A heart failure volume load warning system, method, medium, and device

Abstract

The application provides a heart failure volume load early warning system, method, medium and equipment, and relates to the technical field of heart failure early warning. The system synchronously acquires thoracic impedance, electrocardio, photoplethysmography, body position and electrode contact state signals through a multi-modal acquisition module; generates signal quality evaluation parameters by using a signal quality evaluation module, executes first gating processing and dynamically separates interference to extract a direct current trend component; a signal screening and trend extraction module extracts thoracic impedance trend characteristics; a physiological consistency verification module calculates cross-modal timing correlation and executes second gating verification processing, adjusts risk weight based on two gating results, and outputs corresponding early warning information by a risk output module. The application effectively improves the reliability and accuracy of volume load early warning through signal quality screening and physiological mechanism double gating cooperation.

Description

Technical Field

[0001] This invention relates to the field of heart failure early warning technology, and specifically to a heart failure volume overload early warning system, method, medium, and device. Background Technology

[0002] Patients with heart failure are at high risk of disease fluctuations during home management after discharge, making volume overload monitoring crucial for early warning. Current technologies focus on thoracic impedance measurement, utilizing the principle that increased lung water content leads to decreased impedance. Signals are collected via patch electrodes or smart vests, and after preprocessing such as power frequency filtering, baseline correction, and artifact suppression, impedance trends are extracted. Some solutions combine parameters such as heart rate and blood pressure to generate a comprehensive risk score.

[0003] However, in real-world home environments, non-physiological disturbances such as electrode displacement, skin contact impedance, body position changes, and motion artifacts are significant, and the resulting impedance fluctuations are highly similar to actual volume changes in both time and amplitude. Existing algorithms lack parameter correlation verification based on physiological mechanisms, making it difficult to effectively distinguish between pathological impedance decreases and non-specific drift. This leads to an unstable mapping relationship between chest impedance and volume overload, making it difficult to control the false positive and false negative rates of the early warning system within clinically acceptable ranges, thus limiting the reliable application of this technology.

[0004] Therefore, how to effectively distinguish between physiological capacity load trends and non-physiological disturbances, and how to make full use of the temporal correlation of multimodal physiological signals for intelligent early warning, are technical problems that urgently need to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a heart failure volume overload early warning system, method, computer-readable storage medium, and electronic device to solve the problem of unstable correspondence between changes in chest impedance and actual volume overload in existing home monitoring technologies for heart failure volume overload.

[0006] To address the aforementioned technical problems, this application provides a heart failure volume overload early warning system, comprising: A multimodal acquisition module is used to acquire multimodal physiological signals of a target object; the multimodal physiological signals include at least chest impedance signals and at least one of the following auxiliary physiological signals: electrocardiogram signals, photoplethysmography pulse wave signals, body position signals, and electrode contact status signals; The signal quality evaluation module is used to generate signal quality evaluation parameters that characterize the degree of interference of the chest impedance signal based on at least one of the auxiliary physiological signal and the chest impedance signal. The signal filtering and trend extraction module is used to perform validity filtering on the chest impedance signal according to the signal quality evaluation parameters, and to perform interference separation processing on the chest impedance signal that passes the validity filtering, so as to extract the impedance trend component related to the capacity load state. The physiological consistency verification module is used to extract auxiliary physiological features based on the auxiliary physiological signals, and to verify the physiological rationality of the impedance trend component based on the auxiliary physiological features, thereby generating a consistency evaluation index. The risk output module is used to determine the collaborative control parameters based on the signal quality evaluation parameters and the consistency evaluation index, and output early warning information corresponding to the capacity load status based on the collaborative control parameters.

[0007] Optionally, the auxiliary physiological signals include the body position signal and the electrode contact state signal; the signal quality evaluation module includes: a body movement and position feature extraction unit, used to extract body movement intensity features and body position stability features based on the body position signal; a contact anomaly detection unit, used to extract contact anomaly features based on the electrode contact state signal; an impedance waveform analysis unit, used to extract baseline abrupt change features and respiratory cycle regularity features based on the waveform of the chest impedance signal; and a quality score fusion unit, used to perform a weighted summation of the body movement intensity features, body position stability features, contact anomaly features, baseline abrupt change features, and respiratory cycle regularity features, and use the weighted summation result as the signal quality evaluation parameter.

[0008] Optionally, the signal filtering and trend extraction module includes: a quality gating unit, used to perform validity filtering on the chest impedance signal in the current time window based on the comparison relationship between the signal quality evaluation parameter and a preset quality threshold; when the signal quality evaluation parameter is greater than or equal to a first preset threshold, the chest impedance signal in the current time window is marked as a fully open window; when the signal quality evaluation parameter is less than the first preset threshold and greater than or equal to a second preset threshold, the chest impedance signal in the current time window is marked as a deweighted participation window; when the signal quality evaluation parameter is less than the second preset threshold, the chest impedance signal in the current time window is removed; wherein the first preset threshold is greater than the second preset threshold; and an adaptive filtering unit, used to dynamically adjust at least one of the filtering window length and the filtering update weight for the chest impedance signal that has passed the validity filtering based on the signal quality evaluation parameter, and perform adaptive filtering on the chest impedance signal with the dynamically adjusted parameter to separate and remove interference components from the chest impedance signal, and use the remaining components as the impedance trend components.

[0009] Optionally, the interference components include at least one of non-physiological offset components, motion artifact components, and respiratory communication components; the adaptive filtering unit is specifically used to dynamically adjust at least one of the filter window length and filter update weight according to the signal quality evaluation parameters, and to adaptively filter the chest impedance signal with the dynamically adjusted filter window length and filter update weight, sequentially separating and removing the non-physiological offset components, the motion artifact components, and the respiratory communication components, and using the remaining components after separating and removing the interference components as the impedance trend components.

[0010] Optionally, the auxiliary physiological signals include the electrocardiogram (ECG) signal and the photoplethysmography (PPG) signal; the physiological consistency verification module includes: an ECG feature extraction unit for extracting ECG features from the ECG signal; a pulse wave feature extraction unit for extracting pulse wave features from the PPG signal; a time-series graph network construction unit for constructing a time-series graph network using the impedance trend component, the ECG features, and the pulse wave features as heterogeneous nodes; a causal attention mask generation unit for generating a causal attention mask between each node of the time-series graph network according to a preset causal test rule, wherein the causal attention mask is used to constrain attention to be allocated only on node paths that conform to a preset physiological causal direction and whose lag duration is within a preset lag interval; and a cross-modal association calculation unit for calculating the temporal correlation degree between the impedance trend component and the ECG features and the pulse wave features respectively, based on the time-series graph network constrained by the causal attention mask, and generating the consistency evaluation index based on the temporal correlation degree.

