Method and system for motion compensation of lead ecg monitoring system

Abstract

The present application relates to the technical field of health data quality control, and particularly relates to a motion compensation method and system of a lead electrocardio monitoring system, comprising: collecting multi-lead electrocardio topology signals; identifying artifact contaminated leads and corresponding artifact duration time windows; constructing a lead space motion artifact analysis model, decoupling motion artifacts into common motion artifact components and characteristic contact artifact components; the common motion artifact components represent artifacts with spatial common mode in the lead domain generated by the overall body dynamics disturbance; the characteristic contact artifact components represent artifacts caused by the instability of the specific electrode contact interface, which are concentrated in a single artifact contaminated lead; performing lead domain common mode collaborative suppression on the common motion artifact components, and performing adaptive residual filtering under fidelity constraint on the characteristic contact artifact components, to obtain compensated multi-lead electrocardio data; outputting the compensated multi-lead electrocardio data; the method realizes efficient suppression of artifacts while maintaining the authenticity and accuracy of the signals.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health data quality control, and particularly relates to a motion compensation method and system of a lead electrocardiogram monitoring system. BACKGROUND

[0002] Multi-lead electrocardiogram monitoring systems are widely used in clinical medicine and sports medicine. By arranging multiple electrodes on the surface of the human body, electrocardiogram (ECG) signals are collected to monitor the electrical activity of the heart. These systems can provide multi-angle information on heart health, which is of great significance for early diagnosis and treatment of heart disease. The common 12-lead electrocardiogram is a standard clinical detection tool, but with the advancement of technology, more lead number electrocardiogram monitoring systems have been gradually applied to real-time heart monitoring in a motion state. However, during motion or other physiological activities, factors such as body motion, muscle contraction, and unstable electrode contact can cause motion artifacts, which seriously affect the accuracy and reliability of the electrocardiogram signals. The presence of motion artifacts makes signal analysis more complex, and how to remove these artifacts and restore the true electrocardiogram signals has become a core technical problem faced by electrocardiogram monitoring systems.

[0003] Currently, there are some technical solutions that attempt to solve the problem of motion artifacts, mainly through signal processing and filtering methods to remove artifact interference. For example, based on time-frequency analysis, wavelet transform, and other methods, the signal is processed in the frequency domain, or the blind source separation (BSS) technology is used to separate the artifacts and the effective signal. However, these methods have some shortcomings, especially in the removal of complex motion artifacts. Traditional filtering techniques often cannot effectively distinguish artifact components, especially in cases where there is strong temporal and spatial correlation between leads, the separation effect of artifacts is poor. In addition, some complex algorithms have high demands on computing resources, resulting in poor real-time performance and making it difficult to meet the real-time monitoring needs in high dynamic environments. More seriously, in the process of removing artifacts, existing technologies often cause a certain degree of signal distortion, affecting the authenticity of the electrocardiogram signals, which may lead to diagnostic errors. Therefore, how to achieve accurate and efficient artifact removal, especially to handle complex motion artifacts with strong temporal and spatial correlation, is still a key problem that needs to be solved by existing technologies.

[0004] The information disclosed in this BACKGROUND section is only for the purpose of enriching the understanding of the general background of the application and should not be considered as acknowledging or implying in any form that this information constitutes prior art that is known to those skilled in the art. SUMMARY

[0005] The present application provides a motion compensation method and system of a lead electrocardiogram monitoring system, thereby effectively solving the problems in the background art.

[0006] In order to achieve the above object, the technical scheme adopted by the present application is: a motion compensation method of a lead electrocardiogram monitoring system, comprising the following steps:

[0007] Collecting a multi-lead electrocardiogram topology signal;

[0008] Based on the spatiotemporal correlation characteristics of the multi-lead electrocardiogram topology signal between different leads, identifying a pseudo-contaminated lead and a corresponding pseudo-durational time window;

[0009] Within the pseudo-durational time window, for the signal segment of the pseudo-contaminated lead, a lead space motion pseudo-analysis model is constructed to decouple the mixed motion pseudo into a common motion pseudo component and a characteristic contact pseudo component;

[0010] The common motion pseudo component represents a pseudo component with spatial common mode in the lead domain generated by the overall body dynamics disturbance; the characteristic contact pseudo component represents a pseudo component concentrated in a single pseudo-contaminated lead caused by the instability of a specific electrode contact interface;

[0011] The common motion pseudo component is subjected to lead domain common mode collaborative suppression, and the characteristic contact pseudo component is subjected to adaptive residual filtering under fidelity constraint to obtain compensated multi-lead electrocardiogram data;

[0012] The compensated multi-lead electrocardiogram data is output.

[0013] Further, the lead space motion pseudo-analysis model is constructed by electrocardiogram topology feature orthogonal decomposition to decouple the mixed motion pseudo into a common motion pseudo component and a characteristic contact pseudo component, comprising:

[0014] The multi-lead electrocardiogram signal matrix within the pseudo-durational time window is subjected to centering processing, and a covariance matrix is calculated;

[0015] The covariance matrix is subjected to electrocardiogram topology feature orthogonal decomposition to generate a topology mode vector set arranged in descending order of energy intensity;

[0016] The time sequence signal corresponding to the first topology mode vector with the maximum energy intensity is reconstructed as the common motion pseudo component;

[0017] From the remaining topology mode vectors, the topology mode vector with the maximum absolute weight on the pseudo-contaminated lead is determined, and the time sequence signal corresponding thereto is reconstructed as the characteristic contact pseudo component.

[0018] Further, determining the topology mode vector with the maximum absolute weight on the pseudo-contaminated lead comprises:

[0019] For each of the remaining topological pattern vectors, extract its corresponding vector components at the artifact contaminated lead positions;

[0020] Calculate the absolute values of the extracted vector components as the local weight intensities of the topological pattern vectors for the artifact contaminated leads;

[0021] Compare the local weight intensities of all the remaining topological pattern vectors;

[0022] Determine the topological pattern vector with the largest local weight intensity as the source vector of the characteristic contact artifact component.

