Method and system for processing electrocardiogram signals

Abstract

This application discloses a method and system for processing electrocardiogram (ECG) signals, relating to the field of biomedical signal processing. The method comprises acquiring an ECG signal, segmenting the ECG signal into beats to obtain multiple beats, and obtaining the positional information of the QRS wave, T wave, and ST wave segment of each beat using the Pan-Tompkins algorithm; extracting the dynamic characteristics of each beat based on the positional information; and using a random forest algorithm to assess the importance of the dynamic characteristics, screening key features, and performing serial fusion to obtain comprehensive feature information. This application uses multi-dimensional feature extraction and fusion to more comprehensively capture subtle changes in complex ECG signals, thereby improving the accuracy and efficiency of ECG signal analysis.

Description

Technical Field

[0001] The present application relates to the field of biomedical signal processing, and in particular to a method and system for processing electrocardiogram (ECG) signals. Background Art

[0002] Early diagnosis and treatment of cardiovascular disease are crucial for reducing mortality and improving patients' quality of life. As an important indicator of cardiac electrical activity, electrocardiogram (ECG) signal analysis plays a key role in the diagnosis of cardiovascular disease.

[0003] Currently, ECG signal analysis primarily relies on traditional morphological metrics, such as waveform shape and vector direction. While these methods can provide some information about ECG activity, they are limited in processing complex and variable ECG signals. Related technologies primarily rely on basic features such as waveform angle and amplitude, which struggle to fully capture the complexity and diversity of ECG signals, limiting the extraction and analysis of deeper information about cardiac electrical activity.

[0004] Therefore, there is an urgent need for an ECG signal processing method that can extract valuable information from ECG activities more comprehensively and accurately. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for processing electrocardiogram signals, which can more comprehensively and accurately extract key features related to cardiac electrical activity and effectively integrate these features to reflect the complexity and diversity of cardiac electrical activity.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a method for processing an electrocardiogram signal, comprising:

[0008] Acquire ECG signals based on the Frank lead system;

[0009] The ECG signal is split into beats to obtain multiple beats, and the position information of the QRS wave, T wave, and ST wave of each beat is obtained using the Pan-Tompkins algorithm; each beat contains a preset number of sampling points;

[0010] Based on the position information of the QRS wave, T wave and ST wave segment of each heartbeat, the dynamic characteristics of each heartbeat are extracted respectively; the dynamic characteristics include multi-lead characteristic wave characteristics, plane projection characteristics and three-dimensional space characteristics;

[0011] The importance score of each dynamic feature relative to all dynamic features of the ECG signal is calculated using the random forest algorithm, and the dynamic features corresponding to those with a score greater than a preset importance threshold are regarded as key dynamic features;

[0012] Based on the importance scores corresponding to the key dynamic features, the key dynamic features are fused in series to obtain the comprehensive feature information of the ECG signal.

[0013] In a second aspect, the present application provides an electrocardiogram signal processing system, comprising:

[0014] Signal acquisition module, used to acquire ECG signals based on the Frank lead system;

[0015] The heartbeat segmentation and characteristic wave detection module is used to segment the ECG signal into multiple heartbeats and obtain the position information of the QRS wave, T wave and ST wave segment of each heartbeat using the Pan-Tompkins algorithm; each heartbeat contains a preset number of sampling points;

[0016] A feature extraction module is used to extract the dynamic characteristics of each heartbeat based on the position information of the QRS wave, T wave and ST wave segment of each heartbeat; the dynamic characteristics include multi-lead characteristic wave characteristics, plane projection characteristics and three-dimensional space characteristics;

[0017] A feature screening module is used to calculate the importance score of each dynamic feature relative to all dynamic features of the ECG signal using a random forest algorithm, and to select dynamic features corresponding to values greater than a preset importance score threshold as key dynamic features;

[0018] The feature fusion module is used to fuse the key dynamic features in series based on the importance scores corresponding to the key dynamic features to obtain the comprehensive feature information of the ECG signal.