[0011] Optionally, the preset causality test rule includes Granger causality test or transit entropy test; the causal attention mask generation unit is specifically used to: perform causal significance tests on the relationship between the impedance trend component and the electrocardiogram feature, and between the impedance trend component and the pulse wave feature, at multiple candidate lag durations; mark the lag durations located within the preset lag interval among the candidate lag durations corresponding to paths with significant causal relationships as effective lag durations; and generate the causal attention mask based on the effective lag durations.

[0012] Optionally, the risk output module includes: a collaborative control parameter calculation unit, used to normalize the signal quality evaluation parameters to generate a normalized quality score, normalize the consistency evaluation index to generate a normalized consistency score, and multiply the normalized quality score and the normalized consistency score, using the product as the collaborative control parameter; a risk score weighting unit, used to calculate the weighted capacity load risk score using the collaborative control parameter as the weighting coefficient; a risk level determination unit, used to determine the risk level based on the comparison result between the weighted capacity load risk score and a preset classification threshold; and an alarm control unit, used to cut off the high-level alarm output and generate an active retest command when the collaborative control parameter is lower than the preset collaborative threshold and the risk level reaches a preset alarm level; otherwise, outputting warning information containing the risk level.

[0013] This application also provides a method for early warning of heart failure volume overload, including: Acquire multimodal physiological signals of the target object; the multimodal physiological signals include at least chest impedance signals and at least one of the following auxiliary physiological signals: electrocardiogram signals, photoplethysmography pulse wave signals, body position signals, and electrode contact status signals; Based on at least one of the auxiliary physiological signals and the chest impedance signal, a signal quality evaluation parameter is generated to characterize the degree of interference of the chest impedance signal. The chest impedance signal is subjected to validity screening based on the signal quality evaluation parameters, and interference separation processing is performed on the chest impedance signal that passes the validity screening to extract the impedance trend component related to the capacity load status. Auxiliary physiological features are extracted based on the auxiliary physiological signals, and the physiological rationality of the impedance trend component is verified based on the auxiliary physiological features to generate a consistency evaluation index. Based on the signal quality evaluation parameters and the consistency evaluation index, the coordinated control parameters are determined, and the early warning information corresponding to the capacity load status is output according to the coordinated control parameters.

[0014] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0015] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method described above when it invokes the computer program in the memory.

[0016] This application acquires chest impedance signals and auxiliary physiological signals through a multimodal acquisition module. A signal quality evaluation module generates signal quality evaluation parameters based on these signals. A signal screening and trend extraction module performs validity screening based on these parameters and dynamically adjusts separation parameters to extract impedance trend components, effectively suppressing the influence of non-physiological interference on chest impedance trend estimation and ensuring the reliability of volume load-related feature extraction from the signal source. Furthermore, a physiological consistency verification module extracts auxiliary physiological features from the auxiliary physiological signals, verifies the physiological rationality of the impedance trend components, and generates consistency evaluation indicators. This verifies the rationality of chest impedance changes from a physiological mechanism perspective, effectively eliminating false alarms caused by isolated impedance anomalies. A risk output module integrates the signal quality evaluation parameters and consistency evaluation indicators to determine collaborative control parameters and outputs early warning information based on these parameters. This achieves high-confidence early warning output only when the data is reliable and physiologically consistent, significantly improving the reliability of heart failure volume load early warning.

[0017] This application also provides a method for early warning of heart failure volume overload, a computer-readable storage medium, and an electronic device, which have the above-mentioned beneficial effects, and will not be elaborated here. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application 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 embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the heart failure volume overload early warning system provided in the embodiments of this application; Figure 2 A flowchart of the heart failure volume overload early warning method provided in the embodiments of this application; Figure 3 This is a flowchart of the heart failure volume overload early warning process provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] This embodiment provides a specific implementation of a heart failure volume overload early warning system. For example... Figure 1 As shown, Figure 1 This is a schematic diagram of the heart failure volume overload early warning system provided in this application embodiment. The system includes a multimodal acquisition module, a signal quality evaluation module, a signal screening and trend extraction module, a physiological consistency verification module, and a risk output module.

[0022] The multimodal acquisition module is used to acquire multimodal physiological signals of the target object. The multimodal physiological signals include at least the chest impedance signal and at least one of the following auxiliary physiological signals: electrocardiogram signal, photoplethysmography pulse wave signal, body position signal, and electrode contact status signal.

[0023] The multimodal acquisition module can achieve synchronous acquisition of the above signals by integrating a wearable ECG garment with a chest multi-electrode impedance measurement circuit, surface electrode leads, a reflective photoelectric sensor, a triaxial accelerometer, a gyroscope, and an electrode impedance detection circuit. Each signal is timestamped and aligned within a unified time window; for example, a five-minute time window can be set to ensure the comparability of cross-modal characteristics over time.

[0024] The signal quality assessment module generates signal quality assessment parameters characterizing the degree of interference with the thoracic impedance signal based on at least one of auxiliary physiological signals and thoracic impedance signals. This module comprehensively evaluates the reliability of the thoracic impedance signal within the current time window from multiple dimensions.

[0025] Specifically, the signal quality evaluation module can extract body motion intensity and postural stability features from body position signals; extract contact anomaly features from electrode contact status signals; and extract baseline abrupt change and respiratory cycle regularity features from the waveform of chest impedance signals. Then, the above features are weighted and summed, and the weighted sum is used as the signal quality evaluation parameter. The weights of each feature can be set by those skilled in the art based on implementation experience, determined through regression training using historical data, or weighted using fuzzy comprehensive evaluation methods or lightweight neural networks. The signal quality evaluation parameter can be set from 0 to 1, with higher parameter values ​​indicating higher signal reliability.

[0026] The signal filtering and trend extraction module is used to perform validity filtering on chest impedance signals based on signal quality evaluation parameters, and to perform interference separation processing on chest impedance signals that pass the validity filtering to extract impedance trend components related to capacity load status. This module first performs validity filtering on chest impedance signals in the current time window based on a comparison between the signal quality evaluation parameters and preset quality thresholds. When the signal quality evaluation parameters are greater than or equal to a first preset threshold, the chest impedance signals in the current time window are marked as fully allowed windows; when the signal quality evaluation parameters are less than the first preset threshold but greater than or equal to a second preset threshold, the chest impedance signals in the current time window are marked as deweighted participation windows; when the signal quality evaluation parameters are less than the second preset threshold, the chest impedance signals in the current time window are removed. The first preset threshold is greater than the second preset threshold. For example, the first preset threshold can be set to 0.7, and the second preset threshold can be set to 0.4. These thresholds can be adaptively adjusted based on the distribution of historical signal quality evaluation parameters, or a hysteresis mechanism can be introduced to avoid decision jitter caused by frequent switching near the threshold.