[0023] Further, the time series signal reconstruction comprises:

[0024] Project the centralized multi-lead ECG signal matrix to the target topological pattern vector to obtain a one-dimensional projection coefficient time series;

[0025] Perform an outer product operation between the projection coefficient time series and the target topological pattern vector to reconstruct the complete spatiotemporal distribution signal of the corresponding topological pattern on all leads.

[0026] Further, perform lead domain common mode collaborative suppression on the common motion artifact component and adaptive residual filtering on the characteristic contact artifact component under fidelity constraint to obtain the compensated multi-lead ECG data, comprising:

[0027] Remove the common motion artifact component from the original ECG signal based on the output of the lead space motion artifact analysis model to obtain a primary compensation signal;

[0028] Construct a fidelity constraint adaptive filter, the composite cost function of the fidelity constraint adaptive filter is composed of an error power term and a morphology fidelity constraint term, and the characteristic contact artifact component is taken as the reference noise input, and the signal of the corresponding artifact contaminated lead in the primary compensation signal is taken as the main input;

[0029] Adjust the parameters of the adaptive filter by minimizing the composite cost function to generate the optimal estimate of the characteristic contact artifact component;

[0030] Subtract the optimal estimate of the characteristic contact artifact component from the artifact contaminated lead of the primary compensation signal to obtain the final compensated multi-lead ECG data.

[0031] Further, the adjustment by minimizing the composite cost function comprises:

[0032] computing an instantaneous error signal equal to the signal of the primary compensation signal for the artifact-contaminated lead minus an estimated output of the adaptive filter for the characteristic contact artifact component;

[0033] constructing a composite cost function consisting of an error power term representing the power of the instantaneous error signal, added with a morphology fidelity constraint term multiplied by a trade-off parameter;

[0034] iteratively updating the weight vector of the adaptive filter using a gradient descent based adaptive algorithm to minimize the composite cost function.

[0035] Further, based on the spatio-temporal correlation characteristics of the multi-lead ECG topological signals between different leads, identifying artifact-contaminated leads and corresponding artifact duration time windows, comprising:

[0036] dividing the multi-lead ECG topological signals into a plurality of continuous short-time analysis windows;

[0037] for each of the short-time analysis windows, computing the cross-correlation coefficients between all lead signals pairwise to form a cross-correlation coefficient matrix between leads;

[0038] based on the cross-correlation coefficient matrix, computing the average cross-correlation coefficient of each lead with other leads in the current window as the spatio-temporal correlation measure value of the lead;

[0039] if the spatio-temporal correlation measure value of a lead is lower than a correlation threshold value obtained according to pure ECG signals in a plurality of consecutive short-time analysis windows, determining that the lead is an artifact-contaminated lead, and determining the time period corresponding to the consecutive windows as the artifact duration time window.

[0040] Further, computing the average cross-correlation coefficient of each lead with other leads in the current window as the spatio-temporal correlation measure value of the lead, comprising:

[0041] for the cross-correlation coefficient matrix in the current short-time analysis window, computing the arithmetic mean of the cross-correlation coefficients of the lead with all other leads after removing the autocorrelation term thereof as the initial spatio-temporal consistency measure value of the lead;

[0042] performing exponential weighted moving average filtering on the initial spatio-temporal consistency measure value of the lead in the time domain to smooth transient fluctuations to obtain the final spatio-temporal correlation measure value.

[0043] Further, determining that the lead is an artifact-contaminated lead, comprising:

[0044] comparing the spatio-temporal correlation measure value with a dynamic threshold value;

[0045] The dynamic threshold is determined based on a statistical distribution of the spatiotemporal correlation measure value of the lead in a historical pure signal segment;

[0046] The lead is finally determined as a pseudo-contaminated lead only when the spatiotemporal correlation measure value is lower than the dynamic threshold in the continuous K windows, and the descending slope of the measure value in the corresponding time period exceeds a preset slope threshold.

[0047] The application also includes a motion compensation system of a lead electrocardio monitoring system, the system comprising:

[0048] A signal acquisition module for acquiring multi-lead electrocardio topology signals;

[0049] A motion pseudo identification module for identifying a pseudo-contaminated lead and a corresponding pseudo duration time window based on the spatiotemporal correlation features of the multi-lead electrocardio topology signals between different leads;

[0050] A motion pseudo analysis module for constructing a lead space motion pseudo analysis model for the signal segment of the pseudo-contaminated lead in the pseudo duration time window, so as to decouple the mixed motion pseudo into a common motion pseudo component and a characteristic contact pseudo component;

[0051] The common motion pseudo component represents a pseudo component with spatial common mode in the lead domain generated by the whole body dynamics disturbance; and the characteristic contact pseudo component represents a pseudo component concentrated in a single pseudo-contaminated lead caused by the instability of a specific electrode contact interface;

[0052] A signal motion compensation module for performing lead domain common mode collaborative suppression on the common motion pseudo component, and performing adaptive residual filtering under fidelity constraint on the characteristic contact pseudo component, so as to obtain compensated multi-lead electrocardio data;

[0053] A data output module for outputting the compensated multi-lead electrocardio data.

[0054] The application has the following beneficial effects:

[0055] By using spatiotemporal correlation analysis based on multi-lead ECG signals, motion artifact contamination leads and their duration windows can be accurately identified. An analytical model of lead-space motion artifacts is constructed, decomposing mixed artifacts into common motion artifact components and characteristic contact artifact components, achieving precise differentiation and independent processing of artifact sources. By performing lead-domain common-mode collaborative suppression on common artifacts and adaptive residual filtering under fidelity constraints on characteristic artifacts, interference signals caused by motion or electrode contact instability are effectively removed while preserving the morphological characteristics of the original ECG waveform to the greatest extent. This method not only improves the signal-to-noise ratio and waveform stability of ECG signals but also significantly reduces artifact misjudgment and signal distortion, enhancing the system's robustness and real-time processing capabilities in dynamic environments. Compared with existing filtering or blind source separation methods, this invention achieves higher compensation accuracy and signal fidelity under complex motion conditions, significantly improving the clinical usability and reliability of ECG monitoring results.

