i Layer VIMF components; It is a threshold, which is three times the signal standard deviation, and can be adjusted according to the actual situation; S5: Select layer K2 to reconstruct other detailed information about the cardiac impedance signal in the VIMF component as ICG3, and combine it with ICG1 obtained in step S2 to jointly reconstruct the denoising result s(n) of the cardiac impedance signal: (6) S6: Using the differential threshold and moving integral scheme, the cardiac impedance signal after denoising in step S5 is differentially divided on both sides and the C point is fixed, and then the features of A, B, X and O are extracted in segments.

[0009] Furthermore, in step S1, the NIMF components are sorted from high frequency to low frequency, with the Kth layer as the baseline.

[0010] Furthermore, in step S2, the method for selecting the number of layers K1 is as follows: by finding the minimum value of the energy of each NIMF component except for the first layer and the Kth layer, the layer number close to the low frequency is selected from the minimum value. If there is no minimum value, half of the total number of decomposition layers is taken, which is K1.

[0011] Furthermore, in step S2, the first k Energy of one NIMF component It can be obtained using the following formula: (7) in, N Indicates the number of sampling points. k =1,..., K .

[0012] Furthermore, in step S3, the PSO algorithm is used to iteratively update the decomposition layer number Kvmd (i.e., particle position) and penalty factor α (i.e., particle velocity) in VMD to find the optimal combination of parameters required in the VMD method; where PSO stands for Particle Swarm Optimization Algorithm. The PSO algorithm is as follows: Select the number of decomposition layers Kvmd and the penalty factor α required for VMD decomposition. Record the historical fitness y, particle position Kvmd and particle velocity α for each input ICG signal for VMD decomposition. Select the optimal parameter combination based on the historical fitness to obtain the number of decomposition layers Kvmd and the penalty factor α required for VMD decomposition. The formula for calculating the relevant fitness is as follows; (8) Where y is the corresponding fitness, and Env is the parameter obtained by Hilbert transforming the VIMF component. This parameter is a complex number with a data length of n. This represents the result obtained after modulo Env. i The parameters in the overall n The ratio of each parameter; n This indicates the data length of each VIMF component, which is the same as the length of the input signal.

[0013] Furthermore, in step S5, the method for selecting the layer number K2 is as follows: the VIMF components are mapped to the frequency domain by Fourier transform to obtain the center frequency of each VIMF component, and then the layer corresponding to the first VIMF component with a center frequency less than 25Hz is selected as K2.

[0014] Furthermore, in step S6, the specific steps of the segmented extraction method involving the difference between the two sides and fixing point C include: S61: Perform a difference operation on s(n) once, setting the data less than 0 in the difference signal to 0 and setting it as difficg1; then perform a difference operation on the s(n) signal again, setting the data greater than 0 to 0 and setting it as difficg2; S62: Square the data in difficg1 and difficg2 and use a 3-point smoothing filter. Then perform a moving integral at a sampling rate of 0.05 times fs to obtain the processed signals traicg1 and traicg2. Then, find the peak points of traicg1 and traicg2 with thresholds greater than 0.4*max(traicg1) and 0.4*max(traicg2) respectively. The peak points of the two sets of data are peak1 and peak2 respectively. S63: Subtract each value in peak1 and peak2 to find the distance between each peak point. Then select only the peak points where the difference between the two is greater than 0 and the distance does not exceed 0.2 times the sampling rate fs, and redefine the two sets of peak data as new peak1 and peak2. Find the maximum value of s(n) in the range [peak1, peak2], which is point C, and its amplitude is denoted as C_value. S64: Find the minimum point X1 of s(n) within the range [C, C+0.15*fs] and the minimum point X2 within the range [C, C+0.2*fs]. If they match, X1 is considered to be point X; otherwise, X2 is considered to be point X. S65: Find the minimum point A1 of s(n) within the range [C-0.2*fs,C] and the minimum point A2 within the range [C-0.25*fs,C]. If they match, then A1 is considered point A; otherwise, A2 is considered point A. S66: For s(n), find the point O1 where the first difference value is approximately 0 within the range [X,X+0.25*fs] and the point O2 where the second difference value is approximately 0 within the range [X,X+0.3*fs] (if any). If the range of O2-O1 is within 0.1*fs, then O2 is considered point O; otherwise, O1 is considered point O. S67: Find the position of s(n) with an amplitude of approximately 0.15*C_value within the interval [A,C]. The point corresponding to this position is point B.

