Intelligent cardiovascular health non-inductive monitoring method, device and system

Abstract

The present application relates to the field of heart rate detection, and discloses a kind of intelligent cardiovascular health non-inductive monitoring method, device and system.The present application is based on the motion analysis of entire heart in cardiac cycle, establishes the basic model between the two kinds of one-dimensional signs of body surface, proposes the pre-processing method of sign signal, and the reconstruction method of electrocardiosignal based on fusion mechanism modeling and deep learning method, so that the synthesized electrocardiogram can effectively retain cardiac electrophysiological function information, biological difference and pathological information;Realize the multi-dimensional non-inductive evaluation of cardiac physiology on a single device;At the same time, based on the non-contact acquisition of BCG signal, the detection method is more convenient, safe;Compared with the non-contact device for collecting other modal data, the interference is small, and it is easy to charge;Real-time electrocardiogram data of heart can be captured without direct contact with skin, solving the problem that horizontal heart monitoring can only monitor coarse-grained heart activity.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of heart rate detection, in particular to a smart cardiovascular health non-invasive monitoring method, device and system. BACKGROUND

[0002] Ballistocardiogram (BCG) refers to the mechanical movement signal of the human chest, mainly caused by the contraction and relaxation of the heart and the breathing of the lungs. When the heart contracts, due to the contraction of the heart muscle, blood flow and pressure change, causing slight mechanical vibration of the chest, which can be captured to form a BCG signal. BCG signal is a non-invasive, convenient, repeatable and high-resolution physiological signal, which has many application status.

[0003] The traditional electrocardiogram (ECG) monitoring method needs to paste or clamp the traditional contact electrode on the skin surface, and in order to obtain good electrophysiological information, it is also necessary to apply electrolyte on the skin surface, as shown in patent CN210990273U, which still needs to wear electrode sheet for acquisition. The method is not suitable for patients with damaged skin surface due to burns and other reasons, and the wearing method based on electrode sheet is also not suitable for long-term use in daily life. For some detection devices that do not need to contact the skin, there is the disadvantage of inaccurate detection, as shown in patent CN106404452B, which can only monitor and evaluate the coarse-grained cardiac activity (such as heart rate, in-bed, heart beat interval, etc.). SUMMARY

[0004] To solve at least one of the above problems, the present application first provides a smart cardiovascular health non-invasive monitoring method, comprising the following steps:

[0005] Signal acquisition: lay the flexible detection belt under the chest of the monitored human body, use the characteristic that the flexible detection belt produces micro-deformation after being subjected to pressure, and collect the mechanical movement signal of the chest of the monitored human body;

[0006] Signal preprocessing: use Fourier transform to analyze the frequency distribution of the collected ECG signal, select a specific frequency for filtering; use variational mode decomposition (VMD) technology to decompose and reconstruct the collected BCG signal, and get the stable representation of the overall mechanical movement of the heart through mode matching;

[0007] Feature acquisition: use Pan-Tompkins algorithm to search for R wave in long-time ECG signal, according to the biological mechanism law, take the start and end points of ECG per heartbeat as time calibration to segment and synchronize BCG signal, and determine the position of each feature wave of BCG;

[0008] Modeling: a biological mechanism and blood circulation feedback model is established, the biological mechanism and blood circulation feedback model includes a myocardial cell excitation-contraction coupling equivalent model, a whole heart complex network model, an electrocardiogram-ballistocardiogram coupling model, the electrocardiogram-ballistocardiogram coupling model uses mutual supervision detection and positive feedback correction algorithm based on three-layer iterative regularization algorithm TSI-DTW, accelerates the convergence speed of the network model, and improves the accuracy of waveform detection;

[0009] Non-contact electrocardio signal reconstruction model is established: a multi-domain generative adversarial network Mixed-GAN is established to realize end-to-end reconstruction mapping from BCG to ECG;

[0010] The heart is evaluated: the synthesized ECG and the measured BCG are jointly calculated to evaluate the change speed of cardiac contractility, ventricular valve movement, cardiac contractility, and the fluctuation of cardiac output CO related to aortic blood pressure difference.

[0011]

[0012] Optionally, in the signal preprocessing step, the processing of the collected BCG signal is specifically: using a VMD method to decompose the BCG signal into independent intrinsic mode functions with different center frequencies, and each mode expression is:

[0013]

[0014] In the formula, denotes an envelope, denotes a phase, and a Lagrange multiplier method is used to solve the following expression using a VMD method, wherein the unconstrained variation problem adaptively determines and the center frequency :

[0015] In the formula, is a Dirac function, and the penalty term ensures the accuracy of the decomposition reconstruction in the presence of Gaussian noise, and the Lagrange penalty term can ensure strict constraints on the decomposition process;

[0016] An alternating direction multiplier method ADMM is used to convert an optimization problem with multiple variables into three optimization sub-problems about , and , an augmented Lagrange function is constructed according to the constraint conditions and objective functions of the sub-problems, and then the optimal solutions of the three optimization sub-problems are solved respectively.