[0056] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 A flowchart of the motion compensation method for a lead-based electrocardiogram monitoring system;

[0059] Figure 2 A flowchart for decoupling mixed motion artifacts into common motion artifact components and characteristic contact artifact components;

[0060] Figure 3 A flowchart for determining the topology mode vector with the largest absolute weight on the artifact-contaminated leads;

[0061] Figure 4 A flowchart for obtaining compensated multi-lead ECG data;

[0062] Figure 5 This is a flowchart illustrating the adjustment process by minimizing the composite cost function;

[0063] Figure 6 This is a schematic diagram of the motion compensation system in a lead-based electrocardiogram monitoring system. Detailed Implementation

[0064] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The terminology used herein includes any and all combinations of one or more of the associated listed items.

[0066] Example 1:

[0067] like Figures 1 to 5 As shown, this application provides a motion compensation method for a lead-based electrocardiogram monitoring system, the method comprising:

[0068] S10: Acquire multi-lead ECG topology signals;

[0069] S20: Based on the spatiotemporal correlation characteristics of multi-lead ECG topological signals among different leads, identify artifact-contaminated leads and corresponding artifact duration windows;

[0070] S30: Within the artifact duration window, construct a spatial motion artifact analysis model for the signal segment of the artifact-contaminated lead to decouple the mixed motion artifacts into common motion artifact components and characteristic contact artifact components.

[0071] Among them, the common motion artifact component represents the artifact component that has spatial common mode in the lead domain, generated by the overall dynamics disturbance of the body; the characteristic contact artifact component represents the artifact component that is concentrated in a single artifact contaminating the lead, caused by the instability of a specific electrode contact interface.

[0072] S40: Perform lead-domain common-mode collaborative suppression on common motion artifact components and adaptive residual filtering under fidelity constraints on characteristic contact artifact components to obtain compensated multi-lead ECG data;

[0073] S50: Outputs compensated multi-lead ECG data.

[0074] Specifically, firstly, high-sampling-rate ECG monitoring instruments can be used to acquire multi-lead ECG topological signals. Then, spatiotemporal correlation analysis methods, such as cross-correlation analysis, can be used to identify leads contaminated with artifacts and their duration windows. Next, within the artifact contamination window, a lead spatial motion artifact analytical model is constructed, decoupling motion artifacts into common motion artifact components and characteristic contact artifact components. Common motion artifact components are low-frequency or high-amplitude artifacts appearing simultaneously in multiple leads due to overall movement, such as walking or swaying. Characteristic contact artifact components are artifacts concentrated in only one or a few leads due to unstable electrode contact, such as loose electrodes or sweating. For different artifact components, two methods are used for suppression and filtering: lead-domain common-mode co-suppression and adaptive residual filtering. Lead-domain common-mode co-suppression treats the ECG signals from multiple leads as a whole (a spatial vector), and by mining and utilizing these... Spatial correlation between signals is used to jointly and uniformly suppress common noise components across all leads, effectively removing common motion artifacts caused by body movement that affect all leads while preserving the true ECG signal. Adaptive residual filtering under fidelity constraints, while eliminating characteristic contact artifacts concentrated in a single lead, maximizes the preservation of morphological details of the pure ECG waveform in that lead. Therefore, the above compensation process is achieved through a precise, inverse subtraction-based process: compensated multi-lead ECG data = original multi-lead ECG data - estimated common motion artifact components - estimated characteristic contact artifact components. Finally, the compensated multi-lead ECG data output ensures accuracy and reliability, making it suitable for clinical monitoring and diagnosis. This method improves the quality of ECG monitoring data and avoids signal distortion caused by motion artifact interference in traditional methods.

[0075] It should be noted that the aforementioned spatiotemporal correlation characteristics refer to the statistical correlation characteristics exhibited by multi-lead ECG signals in the spatial and temporal dimensions. Specifically, in the absence of artifact interference, the signals of each lead have a stable cross-correlation matrix pattern, and the signals of each lead are smooth and continuous in the time domain. The aforementioned artifact-contaminated leads are single or multiple specific leads whose ECG signals are interfered with by motion artifacts within a certain time period, while the aforementioned artifact duration window is the specific time period from when the motion artifact begins to affect the signal until its end.

[0076] As a preferred embodiment of the above, in step S30, such as Figure 2 As shown, the analytical model of motion artifacts in lead space is constructed through orthogonal decomposition of ECG topological features, decoupling the mixed motion artifacts into common motion artifact components and characteristic contact artifact components, including:

[0077] S31: Center the multi-lead ECG signal matrix within the artifact duration window and calculate the covariance matrix;

[0078] S32: Perform ECG topological feature orthogonal decomposition on the covariance matrix to generate a set of topological pattern vectors arranged in descending order of energy intensity;

[0079] S33: Reconstruct the time-series signal corresponding to the first topological mode vector with the largest energy intensity into common motion artifact components;

[0080] S34: From the remaining topology pattern vectors, determine the topology pattern vector with the largest absolute weight on the artifact contamination lead, and reconstruct its corresponding time-series signal into the characteristic contact artifact component.