[0019] According to the specific embodiments provided in this application, this application has the following technical effects:

[0020] The present application provides a method and system for processing electrocardiogram (ECG) signals. By using the Pan-Tompkins algorithm to obtain the R-wave peak position information of the ECG signal, the problem of the difficulty of accurately locating the R wave in the ECG signal preprocessing stage in traditional methods is solved, and accurate heartbeat segmentation of the ECG signal is achieved. By improving the Pan-Tompkins algorithm, a preset sliding window, a preset amplitude threshold, and a preset slope threshold are added to solve the accuracy problem when obtaining the T-wave and ST-wave segment position information, and more accurate identification of the QRS wave, T wave, and ST-wave segment position information is achieved. In addition, based on the position information of the QRS wave, T wave, and ST-wave of each heartbeat, dynamic features including multi-lead characteristic wave features, planar projection features, and three-dimensional spatial features are extracted, solving the problem of incomplete feature extraction in the existing technology and achieving a more comprehensive description of cardiac electrical activity. Finally, the feature importance score is calculated by the random forest algorithm, and the key dynamic features are fused in series based on the score, solving the problem of feature selection and fusion, achieving efficient extraction of comprehensive feature information of the ECG signal, and providing new technical support for in-depth analysis of ECG activity. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 This is a diagram of an application environment of an electrocardiogram signal processing method in one embodiment of the present application.

[0023] Figure 2 A flowchart of a method for processing an electrocardiogram signal provided in one embodiment of the present application is provided.

[0024] Figure 3 A schematic diagram of a single heartbeat waveform in a Frank lead system provided in one embodiment of the present application.

[0025] Figure 4 A three-dimensional space ring diagram of the electrocardiogram vector ring and a two-dimensional projection surface diagram thereof are provided in an embodiment of the present application; wherein, Figure 4 (a) Figure 4 (b) and Figure 4 (c) Schematic diagram of the amplitudes of the three leads X, Y, and Z of the ECG vector loop; Figure 4 (d) is the three-dimensional space ring diagram and two-dimensional projection surface diagram of the ECG vector ring.

[0026] Figure 5A diagram of ECG signal processing and characteristic wave detection provided in one embodiment of the present application.

[0027] Figure 6 A complete heartbeat waveform detection diagram provided in an embodiment of the present application; wherein, Figure 6 (a) is a complete heartbeat waveform detection diagram of a healthy individual; Figure 6 (b) is a complete cardiac waveform detection diagram of a patient with myocardial infarction.

[0028] Figure 7 A schematic diagram of the functional modules of an electrocardiogram signal processing system provided in one embodiment of the present application.

[0029] Figure 8 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0032] The method for processing electrocardiogram signals provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the terminal 102 communicates with the server 104 via a network. The data storage system can store data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. Terminal 102 can send an ECG signal acquired based on the Frank lead system to server 104. After receiving the ECG signal, server 104 acquires the ECG signal; performs beat segmentation on the ECG signal to obtain multiple beats, and obtains the position information of the QRS wave, T wave, and ST wave segment of each beat using the Pan-Tompkins algorithm; each beat includes a preset number of sampling points; based on the position information of the QRS wave, T wave, and ST wave segment of each beat, dynamic features of each beat are extracted; the dynamic features include multi-lead characteristic wave features, planar projection features, and three-dimensional spatial features; calculate the importance score of each dynamic feature relative to all dynamic features of the ECG signal using a random forest algorithm, and define the dynamic features corresponding to those greater than a preset importance score threshold as key dynamic features; based on the importance scores corresponding to the key dynamic features, the key dynamic features are serially integrated to obtain comprehensive feature information of the ECG signal. Server 104 can provide feedback to terminal 102 on the obtained comprehensive feature information of the ECG signal. In addition, in some embodiments, the method for processing ECG signals can also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 can directly process the ECG signals to be processed, or the server 104 can obtain the ECG signals to be processed from the data storage system and process the ECG signals to be processed.

[0033] Terminal 102 may include, but is not limited to, various desktop computers, laptops, smartphones, tablet computers, IoT devices, and portable wearable devices. IoT devices may include smart speakers, smart TVs, smart air conditioners, and smart car devices. Portable wearable devices may include smart watches, smart bracelets, and head-mounted devices. Server 104 may be implemented as a standalone server or a server cluster consisting of multiple servers, or may be a cloud server.

[0034] In an exemplary embodiment, Figure 2 As shown, a method for processing an electrocardiogram signal is provided. The method is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the process, including the following steps 201 to 205.

[0035] Step 201: Acquire an ECG signal based on the Frank lead system.

[0036] Step 202 , performing beat segmentation on the ECG signal to obtain multiple beats, and obtaining the position information of the QRS wave, T wave, and ST wave of each beat using the Pan-Tompkins algorithm; each beat contains a preset number of sampling points.

