An electrocardio monitoring system based on internet of things

Abstract

The present application relates to the technical field of remote electrocardio monitoring and signal processing, in particular to an electrocardio monitoring system based on Internet of Things. It comprises: a multi-modal data synchronization unit for synchronously collecting original electrocardio signals and electrode-skin interface dynamic contact impedance at each sampling moment; a dynamic characteristic quantization unit for collecting dynamic contact impedance, calculating interface instability index and morphological deviation; a quality confidence assessment unit for generating signal quality confidence; an adaptive signal repair unit for dynamically repairing original electrocardio signals in response to signal quality confidence generated by the quality confidence assessment unit to generate repaired electrocardio signals; and a clinical decision assistance unit for outputting clinical alarm activation index and executing hierarchical early warning. The present application overcomes the limitation of traditional methods which only rely on electrocardio signal self-analysis, realizes accurate real-time quantization of signal reliability, and provides a solid foundation for subsequent adaptive repair and reliable decision-making.

Description

Technical Field

[0001] This invention relates to the field of remote electrocardiogram (ECG) monitoring and signal processing technology, specifically to an ECG monitoring system based on the Internet of Things (IoT). Background Technology

[0002] IoT-based remote ECG monitoring technology is playing an increasingly important role in personal health management and outpatient monitoring. Traditional remote ECG monitoring systems continuously collect users' ECG signals through portable devices and perform real-time analysis to screen for abnormal events such as arrhythmias. However, in practical applications, such systems generally face the serious challenge of unstable signal quality. Because users are active, the contact interface between electrodes and skin is easily affected by factors such as body movement, sweat secretion, electrode aging, or loose adhesion. These interferences lead to a large number of artifacts in the ECG signal, such as baseline drift, electromyographic interference, and sudden noise. These artifacts are sometimes very similar in shape to the waveforms of real arrhythmias, making it difficult for existing algorithms to accurately distinguish them. Existing technologies cannot trace the cause of artifacts; on the other hand, when artifacts are highly similar to pathological waveforms, misjudgment is very likely to occur. This directly leads to a high false alarm rate in remote ECG monitoring systems, causing alarm fatigue for users, and may even delay treatment due to ignoring real alarms. Therefore, there is an urgent need for a new technology that can accurately assess the real-time quality of ECG signals, effectively suppress artifacts, and reduce the clinical false alarm rate.

[0003] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention discloses an electrocardiogram (ECG) monitoring system based on the Internet of Things (IoT). Specifically, the technical solution of this invention includes: The multimodal data synchronization unit is used to synchronously acquire the raw electrocardiogram signal and the dynamic contact impedance of the electrode skin interface at each sampling moment. The dynamic feature quantization unit is used to calculate the interface instability index based on the dynamic contact impedance acquired by the multimodal data synchronization unit, and to calculate the morphological deviation based on the original electrocardiogram signal. The quality confidence assessment unit is used to fuse the interface instability index and morphological deviation calculated by the dynamic feature quantization unit to generate signal quality confidence. An adaptive signal repair unit is used to dynamically repair the original ECG signal in response to the signal quality confidence generated by the quality confidence assessment unit, so as to generate a repaired ECG signal. It also includes a clinical decision support unit, which combines the analysis results of the repaired ECG signal with the signal quality confidence level to output a clinical alarm activation index and perform graded early warning.

[0005] Preferably, the dynamic feature quantization unit is specifically used for: Obtain the contact impedance measurement values ​​at the current and previous sampling times; Based on the current and previous sampling times of the contact impedance measurement, determine the instantaneous change in contact impedance; The instantaneous change is normalized by using a preset individualized reference impedance to obtain the normalized result. A nonlinear amplification factor is used to amplify the normalization result in order to generate an interface instability index. The individualized reference impedance is the impedance value measured when the user is in a resting state during the initial period of wearing the device.

[0006] Preferably, the dynamic feature quantization unit is specifically used for: Extract morphological feature vectors from the current ECG waveform segment to be analyzed; Based on the preset standard waveform prior mean eigenvector and prior covariance matrix, calculate the Mahalanobis distance from the morphological eigenvector to the center of the standard waveform distribution. The squared Mahalanobis distance is set as the morphological deviation.