[0081] Specifically, firstly, within the duration of artifacts, ECG signals are simultaneously acquired from multiple leads and assembled into a signal set according to lead order. Since each lead may have differences in amplitude, offset, and measurement starting point, signal centering is necessary to ensure data comparability. Centering removes the average offset of each lead signal, making the signal amplitude symmetrically distributed around zero. This is equivalent to comparing all lead signals on the same baseline, avoiding the overall offset effect caused by instrument drift or electrode contact deviation. After centering, the covariance matrix is ​​calculated to describe the relationship between different lead signals, recording whether the signals change synchronously over time and reflecting the degree of correlation between them. For example, if two lead signals rise or fall simultaneously at the same time, their covariance is large; conversely, if the trends are opposite, the covariance is small or even negative. In this way, a comprehensive description of the overall spatiotemporal correlation between all lead signals can be obtained, thus providing a basis for subsequent spatial characterization. Feature analysis provides the foundation; next, the covariance matrix is ​​subjected to ECG topological feature orthogonal decomposition, which breaks down the overall waveform energy formed by the mixing of signals from all leads into several modes representing different spatial distribution characteristics. Each mode describes a specific type of signal change law and is arranged in descending order according to its energy intensity. The energy intensity indicates the contribution of the corresponding mode to the original signal. The first topological mode with the strongest energy reflects the common change feature that appears almost simultaneously in all leads, namely the so-called common motion artifact. After removing the common artifact, the mode with the most significant influence on certain specific leads is further identified from the remaining topological modes. These modes have relatively small energy but have high weights in individual leads, showing local fluctuations. This mode reflects the characteristic contact artifact. By detecting the topological mode with the largest weight in the artifact-contaminated lead and extracting its corresponding signal, the characteristic contact artifact component is obtained. If contact artifacts appear in multiple leads at the same time, the signals of multiple modes are weighted and combined according to the magnitude of the artifacts in each lead in order to more completely recover the artifact structure.

[0082] In this embodiment, in step S34, as Figure 3 As shown, the topological mode vector with the largest absolute weight on the artifact-contaminated leads is determined, including:

[0083] S341: For each topological pattern vector in the remaining set of topological pattern vectors, extract its vector component at the location of the artifact contamination lead.

[0084] S342: Calculate the absolute value of the extracted vector components as the local weight strength of the topology mode vector for artifact-contaminated leads.

[0085] S343: Compare the local weight strengths of all remaining topological pattern vectors;

[0086] S344: The topological pattern vector with the largest local weight intensity is determined as the source vector of the characteristic contact artifact component.

[0087] Specifically, the process begins by extracting the components of each topological pattern vector at the artifact-contaminated lead from the remaining set of topological pattern vectors. These vectors are obtained through orthogonal decomposition of multi-lead ECG signals, with each vector representing a signal characteristic pattern. To identify characteristic contact artifacts, the signal characteristics of each vector need to be matched with the specific artifact-contaminated lead to extract its components. The key to this operation is identifying the specific impact of each topological pattern on the artifact-contaminated lead through the spatiotemporal characteristics of the signal. These components reflect the performance of the topological pattern in the artifact-contaminated lead. After extracting the components of each vector, the next step is to calculate their absolute values. These absolute values ​​serve as the local weight strength of the topological pattern vector in the artifact-contaminated lead, representing the degree of influence of the pattern in that lead. By calculating the absolute values, we can quantify the contribution of each pattern vector to the artifact-contaminated lead and determine which patterns have the greatest impact on the artifacts in that lead. The calculation of the local weight strength not only reflects the influence of a specific pattern vector but also... The contribution of topological pattern vectors to artifact-contaminated leads can effectively distinguish the differences in the influence of different topological patterns on these leads. A larger local weight strength indicates that the topological pattern has a more significant impact on artifact-contaminated leads. By comparing the magnitude of the influence of these topological pattern vectors on artifact-contaminated leads, we can find which topological pattern vector has the largest local weight strength. This comparison step is crucial for identifying the most representative source of artifact contamination. By comparing the local weight strength of different topological pattern vectors, we can accurately identify the main signal characteristics affecting artifact-contaminated leads. This step summarizes the influence of multiple topological pattern vectors into the most prominent pattern, thus providing an accurate basis for subsequent artifact processing. Finally, by comparing the local weight strength, we select the topological pattern vector with the largest local weight strength and determine it as the source vector of the characteristic contact artifact component. This topological pattern vector represents the artifact caused by poor electrode contact, and its timing signal can clearly reflect the characteristics of the artifact. Once the source vector of the characteristic contact artifact component is determined, the timing signal of this pattern can be used to reconstruct the characteristic contact artifact component, thereby achieving accurate compensation of the ECG signal.

[0088] In step S33, the timing signal reconstruction includes:

[0089] S331: Project the centered multi-lead ECG signal matrix onto the target topological pattern vector to obtain a one-dimensional projection coefficient time series;

[0090] S332: Perform an outer product operation between the projection coefficient time series and the target topological pattern vector to reconstruct the complete spatiotemporal distribution signal of the corresponding topological pattern on all leads.

[0091] Specifically, firstly, the spatiotemporal features of the centered multi-lead ECG signal matrix are extracted and projected onto the direction of a pre-obtained target topological pattern vector. This process is equivalent to extracting the dominant change trajectory of the target pattern in the time dimension within a high-dimensional lead space, calculating the amplitude of the original signal's change in this pattern direction time-by-time, forming a one-dimensional time series to represent the activity level or influence intensity of the topological pattern throughout the sampling time. Through this projection operation, the originally complex multi-lead signal is compressed into a single time series, thus clearly revealing the temporal correspondence between the pattern and motion artifacts. Then, the projected coefficient time series and the target topological pattern vector are recombine to extend the one-dimensional time features back into the multi-lead space. Specifically, the system will... The coefficient corresponding to each moment is used as an amplitude weight and applied to each lead component of the target topological pattern vector, so that each lead obtains the response value of the pattern at the current moment. This process is repeated on the entire time axis to reconstruct the complete spatiotemporal distribution signal of the target topological pattern on all leads point by point. In this way, the distribution characteristics of the target pattern in the spatial dimension and its dynamic changes in the temporal dimension can be recovered simultaneously, realizing the accurate restoration of motion artifacts or contact artifacts. After reconstruction, the system can use the generated signal as the estimation result of specific artifact components and use it for artifact compensation processing in subsequent steps. The method, through the design of projection first and then reconstruction, ensures the consistency of lead topology and temporal characteristics, and can accurately separate the artifact components caused by motion or contact without destroying the morphology of the electrocardiogram signal.