[0027] For chest impedance signals that pass the validity screening, the signal screening and trend extraction module dynamically adjusts the separation parameters according to the signal quality evaluation parameters. After separating and removing interference components from the chest impedance signal, it extracts the impedance trend component. Interference components can include at least one of non-physiological offset components, motion artifact components, and respiratory communication components. Non-physiological offset components refer to slow baseline drift caused by non-physiological factors such as electrode attachment position shifts and gradual changes in skin contact impedance. Motion artifact components refer to transient or persistent interference generated by patient limb movements, position changes, coughing, speaking, etc. Respiratory communication components refer to periodic components related to respiratory rhythm superimposed on the chest impedance signal. This module dynamically adjusts at least one of the filter window length and filter update weight according to the signal quality evaluation parameters. When the signal quality evaluation parameters are high, a shorter filter window and a higher update weight can be used, resulting in faster trend estimation response speed; when the signal quality evaluation parameters are low, a longer filter window and a lower update weight can be used, resulting in smoother trend estimation and avoiding being misled by residual interference. The adjustment functions for the filter window length and the filter update weights can be set by those skilled in the art based on implementation experience, ensuring that the filter window length and update weights monotonically change with the signal quality evaluation parameters. Adaptive filtering of the chest impedance signal is performed using the dynamically adjusted filter window length and filter update weights, sequentially separating and removing non-physiological offset components, motion artifact components, and respiratory AC components. The remaining component after separating and removing these interference components is taken as the impedance trend component. After adaptive filtering, a low-pass filter can be used to further smooth the impedance trend component. This impedance trend component represents the chest impedance change trend related to the volume load state.

[0028] The physiological consistency verification module is used to extract auxiliary physiological features from auxiliary physiological signals and to verify the physiological rationality of impedance trend components based on these features, generating consistency evaluation indicators. When the auxiliary physiological signals include ECG signals and photoplethysmography (PPG) signals, the module extracts ECG features from the ECG signal and pulse wave features from the PPG signal. ECG features may include heart rate, heart rate variability, cardiac repolarization interval, and QT dispersion. Heart rate can be obtained through R-wave detection and RR interval statistics. Heart rate variability can be obtained through time-domain or frequency-domain analysis. The cardiac repolarization interval and its dispersion can be calculated by detecting the QRS complex initiation and T-wave termination in the ECG signal. Pulse wave features may include pulse wave conduction time, pulse wave amplitude, and peripheral perfusion variation trends. Pulse wave conduction time can be obtained by calculating the time difference between the R-wave in the ECG signal and the characteristic points in the PPG signal. Pulse wave amplitude can be obtained by detecting the amplitude difference between the peaks and troughs of the PPG signal. The trend of peripheral perfusion variation can be obtained by analyzing the variation trend of the amplitude of the photoplethysmography (PPG) signal.

[0029] The physiological consistency verification module constructs a time-series graph network using impedance trend components, electrocardiogram (ECG) features, and pulse wave features as heterogeneous nodes. The time-series graph network can be implemented using a graph attention network or a Transformer encoder architecture. This module generates causal attention masks between nodes in the time-series graph network according to preset causal testing rules. These masks constrain attention to be allocated only to node paths that conform to a preset physiological causal direction and have lag durations within a preset lag interval. Preset causal testing rules can include Granger causality tests or transitive entropy tests. Specifically, causal significance tests are performed on the relationship between impedance trend components and ECG features, and between impedance trend components and pulse wave features, at multiple candidate lag durations. Taking the Granger causality test as an example, it determines whether a causal relationship exists between two time series by examining whether the past value of one time series can effectively predict the current value of another. For the relationship between impedance trend components and ECG features, it tests whether the past value of the impedance trend component has predictive power for the current value of the ECG feature; for the relationship between impedance trend components and pulse wave features, it tests whether the past value of the impedance trend component has predictive power for the current value of the pulse wave feature. Among the candidate lag durations corresponding to paths with significant causal relationships, those lag durations falling within a preset lag interval are marked as effective lag durations. The preset lag interval can be set according to the physiological lag patterns of ECG and pulse wave changes caused by volume load variations; for example, a minimum lag duration and a maximum lag duration can be set. Paths with effective lag durations within this interval are assigned attention, while attention to other paths is suppressed by a mask. A causal attention mask is generated based on the effective lag durations.

[0030] Based on the time-series graph network constrained by causal attention masks, the physiological consistency verification module calculates the temporal correlation between the impedance trend component and electrocardiogram (ECG) features and pulse wave features within a preset lag interval, and generates a consistency evaluation index based on the temporal correlation. The lag correlation coefficient reflects the correlation between two time series at a specified lag length. The calculated lag correlation coefficients are weighted and summed, and the weighted sum is used as the consistency evaluation index. The weights of each correlation coefficient can be set by those skilled in the art based on implementation experience, or obtained through supervised pre-training using capacity-load labeled data. The consistency evaluation index characterizes the temporal correlation between the impedance trend component and ECG and pulse wave features within the preset lag interval; a higher index value indicates stronger rationality of the impedance trend component at the cross-modal physiological level.

[0031] The physiological consistency verification module also performs gating processing by comparing the consistency evaluation index with a preset consistency threshold. When the consistency evaluation index is greater than or equal to the preset consistency threshold, the current time window is determined to have passed the gating process, indicating that the impedance trend component has cross-modal consistency with the ECG and pulse wave characteristics in accordance with physiological laws, and the impedance change is physiologically reasonable. When the consistency evaluation index is less than the preset consistency threshold, the current time window is determined to have failed the gating process, indicating that the impedance trend component lacks cross-modal physiological verification support, and the impedance change may be caused by non-physiological interference. The preset consistency threshold can be set by those skilled in the art based on implementation experience.

[0032] The risk output module determines the collaborative control parameters based on signal quality evaluation parameters and consistency evaluation indicators, and outputs early warning information corresponding to the capacity load status based on these parameters. This module first normalizes the signal quality evaluation parameters to generate a normalized quality score, and then normalizes the consistency evaluation indicators to generate a normalized consistency score. Normalization maps scores with different dimensions to the same numerical range for easier subsequent calculations. The normalized quality score and the normalized consistency score are multiplied, and the product is used as the collaborative control parameter. In one implementation, for chest impedance signal windows determined to be excluded by validity screening, the collaborative control parameter can be set to zero or multiplied by a very small suppression coefficient. The collaborative control parameter comprehensively reflects both the data reliability and physiological rationality of the current time window; a higher parameter value indicates that the chest impedance signal of the current time window has both high data quality and sufficient cross-modal physiological consistency.

[0033] The risk output module uses the collaborative control parameters as weighting coefficients for the capacity load risk score, calculating a weighted capacity load risk score. The capacity load risk score can include multiple feature items such as impedance drop magnitude, impedance change slope, and offset relative to the baseline. Each feature item is multiplied by the collaborative control parameters and then summed with weights to obtain the weighted capacity load risk score. The risk level is determined by comparing the weighted capacity load risk score with a preset classification threshold. Multiple preset classification thresholds can be included to classify the risk level into low, medium, and high risk levels. Simultaneously, an alarm confidence level is generated based on the collaborative control parameters; the higher the collaborative control parameters, the higher the alarm confidence level.