[0007] Preferably, the quality confidence assessment unit is specifically used for: Multiply the interface instability index by the preset independent influence weights to obtain the first attenuation term; The second attenuation term is obtained by multiplying the morphological deviation by the preset independent influence weights. Multiplying the product of the interface instability index and the morphological deviation by a preset coupling effect weight yields the coupling attenuation term. Summing the first attenuation term, the second attenuation term, and the coupling attenuation term yields the negative penalty term; Signal quality confidence is generated based on the negative penalty term and through exponential function operation.

[0008] Preferably, the adaptive signal repair unit is specifically used for: Subtract the value 1 from the signal quality confidence level to obtain the confidence level difference; The confidence difference is exponentially calculated using a preset sensitivity coefficient to obtain the exponentiation result. The result of the exponentiation operation is multiplied by the preset maximum repair strength limit to generate an adaptive repair strength factor.

[0009] Preferably, the adaptive signal repair unit is further configured to: The raw electrocardiogram signal is subjected to powerful filtering in parallel to generate a reference signal; Based on the adaptive repair intensity factor as the weight, the original ECG signal and the reference signal are linearly superimposed to generate the repaired ECG signal.

[0010] Preferably, the clinical decision support unit is specifically used for: The repaired electrocardiogram signals were analyzed to extract pathological features; Based on the pre-defined scoring criteria in reference clinical guidelines, pathological features are quantified to generate a severity score for pathological events.

[0011] Preferably, the clinical decision support unit is further specifically used for: The severity score of the pathological event is multiplied by the confidence level of the signal quality to generate a clinical alarm activation index; The clinical alarm activation index is compared with preset multi-level thresholds to implement graded early warning.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention, by simultaneously acquiring the dynamic impedance of the electrocardiogram (ECG) signal and the electrode-skin interface, and constructing a fusion evaluation model including coupling effect terms, can quantify the independent and concurrent effects of physical interface instability and signal morphology abnormalities, generating accurate signal quality confidence levels. This overcomes the limitations of traditional methods that rely solely on ECG signal analysis, achieving precise real-time quantification of signal confidence, and providing a solid foundation for subsequent adaptive repair and reliable decision-making.

[0013] 2. This invention establishes an adaptive repair closed loop driven by signal quality confidence. The system dynamically adjusts the repair intensity based on the real-time quantized signal quality confidence, and uses this as a weight to linearly superimpose the original signal and the strongly filtered reference signal. This method can preserve physiological details to the maximum extent in high-quality signal segments, while performing precise, smooth, and controllable artifact suppression in low-quality segments, avoiding waveform distortion or artifact residue problems caused by traditional fixed filters.

[0014] 3. This invention innovatively multiplies the severity score of a pathological event by the quality confidence score of the corresponding original signal segment to generate a clinical alarm activation index. This decision-making mechanism ensures that alarms are not only based on what was detected, but also on the reliability of the detection results. It can fundamentally suppress clinical false alarms triggered by signal artifacts being misidentified as pathological events, significantly improve the specificity of alarms, reduce alarm fatigue among medical staff, and make the monitoring system more clinically practical.

[0015] 4. This invention employs a robust and individualized method in the dynamic feature quantization stage. By introducing an individualized reference impedance for normalization and using nonlinear amplification, it can sensitively capture interface instability events. Simultaneously, it utilizes a probabilistic model based on Mahalanobis distance to quantify ECG waveform deviation, which considers the complex statistical relationships between various features of the standard waveform. This enables the feature quantization process to effectively eliminate individual differences and more accurately distinguish between artifacts and genuine physiological and pathological variations. Attached Figure Description