[0092] As a preferred embodiment of the above, in step S40, as Figure 4 As shown, lead-domain common-mode cooperative suppression is performed on common motion artifact components, and adaptive residual filtering under fidelity constraints is applied to characteristic contact artifact components to obtain compensated multi-lead ECG data. The steps include:

[0093] S41: Based on the output of the lead space motion artifact analysis model, common motion artifact components are removed from the original ECG signal to obtain the primary compensation signal;

[0094] S42: Construct a fidelity-constrained adaptive filter. The composite cost function of the fidelity-constrained adaptive filter is composed of an error power term and a morphological fidelity constraint term. The characteristic contact artifact component is used as the reference noise input, and the signal of the corresponding artifact contamination lead in the primary compensation signal is used as the main input.

[0095] S43: By minimizing the composite cost function and adjusting the parameters of the adaptive filter, the optimal estimate of the characteristic contact artifact component is generated;

[0096] S44: Subtract the optimal estimate of the characteristic contact artifact component from the artifact-contaminated leads of the primary compensation signal to obtain the final compensated multi-lead ECG data.

[0097] Specifically, firstly, based on the output of the aforementioned analytical model of lead spatial motion artifacts, most of the low-frequency drift and baseline shift caused by overall motion in the original ECG signal are effectively eliminated, resulting in a relatively stable primary compensation signal, laying the foundation for subsequent local artifact filtering. After completing the suppression of common artifacts, a fidelity-constrained adaptive filter is constructed for the remaining characteristic contact artifacts. This fidelity-constrained adaptive filter is used to achieve artifact suppression while preserving the morphological and amplitude characteristics of the real ECG waveform to the maximum extent. Specifically, the composite cost function of the fidelity-constrained adaptive filter consists of two parts: an error power term, used to minimize the mean square error between the output signal and the reference input; and a morphological fidelity constraint term, used to limit the morphological distortion of the dominant components of the ECG waveform during the filtering process. The dominant components of the ECG waveform can be P waves, QRS complexes, and T waves, etc. The characteristic contact artifact component obtained through the aforementioned analytical model is used as the reference noise input signal, and the signal of the corresponding artifact-contaminated lead in the primary compensation signal is used as the main input signal. During filtering, the weight parameters of the filter are dynamically adjusted to make the filtering result more stable. The method effectively suppresses artifacts while maintaining the continuity of the original ECG characteristic morphology. During the filtering process, the filtering parameters are adjusted in real time based on the principle of minimizing the composite cost function. This process can be achieved through an iterative optimization algorithm. In each iteration cycle, the filtering weights are updated according to the error output, so that the output signal gradually approaches the ideal artifact-free state. When the cost function converges to a stable value, the optimal estimation signal for the characteristic contact artifact component is generated. This optimal estimation signal contains the characteristic noise components caused by local lead contact instability, lead compression changes, or electrode loosening, and can be regarded as a high-confidence prediction of artifact interference. Finally, the above-mentioned optimal artifact estimation signal is subtracted from the corresponding artifact-contaminated leads in the primary compensation signal to obtain the final compensated multi-lead ECG data. The compensation result is more stable in signal morphology, significantly reduces artifact interference, and completely preserves the ECG waveform structure. Cross-validation between multiple leads confirms that the phase relationship, waveform correspondence, and energy distribution of the compensated signal between each lead are restored, thus providing high-quality input data for clinical observation and subsequent automatic diagnostic algorithms.

[0098] In this embodiment, in step S43, as Figure 5 As shown, adjustments are made by minimizing the composite cost function, including:

[0099] S431: Calculate the instantaneous error signal, which is equal to the signal in the primary compensation signal corresponding to the artifact contamination lead minus the estimated output of the adaptive filter for the characteristic contact artifact component;

[0100] S432: Construct a composite cost function, which is composed of an error power term representing the instantaneous error signal power and a morphological fidelity constraint term, wherein the morphological fidelity constraint term is multiplied by a tradeoff parameter.

[0101] S433: An adaptive algorithm based on gradient descent is used to iteratively update the weight vector of the adaptive filter in order to minimize the composite cost function.

[0102] Specifically, in each update cycle of the adaptive filter, the ECG signal of the corresponding artifact-contaminated lead in the primary compensation signal is used as the main input, and the output generated by the adaptive filter based on its characteristics to remove artifact components is used as the artifact estimation signal. The difference between the two is calculated to obtain the instantaneous error signal. The instantaneous error signal reflects the accuracy of the current filter's artifact estimation, and its magnitude characterizes the deviation between the filter output and the ideal net signal. When the error is large, it indicates that the current filter parameters have not yet converged; when the error approaches zero, it indicates that the artifact has been effectively suppressed and the signal recovery is high. Then, a composite cost function is constructed to simultaneously constrain the filtering error and waveform fidelity. The composite cost function consists of two parts: an error power term and a weighting parameter is introduced between the two for dynamic adjustment. The weighting parameter is used to balance the artifact suppression strength and waveform fidelity. When the motion artifact is strong, the weight of the error power term can be increased to preferentially remove the artifact. When the artifact is weak but the ECG morphology requirement is high, the weight of the fidelity constraint term can be increased to maintain waveform stability. The trade-off parameter can be dynamically controlled by the signal quality index. Finally, an adaptive algorithm based on gradient descent is used to iteratively update the filter weights. In each iteration, the gradient direction of the cost function with respect to the current filter weights is calculated, and the weights are adjusted in the opposite direction of the gradient to gradually reduce the cost function. To improve stability under complex motion conditions, the step size is automatically adjusted according to the severity of error changes. When the error fluctuation is large, the step size is reduced to avoid over-adjustment and signal oscillation. When the signal tends to stabilize, the step size is increased to accelerate the convergence speed, i.e., real-time correction of the filter output. By continuously correcting the parameters, the artifact estimation signal gradually approaches the optimal solution. After multiple iterations, when the cost function tends to stabilize and the error signal remains below the preset threshold, the filter is considered to have converged. At this time, the filter output is the optimal estimation signal for the characteristic contact artifact component.