[0017] Optionally, the VMD parameters are optimized, and the specific optimization steps are as follows: first, determine the range of the parameters, then perform VMD decomposition on the original signal and calculate the relative entropy under each parameter combination, iteratively update the parameters and calculate the relative entropy, and finally select the parameters corresponding to the minimum relative entropy as the optimal solution; preset the cardiac pattern matching score threshold, discard the data segments whose distance from the best template matching result LDTW is higher than the threshold, and reduce the impact of body motion interference of continuous data on the BCG signal.

[0018] Optionally, the best influencing parameter combination of VMD is searched, and the penalty parameter and the number of components are selected according to the minimum relative entropy. First, the original signal is decomposed into Narrowband modal functions are divided into low-frequency modes and high-frequency modes according to the size of the center frequency. The initial value is 5, the step size is 1, and the range is [3, 20]; the penalty parameter The initial value is 2000, the step size is 50, the range is [100, 3000], and the convergence tolerance is .

[0019] Optionally, the specific method for establishing the myocardial cell excitation-contraction coupling equivalent model is as follows:

[0020] At the i-th time point in a cardiac cycle, the electrical signal is transmitted sequentially along the excitation conduction system to the n-th myocardial cell, and the cardiac electrical activity is obtained. , mechanical activities As the observation value, a two-port network is established. At the i-th time point of the cardiac cycle, the s-domain expression of the electrical-mechanical signal relationship of the "empty heart" is:

[0021]

[0022] Where, represents the transfer function of the ECC mechanism, represents the absolute refractory period ERP, Indicates the mechanical activity of the heart, Indicates the electrical activity of the heart. Indicates electrocardiogram readings;

[0023]

[0024] Where, represents the mechanotransduction transfer function, Represents force sensor results.

[0025] Optionally, the specific method for establishing the whole heart composite network model is as follows:

[0026] In a cardiac cycle Time points ( ), sequentially transmit electrical signals to the first cardiac muscle cell along the excitation conduction system; when the blood pumping event occurs, the electrical-mechanical movement of the whole heart is a composite port network ; then at the first time point of the cardiac cycle, the s-domain expression of the electrical-mechanical movement signal relationship of the whole heart is:

[0027]

[0028] In the formula, represents the transfer function of the electrical-mechanical movement of the whole heart, represents the absolute refractory period ERP, represents the mechanical activity of the heart, represents the electrical activity of the heart, represents the electrocardiogram index;

[0029]

[0030] Therefore, the s-domain expression of the electrical-mechanical relationship of the whole heart in the whole cardiac cycle is:

[0031] In the formula, , represents the length of the cardiac cycle, represents the transfer function of the ECC mechanism, represents the transfer function of the electrical-mechanical movement of the whole heart.

[0032] Alternatively, the specific method for establishing the electrocardiogram- ballistocardiogram coupling model is as follows:

[0033] Taking the periodic movement of the whole heart as the observation object, the depolarization and repolarization of the myocardial cells along the excitation conduction system sequentially from the sinoatrial node can be equivalent to the potential change of the sinoatrial node at different phases, that is: ; by introducing the blood circulation feedback system, the relationship between the ballistocardiogram and the electrocardiogram is described;

[0034] The electrocardiogram-ballistocardiogram transfer function represented by the electrical and mechanical activities of the heart is:

[0035]

[0036] In the formula, is the system output, which is used to describe the mechanical observation signal, and the unit is Newton (N); is the system input, which is used to describe the electrical observation signal, and the unit is ampere (A); is the transfer function of the excitation-contraction coupling; is the transfer function of the blood circulation; is a unit step function, which means that The blood circulation events occur after the moment begins; this formula describes the changes of BCG over time, including the open-loop response of the myocardial cell two-port network system and the positive feedback of blood circulation. The impact of moment superposition:

[0037] A mutually supervised BCG detection method is proposed, which includes feedback correction, detection algorithm constraints, and a three-layer iterative time dynamic warping algorithm TSI-DTW. The waveform detection results are used to feed back DTW template training. The mutually supervised learning method of DTW template for feature extraction and labeling can accelerate the convergence of the network model. Moreover, the DTW template corrected by mutual supervision learning can correct the shortcomings of manual experience, reduce poorly labeled heartbeats, and further improve the accuracy of waveform detection.

[0038] Optionally, the specific method for establishing the non-contact ECG signal reconstruction model is as follows:

[0039] Multi-domain generative adversarial network Mixed-GAN includes generator and time domain discriminator and frequency domain discriminator , generator The input is the measured ECG and BCG, and the output is the synthesized , time domain discriminator The input is and , frequency domain discriminator The input is the spectrum amplitude of the real ECG synthesized after Fourier transform, denoted as and ;

[0040] The generator is used to generate synthetic data similar to real ECG signals from noise; the generator consists of five fully connected layers. The first four fully connected layers perform batch normalization (BN) and ReLU activation functions to enhance the nonlinear representation capability of the model; the last fully connected layer uses the Tanh function to ensure that the generated ECG signal falls within the range of [-1, 1].