[0037] Step 203 , based on the position information of the QRS wave, T wave and ST wave of each heartbeat, respectively extract the dynamic characteristics of each heartbeat; the dynamic characteristics include multi-lead characteristic wave characteristics, plane projection characteristics and three-dimensional space characteristics.

[0038] Step 204 , calculating the importance score of each dynamic feature relative to all dynamic features of the ECG signal using a random forest algorithm, and taking the dynamic features corresponding to values greater than a preset importance score threshold as key dynamic features.

[0039] Step 205 : Based on the importance scores corresponding to the key dynamic features, the key dynamic features are serially fused to obtain comprehensive feature information of the ECG signal.

[0040] By implementing steps 201 to 205 above, the present application can effectively extract and fuse key dynamic features from ECG signals, thereby improving the accuracy and reliability of ECG signal analysis. By using the Pan-Tompkins algorithm and its improved version to accurately locate key waveforms in ECG signals, the present application can more comprehensively capture the complex changes in cardiac electrical activity. In addition, by evaluating and screening the importance of features through the random forest algorithm, the present application ensures that only the most relevant features are used for the final feature fusion, which helps to reduce noise and improve diagnostic accuracy. Ultimately, by fusing key dynamic features in series, the present application provides a comprehensive feature information that can be used to more accurately analyze and diagnose cardiovascular diseases, providing strong support for clinical decision-making.

[0041] In another exemplary embodiment of the present application, before step 202 , the step further includes: performing denoising processing on the ECG signal using a discrete wavelet transform method.

[0042] In another exemplary embodiment of the present application, step 202 specifically includes:

[0043] Step 301 : Perform median filtering and bandpass filtering on the heartbeat signal of each heartbeat to obtain a denoised heartbeat signal of each heartbeat.

[0044] like Figure 5As shown in the figure, median filtering and discrete wavelet decomposition methods are used to denoise the ECG signals to effectively remove noise interference. Subsequently, the Pan-Tompkins algorithm is used to locate the R-wave peak of the denoised ECG signal, and the signal is segmented according to the position of the R-wave peak. Specifically, with each R-wave peak as the center, 250 sampling points are taken forward and 400 sampling points are taken backward, resulting in a complete heartbeat containing 651 sampling points. This method can effectively preserve the morphological characteristics of the ECG signal, while providing a high-quality data foundation for subsequent heartbeat analysis and feature extraction.

[0045] In another exemplary embodiment of the present application, Figure 5 As shown, the ECG signal undergoes median and bandpass filtering to remove low- and high-frequency noise and improve signal quality. The bandpass filter removes both low- and high-frequency noise, preserving the primary signal components of the QRS complex. Typically, the bandpass filter frequency range is set to 0.5 Hz to 50 Hz. Low-frequency noise (such as baseline wander) typically occurs below 0.5 Hz, while high-frequency noise (such as myoelectric noise) typically occurs above 50 Hz. The signal is then passed through a median filter with cutoff frequencies of 5 Hz and 100 Hz to attenuate the T and P waves, allowing the QRS complex shape to be more clearly visualized.

[0046] Step 302: Perform differentiation and square operations on the denoised heartbeat signal of each heartbeat to obtain a squared heartbeat signal.

[0047] Step 303 : Determine the position corresponding to the sampling point in the squared heart beat signal that is greater than a preset R wave peak amplitude threshold and is within a preset time threshold range as the R wave peak position.

[0048] In step 304, the first local minimum before the R wave peak is determined as the Q wave starting point; the first local minimum after the R wave peak is determined as the S wave starting point; and the first local minimum after the first local maximum after the S wave starting point is determined as the S wave end point.

[0049] Use differential operations to highlight the rapid changes of the QRS wave in order to more accurately detect the QRS wave; use square operations to remove negative values in the signal so that all peaks are positive, thereby enhancing the peak of the R wave; the Q wave is the first local minimum before the R wave, and the S wave is the second local minimum after the R wave. Perform S wave detection within a specified time window after the R wave, limiting the search range to effectively identify the exact position of the S wave; adaptively detect the starting and ending points of the T wave based on the position and timing characteristics of the QRS wave. By analyzing the proportional relationship between the duration of the QRS wave and the T wave process, determine the position of the T wave end point, and determine the ST segment based on the QRS wave end point and the T wave starting point. The schematic diagram of a single heartbeat waveform in the Frank lead is shown in the figure below. Figure 3 shown.