[0103] As a preferred embodiment of the above, in step S20, based on the spatiotemporal correlation characteristics of multi-lead ECG topological signals among different leads, artifact-contaminated leads and corresponding artifact duration windows are identified, including:

[0104] S21: Divide the multi-lead ECG topology signal into multiple consecutive short-time analysis windows;

[0105] S22: For each short-time analysis window, calculate the cross-correlation coefficients between all lead signals to form a cross-correlation coefficient matrix between leads;

[0106] S23: Based on the cross-correlation matrix, calculate the average cross-correlation coefficient of each lead with other leads in the current window, and use it as a measure of the spatiotemporal correlation of the leads.

[0107] S24: If the spatiotemporal correlation metric of a certain lead is lower than the correlation threshold obtained from the pure electrocardiogram signal statistics in multiple consecutive short-term analysis windows, the lead is determined to be an artifact-contaminated lead, and the time period corresponding to the multiple consecutive windows is determined as the artifact persistence window.

[0108] Specifically, the continuously acquired multi-lead ECG signals are first divided into multiple short-time analysis windows that are adjacent in time and of equal length. Each window contains multi-lead signals within a certain time range for local analysis of the correlation characteristics between leads. Adjacent windows can overlap appropriately to ensure that edge information is not missed when identifying the start and end of artifacts. Next, within each short-time analysis window, all lead signals are compared pairwise to calculate their similarity in waveform morphology, phase, and trend. The similarity between each pair of lead signals is expressed as a correlation coefficient, with a value ranging between perfect similarity and complete dissimilarity. The signals are arranged in rows and columns to form a correlation matrix between leads, which is used to characterize the spatiotemporal consistency among the leads as a whole. This matrix reflects the synergistic characteristics of different leads within the same time period. When the monitored subject is at rest or has slight movement, the correlation coefficient between most leads is high. When the body makes significant movements or individual electrode contacts are unstable, the correlation of the corresponding leads will decrease significantly, thus providing clues for artifact detection. Then, the average correlation degree of each lead with other lead signals within the current time window is calculated, i.e., the spatiotemporal correlation metric, which reflects whether the lead keeps pace with other leads within the current window. Specifically, when a certain lead... When a lead is undisturbed, its waveform characteristics typically maintain high consistency with other leads, resulting in a high spatiotemporal correlation metric. However, if the lead is affected by artifacts, such as due to electrode loosening or sudden changes in contact impedance caused by localized body movements, its waveform characteristics will significantly deviate from the overall trend, leading to a decrease in the metric. Finally, the spatiotemporal correlation metric curves of each lead are continuously monitored. When the metric value of a lead remains below a set threshold for multiple adjacent time windows, the lead is determined to be affected by artifacts during that time period. The time range covered by these consecutive windows is defined as the artifact duration window for that lead. The threshold is set based on... Based on the statistical characteristics analysis of pure ECG signals, a multi-lead ECG signal in an artifact-free state can be acquired during the calibration phase. The average correlation level between each lead is calculated, and a lower limit is set as the judgment standard according to the actual situation. When the correlation of a lead is detected to be lower than this lower limit during the monitoring process, it is considered to be affected by artifacts. In order to avoid misjudging instantaneous fluctuations or interference as artifacts, a time continuity constraint is introduced during the judgment. Only when the duration of the low correlation state exceeds the set minimum duration threshold is the existence of artifacts confirmed. This enables accurate judgment of artifact-contaminated leads and their duration windows in multi-lead ECG signals under dynamic motion environment.

[0109] In this embodiment, in step S23, the average cross-correlation coefficient of each lead with other leads within the current window is calculated as a measure of the spatiotemporal correlation of the leads. This step includes:

[0110] S231: For the cross-correlation matrix within the current short-time analysis window, calculate the arithmetic mean of the cross-correlation coefficients of the lead after removing its autocorrelation terms with all other leads, and use it as the initial spatiotemporal consistency measure of the lead.

[0111] S232: Apply an exponentially weighted moving average filter in the time domain to the initial spatiotemporal consistency metric of the leads to smooth out instantaneous fluctuations and obtain the final spatiotemporal correlation metric.

[0112] Specifically, firstly, for each lead in the matrix, the average correlation between it and all other lead signals is calculated. To avoid the interference of autocorrelation on the overall calculation results, the correlation term between the lead and itself is removed when calculating the average value. Only the correlation data between the lead and the other leads is considered. The average value obtained in this way can more accurately reflect the overall synchronous change characteristics of the lead and other leads within the time window, reducing the false high correlation phenomenon caused by the energy fluctuation of the lead itself. Then, the aforementioned initial spatiotemporal consistency metric is smoothed in the time domain using an exponentially weighted moving average filter. That is, when calculating the final spatiotemporal correlation metric of the current window, not only the initial consistency result of the current window is considered, but also the historical results of the previous time window are weighted and superimposed. Newer data has a higher weight, and older data has a lower weight, so that the output result has smoothness while maintaining real-time performance. This process can effectively weaken the impact of sudden fluctuations within a single time window and ensure that the lead correlation assessment has continuity and trend stability.

[0113] In step S24, determining whether a lead is contaminated by an artifact involves the following steps:

[0114] S241: Compare the spatiotemporal correlation metric with the dynamic threshold;

[0115] S242: The dynamic threshold is determined based on the statistical distribution of the spatiotemporal correlation metric of leads within historical clean signal segments;

[0116] S243: A lead is finally determined to be an artifact-contaminated lead only when the spatiotemporal correlation metric is lower than the dynamic threshold for K consecutive windows and the slope of the metric decrease exceeds the preset slope threshold in the corresponding time period.