[0034] When the gating processes for effectiveness screening and physiological consistency verification both pass, and the risk level corresponding to the capacity load risk score reaches the preset alarm level, a high-confidence warning message containing the risk level and alarm confidence level is output. When the risk level corresponding to the capacity load risk score reaches the preset alarm level, but the collaborative control parameter is lower than the preset collaborative threshold, the high-level alarm output is truncated and an active retest instruction is generated. The warning message can be pushed to the nursing terminal via wired or wireless network. The warning data packet can contain information such as risk level, confidence level, key physiological data segments, and major contributing modality identifiers.

[0035] Based on the above embodiments, in one feasible implementation, the system further includes a baseline update module. The baseline update module updates the individualized chest impedance baseline using the chest impedance signal of the current time window when the signal quality evaluation parameter is greater than or equal to a preset scoring threshold, no abnormal movement or contact is detected, the consistency evaluation index is lower than the abnormality judgment level, and the risk level is lower than the baseline update risk threshold. The baseline update module can update the baseline using an exponentially weighted moving average, allowing the baseline to slowly adjust with long-term changes in the patient's physiological state. When a suspected artifact, suspected wearing abnormality, or a risk level higher than the baseline update risk threshold is identified, the baseline update module freezes the baseline update, maintaining the current baseline unchanged. Upon first use, the baseline update module collects the DC trend of chest impedance during an initial period when the patient is at rest, in a stable position, and the signal quality evaluation parameter meets high-quality conditions, using its mean as the initial individualized chest impedance baseline.

[0036] This embodiment utilizes a multimodal acquisition module to simultaneously acquire and time-align multiple signals, including chest impedance, electrocardiogram (ECG), photoplethysmography (PPG), body position, and electrode contact status, providing a complete multimodal data foundation for subsequent processing. The signal quality evaluation module comprehensively assesses chest impedance signal quality from multiple dimensions, including body position stability, body movement intensity, contact quality, baseline stability, and respiratory regularity, ensuring that the signal quality evaluation parameters fully reflect the signal reliability in a real home environment. The signal filtering and trend extraction module uses a three-level gating mechanism to filter signal quality, effectively eliminating low-quality signal windows that interfere with subsequent analysis. Furthermore, this module dynamically adjusts adaptive filtering parameters based on the signal quality evaluation parameters, enabling the trend separation process to adaptively balance response speed and smoothness, ensuring sensitivity to changes in actual volume load while reducing sensitivity to non-physiological interference. The physiological consistency verification module uses a time-series graph network and a causal attention masking mechanism to quantitatively assess cross-modal physiological consistency, providing a physiological mechanism-level basis for judging the rationality of chest impedance changes. The risk output module integrates the results of the two gating methods through normalized multiplication and dynamically adjusts the risk weight and alarm confidence based on the collaborative control parameters. This achieves end-to-end quality control from signal acquisition to early warning output, effectively improving the specificity and clinical usability of the heart failure volume overload early warning system.

[0037] This application also provides a method for early warning of heart failure volume overload, such as Figure 2 As shown, Figure 2 The flowchart illustrates a heart failure volume overload early warning method provided in this application embodiment, which includes the following steps.

[0038] Step S101: Obtain multimodal physiological signals of the target object.

[0039] In this step, multimodal physiological signals are simultaneously acquired using wearable devices. These wearable devices can be ECG vests, chest monitoring straps, or other home monitoring terminals with multi-sensor integration capabilities. The multimodal physiological signals include at least chest impedance signals and at least one of the following auxiliary physiological signals: ECG signal, photoplethysmography (PPG) signal, body position signal, and electrode contact status signal. Chest impedance signals are obtained by applying a weak, high-frequency alternating current to the patient's chest using chest patch electrodes or a wearable smart vest and measuring the impedance response. ECG signals are acquired via surface electrode leads. PPG signals are acquired using a reflective photoelectric sensor, which can be placed on the wrist, fingertips, or chest. Body position signals are acquired using a triaxial accelerometer and gyroscope to reflect the patient's posture and body movements. Electrode contact status signals are acquired using an electrode impedance detection circuit to reflect the contact quality between the electrode patch and the skin. All signals are timestamped and aligned according to a unified time window, for example, every five minutes can be set as a time window to ensure the comparability of cross-modal characteristics over time. After the above acquisition and alignment processes, multimodal physiological signals can be obtained for subsequent processing.

[0040] Step S102: Generate signal quality evaluation parameters to characterize the degree of interference of the chest impedance signal based on at least one of the auxiliary physiological signal and the chest impedance signal.

[0041] In this step, a signal quality evaluation parameter for the chest impedance signal is generated based on at least one of the auxiliary physiological signals and the chest impedance signal from the multimodal physiological signals. This parameter characterizes the reliability of the chest impedance signal within the current time window. Specifically, body motion intensity features and postural stability features can be extracted from the postural state signal. Body motion intensity features reflect the intensity of the patient's activity within the time window, for example, by calculating the amplitude change or root mean square value of the acceleration signal output by a triaxial accelerometer. Postural stability features reflect the stability of the patient's posture, for example, by calculating the statistical measure of the rate of change of posture angles. Contact abnormality features can be extracted from the electrode contact state signal. These features reflect whether there is a sudden change or exceedance of the normal range in the contact impedance between the electrode patch and the skin, for example, by statistically analyzing the amount of contact impedance abrupt change or the number of times it exceeds the limit. Baseline abrupt change features and respiratory cycle regularity features can be extracted from the waveform of the chest impedance signal. Baseline abrupt change features reflect whether there is a sudden jump in the baseline of the chest impedance signal, for example, by calculating the extreme value of the first-order difference of the chest impedance signal. The regularity of the respiratory cycle can reflect the regularity of the respiratory component in the thoracic impedance signal, for example, by analyzing the concentration of the respiratory component spectrum.

[0042] After extracting the above features, the body motion intensity features, postural stability features, contact anomaly features, baseline abrupt change features, and respiratory cycle regularity features are weighted and summed. The weighted sum is used as the signal quality evaluation parameter. The weights of each feature can be set by those skilled in the art based on implementation experience, or determined through regression training using historical data, or weight allocation can be performed using fuzzy comprehensive evaluation or lightweight neural networks. The value range of the signal quality evaluation parameter can be set from 0 to 1, with higher parameter values ​​indicating higher signal reliability.

[0043] Step S103: Perform validity screening on the chest impedance signal according to the signal quality evaluation parameters, and perform interference separation processing on the chest impedance signal that passes the validity screening to extract the impedance trend component related to the capacity load status.