[0050] Reference points in an ECG signal are the peak, onset, and offset locations of the waveform. The QRS complex, ST segment, and T wave all indicate the location of cardiac electrical activity. Therefore, they can be used to diagnose heart disease. Using the Pan-Tompkins algorithm, which already locates the QRS complex, and adaptively locates the T wave endpoint, this preprocesses the ECG signal and accurately detects the characteristic waves in all three Frank leads.

[0051] Step 305: Determine the minimum duration of the T wave and the maximum duration of the T wave.

[0052] As an optional implementation, the calculation formulas for the minimum T wave duration and the maximum T wave duration are respectively:

[0053] .

[0054] .

[0055] in, Indicates the minimum duration of the T wave; Indicates the actual total duration of the QRS complex; Indicates the maximum upper limit of the QRS complex duration; Indicates the minimum duration of the initial T wave; Indicates the maximum duration of the T wave; Indicates the maximum duration of the initial T wave.

[0056] Step 306: The S wave end position is used as the starting point of the first preset sliding window, and the position corresponding to the sampling point within the first preset sliding window that is greater than the first preset T wave amplitude threshold and the first preset T wave slope threshold is used as the starting point of the T wave; the size of the first preset sliding window is twice the maximum duration of the T wave.

[0057] Step 307: The starting point of the T wave is used as the midpoint of a second preset sliding window, and the positions corresponding to sampling points within the second preset sliding window that are greater than a second preset T wave amplitude threshold and a second preset T wave slope threshold are used as first screening positions for the T wave; the size of the second preset sliding window is twice the minimum duration of the T wave;

[0058] Step 308: The position obtained by adding the QRS wave end position to the minimum T wave duration is used as the second screening position of the T wave.

[0059] Step 309: Select the position corresponding to the sampling point with the largest time between the first and second screening positions as the end position of the T wave. Select the point with the later time, i.e., the one with the larger value, as the final optimized T wave end position.

[0060] In step 310 , the end point of the S wave and the start point of the T wave are used as the start point and end point of the ST segment, respectively.

[0061] The Pan-Tompkins algorithm is used to realize the positioning of the adaptive T wave starting and ending points on the basis of QRS wave positioning. Then, the position of the ST segment can be determined by the QRS wave ending point and the T wave starting point, thereby completing the preprocessing of the ECG signal and accurately detecting the characteristic waves of the ECG signal.

[0062] In the Pan-Tompkins algorithm, the QRS complex exhibits rapid rise and fall in cardiac electrical activity, compared to the more gradual changes in the P and T waves. Differentiation is used to enhance the rapid changes in the QRS complex. This significantly amplifies the amplitude of the QRS complex while effectively reducing the interference from the P and T waves, thereby highlighting the unique characteristics of the QRS complex within the complex ECG signal. The differentiated signal is then squared. This operation converts negative values in the signal to positive values, making all peaks appear positive, significantly enhancing the peak value of the QRS complex and particularly highlighting the characteristic features of the R wave. The QRS complex plays a central role in the conduction of cardiac electrical activity and possesses unique morphological and temporal characteristics. Based on these characteristics, the PT (Pan-Tompkins) algorithm filters the processed signal by setting time and amplitude thresholds to identify possible QRS complexes.

[0063] As an optional embodiment, the T wave represents the ventricular repolarization process, which normally lasts 0.10-0.25 seconds. Ventricular depolarization time and the repolarization process are somewhat correlated. Therefore, in this embodiment, T wave adaptive adjustment is initiated when a QRS wave is detected to exceed 0.11 seconds.

[0064] As an optional implementation, taking the end point of the QRS wave as the starting point, in view of the fact that the duration of the ST band is usually about 0.15 seconds, this embodiment selects 0.15 seconds after the end point of the QRS wave as the starting point, and extends 0.15 seconds forward and backward to define the first preset sliding window, which is used as the candidate range of the T wave starting point. The first preset T wave amplitude threshold and the first preset T wave slope threshold are used for screening. By counting the signal amplitudes within the candidate range of the T wave starting point, the mean and standard deviation are calculated, and then the average amplitude is used as the first preset T wave amplitude threshold. Similarly, the signal slope values within the candidate range of the T wave starting point are counted, the mean and standard deviation are calculated, and the average slope is set as the first preset T wave slope threshold. By calculating the amplitude mean and amplitude standard deviation within the candidate range, the first point in the candidate area of the T wave starting point that is greater than these two values is screened and determined as the T wave starting point.

[0065] As an optional implementation, the first preset T wave amplitude threshold is:

[0066] .