[0041] In generative adversarial training, the temporal discriminator is updated based on the consistency of the temporal waveform. The objective function is expressed as follows:

[0042]

[0043] Where, express expectations; represents the time domain discriminator Output for generating ECG; express Output of real ECG;

[0044] updating the frequency domain discriminator based on the frequency domain consistency The objective function of the generator is expressed as follows:

[0045]

[0046] wherein, denotes expectation; denotes the frequency domain discriminator outputting the ECG spectrum amplitude; denotes outputting the real ECG spectrum amplitude;

[0047] updating the generator The loss function of the generator is expressed as follows:

[0048]

[0049] Based on the deep learning network architecture of the learned cross-domain mapping of the heart mechanical activity and the heart electrical activity, the heart mechanical activity data extracted at the current moment is input, and the ECG measurement result at the current moment is output, so as to realize non-contact electrocardiogram monitoring. After 25 healthy user experimental data machine training, the Pearson correlation coefficient of the method on the test set data is 0.963, and the root mean square error is 0.058.

[0050] Compared with the prior art, the intelligent cardiovascular health non-sensing monitoring method in the application proposes a domain mapping algorithm for the specified one-dimensional mechanical signal of the heart, which can realize electrocardiogram based on the one-dimensional mechanical signal of the heart, and can realize fine-grained evaluation of the heart based on the electrocardiogram. Real-time electrocardiogram data of the heart can be captured without direct contact with the skin, and the deficiency that the horizontal heart monitoring can only monitor the coarse-grained heart activity is solved.

[0051] In addition, the application provides an intelligent cardiovascular health non-sensing monitoring device, comprising:

[0052] A flexible detection belt is used to collect the mechanical motion signal of the chest of the human body to be monitored, and the flexible detection belt comprises a piezoelectric sensor, a piezoresistive sensor, a dynamic range controllable data acquisition circuit, a feedback control module, an AI autonomous data analysis module, a second data analysis module, a minute data analysis module, a daily report analysis module, a data storage module, a WIFI Internet of Things protocol module, a router, a cloud server, a background management, and a user terminal.

[0053] A general processor is used for data processing and comprehensive evaluation according to the intelligent cardiovascular health non-sensing monitoring method.

[0054] The intelligent cardiovascular health non-sensing monitoring device has the advantages that the non-contact key life feature signal acquisition is invisible to the patient, no object needs to be connected to the patient, the device is small in size, light in weight, and thin and can be curled and carried conveniently, and the device is simple in structure and convenient to install.

[0055] In addition, the application provides an intelligent cardiovascular health non-sensing monitoring system, which comprises an electronic terminal, a WIFI module is arranged on the electronic terminal, the WIFI module is wirelessly connected with a flexible detection belt, the electronic terminal stores instructions executable by a general processor, and the instructions are executed by the general processor to realize the intelligent cardiovascular health non-sensing monitoring method.

[0056] Compared with the prior art, the intelligent cardiovascular health non-sensing monitoring system has the advantages that open design ideas, equipment and cloud services are adopted, and the system has flexible expansion capability; the system has the hardware data backtracking function for delay / missing transmission caused by various conditions by adopting the idea of long and short storage; the device has the network outage back transmission function, and the first 60 seconds of the second data and the first 30 minutes of the data can be retrieved; the local 15-day daily report storage is adopted to exclude unexpected interference and ensure that data is not lost. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 A flowchart of an intelligent cardiovascular health non-sensing monitoring method according to an embodiment of the application;

[0058] Figure 2 A schematic diagram of a multi-module hardware sensor;

[0059] Figure 3 A device architecture diagram of an integrated multi-module sensing intelligent fusion module;

[0060] Figure 4 A filtering result and a frequency spectrum of ECG before and after filtering;

[0061] Figure 5 A flowchart of optimization calculation of VMD parameters;

[0062] Figure 6 A BCG preprocessing result;

[0063] Figure 7 A result of signal segmentation and feature extraction;

[0064] Figure 8 An ECC mechanism of a cell and an equivalent circuit thereof;

[0065] Figure 9 A whole heart composite network model;

[0066] Figure 10 A blood circulation feedback system;

[0067] Figure 11 TSI-DTW mutual supervision algorithm mode matching result;

[0068] Figure 12 Multi-domain generative adversarial network structure diagram;

[0069] Figure 13 Result of the reconstruction model based on non-contact electrocardiogram signals;

[0070] Figure 14 Expansion application result of the synthesized ECG and the measured BCG;

[0071] Figure 15 Intelligent cardiovascular health non-inductive monitoring system communication layer design idea;

[0072] Figure 16 Mobile phone application interface. DETAILED DESCRIPTION

[0073] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0074] In the description of the present application, it should be understood that the terms "upper", "lower" and the like indicate the orientation or positional relationship based on the orientation or positional relationship when the product is normally used.

[0075] The terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features.

[0076] The embodiment of the present application provides an intelligent cardiovascular health non-inductive monitoring method, which comprises the steps as shown in the figure: Figure 1 The flexible sensor is used to collect the mechanical motion signal related to the heart on the surface of the body; the collected signal is preprocessed, including signal filtering, segmentation and feature extraction; based on the biological mechanism of myocardial cells, whole heart activity and blood circulation, a electrocardiogram- ballistocardiogram coupling model is proposed; a mutual supervision detection and positive feedback correction algorithm based on three-layer iterative regularization algorithm TSI-DTW is proposed: after preprocessing, the BCG signal is matched by mode matching to obtain a stable representation of the whole heart mechanical movement synthesis module; an improved generative adversarial network (Mixed-GAN) is proposed to realize the end-to-end reconstruction mapping from BCG to ECG; finally, based on the deep learning network architecture of the learned cross-domain mapping of the heart mechanical activity and the heart electrical activity, the current time heart mechanical activity data is input, and the current time reconstruction ECG is output, and then the non-contact electrocardiogram monitoring is realized.