[0016] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a system structure block diagram of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0018] Example 1: Please see Figure 1 An IoT-based electrocardiogram monitoring system includes: The multimodal data synchronization unit is used to synchronously acquire the raw electrocardiogram signal and the dynamic contact impedance of the electrode skin interface at each sampling moment. The dynamic feature quantization unit is used to calculate the interface instability index based on the dynamic contact impedance acquired by the multimodal data synchronization unit, and to calculate the morphological deviation based on the original electrocardiogram signal. The quality confidence assessment unit is used to fuse the interface instability index and morphological deviation calculated by the dynamic feature quantization unit to generate signal quality confidence. An adaptive signal repair unit is used to dynamically repair the original ECG signal in response to the signal quality confidence generated by the quality confidence assessment unit, so as to generate a repaired ECG signal. And a clinical decision support unit, which combines the analysis results of the repaired ECG signal with the signal quality confidence level, outputs a clinical alarm activation index, and performs graded early warning; An IoT-based electrocardiogram (ECG) monitoring system aims to address the core technical problem in traditional ECG monitoring: signal artifacts caused by electrode-skin interface instability are difficult to distinguish from actual pathological waveforms, leading to high false alarm rates and low diagnostic efficiency. The system constructs a complete technical closed loop, from dynamic perception of the physical interface and multimodal feature fusion assessment to adaptive signal repair and confidence-guided decision-making, to achieve accurate quantification of ECG signal quality and highly reliable clinical early warning. The system comprises the following core units: The multimodal data synchronization unit aims to provide time-aligned, different but related raw data streams for subsequent coupled analysis. This unit uses the built-in sensing module of the ECG monitor to synchronize data at each sampling time. Simultaneously, two acquisition tasks are performed: acquiring raw electrocardiogram (ECG) signals reflecting cardiac electrophysiological activity, expressed as voltage values. By applying a weak high-frequency alternating current between the electrode pairs, the dynamic contact impedance at the electrode-skin interface is measured and acquired. The synchronization of the two data streams is crucial to ensuring accurate subsequent analysis of changes in the physical interface. Reflecting changes in signal morphology The basis for reflecting the causal relationship between them; The dynamic feature quantization unit aims to transform the two acquired raw time-series data streams into quantization metrics that characterize the physical interface stability and signal morphology standardization, respectively. This unit receives data from the multimodal data synchronization unit. and It is based on dynamic contact impedance. The time series data is used to calculate the interface instability index using a nonlinear change detection model. This index is used to sensitively detect physical instability events such as minute electrode displacement or poor contact; it is based on raw electrocardiogram signals. The waveform segment was used to calculate the morphological deviation using a probabilistic model based on statistical pattern recognition. This metric measures the degree of difference between the current electrocardiogram waveform and its standard template. The quality confidence assessment unit aims to establish a correlation between physical layer characteristic interface instability and signal layer characteristic morphological deviation, generating a unified index that comprehensively reflects the current signal confidence level. This unit receives the output from the dynamic feature quantization unit. and It uses an exponential decay model that includes independent influence terms and coupling effect terms to nonlinearly fuse these two indicators, ultimately generating a signal quality confidence score in the range (0,1). The core design of this model is that when both interface instability and shape deviation increase simultaneously, a penalty much greater than the independent influence of the two will be imposed, thereby drastically reducing the confidence level and accurately corresponding to scenarios where poor physical contact leads to severe signal distortion. The adaptive signal restoration unit aims to dynamically and intelligently suppress artifacts in the original ECG signal using the signal quality confidence level generated in the preceding steps. This unit responds to the signal quality confidence level generated by the quality confidence level assessment unit. It is based on Calculate an adaptive repair strength factor This factor and It is inversely proportional; it uses this factor as a weight for the raw electrocardiogram signal. A reference signal that has been strongly filtered Linear superposition is performed to generate the repaired ECG signal. When the confidence level is high, the repair strength approaches zero, preserving the original signal details to the maximum extent; when the confidence level is low, the repair strength is enhanced, effectively suppressing artifacts. The clinical decision support unit aims to combine signal processing results with signal quality assessment results to output a final clinical alarm command that balances pathological severity and evidence reliability. This unit processes the repaired ECG signal output by the adaptive signal repair unit. Pathological analysis was performed to obtain a severity score for the pathological event. It will rate this. Signal quality confidence level corresponding to this signal Multiply to obtain the clinical alarm activation index. The activation index is compared with a preset multi-level threshold to execute a graded warning. This design ensures that a high-level alarm is triggered only when the pathological event itself is severe and the quality of its corresponding original signal is reliable. The IoT-based ECG monitoring system in this embodiment synchronously acquires physical interface impedance and ECG signals, and innovatively establishes a coupling evaluation model between the two. This enables real-time and accurate quantification of the quality confidence of ECG signals. Based on this confidence, the system achieves adaptive adjustment of signal repair intensity and confidence weighting for clinical alarm decisions. This preserves the true pathological characteristics while suppressing false alarms caused by dynamic artifacts, significantly improving the accuracy and clinical application value of remote ECG monitoring.