[0067] .

[0068] in, represents the first preset T wave amplitude threshold, that is, the average value of the amplitudes of all sampling points in the first preset sliding window; N represents the total number of sampling points in the first preset sliding window; represents the signal amplitude of the nth sampling point in the first preset sliding window, Represents the standard deviation of the signal amplitude, which is used to measure the dispersion of the ECG signal amplitude within the candidate range.

[0069] The first preset T wave slope threshold is:

[0070] .

[0071] .

[0072] in, represents the first preset T wave slope threshold, that is, the average value of the slopes of all sampling points in the first preset sliding window; represents the slope of the nth sampling point in the first preset sliding window, Represents the slope standard deviation, which is used to measure the dispersion of the signal slope within the candidate range.

[0073] In this embodiment, after determining the starting point of the T wave, in order to accurately screen out the T wave end point, the following optimization strategy can be adopted. Based on the minimum duration of the T wave of 0.1 seconds, the interval is extended by 0.1 seconds before and after the starting point of the T wave, and the interval thus delineated is used as the second preset sliding window, that is, the candidate range of the T wave end point. Within the candidate range of the T wave end point, the above-mentioned method based on amplitude and slope thresholds is used to calculate the second preset T wave amplitude threshold and the second preset T wave slope threshold to perform preliminary screening of the T wave end point. Specifically, the threshold conditions set by combining the mean, standard deviation and relevant statistical characteristics of the amplitude and slope are used to analyze the electrocardiogram signals within the candidate range of the T wave end point point by point, thereby determining the preliminarily screened T wave end point.

[0074] The detection of Q, S, R, and T waves in each heart beat was completed, and the waveform of the detection results was drawn using a visualization tool, such as Figure 6 As shown in the figure, the distribution visualizes the QRS wave, ST wave and T wave. Figure 6 shown; among them, Figure 6 (a) is a waveform detection diagram of a complete heartbeat signal of a healthy individual; Figure 6 (b) is a waveform detection diagram of a complete heartbeat signal of a patient with myocardial infarction. Figure 6 (a) and Figure 6 In (b), the yellow background on the far left is the QRS wave, the orange background in the middle is the ST wave, and the green background on the far right is the T wave detected by the adaptive T wave detection method. The method of combining differentiation and squaring is used to locate the characteristic band, and the threshold window method is used to accurately detect the Q point and S point. The T wave is detected by the adaptive detection method. Figure 6 The position distribution of each wave point can be clearly seen in the figure, which verifies the effectiveness of the algorithm. Figure 6 (a) and Figure 6 (b) The differences in heartbeat signals between healthy individuals and myocardial infarction patients can be observed, further verifying the applicability and accuracy of the method in different situations.

[0075] In another exemplary embodiment of the present application, the dynamic information of different dimensions in the ECG signal can reflect the dynamic changes of the heart's electrical activity. For example, the maximum vector change of the QRS loop exceeds the healthy range, the loop is affected by the change of the heart's electrical activity, causing the ECG vector to be asymmetric, and the disappearance of the ECG vector at a certain position will cause the area to completely lose normal electrical activity, and the overall electrical vector will be biased toward the place where the ECG vector disappears, thereby causing a change in dynamic information. Feature changes are mainly concentrated in the QRS loop, T loop, and ST band. Therefore, the length of the QRS loop and T loop, the maximum vector and angle of the QRS loop and T loop, the angle between the maximum vector of the QRS loop and T loop, the octant-related features, the mean of the ST band, and the projected surface area and area ratio are extracted to show the dynamic changes of the heart's electrical vector, and in order to evaluate the signal complexity from the QRS loop to the T loop and detect abnormal signals, the sample entropy for the QRS wave to the T wave is specially added. The features are divided into the following three dimensions:

[0076] The multi-lead characteristic wave features in step 203 specifically include: the signal amplitude mean of all sampling points in the ST band, the T wave peak value, and the sample entropy from the QRS wave to the T wave.

[0077] The maximum value of the T wave, that is, the T wave peak, usually indicates the direction of the strongest current during cardiac repolarization, that is, the dominant direction of cardiac repolarization. The T wave may appear inverted or flat, or even form an inconsistent distribution on the projection surface.

[0078] .

[0079] in, Indicates the peak value of the T wave, Indicates the first The maximum absolute value of the T wave potential among the sampling points is the T wave peak value.