[0077] Signal acquisition:

[0078] As shown in Figure 2 and 3 , a multi-mode body surface motion sensor is used to capture the small movements of the human heart, breathing, etc. The sensor locally runs an AI adaptive algorithm, automatically adjusts the optimal working parameters according to the user and the environment, thereby avoiding the adaptability problem under different people and different use conditions; using a fast-response autonomous Internet of Things protocol, the sensor to the server to the terminal is transmitted in seconds without delay; using an open design idea, the sensor and cloud service have flexible expansion capability; the sensor has a network disconnection return function, using the idea of long and short storage, it has a hardware data backtracking function for delay / missing transmission caused by various conditions, it can retrieve the previous 60 seconds of second data and the previous 30 minutes of data; using local daily report storage for up to 15 days, excluding unexpected interference, to ensure that data will not be lost.

[0079] ECG preprocessing:

[0080] The frequency distribution of the ECG signal is analyzed using Fourier transform, and it is observed that the frequency of the ECG signal is mainly below 45Hz. Since the useful information of the ECG signal is below 30Hz, a Butterworth band-pass filter of 1-30Hz is used to filter the ECG. The filtering result and the frequency spectrum of the ECG before and after filtering are shown in Figure 4 .

[0081] BCG preprocessing:

[0082] The observed BCG mainly contains respiratory signals, noise and BCG itself. After spectral analysis, it is found that the respiratory signal behaves as a low-frequency harmonic in the 0.2-0.9Hz frequency band, corresponding to a mode with a low center frequency; while the ballistocardiogram signal energy is mainly concentrated in the 1-10Hz frequency band. Therefore, the VMD method is used to separate the respiratory signal and noise interference in the observed BCG, which can adaptively determine the center frequency and bandwidth. After VMD decomposition, the original signal is decomposed into independent intrinsic mode functions with different center frequencies, and the expression of each mode is:

[0083]

[0084] In the formula, represents the envelope, represents the phase. Combined with the Lagrange multiplier method, the VMD algorithm adaptively determines and the center frequency by solving the unconstrained variational problem of the following formula.

[0085]

[0086] wherein, is the Dirac function, the penalty term ensures the accuracy of the decomposition reconstruction in the presence of Gaussian noise, the Lagrange penalty term can ensure the strict constraints of the decomposition process.

[0087] Based on the alternating direction multiplier method ADMM, the optimization problem with multiple variables is converted into three optimization sub-problems about , and λ(t). Specifically, the augmented Lagrange function is constructed according to the constraint conditions and objective function of the sub-problems, and then the optimal solutions of the three optimization sub-problems are solved respectively.

[0088] Further, the optimization calculation process of the VMD parameters is as shown in Figure 5 . When optimizing the VMD parameters, first determine the range of the parameters, then perform VMD decomposition on the original signal and calculate the relative entropy under each parameter combination. Update the parameters iteratively and calculate the relative entropy, and finally select the parameters corresponding to the minimum relative entropy as the optimal solution. Considering the influence of body motion interference on continuous data on BCG signals, a heart mode matching score threshold is preset to discard data segments with LDTW distance higher than the threshold from the best template matching result.

[0089] Search for the best parameter combination of VMD, select the penalty parameter and the number of components according to the minimum value of the relative entropy. Specifically, first decompose the original signal into a narrowband modal function, which is divided into low-frequency modal and high-frequency modal according to the size of the center frequency, and the number of decomposition layers is 5, the step is 1, and the range is [3, 20]; the penalty parameter is 2000, the step is 50, and the range is [100, 3000]. The convergence tolerance size is . As shown in Figure 6 , it is the BCG decomposition reconstruction result, which conforms to the frequency spectrum distribution rule.

[0090] Feature extraction:

[0091] After searching for R waves in long ECG signals using the Pan-Tompkins algorithm, the ECG signal is segmented according to the biological mechanism rule; the start and end points of each heartbeat ECG are used as time markers to segment and synchronize the BCG signal. As shown in Figure 7 , the positions of the characteristic waves of BCG are determined according to the generation mechanism of each wave of BCG and the length of the cardiac cycle event.

[0092] Excitation-contraction coupling equivalent model of myocardial cells:

[0093] At the ith time point in a cardiac cycle, the electrical signal is sequentially transmitted to the nth myocardial cell along the excitation conduction system. The ECC mechanism and its equivalent circuit of the cell are shown in Figure 8 .

[0094] In order to observe the mapping relationship of the cardiac electrical-mechanical signals without discussing the process, the cardiac electrical activity , mechanical activity are taken as observation values, and a two-port network is established. Then, at the ith time point in a cardiac cycle, the s-domain expression of the electrical-mechanical signal relationship of the "empty heart" is:

[0095]

[0096] In the formula, represents the transfer function of the ECC mechanism, represents the absolute refractory period ERP, represents the cardiac mechanical activity, represents the cardiac electrical activity, represents the electrocardiogram.