[0019] Example 2: The dynamic feature quantization unit is specifically used for: Obtain the contact impedance measurement values ​​at the current and previous sampling times; Based on the current and previous sampling times of the contact impedance measurement, determine the instantaneous change in contact impedance; The instantaneous change is normalized by using a preset individualized reference impedance to obtain the normalized result. A nonlinear amplification factor is used to amplify the normalization result in order to generate an interface instability index. Among them, the individualized reference impedance is the impedance value measured in the resting state during the initial period of user wearing the device; This embodiment is a further specification of the dynamic feature quantization unit in solving the interface instability index, based on embodiment 1. The dynamic feature quantization unit is implemented as follows when quantizing the interface instability index: The unit at the current sampling time Obtain the current contact impedance measurement value And retrieve the contact impedance measurement value from the storage at the previous sampling time. ; The unit determines the instantaneous change in contact impedance based on the current and previous sampling times, i.e., it calculates the absolute value of the difference between the two measurements. This step aims to capture real-time fluctuations in impedance. The unit uses a preset individualized reference impedance to normalize the instantaneous changes and obtain the normalized result. Individualized reference impedance, denoted here as This is a stable contact impedance value measured and stored during the initial resting state when the user first wears this monitoring device; its function is to serve as a personalized calibration baseline to eliminate the influence of differences in individual skin characteristics, electrode cream application, etc., so that subsequent calculations are comparable to individuals; it is automatically calibrated during the device initialization phase through the measurement of a stable signal. The unit uses a nonlinear amplification factor to amplify the normalization result in order to generate an interface instability index. To achieve the above steps, this embodiment introduces an interface instability index. The computational model is defined as follows: in, The interface instability index is a dimensionless value used to characterize... The degree of physical instability at the electrode-skin interface at any given time; and The contact impedance measurement value is acquired in real time by the multi-mode data synchronization unit. The individualized reference impedance is obtained through device initialization calibration; before calculation, the system should... The validity of the test shall be checked; if If calibration fails or its value falls below a preset minimum physical effective threshold, the system should adopt a universal default reference impedance value obtained through extensive statistical analysis and record an initialization anomaly to ensure robust system operation. This is a nonlinear amplification factor, with a value greater than 1, typically set between 1.5 and 2.5. It nonlinearly amplifies the normalized instantaneous change in impedance, thereby suppressing small, gradual changes caused by physiological activities such as respiration, while significantly amplifying dramatic instantaneous jumps caused by physical instability such as electrode displacement. As a pre-determined model constant, this parameter is selected on a calibration dataset containing known dynamic decoupling events to maximize... The distinction between decoupled events and normal body movement is the optimization objective, and it is obtained by training a machine learning optimization algorithm. By introducing an individualized reference impedance for normalization and using a nonlinear amplification factor for processing, the quantization method in this embodiment not only eliminates individual differences but also greatly enhances the sensitivity to drastic impedance changes caused by interface instability. At the same time, it effectively ignores benign physiological fluctuations, thereby more accurately and robustly quantifying the physical root causes of signal artifacts.