[0015] The beneficial effects of this invention are as follows: (1) The signal processing method for separating cardiac impedance signal details designed in this invention has better robustness and stable noise reduction effect on the basis of preserving signal detail features. Based on the combination of human physiological significance and the temporal relationship of feature points, the extracted feature points related to the contraction and relaxation relationship of cardiac function have higher accuracy.

[0016] (2) The method of the present invention achieves feature preservation and stable denoising of cardiac impedance signals, solves the problems of strong subjectivity and poor adaptability in signal processing, and provides a solid technical foundation for the field of non-invasive hemodynamics.

[0017] (3) The present invention is based on a method for realizing detailed separation and reconstruction of cardiac impedance signals and feature extraction. It has high stability and robustness, and provides a high-quality signal guarantee for obtaining more accurate hemodynamic parameters.

[0018] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0019] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1This is a schematic diagram of the overall process of the method for realizing detailed separation and reconstruction and feature extraction of cardiac impedance signals according to the present invention; Figure 2 A schematic diagram of the waveform feature points of the cardiac impedance signal in the present invention; Figure 3 This invention provides a time-domain data graph of cardiac impedance at a sampling rate of 1000Hz for 5 seconds. Figure 4 This is a schematic diagram illustrating the time-domain and spectral analysis of NIMF modal components based on ICEEMDAN decomposition, as implemented in this invention. Figure 5 This is a schematic diagram of the energy distribution of the NIMF component in the K1 layer of the present invention, which combines signal energy to select cardiac impedance detail reconstruction. Figure 6 This is a schematic diagram illustrating the iterative update process of the parameters required by the VMD algorithm using the PSO algorithm in the implementation of this invention. Figure 7 This is a schematic diagram of the time-domain and spectral analysis of VIMF modal components based on VMD decomposition according to an embodiment of the present invention; Figure 8 This is a schematic diagram showing the time domain and spectrum comparison of cardiac impedance signal before and after denoising according to an embodiment of the present invention. Figure 9 This is a flowchart illustrating an embodiment of the present invention based on two-sided differential and segmented positioning feature points. Figure 10 This is a schematic diagram of the differential peak value and cardiac impedance signal point C for bilateral localization of cardiac impedance differential in an embodiment of the present invention. Figure 11 This is a schematic diagram of a cardiac impedance signal for locating five feature points (A, B, C, X, and O) after denoising, as implemented in this invention. Detailed Implementation

[0020] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0021] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0022] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0023] Please see Figures 1 to 11 This invention provides a method for detailed separation, reconstruction, and feature extraction of cardiac impedance signals. The so-called cardiac impedance signal feature points refer to the A wave (symbolizing atrial systole), the B point (representing aortic opening), the C wave (highest amplitude in the impedance diagram), the X point (aortic valve closure), and the O wave (representing left atrioventricular valve opening). These are used to calculate non-invasive hemodynamic parameters, such as calculating stroke volume (SV) using the time from B to X (left ventricular ejection time, LVET) and the C wave amplitude. They can represent the systolic and diastolic functions of the human heart, as well as other physiological characteristics, and have broad prospects for medical applications. Based on this objective, this invention addresses the problems of error in feature point identification, waveform distortion, and amplitude attenuation in existing technologies. The practical application and effects of this invention are further illustrated below through specific embodiments.

[0024] Example: Figure 3 This is a 5-second cardiac impedance signal acquired by a clinical cardiac impedance device at a sampling frequency of 1000Hz. The signal contains a significant amount of noise that is not part of the cardiac impedance information. Directly locating features from this signal would lead to serious detection biases and make the calculation of non-invasive hemodynamic parameters difficult.

[0025] This implementation employs a method for detail separation, reconstruction, and feature extraction of cardiac impedance signals. The overall process is as follows: Figure 1 As shown, the collected cardiac impedance signal is input into the computer, and the processing steps include: S1: Using the ICEEMDAN method, the cardiac impedance signal ICG is decomposed into a set of NIMF eigenmode components representing different frequency and amplitude components of the signal:

[0026] in, This represents the first decomposition by ICEEMDAN. k Layer NIMF eigenmode functions, k =1,..., K .