[0044] In this step, validity screening is first performed based on the comparison between the signal quality evaluation parameters and preset quality thresholds. Validity screening can employ a three-level gating mechanism: when the signal quality evaluation parameters are greater than or equal to the first preset threshold, the chest impedance signal in the current time window is marked as a fully open window, indicating that the signal quality in this window is good and can be fully trusted; when the signal quality evaluation parameters are less than the first preset threshold but greater than or equal to the second preset threshold, the chest impedance signal in the current time window is marked as a deweighted participation window, indicating that the signal quality in this window is moderate and can participate in subsequent analysis, but its weight should be reduced; when the signal quality evaluation parameters are less than the second preset threshold, the chest impedance signal in the current time window is removed, indicating that the signal quality in this window is too poor and should not participate in subsequent analysis. The first preset threshold is greater than the second preset threshold. For example, the first preset threshold can be set to 0.7, and the second preset threshold can be set to 0.4. These thresholds can be adaptively adjusted based on historical score distribution, or a hysteresis mechanism can be introduced to avoid decision jitter caused by frequent switching near the thresholds.

[0045] For chest impedance signals that pass the validity screening (i.e., those with signal quality evaluation parameters not lower than the second preset threshold), the separation parameters are dynamically adjusted according to the signal quality evaluation parameters. After separating and removing interference components from the chest impedance signal, the impedance trend component is extracted. Interference components can include three categories: non-physiological offset components, motion artifact components, and respiratory communication components. Non-physiological offset components refer to slow baseline drift caused by non-physiological factors such as electrode attachment position deviation and gradual changes in skin contact impedance. Motion artifact components refer to transient or continuous interference generated by patient limb movements, position changes, coughing, speaking, etc. Respiratory communication components refer to periodic components related to respiratory rhythm superimposed on the chest impedance signal.

[0046] In this step, at least one of the filter window length and filter update weight is dynamically adjusted based on the signal quality evaluation parameters. When the signal quality evaluation parameters are high, a shorter filter window and a higher update weight can be used to achieve a faster response speed in trend estimation; when the signal quality evaluation parameters are low, a longer filter window and a lower update weight can be used to achieve a smoother trend estimation and avoid being misled by residual interference. The adjustment functions for the filter window length and the filter update weight can be set by those skilled in the art based on implementation experience. Adaptive filtering is performed on the chest impedance signal using the dynamically adjusted filter window length and filter update weight, sequentially separating and removing non-physiological offset components, motion artifact components, and respiratory communication components. The remaining component after separating and removing the aforementioned interference components is taken as the impedance trend component. After adaptive filtering, a low-pass filter can be used to further smooth the impedance trend component. This impedance trend component represents the chest impedance change trend related to the volume load state.

[0047] Step S104: Extract auxiliary physiological features based on auxiliary physiological signals, and verify the physiological rationality of impedance trend components based on auxiliary physiological features to generate consistency evaluation indicators.

[0048] In this step, with auxiliary physiological signals including electrocardiogram (ECG) signals and photoplethysmography (PPG) signals, ECG features are extracted from the ECG signals, and pulse wave features are extracted from the PPG signals. ECG features may include heart rate, heart rate variability, cardiac repolarization interval, and QT dispersion. Heart rate can be obtained through R-wave detection and RR interval statistics. Heart rate variability can be obtained through time-domain or frequency-domain analysis. Cardiac repolarization interval and its dispersion can be calculated by detecting the QRS complex initiation and T-wave termination in the ECG signal. Pulse wave features may include pulse wave conduction time, pulse wave amplitude, and peripheral perfusion variation trends. Pulse wave conduction time can be obtained by calculating the time difference between the R wave in the ECG signal and the characteristic points in the PPG signal. Pulse wave amplitude can be obtained by detecting the amplitude difference between the peaks and troughs of the PPG signal. Peripheral perfusion variation trends can be obtained by analyzing the variation trend of the PPG signal amplitude.

[0049] In this step, impedance trend components, electrocardiogram (ECG) features, and pulse wave features are used as heterogeneous nodes to construct a time series graph network. The time series graph network can be implemented using a graph attention network or a Transformer encoder architecture. Causal attention masks are generated between the nodes of the time series graph network according to preset causality testing rules. These preset causality testing rules can include Granger causality tests or transitive entropy tests. Specifically, causal significance tests are performed on the relationship between impedance trend components and ECG features, and between impedance trend components and pulse wave features, at multiple candidate lag durations. Taking the Granger causality test as an example, the existence of a causal relationship between two time series is determined by examining whether the past values ​​of one time series can effectively predict the current values ​​of another time series. For the relationship between impedance trend components and ECG features, the test examines whether the past values ​​of the impedance trend components have predictive power for the current values ​​of the ECG features; for the relationship between impedance trend components and pulse wave features, the test examines whether the past values ​​of the impedance trend components have predictive power for the current values ​​of the pulse wave features. Among the candidate lag durations corresponding to paths with significant causal relationships, those lag durations falling within a preset lag interval are marked as effective lag durations. The preset lag interval can be set according to the physiological lag patterns of ECG and pulse wave changes caused by volume load variations; for example, a minimum lag duration and a maximum lag duration can be set. A causal attention mask is generated based on the effective lag durations. This causal attention mask is used to constrain attention to be allocated only to node paths that conform to the preset physiological causal direction and whose lag durations fall within the preset lag interval.

[0050] Based on the time-series graph network constrained by causal attention masks, the temporal correlation between the impedance trend component and electrocardiogram (ECG) features and pulse wave features within a preset lag interval is calculated. The temporal correlation degree can be characterized by the lag correlation coefficient, which reflects the correlation between two time series at a specified lag length. The calculated lag correlation coefficients are weighted and summed, and the weighted sum is used as a consistency evaluation index. The weights of each correlation coefficient can be set by those skilled in the art based on implementation experience, or obtained through supervised pre-training using capacity-load labeled data. The consistency evaluation index is used to characterize the temporal correlation between the impedance trend component and ECG and pulse wave features within the preset lag interval; a higher index value indicates stronger rationality of the impedance trend component at the cross-modal physiological level.

[0051] In this step, a gating process is performed by comparing the consistency evaluation index with a preset consistency threshold. When the consistency evaluation index is greater than or equal to the preset consistency threshold, the current time window is determined to have passed the gating process, indicating that the impedance trend component has cross-modal consistency with the ECG and pulse wave characteristics in accordance with physiological laws, and the impedance change is physiologically reasonable. When the consistency evaluation index is less than the preset consistency threshold, the current time window is determined to have failed the gating process, indicating that the impedance trend component lacks cross-modal physiological verification support, and the impedance change may be caused by non-physiological interference. The preset consistency threshold can be set by those skilled in the art based on implementation experience.

[0052] Step S105: Determine the coordinated control parameters based on the signal quality evaluation parameters and consistency evaluation indicators, and output early warning information corresponding to the capacity load status based on the coordinated control parameters.