[0097]

[0098] In the formula, represents the mechanical conduction transfer function, represents the mechanical sensor result.

[0099] The overall composite network model of the heart:

[0100] As shown in Figure 9 , at the ith time point in a cardiac cycle, the electrical signal is sequentially transmitted to the nth myocardial cell along the excitation conduction system. When the blood pumping event occurs, the electrical-mechanical movement of the entire heart is a composite port network . Then, at the ith time point in a cardiac cycle, the s-domain expression of the electrical-mechanical signal relationship of the entire heart is: In the formula,

[0101] represents the transfer function of the electrical-mechanical movement of the entire heart, represents the absolute refractory period ERP, represents the cardiac mechanical activity,

[0102] represents the cardiac electrical activity, represents the electrocardiogram.

[0103] ​​​​In summary, the s-domain expression of the electro-mechanical relationship of the heart as a whole during the entire cardiac cycle is:

[0104]

[0105] Where, , represents the length of the cardiac cycle, represents the transfer function of the ECC mechanism, A transfer function that represents the electromechanical motion of the entire heart.

[0106] ECG-Ballistocardiogram coupling model:

[0107] To obtain an accurate and complete electrocardiogram (ECG), it is necessary to know the mechanical motion of the heart. However, this state cannot be directly monitored. The ballistocardiogram (Ballistocardiogram), a mechanical signal from a specific monitoring location on the body surface, is selected as the observation signal of mechanical motion, and the ECG as the observation signal of electrical activity.

[0108] like Figure 10 As shown in the figure, the overall cyclical motion of the heart is taken as the observation object. Starting from the sinoatrial node, the myocardial cells that sequentially depolarize and repolarize along the excitation conduction system can be equivalent to the potential changes of different phases of the sinoatrial node electrical activity, that is: Similarly, the cardiac mechanical activity can be divided into piecewise functions. Figure 10 The blood circulation feedback system shown describes the relationship between the ballistocardiogram and the electrocardiogram.

[0109] Note: For the electrical excitation and mechanical response of the heart, is the product of the transfer functions of each blood circulation event, It is a delayed switch for each blood circulation function. It represents the phase difference between BCG and ECG signals after all delays are added together.

[0110] The electrocardiogram-ballistocardiogram transfer function, which represents the electrical and mechanical activity of the heart, is:

[0111] (*)

[0112] Where, is the system output, describing the mechanical observation signal, in Newton (N); It is the system input, describing the electrical observation signal, and its unit is ampere (A). is the transfer function of excitation-contraction coupling; is the transfer function of blood circulation; is a unit step function, which means that Blood circulation events occur after the moment begins.

[0113] The formula describes the change of BCG over time, which includes the open-loop response of the two-port network system of myocardial cells and the influence of the positive feedback of blood circulation superimposed at time: on the one hand, the positive feedback of blood circulation makes the overall response of the system stronger in the excitation-contraction coupling mechanism; but on the other hand, due to the intensification of oscillation, the instability of the system is also increased.

[0114] Cardiac pattern matching model:

[0115] BCG and ECG signals are typical time series biological signals with randomness and instability. However, continuous cardiac signals have stable periodicity, and for the same subject, the electrical-mechanical relationship described by continuous ECG and BCG has stable periodicity. A mutual supervision BCG detection method is proposed, including feedback correction, detection algorithm constraint and three-layer iterative time dynamic regularization algorithm TSI-DTW. As Figure 11 shown in the pattern matching result, the non-overlapping segments after correction have regularity in shape, and the characteristic waves are consistent with the typical waveforms.

[0116] Table 1 Comparison of BCG-based heart rate calculation results

[0117] Feature name Recall Precision TSI-DTW 97.79% 97.91% CEEMDAN 96.72% 97.31% PDA 79.57% 76.63%

[0118] Table 1 summarizes the results of TSI-DTW, CEEMDAN and PDA algorithms in extracting heart rate from BCG on the public data set. Through comparison, it can be found that the mutual supervision learning method of using waveform detection results to feedback DTW template training and DTW template to feature extraction and labeling can accelerate the convergence speed of the network model, and the DTW template corrected by mutual supervision learning can correct the shortcomings of artificial experience, reduce the labeling of poor heartbeats, and the excellent training template will further improve the accuracy of waveform detection.

[0119] Non-contact electrocardiogram signal reconstruction model:

[0120] Algorithm architecture:

[0121] Figure 13 The structure diagram of the multi-domain generative adversarial network is shown, specifically, the Mixed-GAN includes a generator and a time domain discriminator and a frequency domain discriminator . The input of the generator is the measured ECG and BCG, and the output is the synthesized . The input of the time domain discriminator and , and the frequency domain discriminator ​The input of the generator is the spectrum amplitude of the synthesized, real ECG after Fourier transform, denoted as and .

[0122] Note: : Generator, : Time-domain discriminator network, : Frequency-domain discriminator network. : Real ECG time series; : Synthetic ECG time series; : Real ECG frequency domain amplitude; : Synthetic ECG frequency domain amplitude.