[0027] In this embodiment, the cardiac impedance signal and its interference are distributed in different frequency bands, such as Figure 4 As shown, the NIMF components are sorted from high frequency to low frequency according to the characteristics of the ICEEMDAN algorithm, resulting in K=10 layers of NIMF components, where the 10th layer is the baseline.

[0028] S2: Select K1 to divide the detailed components of the NIMF component related to cardiac impedance signal and reconstruct them into ICG1; and reconstruct the remaining components that still contain some cardiac impedance signal information into ICG2:

[0029]

[0030] The minimum values ​​of the energy of each NIMF component, excluding the first and Kth layers, are obtained by using the following formula. Then, the layer number closest to the low frequency is selected from the minimum values. If there is no minimum value, half of the total number of decomposition layers is taken, which is K1.

[0031]

[0032] in, Indicates the first k Energy of one NIMF component N Indicates the number of sampling points. k =1,..., K .

[0033] In this embodiment, the energy of each layer of NIMF(k,n) is calculated as follows: Figure 5 As shown, K1=6 is obtained according to the method described above.

[0034] S3: Select the decomposition level Kvmd and the penalty factor α to perform VMD decomposition on ICG2, and obtain the VIMF components of the Kvmd group:

[0035] Please see Figure 6The optimal combination of parameters required for the VMD method is found by iteratively updating the decomposition level Kvmd (particle position) and penalty factor α (particle velocity) in the VMD decomposition using the Particle Swarm Optimization (PSO) algorithm. Specifically, by recording the historical fitness (y), particle position (Kvmd), and particle velocity (α) for each input ICG signal in VMD decomposition, the optimal parameter combination is selected based on the historical fitness to obtain the required decomposition level Kvmd and penalty factor α for VMD decomposition. The relevant fitness calculation formula is as follows.

[0036]

[0037] Where y is the corresponding fitness, and Env is the parameter obtained by Hilbert transforming the VIMF component. This parameter is a complex number with a data length of n. This represents the result obtained after modulo Env. i The parameters in the overall n The ratio of each parameter, n This indicates the data length of each VIMF component, which is the same as the length of the input signal.

[0038] In this example, the upper and lower boundary values ​​for particle position and particle velocity are set to [5,20] and [500,5000], respectively, and only integers are used. The VMD algorithm yields a decomposition level Kvmd of 10 and a penalty factor α of 3645. The ICG2 decomposition yields the following results: Figure 7 The variational mode components shown show that the frequencies of VIMF8-VIMF10 are all within the frequency range of the cardiac impedance signal and still contain a lot of detailed information, while VIMF1 to VIMF7 gather high-frequency information above 50Hz.

[0039] S4: Adaptive thresholding of VIMF components is performed using the following formula:

[0040] in, Indicates the first i Layer VIMF components, Indicates the number of thresholds after threshold processing i Layer VIMF components; It is the threshold, which is three times the standard deviation of the signal.

[0041] In this embodiment, the threshold is 4.438.

[0042] S5: Select ICG3, which is the other detailed information about the cardiac impedance signal in the VIMF component reconstructed by K2, and combine it with ICG1 obtained in step S2 to jointly reconstruct the denoising result s(n) of the cardiac impedance signal:

[0043] The center frequency of each VIMF component is obtained by mapping the VIMF components to the frequency domain through Fourier transform, and then the layer corresponding to the first VIMF component with a center frequency of less than 25Hz is selected as K2.

[0044] In this embodiment, based on Figure 7 The spectrum also shows that K2=8. The final time-domain and spectrum diagrams of the cardiac impedance signal before and after denoising are shown below. Figure 8 As shown.

[0045] S6: Please refer to Figure 9 The flowchart describes a method using differential thresholding and moving integrals to differentially analyze the cardiac impedance signal from both sides, fix point C, and then extract features from A, B, X, and O segments. The basic steps include: S61: Perform a differential operation on s(n) once, setting the data less than 0 in the differential signal to 0 and setting it as difficg1; then perform a differential operation on the s(n) signal again, setting the data greater than 0 to 0 and setting it as difficg2.