[0053] In this step, the results of the validity screening and gating process are combined to generate collaborative control parameters. First, the signal quality evaluation parameters are normalized to generate a normalized quality score, and the consistency evaluation index is normalized to generate a normalized consistency score. Normalization maps scores with different dimensions to the same numerical range for easier subsequent calculations. The normalized quality score and the normalized consistency score are multiplied, and the product is used as the collaborative control parameter. In one implementation, for chest impedance signal windows determined to be excluded by the validity screening, the collaborative control parameter can be set to zero or multiplied by a very small suppression coefficient. The collaborative control parameter comprehensively reflects both the data reliability and physiological rationality of the current time window; a higher parameter value indicates that the chest impedance signal of the current time window has both high data quality and sufficient cross-modal physiological consistency.

[0054] In this step, the coordinated control parameters are used as weighting coefficients for the capacity load risk score, and a weighted capacity load risk score is calculated. The capacity load risk score may include multiple feature items such as impedance drop magnitude, impedance change slope, and offset relative to the baseline. Each feature item is multiplied by the coordinated control parameters and then summed with weights to obtain the weighted capacity load risk score. The risk level is determined by comparing the weighted capacity load risk score with a preset classification threshold. Multiple preset classification thresholds may be included to classify the risk level into low risk, medium risk, high risk, and other levels. Simultaneously, an alarm confidence level is generated based on the coordinated control parameters; the higher the coordinated control parameters, the higher the alarm confidence level.

[0055] When both validity screening and gating processing pass, and the risk level corresponding to the capacity load risk score reaches the preset alarm level, a high-confidence early warning message containing the risk level and alarm confidence level is output. When the risk level corresponding to the capacity load risk score reaches the preset alarm level, but the collaborative control parameter is lower than the preset collaborative threshold, the high-level alarm output is truncated and an active retest instruction is generated. The early warning message can be pushed to the nursing terminal via wired or wireless network. The early warning data packet can contain information such as risk level, confidence level, key physiological data segments, and major contributing modality identifiers.

[0056] Based on the above embodiments, in one feasible implementation, the method further includes a baseline update step. When the signal quality evaluation parameter is greater than or equal to a preset scoring threshold, no abnormal movements or contact abnormalities are detected, the consistency evaluation index is lower than the abnormality judgment level, and the risk level is lower than the baseline update risk threshold, the individualized chest impedance baseline is updated using the chest impedance signal of the current time window. Specifically, an exponentially weighted moving average can be used to update the baseline, allowing it to slowly adjust to long-term changes in the patient's physiological state. When a suspected artifact, suspected wearing abnormality, or a risk level higher than the baseline update risk threshold is identified, the baseline update is frozen, maintaining the current baseline unchanged. Upon first use, the chest impedance DC trend can be collected during the initial period when the patient is at rest, in a stable position, and the signal quality evaluation parameter meets high-quality conditions, and its mean value is used as the initial individualized chest impedance baseline.

[0057] This embodiment provides a comparable data foundation in the time dimension for subsequent cross-modal feature analysis by synchronously acquiring multimodal physiological signals and aligning them with a unified time window. Signal quality evaluation parameters generated based on body position, electrode contact status, and chest impedance waveform characteristics comprehensively assess the reliability of chest impedance signals from multiple dimensions, providing a quantitative basis for validity screening. Validity screening employs a three-level gating mechanism to filter chest impedance signals, effectively eliminating low-quality signal windows and reducing the probability of non-physiological interference entering subsequent analysis stages. Furthermore, adaptive filtering is performed by dynamically adjusting the filter window length and update weights according to the signal quality evaluation parameters. This ensures that the response speed and smoothness of trend estimation are linked to signal reliability, guaranteeing a rapid response to actual volume load changes when signal quality is good, while avoiding artifacts being misjudged as volume load trends when signal quality is poor. Multidimensional features extracted from ECG and photoplethysmography (PPG) pulse wave signals provide rich physiological information for cross-modal physiological consistency verification. By employing a time-series graph network and a causal attention masking mechanism, cross-modal correlations consistent with physiological causality are used as the verification basis for the rationality of impedance trend component changes, effectively suppressing false alarms caused by isolated impedance anomalies. Cooperative control parameters are generated by integrating data reliability and physiological consistency, subjecting early warning decisions to the dual constraints of signal quality screening and physiological consistency gating. Finally, risk weights and alarm confidence levels are dynamically adjusted based on the cooperative control parameters, outputting high-confidence warnings only when the data is reliable, physiologically consistent, and deviates from the individual baseline, significantly improving the specificity and reliability of heart failure volume overload warnings.

[0058] Based on the above embodiments, in one feasible implementation, when the signal quality evaluation parameter is greater than or equal to the preset scoring threshold, no abnormal action or contact is detected, the cross-modal consistency score is lower than the abnormal judgment level and the risk level is lower than the baseline update risk threshold, the individualized chest impedance baseline is updated using the chest impedance signal of the current time window.

[0059] Specifically, an exponentially weighted moving average can be used to update the baseline, allowing it to adjust slowly over time according to changes in the patient's physiological state. When a suspected artifact, suspected wearing abnormality, or risk level exceeding the baseline update risk threshold is identified, the baseline update is frozen, maintaining the current baseline unchanged. For initial use, the DC trend of chest impedance can be collected during the initial period when the patient is at rest, in a stable position, and the signal quality evaluation parameters meet high-quality conditions. The mean value of this data is used as the initial individualized chest impedance baseline.

[0060] This embodiment provides a comparable data foundation in the time dimension for subsequent cross-modal feature analysis by synchronously acquiring multimodal physiological signals and aligning them with a unified time window. Signal quality evaluation parameters generated based on body position, electrode contact status, and chest impedance waveform characteristics comprehensively assess the reliability of the chest impedance signal from multiple dimensions, providing a quantitative basis for the first gating process. The first gating process filters the chest impedance signal through a three-level gating mechanism, effectively eliminating low-quality signal windows and reducing the probability of non-physiological interference entering subsequent analysis stages.

[0061] Based on this, adaptive filtering is performed by dynamically adjusting the filter window length and update weights according to signal quality evaluation parameters. This makes the response speed and smoothness of trend estimation linked to signal reliability, ensuring a rapid response to real volume overload changes when signal quality is good, while avoiding artifacts being misjudged as volume overload trends when signal quality is poor. Multidimensional features extracted from ECG and photoplethysmography (PPG) signals provide rich physiological information for cross-modal physiological consistency verification. The physiological consistency verification module uses a time-series graph network and causal attention masking mechanism to use cross-modal correlations consistent with physiological causality as the verification basis for the rationality of chest impedance changes, effectively suppressing false alarms caused by isolated impedance anomalies. The risk output module generates collaborative control parameters by integrating data reliability and physiological consistency, subjecting early warning decisions to dual constraints of signal quality gating and causal consistency gating. Finally, risk weights and alarm confidence are dynamically adjusted based on the collaborative control parameters, outputting high-confidence warnings only when the data is reliable, physiologically consistent, and deviates from the individual baseline, significantly improving the specificity and reliability of heart failure volume overload warnings.