[0123] Generator and discriminator:

[0124] The generator model is the core part of the entire algorithm, responsible for generating synthetic data similar to real ECG signals from noise. The generator model proposed in this application consists of five fully connected layers. The first four fully connected layers perform batch normalization BN and ReLU activation function to enhance the non-linear representation ability of the model. The last layer uses the Tanh function to ensure that the generated ECG signal falls within the range of [-1, 1].

[0125] Loss function:

[0126] In the generative adversarial training, based on the consistency of the time domain waveform, the objective function of the time domain discriminator is updated, expressed as follows:

[0127]

[0128] In the formula, represents the expectation; represents the output of the time domain discriminator to the generated ECG; represents the output of to the real ECG.

[0129] Based on the frequency domain consistency, the objective function of the frequency domain discriminator is updated, expressed as follows:

[0130]

[0131] In the formula, represents the expectation; represents the output of the frequency domain discriminator to the generated ECG spectrum amplitude; represents the output of to the real ECG spectrum amplitude.

[0132] The generator The loss function of the model is expressed as follows:

[0133]

[0134] Based on the deep learning network architecture of the learned cross-domain mapping of the heart mechanical activity and the heart electrical activity, the current time heart mechanical activity data is input, and the output is the ECG measurement result of the current time as shown in the formula (1), and then the non-contact electrocardiogram monitoring is realized. Through the training data of 25 healthy user experimental data, the Pearson correlation coefficient of the method on the test set data is 0.963, and the root mean square error is 0.058. Figure 14

[0135] Extended application:

[0136] The following parameters are taken as examples to describe the extended application potential of the synthesized ECG and the measured BCG.

[0137] The following parameters are calculated based on the synthesized ECG and the measured BCG, and the calculation results are shown in the formula (2). Figure 15

[0138] Feature Description R J: reflects the speed of the change in the contractility of the heart R-wave to J-wave time interval R J Amp: J-wave corresponds to ventricular valve movement R-wave to J-wave amplitude difference R I: classical BCG measure of the contractility of the heart R-wave to I-wave time interval I J Amp: IJ-wave amplitude difference reflects the fluctuations in cardiac output (CO) related to the difference in aortic blood pressure I-wave to J-wave amplitude difference

[0139] Compared with the prior art, the intelligent cardiovascular health non-sensing monitoring method in the application proposes a domain mapping algorithm for the specified one-dimensional mechanical signal of the heart, which can realize the electrocardiogram based on the one-dimensional mechanical signal of the heart, and can realize the fine-grained evaluation of the heart based on the electrocardiogram. The real-time electrocardiogram data of the heart can be captured without direct contact with the skin, and the deficiency that the horizontal heart monitoring can only perform coarse-grained heart activity monitoring is solved.

[0140] Another embodiment of the application provides an intelligent cardiovascular health non-sensing monitoring device, which comprises:

[0141] The flexible detection belt is used for collecting the mechanical motion signal of the chest of the human body to be monitored, and comprises a piezoelectric sensor, a piezoresistive sensor, a dynamic range controllable data acquisition circuit, a feedback control module, an AI autonomous data analysis module, a second data analysis module, a minute data analysis module, a daily report analysis module, a data storage module, a WIFI Internet of Things protocol module, a router, a cloud server, a background management, and a user terminal.

[0142] The total processor is used for data processing and comprehensive evaluation according to the intelligent cardiovascular health non-sensing monitoring method.

[0143] Compared with the prior art, the intelligent cardiovascular health non-sensing monitoring device has the advantages that the non-contact key life feature signal acquisition is invisible to the patient, and does not need to connect the object to the patient's body; the device is small in size, light in weight, and can be curled, and is convenient to carry; the structure is simple, and the installation is convenient.​​

[0144] Another embodiment of the present invention provides an intelligent cardiovascular health non-sensing monitoring system, including an electronic terminal, on which a WIFI module is provided, which is wirelessly connected to a flexible detection belt. The electronic terminal stores instructions that can be executed by the main processor, and the instructions are executed by the main processor to implement the intelligent cardiovascular health non-sensing monitoring method as described above.

[0145] Heart monitoring system:

[0146] Furthermore, based on the algorithm results of the above method and device, Figure 15 The communication architecture shown in FIG. 1 and the results are submitted to the terminal shown in FIG. 1 . The system also includes Figure 16 The mobile phone application shown and the Wi-Fi module in conjunction with the local gateway can also be connected to the integrated management system.

[0147] Compared with the existing technology, the advantages of the intelligent cardiovascular health non-contact monitoring system described in the present invention are that it adopts open design ideas, equipment and cloud services, and has flexible expansion capabilities; it adopts the idea of ​​both long and short storage, and has hardware data backtracking function for delays / missing transmissions caused by various situations: the equipment has a network disconnection backhaul function, which can recover the previous 60 seconds of second data and the previous 30 minutes of minute data; it adopts local daily storage of up to 15 days to eliminate accidental interference and ensure that data is not lost.