[0020] Example 3: The dynamic feature quantization unit is specifically used for: Extract morphological feature vectors from the current ECG waveform segment to be analyzed; Based on the preset standard waveform prior mean eigenvector and prior covariance matrix, calculate the Mahalanobis distance from the morphological eigenvector to the center of the standard waveform distribution. The squared Mahalanobis distance is set as the morphological deviation. This embodiment is a further specification of the dynamic feature quantization unit in solving morphological deviation, based on embodiment 1. The dynamic feature quantization unit performs morphological deviation quantization as follows: The unit extracts morphological feature vectors from the current ECG waveform segment to be analyzed; this ECG waveform segment can be a single cardiac cycle, and the morphological feature vector is denoted here as... , is a multidimensional vector that can represent the morphological characteristics of the waveform segment; its function is to transform the time-domain waveform into a feature point that is mathematically easy to statistically compare; it is the result of the signal processing module directly using the time-normalized waveform sampling point sequence within a cardiac cycle as a vector, or extracting a set of coefficients under a specific wavelet basis as a vector, in order to capture the morphological and frequency characteristics of the waveform. The unit calculates the Mahalanobis distance from the morphological feature vector to the center of the standard waveform distribution based on the preset standard waveform prior mean feature vector and prior covariance matrix. The eigenvector of the prior mean of the standard waveform is denoted here as... It is the average feature vector calculated by statistically learning from a large number of high-quality, standard ECG waveform samples annotated by medical experts in authoritative ECG databases such as the MIT-BIH arrhythmia database; it represents the most standard waveform morphology. The prior covariance matrix, denoted here as Similarly, it is obtained through statistical learning of the above standard sample set; it describes the correlation between various feature dimensions of the standard waveform and their respective range of variation; The unit sets the calculated Mahalanobis distance as the morphological deviation. To achieve the above steps, this embodiment introduces morphological deviation. The computational model is defined as follows: in, Morphological deviation is a dimensionless scalar value used to measure the current waveform vector. Statistical distance from the standard waveform distribution; The morphological feature vector is extracted in real time from the current ECG signal segment; and The prior parameters are pre-generated and stored in the system through offline training on an authoritative database. The technical motivation of this formula is that Mahalanobis distance not only considers the Euclidean distance between the current waveform and the standard mean, but also uses the covariance matrix to standardize and decorrelate the data, so that the measurement is not affected by feature scale and correlation, and can more robustly assess the degree of morphological anomaly. By employing a probabilistic model based on Mahalanobis distance, this embodiment can quantify morphological abnormalities of ECG waveforms in a statistically more robust manner. Compared to simple template matching or threshold comparison, this method considers the complex statistical relationships between various features within a normal waveform, thus more accurately identifying true morphological variations caused by artifacts or pathology, providing a more reliable input for subsequent quality assessment.