[0117] Specifically, a dynamic threshold related to the lead state is first introduced. The spatiotemporal correlation metric of each lead is compared in real time to determine whether it is in an abnormal state. Traditional methods typically use a fixed threshold for judgment. However, in actual monitoring, the statistical characteristics of ECG signals fluctuate with individual differences, body movement intensity, electrode position changes, and skin contact conditions. Fixed thresholds are difficult to adapt to diverse scenarios and are prone to misjudgment or missed judgment. Therefore, this implementation adopts a dynamic threshold mechanism, allowing the threshold to adaptively adjust according to lead characteristics and the current signal state. Then, the dynamic threshold is determined by the system based on the historical clean signal data of the lead. The statistical characteristics within a segment are automatically determined. At the beginning of monitoring or when the monitored subject is at rest, a relatively artifact-free ECG signal segment is automatically identified as a baseline sample segment. Statistical analysis is performed on the spatiotemporal correlation metric of each lead within this baseline segment to obtain the average level and fluctuation range of that lead, and based on this, the upper and lower limits of the threshold are determined. Then, based on the low correlation state, time continuity constraints and trend change constraints are introduced to further improve the accuracy of the judgment. Specifically, the time continuity constraint means that when the spatiotemporal correlation metric of a lead is lower than the dynamic threshold for several consecutive short time windows, the lead is considered to be in an abnormal state. The state exhibits a persistent characteristic. A short-term drop in a single window may be caused by transient electromyographic interference or micro-movements of the body surface and does not represent a true artifact. A continuous low correlation state indicates that the lead signal deviates from the population pattern for a long time, and is more likely to be affected by electrode loosening or strenuous body movement. The trend change constraint means that in addition to the absolute value being below the threshold, the decreasing trend of the spatiotemporal correlation metric of the lead is also calculated. If the metric shows a significant decreasing trend within a continuous window, and the rate of decrease exceeds a preset slope threshold, it indicates that the artifact is exacerbating the damage to signal quality. Only when both of the above conditions are met simultaneously is it finally confirmed that the lead is contaminated by artifacts. The introduction of a fixed strategy can effectively distinguish between short-term fluctuations and genuine artifact interference. For example, when a subject moves their arm slightly, the correlation of a certain lead may drop momentarily, but this will not last for multiple windows or show a significant downward trend. However, during running or strenuous activity, the correlation of the lead will continue to decrease and show a significant downward trend. Based on this, artifact-contaminated leads can be accurately identified and labeled. The proposed dynamic threshold determination and trend dual constraint mechanism breaks through the limitations of fixed threshold and single time window judgment in traditional artifact identification, realizing intelligent and high-precision identification of artifact-contaminated leads, and providing a reliable foundation for subsequent motion artifact compensation.

[0118] Example 2:

[0119] The present invention also includes a motion compensation system for a lead-based electrocardiogram monitoring system, such as Figure 6 As shown, the system includes:

[0120] The signal acquisition module is used to acquire multi-lead ECG topology signals;

[0121] The motion artifact recognition module is used to identify artifact-contaminated leads and corresponding artifact duration windows based on the spatiotemporal correlation characteristics of multi-lead ECG topological signals between different leads.

[0122] The motion artifact analysis module is used to construct a lead space motion artifact analysis model for the signal segment of the lead contaminated by artifacts within the artifact duration window, so as to decouple the mixed motion artifacts into common motion artifact components and characteristic contact artifact components.

[0123] Among them, the common motion artifact component represents the artifact component that has spatial common mode in the lead domain, generated by the overall dynamics disturbance of the body; the characteristic contact artifact component represents the artifact component that is concentrated in a single artifact contaminating the lead, caused by the instability of a specific electrode contact interface.

[0124] The signal motion compensation module is used to perform lead-domain common-mode collaborative suppression on common motion artifact components and adaptive residual filtering under fidelity constraints on characteristic contact artifact components to obtain compensated multi-lead ECG data.

[0125] The data output module is used to output compensated multi-lead ECG data.

[0126] The adjustment system described above in this invention can effectively realize the motion compensation method of the lead electrocardiogram monitoring system, and the technical effects it can achieve are as described in the above embodiments, which will not be repeated here.

[0127] Similarly, the above-mentioned optimization schemes for the system can also achieve the optimization effects corresponding to the methods in Embodiment 1, which will not be repeated here.

[0128] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and accompanying drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A motion compensation method for a lead-based electrocardiogram monitoring system, characterized in that, The method includes: Acquire multi-lead ECG topology signals; Based on the spatiotemporal correlation characteristics of the multi-lead ECG topology signal among different leads, artifact-contaminated leads and corresponding artifact duration windows are identified. Within the duration window of the artifact, a spatial motion artifact analysis model is constructed for the signal segment of the lead contaminated by the artifact, so as to decouple the mixed motion artifacts into common motion artifact components and characteristic contact artifact components. The common motion artifact component represents the artifact component that has spatial common mode within the lead domain, generated by the overall dynamics disturbance of the body; the characteristic contact artifact component represents the artifact component that is concentrated in a single artifact contaminating the lead, caused by the instability of a specific electrode contact interface. Lead-domain common-mode cooperative suppression is performed on the common motion artifact components, and adaptive residual filtering under fidelity constraints is performed on the characteristic contact artifact components to obtain compensated multi-lead ECG data. Output the compensated multi-lead ECG data.

2. The motion compensation method for the lead-based electrocardiogram monitoring system according to claim 1, characterized in that, The analytical model for motion artifacts in the lead space is constructed through orthogonal decomposition of ECG topological features, decoupling the mixed motion artifacts into common motion artifact components and characteristic contact artifact components, including: The multi-lead ECG signal matrix within the duration window of the artifact is centered, and the covariance matrix is ​​calculated. The covariance matrix is ​​subjected to ECG topological feature orthogonal decomposition to generate a set of topological pattern vectors arranged in descending order of energy intensity; The time-series signal corresponding to the first topological mode vector with the highest energy intensity is reconstructed into the common motion artifact component; From the remaining topology pattern vectors, determine the topology pattern vector with the largest absolute weight on the artifact contamination lead, and reconstruct its corresponding time-series signal into the characteristic contact artifact component.