[0148] This application is based on the analysis of the relationship between the two types of movements of the entire heart during the cardiac cycle, and establishes a basic model between two types of one-dimensional representations of the body surface; proposes a device that integrates a multi-module sensor intelligent fusion module; a series of two-sign signal preprocessing methods; and a method for reconstructing ECG signals that integrates mechanism modeling and deep learning methods. The electrocardiogram synthesized based on this method can effectively retain cardiac electrophysiological function information, biological differences, and pathological information; it breaks through the dilemma of the conflict between collection and evaluation of existing cardiovascular mobile assessment equipment, and realizes multi-dimensional non-sensing assessment of cardiac physiological function on a single device; the model is based on non-contact BCG signals, and its detection method is more convenient and safe; at the same time, compared with non-contact devices that collect other modal data, it is less subject to interference and easy to charge.

[0149] The above embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patented invention. It should be noted that those skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention. These variations and improvements are equivalent modifications and improvements to the above embodiments based on the essential technology of the present invention and fall within the scope of protection of the present invention.

Claims

1. A smart cardiovascular health non-invasive monitoring method, characterized in that, The method comprises the following steps: Signal acquisition: lay the flexible detection belt under the chest of the person to be monitored, and collect the mechanical movement signals of the chest of the person to be monitored by using the characteristic that the flexible detection belt produces micro-deformation after being subjected to pressure; Signal preprocessing: analyze the frequency distribution of the collected ECG signals by using Fourier transform, select a specific frequency for filtering, and use VMD (Variational Mode Decomposition) technology to decompose and reconstruct the collected BCG signals, and obtain the stable representation of the overall mechanical movement of the heart through mode matching; Feature acquisition: search for R waves in the long-time ECG signal by using the Pan-Tompkins algorithm, divide and synchronize the BCG signal according to the start and end points of the ECG per heartbeat as the time calibration, and determine the positions of the characteristic waves of the BCG; Model establishment: establish a biological mechanism and blood circulation feedback model, which comprises a myocardial cell excitation-contraction coupling equivalent model, a heart overall composite network model and an electrocardiogram- ballistocardiogram coupling model, the electrocardiogram- ballistocardiogram coupling model uses a mutual supervision detection and positive feedback correction algorithm based on a three-layer iterative regularization algorithm (TSI-DTW) to accelerate the convergence speed of the network model and improve the accuracy of waveform detection; Non-contact electrocardio signal reconstruction model establishment: establish a mixed generative adversarial network (Mixed-GAN) to realize the end-to-end reconstruction mapping from BCG to ECG; Cardiac evaluation: jointly calculate the synthesized ECG and the measured BCG to evaluate the change speed of the cardiac contractility, the ventricular valve movement, the cardiac contractility, and the fluctuation of the cardiac output (CO) related to the aortic blood pressure difference.

2. The intelligent cardiovascular health and non-invasive monitoring method of claim 1, wherein, In the signal preprocessing step, the processing of the collected BCG signal is specifically as follows: the VMD method is used to decompose the BCG signal into independent intrinsic mode functions with different center frequencies, and the mode expression is: wherein denotes the envelope, denotes the phase, in combination with the Lagrange multiplier method, the VMD method is used to solve the following expression, in which the unconstrained variational problem adaptively determines and the center frequency : wherein is the Dirac function, the penalty term ensures the accuracy of the decomposition reconstruction in the presence of Gaussian noise, the Lagrangian penalty term can ensure the strict constraints of the decomposition process; Using the alternating direction method of multipliers (ADMM), an optimization problem with multiple variables is converted into three optimization sub-problems with respect to , and , an augmented Lagrangian function is constructed according to the constraint conditions and objective functions of the sub-problems, and the optimal solutions of the three optimization sub-problems are solved respectively.

3. The intelligent cardiovascular health and non-invasive monitoring method of claim 2, wherein, The VMD parameters are optimized, and the specific optimization steps are as follows: first, determine the range of the parameters, then perform VMD decomposition on the original signal and calculate the relative entropy under each parameter combination, iteratively update the parameters and calculate the relative entropy, and finally select the parameters corresponding to the minimum relative entropy as the optimal solution; A preset cardiac mode matching score threshold is set, and data segments with a LDTW distance higher than the threshold from the best template matching result are discarded to reduce the influence of body motion interference on the BCG signal.

4. The intelligent cardiovascular health and non-invasive monitoring method of claim 3, wherein, The best parameter combination of VMD is searched, the penalty parameter and the number of components are selected according to the minimum relative entropy, and the original signal is first decomposed into A narrow-band modal function, which is divided into low-frequency modal and high-frequency modal according to the size of the center frequency, the number of decomposition layers The initial value is 5, the step is 1, and the range is [3, 20]; the penalty parameter The initial value is 2000, the step is 50, the range is [100, 3000], and the convergence tolerance size is .

5. The intelligent cardiovascular health and non-invasive monitoring method of claim 1, wherein, The specific method for establishing the myocardial cell excitation-contraction coupling equivalent model is as follows: At the ith time point in a cardiac cycle, the electrical signal is sequentially transmitted to the nth myocardial cell along the excitation conduction system, and the cardiac electrical activity , mechanical activity As an observation value, a two-port network is established, and the s-domain expression of the electrical-mechanical signal relationship of the "empty heart" at the ith time point in a cardiac cycle is: wherein represents a transfer function of the ECC mechanism, represents an absolute refractory period ERP, represents a mechanical activity of the heart, represents an electrical activity of the heart, represents an electrocardiogram reading; wherein represents the mechanical conduction transfer function, represents the mechanical sensor result.