[0021] Example 4: The quality confidence assessment unit is specifically used for: Multiply the interface instability index by the preset independent influence weights to obtain the first attenuation term; The second attenuation term is obtained by multiplying the morphological deviation by the preset independent influence weights. Multiplying the product of the interface instability index and the morphological deviation by a preset coupling effect weight yields the coupling attenuation term. Summing the first attenuation term, the second attenuation term, and the coupling attenuation term yields the negative penalty term; Based on the negative penalty term, the signal quality confidence is generated through exponential function operation; This embodiment is a further specification of the quality confidence assessment unit based on Embodiment 1; The quality confidence assessment unit combines the interface instability index and morphological deviation, and its implementation is as follows: This unit receives the interface instability index calculated by the dynamic feature quantization unit. and morphological deviation ; The unit will use the interface instability index Compared with the preset independent influence weight Multiplying them together yields the first attenuation term. ; morphological deviation Compared with the preset independent influence weight Multiplying them together yields the second attenuation term. These two weights and These represent the degree of negative impact on signal quality caused by physical interface instability and independent signal morphology abnormalities, respectively. A key step is that the cell will use the interface instability index. Deviation from morphology The product of these factors, multiplied by a preset coupling effect weight. The coupling attenuation term is obtained. The technical motivation behind this design lies in the fact that when physical interface instability and signal morphology abnormalities occur simultaneously... and At the same time, a large value indicates a high probability of physical artifacts contaminating the signal. In this case, the unreliability of the signal should be much greater than the sum of the effects of the two independent factors. The coupling term is precisely to impose additional, punitive attenuation on such events. The unit sums the first attenuation term, the second attenuation term, and the coupling attenuation term to obtain the negative penalty term: ; The unit generates signal quality confidence based on a negative penalty term and through exponential function calculation; To implement the above fusion logic, this embodiment introduces a signal quality confidence level. The computational model is defined as follows: in, The confidence score for signal quality is an evaluation value mapped to the interval (0,1]. The closer the value is to 1, the higher the signal quality and the more reliable it is. and The input features are calculated from the preceding steps; , , The model weights are predetermined constants and are trained using machine learning optimization algorithms such as gradient descent on a mixed ECG calibration dataset with various artifacts artificially injected, such as electrode movement and electromyography interference. The optimization goal is to maximize the area under the receiver operating characteristic curve (AUC) that the model can use to distinguish between real pathological waveforms and artifact-contaminated waveforms. By designing an exponential fusion model that includes independent attenuation terms and coupled attenuation terms, this embodiment can accurately model signal quality; it is an innovative coupling term. This enables the assessment model to identify and severely punish high-artifact-risk scenarios involving both physical instability and signal anomalies, thereby improving the accuracy of signal quality assessment and the ability to identify complex artifacts, laying a solid foundation for subsequent adaptive repair and reliable decision-making.

[0022] Example 5: The adaptive signal repair unit is specifically used for: Subtract the value 1 from the signal quality confidence level to obtain the confidence level difference; The confidence difference is exponentially calculated using a preset sensitivity coefficient to obtain the exponentiation result. The result of the exponentiation operation is multiplied by the preset maximum repair strength limit to generate an adaptive repair strength factor; The adaptive signal repair unit is also specifically used for: The raw electrocardiogram signal is subjected to powerful filtering in parallel to generate a reference signal; Based on the adaptive repair intensity factor as the weight, the original ECG signal and the reference signal are linearly superimposed to generate the repaired ECG signal; This embodiment, based on Embodiment 1, further specifies the adaptive signal repair unit, describing how it generates a repair intensity factor and applies the factor to perform signal repair. The adaptive signal repair unit receives the signal quality confidence generated by the quality confidence assessment unit. Based on this, the original electrocardiogram signal was analyzed. Perform dynamic repair; This unit is used to generate adaptive repair strength factors. The calculation steps are as follows: Compare the value 1 with the signal quality confidence level Perform the subtraction operation to obtain the confidence difference. This difference is directly proportional to the unreliability of the signal. The confidence difference is exponentially calculated using a preset sensitivity coefficient to obtain the exponentiation result. ; The result of the exponentiation operation is multiplied by the preset maximum repair strength limit to generate an adaptive repair strength factor; This process is driven by the following adaptive repair strength factor. Defined by a computational model: in, The adaptive repair strength factor is a factor in the range [0, 1]. The dynamic adjustment parameters between [ ] are used to control the strength of subsequent repair algorithms; The signal quality confidence level is provided by the quality confidence assessment unit. The maximum repair intensity limit is a constant preset according to the application scenario, usually set to 1, which means performing the strongest repair. The sensitivity coefficient is a preset constant greater than 1, used to adjust the nonlinear relationship between repair intensity and confidence level. Its function is to make the change in repair intensity relatively gradual in the high confidence region, while the repair intensity will rapidly climb to the maximum value in the low confidence region. The value of this parameter is determined by conducting experiments on a preset test dataset to optimize the balance between repair response speed and stability. In generation Subsequently, the adaptive signal repair unit uses this factor to repair the original ECG signal, and its application is as follows: Unit for raw electrocardiogram signals Parallel intensive filtering is performed to generate a reference signal. The powerful filter here can be a low-pass or band-pass filter with fixed parameters, such as a fourth-order Butterworth low-pass filter with a cutoff frequency of 25 Hz, designed to suppress common artifacts such as electromyography interference and baseline drift to the greatest extent possible, even if this may smooth out some high-frequency details of the ECG waveform. Unit based on adaptive repair strength factor As a weight, the raw electrocardiogram signal With reference signal Linear superposition was performed to generate the repaired ECG signal. ; The mathematical implementation of this repair process is as follows: Among them, when the signal quality confidence level At a very high level, Approaching 0, the repaired signal Almost exactly equal to the original signal This preserves all the physiological details; conversely, when At extremely low temperatures, Approaching For example, 1, the repaired signal It mainly consists of the reference signal after strong filtering. This structure effectively suppresses artifacts. By nonlinearly correlating the repair intensity with the signal quality confidence level, and using this intensity factor to dynamically and linearly superimpose the original signal and the filtered reference signal, this embodiment constructs a truly adaptive repair closed loop. It can accurately, smoothly, and with controllable intensity repair of low-quality signal segments while ensuring the fidelity of high-quality signal segments, avoiding the information loss or artifact residue problems caused by the one-size-fits-all processing of traditional fixed-intensity filters.