3. The motion compensation method for the lead-based electrocardiogram monitoring system according to claim 2, characterized in that, Determine the topological mode vector with the largest absolute weight on the artifact-contaminated leads, including: For each topological pattern vector in the remaining set of topological pattern vectors, extract its corresponding vector component at the artifact contamination lead location; Calculate the absolute value of the extracted vector components as the local weight strength of the topological mode vector for the artifact contamination lead; Compare the local weight strengths of all the remaining topological pattern vectors; The topology pattern vector with the largest local weight intensity is determined as the source vector of the characteristic contact artifact component.

4. The motion compensation method for the lead-based electrocardiogram monitoring system according to claim 2, characterized in that, The time-series signal reconstruction includes: The centered multi-lead ECG signal matrix is ​​projected onto the target topological pattern vector to obtain a one-dimensional projection coefficient time series. The projection coefficient time series is multiplied by the target topology pattern vector to reconstruct the complete spatiotemporal distribution signal of the corresponding topology pattern on all leads.

5. The motion compensation method for the lead-based electrocardiogram monitoring system according to claim 1, characterized in that, Perform lead-domain common-mode cooperative suppression on the common motion artifact components and adaptive residual filtering under fidelity constraints on the characteristic contact artifact components to obtain compensated multi-lead ECG data. The steps include: Based on the output of the lead spatial motion artifact analysis model, the common motion artifact component is removed from the original electrocardiogram signal to obtain the primary compensation signal; A fidelity-constrained adaptive filter is constructed. The composite cost function of the fidelity-constrained adaptive filter is composed of an error power term and a morphological fidelity constraint term. The characteristic contact artifact component is used as the reference noise input, and the signal corresponding to the artifact contamination lead in the primary compensation signal is used as the main input. By minimizing the composite cost function and adjusting the parameters of the adaptive filter, an optimal estimate of the characteristic contact artifact component is generated. The optimal estimate of the characteristic contact artifact component is subtracted from the artifact-contaminated leads of the primary compensation signal to obtain the final compensated multi-lead ECG data.

6. The motion compensation method for the lead-based electrocardiogram monitoring system according to claim 5, characterized in that, Adjustments are made by minimizing the composite cost function, including: Calculate the instantaneous error signal, which is equal to the signal in the primary compensation signal corresponding to the artifact contamination lead minus the estimated output of the adaptive filter for the characteristic contact artifact component; Construct a composite cost function, which is formed by adding an error power term representing the instantaneous error signal power to a morphological fidelity constraint term, wherein the morphological fidelity constraint term is multiplied by a tradeoff parameter. An adaptive algorithm based on gradient descent is used to iteratively update the weight vector of the adaptive filter in order to minimize the composite cost function.

7. The motion compensation method for the lead-based electrocardiogram monitoring system according to claim 1, characterized in that, Based on the spatiotemporal correlation characteristics of the multi-lead ECG topological signals among different leads, artifact-contaminated leads and corresponding artifact duration windows are identified, including: The multi-lead ECG topology signal is divided into multiple consecutive short-time analysis windows; For each short-time analysis window, calculate the cross-correlation coefficients between all lead signals to form a cross-correlation coefficient matrix between leads; Based on the cross-correlation matrix, the average cross-correlation coefficient of each lead with other leads within the current window is calculated, which serves as a measure of the spatiotemporal correlation of the lead. If the spatiotemporal correlation metric of a certain lead is lower than the correlation threshold obtained from the pure electrocardiogram signal statistics in multiple consecutive short-term analysis windows, the lead is determined to be an artifact-contaminated lead, and the time period corresponding to the multiple consecutive windows is determined as the artifact duration window.

8. The motion compensation method for the lead-based electrocardiogram monitoring system according to claim 7, characterized in that, Calculate the average cross-correlation coefficient of each lead with other leads within the current window, using this as a measure of the spatiotemporal correlation of the leads. The steps include: For the cross-correlation matrix within the current short-term analysis window, calculate the arithmetic mean of the cross-correlation coefficients of the lead and all other leads after removing its autocorrelation terms, and use it as the initial spatiotemporal consistency measure of the lead. The initial spatiotemporal consistency metric of the leads is filtered by an exponentially weighted moving average in the time domain to smooth out instantaneous fluctuations and obtain the final spatiotemporal correlation metric.

9. The motion compensation method for the lead-based electrocardiogram monitoring system according to claim 7, characterized in that, The steps to determine if a lead is an artifact-contaminated lead include: The spatiotemporal correlation metric is compared with a dynamic threshold. The dynamic threshold is determined based on the statistical distribution of the spatiotemporal correlation metric of the leads within historical clean signal segments; The lead is ultimately determined to be an artifact-contaminated lead only if the spatiotemporal correlation metric is lower than the dynamic threshold for K consecutive windows, and the rate of decline of the metric exceeds a preset slope threshold in the corresponding time period.

10. A motion compensation system for a lead-based electrocardiogram monitoring system, characterized in that, The system includes: The signal acquisition module is used to acquire multi-lead ECG topology signals; The motion artifact recognition module is used to identify artifact-contaminated leads and corresponding artifact duration windows based on the spatiotemporal correlation characteristics of the multi-lead ECG topological signals between different leads. The motion artifact analysis module is used to construct a lead space motion artifact analysis model for the signal segment of the lead contaminated by the artifact within the duration window of the artifact, so as to decouple the mixed motion artifacts into common motion artifact components and characteristic contact artifact components. The common motion artifact component represents the artifact component that has spatial common mode within the lead domain, generated by the overall dynamics disturbance of the body; the characteristic contact artifact component represents the artifact component that is concentrated in a single artifact contaminating the lead, caused by the instability of a specific electrode contact interface. The signal motion compensation module is used to perform lead-domain common-mode collaborative suppression on the common motion artifact components and to perform adaptive residual filtering under fidelity constraints on the characteristic contact artifact components to obtain compensated multi-lead ECG data. The data output module is used to output the compensated multi-lead ECG data.

Images