6. The intelligent cardiovascular health and non-invasive monitoring method of claim 1, wherein, The specific method for establishing the heart overall composite network model is as follows: In a cardiac cycle Time points ( ), transmits electrical signals in sequence along the excitation conduction system to the When a pumping event occurs, the electromechanical movement of the entire heart is a complex port network. ; Then in the first At each time point, the s-domain expression of the electrical-mechanical motion signal relationship of the entire heart is: wherein represents a transfer function of the electrical-mechanical movement of the whole heart, represents an absolute refractory period ERP, represents a mechanical activity of the heart, represents an electrical activity of the heart, represents an electrocardiogram representation; In summary, the s-domain expression of the electrical-mechanical relationship of the overall heart in the entire cardiac cycle is as follows: wherein , denotes the length of the cardiac cycle, denotes the transfer function of the ECC mechanism, denotes the transfer function of the electrical-mechanical movement of the whole heart.

7. The intelligent cardiovascular health and non-invasive monitoring method of claim 1, wherein, The specific method for establishing the electrocardiogram- ballistocardiogram coupling model is as follows: The periodic motion of the whole heart is taken as the observation object, and the depolarization and repolarization of the myocardial cells along the excitation conduction system from the sinoatrial node in sequence can be equivalent to the potential changes of different phases of the sinoatrial node electrical activity, that is: ; The relationship between the ballistocardiogram and the electrocardiogram is described by introducing a blood circulation feedback system; The electrocardiogram- ballistocardiogram transfer function represented by the electrical and mechanical activities of the heart is as follows: where is the system output, describing the mechanical observation signal, in Newton (N); is the system input, describing the electrical observation signal, in Ampere (A); is the transfer function of the excitation-contraction coupling; is the transfer function of the blood circulation; is the unit step function, indicating the occurrence of the blood circulation event after time instant; this equation describes the BCG over time, including the open-loop response of the myocardial cell two-port network system and the influence of the blood circulation positive feedback superimposed at time instant: A mutual supervision BCG detection method is proposed, including feedback correction, detection algorithm constraint and three-layer iterative time dynamic regularization algorithm TSI-DTW; the waveform detection result is used to feedback the DTW template training, and the mutual supervision learning mode of feature extraction and label of the DTW template can accelerate the convergence speed of the network model, and the DTW template corrected by mutual supervision learning can correct the shortcomings of artificial experience, reduce the labeling of poor heartbeat, and further improve the accuracy of waveform detection.

8. The intelligent cardiovascular health and non-invasive monitoring method of claim 1, wherein, The specific method for establishing the non-contact ECG signal reconstruction model is as follows: Multi-domain generative adversarial network Mixed-GAN includes generator G and time domain discriminator and frequency domain discriminator , generator The input is the measured ECG and BCG, and the output is the synthesized , time domain discriminator The input is and , frequency domain discriminator The input is the spectrum amplitude of the real ECG synthesized after Fourier transform, denoted as and ; The generator is used to generate synthetic data similar to the real ECG signal from noise; the generator includes five fully connected layers, the first four fully connected layers are batch normalized BN and ReLU activation function to enhance the nonlinear representation ability of the model; the last fully connected layer adopts Tanh function to ensure that the generated ECG signal falls within the range of [-1, 1]; In the generative adversarial training, based on the consistency of the time domain waveform, the time domain discriminator is updated The objective function of the target is expressed as follows: In the formula, represents expectation; represents a time domain discriminator to generate an output ECG; represents to a real ECG; updating the frequency domain discriminator based on frequency domain consistency The objective function is expressed as follows: wherein represents desired; represents a frequency domain discriminator to generate an output of ECG spectral magnitudes; represents to generate an output of true ECG spectral magnitudes; Update generator The loss function of the generator G is expressed as follows: Based on the deep learning network architecture of the learned cross-domain mapping of heart mechanical activity and heart electrical activity, the heart mechanical activity data at the current time is input, and the ECG measurement result at the current time is output, thereby realizing non-contact electrocardiogram monitoring; After 25 healthy user experimental data machine training, the Pearson correlation coefficient of this method on the test set data is 0.963, and the root mean square error is 0.

058.

9. An intelligent cardiovascular health and non-invasive monitoring device characterized in that, It comprises: A flexible detection belt is used to collect the mechanical motion signals of the chest of the human body to be monitored, and the flexible detection belt comprises a piezoelectric sensor, a piezoresistive sensor, a dynamic range controllable data acquisition circuit, a feedback control module, an AI autonomous data analysis module, a second data analysis module, a minute data analysis module, a daily report analysis module, a data storage module, a WIFI Internet of Things protocol module, a router, a cloud server, a background management, and a user terminal. A total processor is used for data processing and comprehensive evaluation according to the intelligent cardiovascular health non-sensing monitoring method of any one of claims 1-8.

10. An intelligent cardiovascular health and non-invasive monitoring system characterized in that, An electronic terminal is provided, and the WIFI module is arranged on the electronic terminal and wirelessly connected with the flexible detection belt; the electronic terminal stores instructions for realizing the intelligent cardiovascular health non-sensing monitoring method of any one of claims 1-8.

Images