[0023] Example 6: The clinical decision support unit is specifically used for: The repaired electrocardiogram signals were analyzed to extract pathological features; Based on the pre-defined scoring criteria in reference clinical guidelines, pathological features are quantified to generate a severity score for pathological events. The clinical decision support unit is also specifically used for: The severity score of the pathological event is multiplied by the confidence level of the signal quality to generate a clinical alarm activation index; The clinical alarm activation index is compared with preset multi-level thresholds to execute tiered early warnings; This embodiment, based on Embodiment 1, further specifies the clinical decision support unit and describes how it integrates the analysis results of the repaired signal with the original quality confidence level of the signal to achieve highly reliable graded early warning. The clinical decision support unit is the final output of the system, and its workflow is as follows: Unit for repaired ECG signals Analyze the data to extract pathological features; It is the output from the adaptive signal repair unit, and its artifacts have been effectively suppressed; the extraction of pathological features may include algorithms such as heart rate variability (HRV) analysis, ST segment deviation detection, and arrhythmia pattern recognition such as premature ventricular contractions and atrial fibrillation. The unit quantifies pathological features based on the pre-defined assignment criteria in reference clinical guidelines to generate a pathological event severity score; The severity score of the pathological event is denoted here as , is a quantized value used to represent in The clinical risk level of physiological abnormal events detected at any time; the assignment criteria are pre-set. For example, according to the ACC / AHA / HRS guidelines for the management of ventricular arrhythmias and the prevention of sudden cardiac death, the severity score of detected ventricular tachycardia is set higher than that of occasional premature ventricular contractions. To avoid misjudgments due to relying solely on signal analysis results, the unit performs a crucial fusion step: scoring the severity of pathological events. With signal quality confidence Multiply to generate a clinical alarm activation index; Signal quality confidence Specifically refers to calculation The original confidence level generated during the quality assessment phase, corresponding to the segment of signal used; This decision-making process is driven by the following clinical alarm activation index. Defined using a decision model: in, The Clinical Alarm Activation Index is a final risk score that integrates the severity of an event with the reliability of the evidence. For the repaired signal The analysis results It is the quality confidence assessment unit for the original signal The assessment results; this multiplicative relationship embodies the core principle of decision theory: the ultimate risk of an event is the product of its inherent severity and the certainty of evidence; for example, even if the system detects an event with an extremely high severity level... A pathological event with a high signal value, but if the original signal segment in which the event occurs has extremely poor quality. The value is extremely low, the final It will also be very small, thus effectively suppressing high-level alarms triggered by artifacts being misidentified as pathological events; The unit will activate the clinical alarm index. The system compares the data with preset multi-level thresholds to execute tiered alerts; for example, the system can preset three thresholds. ,when When a low-risk warning is triggered, When a medium-risk warning is triggered, High-risk emergency alarms are triggered at certain times; these thresholds are set based on the needs of clinical risk management and the risk tolerance for different levels of events, and are determined through statistical analysis of historical data or expert consultation. This embodiment creates a novel clinical alarm activation index by multiplying the severity score of a pathological event by the quality confidence score of the corresponding signal segment. This mechanism ensures that alarm decisions are based not only on what is detected, but also on how reliable the detection results are, thereby fundamentally solving the problem of clinical false alarms caused by signal artifacts. This greatly improves the specificity of the alarm system, reduces alarm fatigue, and allows medical staff to focus their attention on high-confidence, high-risk real clinical events, thus enhancing the clinical practicality and reliability of the entire monitoring system.