[0062] See Figure 3 , Figure 3 This is a flowchart of the heart failure volume overload early warning process provided in the embodiments of this application. Figure 3 In, such as Figure 3 As shown in the flowchart, from left to right and from top to bottom, the complete processing flow includes acquisition of multimodal physiological signals, signal quality evaluation, first gating processing and DC trend extraction, multimodal feature extraction, cross-modal causal consistency verification, dual-gating collaborative decision-making, and early warning output.

[0063] First, we enter the multimodal physiological signal acquisition node. This node indicates that the multimodal acquisition module synchronously acquires the multimodal physiological signals of the target object. The above signals are marked with a unified timestamp and aligned according to the time window to provide a multimodal data foundation for subsequent processing.

[0064] The body position signal, electrode contact status signal, and waveform stability characteristics of the chest impedance signal serve as input sources, all pointing to the signal quality evaluation parameter generation node. This indicates that the signal quality evaluation module generates signal quality evaluation parameters based on the aforementioned signals to characterize the degree of interference affecting the chest impedance signal. Figure 3 Signal quality score is used as an example of signal quality evaluation parameter. Specifically, body motion intensity features and body position stability features are extracted from body position state signals, contact abnormality features are extracted from electrode contact state signals, and baseline abrupt change features and respiratory cycle regularity features are extracted from the waveform of chest impedance signals. The above features are weighted and summed, and the weighted sum is used as the signal quality score.

[0065] The signal quality evaluation parameter node points to both the first gating processing execution node and the DC trend component extraction node. In the first gating processing execution node, the quality gating unit performs validity screening on the chest impedance signal within the current time window based on a comparison between the signal quality evaluation parameter and a preset quality threshold. When the signal quality evaluation parameter is greater than or equal to the first preset threshold, the chest impedance signal within the current time window is marked as a fully allowed window; when the signal quality evaluation parameter is less than the first preset threshold but greater than or equal to the second preset threshold, it is marked as a deweighted participation window; when the signal quality evaluation parameter is less than the second preset threshold, the chest impedance signal within the current time window is removed.

[0066] The chest impedance signal processed by the first gate enters the DC trend component extraction node, indicating that the adaptive filtering unit dynamically adjusts at least one of the filter window length and filter update weights according to the signal quality evaluation parameters to separate and remove interference components from the chest impedance signal and extract the DC trend component. This DC trend component is then input to the signal filtering and trend extraction module.

[0067] Both the electrocardiogram (ECG) signal and the photoplethysmography (PPG) signal point to the feature extraction node, indicating that the signal filtering and trend extraction module extracts ECG features from the ECG signal, extracts PPG features from the PPG signal, and extracts chest impedance trend features, i.e., impedance trend components, based on the DC trend component.

[0068] The chest impedance trend feature, electrocardiogram feature, and pulse wave feature all point to the cross-modal causal consistency verification node, indicating that the physiological consistency verification module constructs a time-series graph network using the above three features as heterogeneous nodes. According to the preset causal verification rules, a causal attention mask is generated between each node in the time-series graph network to constrain attention to be allocated only on node paths that conform to the preset physiological causal direction and whose lag duration is within the preset lag interval. Based on the time-series graph network constrained by the causal attention mask, the temporal correlation degree between the chest impedance trend feature and the electrocardiogram feature and the pulse wave feature within the preset lag interval is calculated to generate a cross-modal consistency result. The second gating process is then performed based on the relationship between the cross-modal consistency result and the preset consistency threshold.

[0069] The two nodes, the result state of the first gating process and the cross-modal consistency result, both point to the dual-gating collaborative decision node, which is used to generate collaborative control parameters based on the results of the first and second gating processes. Specifically, the signal quality evaluation parameters are normalized to generate a normalized quality score, the cross-modal consistency score is normalized to generate a normalized consistency score, and the normalized quality score and the normalized consistency score are multiplied together, with the product used as the collaborative control parameter.

[0070] The collaborative control parameter node points to the weight and confidence level dynamic adjustment node. The risk output module uses the collaborative control parameters as the weight coefficients for the capacity load risk score, calculates the weighted capacity load risk score, determines the risk level based on the comparison between the weighted capacity load risk score and the preset classification threshold, and generates an alarm confidence level based on the collaborative control parameters.

[0071] Ultimately, the system outputs warning information to the warning information output node. The risk output module outputs warning information corresponding to the capacity load status based on the collaborative control parameters. When both the first and second gating processes pass and the risk level corresponding to the capacity load risk score reaches the preset alarm level, a high-confidence warning message containing the risk level and alarm confidence is output. When the risk level corresponding to the capacity load risk score reaches the preset alarm level, but the collaborative control parameters are lower than the preset collaborative threshold, the high-level alarm output is truncated and an active retest command is generated.

[0072] This application also provides an embodiment of a computer-readable storage medium and a computer program product. Both the computer-readable storage medium and the computer program product may store a computer program that, when executed by a processor, implements the steps of the method described in the above method embodiments.

[0073] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0074] The computer-readable storage medium provided in this embodiment includes computer programs that implement the corresponding embodiments of the above methods, and the effects are the same as above.

[0075] This application also provides an electronic device, see [link to document]. Figure 4 The present application provides a structural diagram of an electronic device, such as... Figure 4 As shown, it may include a processor 1410 and a memory 1420.

[0076] The processor 1410 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 1410 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1410 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 1410 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 1410 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0077] The memory 1420 may include one or more computer-readable storage media, which may be non-transitory. The memory 1420 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 1420 is used to store at least the following computer program 1421, which, after being loaded and executed by the processor 1410, is capable of implementing the relevant steps in the methods executed by the electronic device side as disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 1420 may also include an operating system 1422 and data 1423, etc., and the storage method may be temporary storage or permanent storage. The operating system 1422 may include Windows, Linux, Android, etc.

[0078] In some embodiments, the electronic device may further include a display screen 1430, an input / output interface 1440, a communication interface 1450, a sensor 1460, a power supply 1470, and a communication bus 1480.

[0079] certainly, Figure 4 The structure of the electronic device shown does not constitute a limitation on the electronic device in the embodiments of this application. In practical applications, the electronic device may include more than [other components]. Figure 4 More or fewer components as shown, or combinations of certain components.

[0080] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. As the system provided in the embodiments corresponds to the method provided in the embodiments, the description is relatively simple; relevant parts can be found in the method section.