[0046] S62: Square the data in difficg1 and difficg2 and use a 3-point smoothing filter. Then perform a moving integral at 0.05 times the sampling rate (fs) to obtain the processed signals traicg1 and traicg2. Then, find the peak points of traicg1 and traicg2 with thresholds greater than 0.4*max(traicg1) and 0.4*max(traicg2) respectively. The peak points of the two sets of data correspond to peak1 and peak2 respectively.

[0047] S63: Subtract each value from peak1 and peak2 to find the distance between each peak point. Then select only the peak points where the difference between the two is greater than 0 and the distance does not exceed 0.2 times the sampling rate (fs), and redefine the two sets of peak data as new peak1 and peak2. Find the maximum value of s(n) in the range [peak1, peak2], which is point C, and its amplitude is denoted as C_value.

[0048] S64: Find the minimum point X1 of s(n) within the range [C, C+0.15*fs] and the minimum point X2 within the range [C, C+0.2*fs]. If they match, X1 is considered to be point X; otherwise, X2 is considered to be point X.

[0049] S65: Find the minimum point A1 of s(n) within the range [C-0.2*fs,C] and the minimum point A2 within the range [C-0.25*fs,C]. If they are the same, then A1 is considered point A; otherwise, A2 is considered point A.

[0050] S66: For s(n), find the point O1 whose first difference value is approximately 0 within the range [X,X+0.25*fs] and the point O2 whose second difference value is approximately 0 within the range [X,X+0.3*fs] (if any). If the range of O2-O1 is within 0.1*fs, then O2 is considered to be point O; otherwise, O1 is considered to be point O.

[0051] S67: Find the position of s(n) with an amplitude of approximately 0.15*C_value within the interval [A,C]. The point corresponding to this position is point B.

[0052] In this embodiment, as Figure 10 As shown, several peaks 1 and 2 are obtained from the differential thresholds on both sides. By mapping these peaks to the cardiac impedance signal, the accurate location of point C is obtained. Furthermore, based on the above steps, the following is obtained: Figure 11 The positions of the other four feature points A, B, X, and O shown.

[0053] The algorithm of this invention can further improve the robustness of signal processing algorithms while ensuring the amplitude, waveform and other characteristics of cardiac impedance. It can stably denoise and extract feature points from cardiac impedance signals, laying the foundation for further calculation of hemodynamic parameters and clinical applications.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for detail separation, reconstruction, and feature extraction of cardiac impedance signals, characterized in that, The method specifically includes the following steps: S1: Using the ICEEMDAN method, the cardiac impedance signal (ICG) is decomposed into NIMF components representing different frequency and amplitude components of the signal: in, This represents the first decomposition by ICEEMDAN. k Layer NIMF eigenmode functions, k =1,..., K , K This indicates the number of decomposition layers of the ICG signal after ICEEMDAN decomposition. n The length of each NIMF component is equal to that of the input signal; ICEEMDAN represents the improved adaptive noise complete set empirical mode decomposition; and NIMF represents the mode components obtained after ICEEMDAN decomposition of the ICG signal. S2: Select layer K1 to divide the NIMF component into detailed components related to cardiac impedance signal and reconstruct them into ICG1, and reconstruct the remaining components that still contain some cardiac impedance signal information into ICG2: S3: Select the number of decomposition layers Kvmd VMD decomposition of ICG2 with penalty factor α yields... Kvmd Group VIMF components: Wherein, VMD represents variational mode decomposition, and VIMF represents the modal components obtained after VMD decomposition of the ICG signal; S4: Adaptive thresholding of VIMF components is performed using the following formula: in, Indicates the first i Layer VIMF components, Indicates the number of th values ​​after threshold processing i Layer VIMF components; This is a threshold value, which can be adjusted based on the actual situation. S5: Select layer K2 to reconstruct other detailed information about the cardiac impedance signal in the VIMF component as ICG3, and combine it with ICG1 obtained in step S2 to jointly reconstruct the denoising result s(n) of the cardiac impedance signal: S6: Using the differential threshold and moving integral scheme, the cardiac impedance signal after denoising in step S5 is differentially divided on both sides and the C point is fixed, and then the features of A, B, X, and O are extracted in segments. The specific steps of the segmented extraction method using the two-sided difference and fixed point C include: S61: Perform a difference operation on s(n) once, setting the data less than 0 in the difference signal to 0 and setting it as difficg1; then perform a difference operation on the s(n) signal again, setting the data greater than 0 to 0 and setting it as difficg2; S62: Square the data in difficg1 and difficg2 and apply a 3-point smoothing filter. Then perform a moving integral at 0.05 times the sampling rate fs to obtain the processed signals traicg1 and traicg2. Finally, calculate the threshold values ​​for traicg1 and traicg2 respectively. and The peak points of the two sets of data are peak1 and peak2, respectively; S63: Subtract each value in peak1 and peak2 to find the distance between each peak point. Then select only the peak points where the difference between the two is greater than 0 and the distance does not exceed 0.2 times the sampling rate fs, and redefine the two sets of peak data as new peak1 and peak2. Find the maximum value of s(n) in the range [peak1, peak2], which is point C, and its amplitude is denoted as C_value. S64: The range of s(n) is within The minimum point X1 within the range and the range within If the minimum value point X2 within the range is the same, then X1 is considered to be point X; otherwise, X2 is considered to be point X. S65: The range of s(n) is within The minimum point A1 within the range is within If the minimum value point A2 is the same as that of A1, then A1 is considered point A; otherwise, A2 is considered point A. S66: The range of s(n) is within The point O1 where the first difference value is approximately 0 and the range is within The second difference value is approximately 0 at point O2, and the range of O2-O1 is within... If the value is within the range of O2, then O2 is considered point O; otherwise, O1 is considered point O. S67: For s(n), the range is approximately [A, C], and the amplitude is approximately... The location of the point is point B.