[0024] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An electrocardiogram (ECG) monitoring system based on the Internet of Things (IoT), characterized in that, include: The multimodal data synchronization unit is used to synchronously acquire the raw electrocardiogram signal and the dynamic contact impedance of the electrode skin interface at each sampling moment. The dynamic feature quantization unit is used to calculate the interface instability index based on the dynamic contact impedance acquired by the multimodal data synchronization unit, and to calculate the morphological deviation based on the original electrocardiogram signal. The quality confidence assessment unit is used to fuse the interface instability index and morphological deviation calculated by the dynamic feature quantization unit to generate signal quality confidence. An adaptive signal repair unit is used to dynamically repair the original ECG signal in response to the signal quality confidence generated by the quality confidence assessment unit, so as to generate a repaired ECG signal. And a clinical decision support unit, which combines the analysis results of the repaired ECG signal with the signal quality confidence level, outputs a clinical alarm activation index, and performs graded early warning; The dynamic feature quantization unit is specifically used for: Obtain the contact impedance measurement values ​​at the current and previous sampling times; Based on the current and previous sampling times of the contact impedance measurement, determine the instantaneous change in contact impedance; The instantaneous change is normalized by using a preset individualized reference impedance to obtain the normalized result. A nonlinear amplification factor is used to amplify the normalization result in order to generate an interface instability index. Among them, the individualized reference impedance is the impedance value measured in the resting state during the initial period of user wearing the device; The dynamic feature quantization unit is specifically used for: Extract morphological feature vectors from the current ECG waveform segment to be analyzed; Based on the preset standard waveform prior mean eigenvector and prior covariance matrix, calculate the Mahalanobis distance from the morphological eigenvector to the center of the standard waveform distribution. The squared Mahalanobis distance is set as the morphological deviation. The quality confidence assessment unit is specifically used for: Multiply the interface instability index by the preset independent influence weights to obtain the first attenuation term; The second attenuation term is obtained by multiplying the morphological deviation by the preset independent influence weights. Multiplying the product of the interface instability index and the morphological deviation by a preset coupling effect weight yields the coupling attenuation term. Summing the first attenuation term, the second attenuation term, and the coupling attenuation term yields the negative penalty term; Based on the negative penalty term, the signal quality confidence is generated through exponential function operation; The clinical decision support unit is specifically used for: The repaired electrocardiogram signals were analyzed to extract pathological features; Based on the pre-defined scoring criteria in reference clinical guidelines, pathological features are quantified to generate a severity score for pathological events. The clinical decision support unit is also specifically used for: The severity score of the pathological event is multiplied by the confidence level of the signal quality to generate a clinical alarm activation index; The clinical alarm activation index is compared with preset multi-level thresholds to implement graded early warning.

2. The electrocardiogram monitoring system based on the Internet of Things according to claim 1, characterized in that, The adaptive signal repair unit is specifically used for: Subtract the value 1 from the signal quality confidence level to obtain the confidence level difference; The confidence difference is exponentially calculated using a preset sensitivity coefficient to obtain the exponentiation result. The result of the exponentiation operation is multiplied by the preset maximum repair strength limit to generate an adaptive repair strength factor.

3. The electrocardiogram monitoring system based on the Internet of Things according to claim 2, characterized in that, The adaptive signal repair unit is also specifically used for: The raw electrocardiogram signal is subjected to powerful filtering in parallel to generate a reference signal; Based on the adaptive repair intensity factor as the weight, the original ECG signal and the reference signal are linearly superimposed to generate the repaired ECG signal.

Images