[0081] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

[0082] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A heart failure volume overload early warning system, characterized in that, include: The multimodal acquisition module is used to acquire multimodal physiological signals of the target object; The multimodal physiological signals include at least chest impedance signals and at least one of the following auxiliary physiological signals: electrocardiogram signals, photoplethysmography pulse wave signals, body position signals, and electrode contact status signals. The signal quality evaluation module is used to generate signal quality evaluation parameters that characterize the degree of interference of the chest impedance signal based on at least one of the auxiliary physiological signal and the chest impedance signal. The signal filtering and trend extraction module is used to perform validity filtering on the chest impedance signal according to the signal quality evaluation parameters, and to perform interference separation processing on the chest impedance signal that passes the validity filtering, so as to extract the impedance trend component related to the capacity load state. The physiological consistency verification module is used to extract auxiliary physiological features based on the auxiliary physiological signals, and to verify the physiological rationality of the impedance trend component based on the auxiliary physiological features, thereby generating a consistency evaluation index. The risk output module is used to determine the collaborative control parameters based on the signal quality evaluation parameters and the consistency evaluation index, and output early warning information corresponding to the capacity load status based on the collaborative control parameters.

2. The system according to claim 1, characterized in that, The auxiliary physiological signals include the body position signal and the electrode contact status signal; Accordingly, the signal quality evaluation module includes: A body movement and position feature extraction unit is used to extract body movement intensity features and body position stability features based on the body position state signal; A contact anomaly detection unit is used to extract contact anomaly features based on the electrode contact state signal. The impedance waveform analysis unit is used to extract baseline abrupt change features and respiratory cycle regularity features based on the waveform of the chest impedance signal. The quality scoring fusion unit is used to perform a weighted summation of the body motion intensity characteristics, the body position stability characteristics, the contact abnormality characteristics, the baseline abrupt change characteristics, and the respiratory cycle regularity characteristics, and use the weighted summation result as the signal quality evaluation parameter.

3. The system according to claim 1, characterized in that, The signal filtering and trend extraction module includes: A quality gating unit is used to perform validity screening on the chest impedance signal in the current time window based on the comparison relationship between the signal quality evaluation parameter and a preset quality threshold; when the signal quality evaluation parameter is greater than or equal to a first preset threshold, the chest impedance signal in the current time window is marked as a fully open window; when the signal quality evaluation parameter is less than the first preset threshold but greater than or equal to a second preset threshold, the chest impedance signal in the current time window is marked as a deweighted participation window; when the signal quality evaluation parameter is less than the second preset threshold, the chest impedance signal in the current time window is removed; wherein the first preset threshold is greater than the second preset threshold. An adaptive filtering unit is used to dynamically adjust at least one of the filter window length and filter update weight according to the signal quality evaluation parameters of the chest impedance signal that has passed the validity screening, and to perform adaptive filtering on the chest impedance signal with the dynamically adjusted parameters, to separate and remove interference components from the chest impedance signal, and to use the remaining components as the impedance trend components.

4. The system according to claim 1, characterized in that, The interference components include at least one of the following: non-physiological deviation components, motion artifact components, and respiratory communication components. The adaptive filtering unit is used to dynamically adjust at least one of the filter window length and filter update weight according to the signal quality evaluation parameters, and to adaptively filter the chest impedance signal with the dynamically adjusted filter window length and filter update weight, thereby separating and removing the non-physiological offset component, the motion artifact component and the respiratory communication component in sequence, and taking the remaining component after separating and removing the interference components as the impedance trend component.

5. The system according to claim 1, characterized in that, The auxiliary physiological signals include the electrocardiogram signal and the photoplethysmography (PPG) signal; The physiological consistency verification module includes: An electrocardiogram (ECG) feature extraction unit is used to extract ECG features from the ECG signal; A pulse wave feature extraction unit is used to extract pulse wave features from the photoplethysmography pulse wave signal. A time-series graph network construction unit is used to construct a time-series graph network by using the impedance trend component, the electrocardiogram feature and the pulse wave feature as heterogeneous nodes. The causal attention mask generation unit is used to generate causal attention masks between each node of the time sequence graph network according to a preset causal verification rule. The causal attention mask is used to constrain attention to be allocated only on node paths that conform to a preset physiological causal direction and whose lag duration is within a preset lag interval. The cross-modal correlation calculation unit is used to calculate the temporal correlation degree between the impedance trend component and the electrocardiogram feature and the pulse wave feature respectively, based on the temporal graph network constrained by the causal attention mask, and to generate the consistency evaluation index based on the temporal correlation degree.

6. The system according to claim 5, characterized in that, The preset causality test rules include Granger causality test or transit entropy test; The causal attention mask generation unit is specifically used for: Causal significance tests were performed on the relationship between the impedance trend component and the electrocardiogram feature, and between the impedance trend component and the pulse wave feature, at multiple candidate hysteresis durations. Among the candidate lag times corresponding to paths with significant causal relationships, the lag times that fall within the preset lag interval are marked as effective lag times; The causal attention mask is generated based on the effective lag duration.

7. The system according to claim 1, characterized in that, The risk output module includes: The collaborative control parameter calculation unit is used to normalize the signal quality evaluation parameters to generate a normalized quality score, normalize the consistency evaluation index to generate a normalized consistency score, multiply the normalized quality score and the normalized consistency score, and use the product result as the collaborative control parameter. The risk scoring weighting unit is used to calculate the weighted capacity load risk score by using the coordinated control parameters as weighting coefficients for the capacity load risk score. The risk level determination unit is used to determine the risk level based on the comparison result between the weighted capacity load risk score and the preset classification threshold. An alarm control unit is used to cut off the high-level alarm output and generate an active retest command when the collaborative control parameter is lower than a preset collaborative threshold and the risk level reaches a preset alarm level; otherwise, it outputs a warning message containing the risk level.

8. A method for early warning of heart failure volume overload, characterized in that, include: Acquire multimodal physiological signals of the target object; The multimodal physiological signals include at least chest impedance signals and at least one of the following auxiliary physiological signals: electrocardiogram signals, photoplethysmography pulse wave signals, body position signals, and electrode contact status signals. Based on at least one of the auxiliary physiological signal and the chest impedance signal, generate signal quality evaluation parameters to characterize the degree of interference to the chest impedance signal; The chest impedance signal is subjected to validity screening based on the signal quality evaluation parameters, and interference separation processing is performed on the chest impedance signal that passes the validity screening to extract the impedance trend component related to the capacity load status. Auxiliary physiological features are extracted based on the auxiliary physiological signals, and the physiological rationality of the impedance trend component is verified based on the auxiliary physiological features to generate a consistency evaluation index. Based on the signal quality evaluation parameters and the consistency evaluation index, the coordinated control parameters are determined, and the early warning information corresponding to the capacity load status is output according to the coordinated control parameters.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for implementing the steps of the method as described in claim 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the steps of the method as described in claim 8.

Images