2. The method for detail separation, reconstruction, and feature extraction of cardiac impedance signals according to claim 1, characterized in that, In step S1, the NIMF components are sorted from high frequency to low frequency, with the Kth layer being the baseline.

3. The method for detail separation, reconstruction, and feature extraction of cardiac impedance signals according to claim 1, characterized in that, In step S2, the method for selecting the number of layers K1 is as follows: by finding the minimum value of the energy of each NIMF component except for the first layer and the Kth layer, the layer number close to the low frequency is selected from the minimum value. If there is no minimum value, half of the total number of decomposition layers is taken, which is K1.

4. The method for detail separation, reconstruction, and feature extraction of cardiac impedance signals according to claim 3, characterized in that, In step S2, the first k Energy of one NIMF component It can be obtained using the following formula: in, N Indicates the number of sampling points. k =1,..., K .

5. The method for detail separation, reconstruction, and feature extraction of cardiac impedance signals according to claim 1, characterized in that, In step S3, the PSO algorithm is used to iteratively update the decomposition layer Kvmd (particle position) and penalty factor α (particle velocity) in VMD to find the optimal combination of parameters required in the VMD method; where PSO stands for Particle Swarm Optimization Algorithm. The PSO algorithm is as follows: Select the number of decomposition layers Kvmd and the penalty factor α required for VMD decomposition. Record the historical fitness y, particle position Kvmd and particle velocity α for each input ICG signal for VMD decomposition. Select the optimal parameter combination based on the historical fitness to obtain the number of decomposition layers Kvmd and the penalty factor α required for VMD decomposition. The formula for calculating the relevant fitness is as follows; Where y is the corresponding fitness, and Env is the parameter obtained by Hilbert transforming the VIMF component. This parameter is a complex number with a data length of n. This represents the result obtained after modulo Env. i The parameters in the overall n The ratio of each parameter; n This indicates the data length of each VIMF component, which is the same as the length of the input signal.

6. The method for detail separation, reconstruction, and feature extraction of cardiac impedance signals according to claim 1, characterized in that, In step S5, the method for selecting the layer number K2 is as follows: the VIMF components are mapped to the frequency domain by Fourier transform to obtain the center frequency of each VIMF component, and then the layer corresponding to the first VIMF component with a center frequency less than 25Hz is selected as K2.

Images