[0080] The increase or decrease of the ST wave band is usually a sign of cardiac ischemia or other heart problems. Therefore, the average value of the ST wave band is calculated to reflect whether the heart is abnormal. The signal amplitude average of all sampling points in the ST wave band is calculated using the following formula:

[0081] .

[0082] in, Represents the mean signal amplitude of all sampling points in the ST band, Indicates the starting point of the ST wave segment. Indicates the end point of the ST wave. Indicates the ST wave The signal amplitude at each sampling point.

[0083] Sample entropy is a measure of the complexity and self-similarity of a time series. Smaller values indicate greater signal regularity, while larger values indicate greater signal complexity. The calculation of sample entropy relies on the time series of the data and a tolerance value, often called r, to assess the number of similar patterns in the sequence. Sample entropy measures the complexity of a time series by comparing the matching probabilities for template vectors of length m and m+1. Sample entropy from QRS to T waves is calculated using the following formula:

[0084] .

[0085] in, Represents the sample entropy from QRS wave to T wave; l Indicates the length of the template vector; r Indicates the preset tolerance value; Indicates the total number of sampling points from the starting point of the QRS wave to the end point of the T wave; Indicates r Under , the matching probability between any two vectors of length m from QRS wave to T wave; Indicates r Under this condition, the matching probability between any two vectors of length m+1 in QRS wave to T wave.

[0086] Approximate variance is an estimate of the variance of sample data, often used to describe the degree of dispersion or volatility of the data. The formula for calculating the approximate variance of the QRS wave to T wave is as follows.

[0087] .

[0088] in, Indicates the approximate variance from QRS wave to T wave; Indicates the total number of sampling points from the starting point of the QRS wave to the end point of the T wave. Indicates the first wave from QRS wave to T wave sampling points, Indicates the first The amplitude of the sampling points, Indicates the mean amplitude of all sampling points from the QRS wave to the T wave.

[0089] In wavelet transform, a signal is decomposed into components of different frequencies, where detail coefficients represent high-frequency information. The variance of detail coefficients is used to describe the variability or complexity of the high-frequency portion of the signal. Assuming that the sequence of sampling points from the QRS wave to the T wave is subjected to wavelet transform, the detail coefficients are obtained. The formula for calculating the variance of the detail coefficients is as follows:

[0090] .

[0091] in, Indicates the variance of the QRS wave to T wave detail coefficient; The sampling point sequence from QRS wave to T wave is transformed by wavelet transform. The detail coefficient of each sampling point, It represents the mean value of the detail coefficients of all sampling points after the sampling point sequence from QRS wave to T wave is transformed by wavelet transform.

[0092] The plane projection features of step 203 specifically include: the maximum QT vector angle, the maximum vector length, angle, and 12-quadrant area ratio of the projection surface loops of the QRS loop and the T loop on the two-dimensional plane; the QRS loop is formed by the potential values of the QRS wave of each heartbeat on the X, Y, and Z leads based on the Frank lead system; the T loop is formed by the potential values of the T wave of each heartbeat on the X, Y, and Z leads based on the Frank lead system.

[0093] The QRS complex for each heartbeat is formed by the potential values in leads X, Y, and Z of the Frank lead system. The specific steps are as follows: During each heartbeat, the potential values of the QRS complex are extracted from leads X, Y, and Z. These potential values reflect the voltage changes in the heart's electrical activity in the X, Y, and Z directions during the QRS complex. For each sampling point in a heartbeat, during the QRS complex, the potential values of leads X, Y, and Z are combined to form a three-dimensional QRS vector. The three-dimensional QRS vectors of all sampling points in each heartbeat are then connected in chronological order to form a closed loop, the QRS loop. This loop reflects the trajectory of the electrical activity of the QRS complex in three dimensions during a single heartbeat.

[0094] The T loop is formed in a similar process to the QRS loop, but focuses on the electrical activity during the T wave. During each heartbeat, the T wave potential values are extracted from leads X, Y, and Z. These potential values reflect the voltage changes in the heart's electrical activity in the X, Y, and Z directions during the T wave. For each sampling point in a heartbeat, during the T wave, the potential values of leads X, Y, and Z are combined into a three-dimensional T vector. The three-dimensional T vectors of all sampling points in each heartbeat are then connected in chronological order to form a closed loop, the T loop. This loop reflects the trajectory of the T wave's electrical activity in three dimensions during a single heartbeat.

[0095] Compared to the QRS loop, the T loop better reflects the spatial distribution and dynamic changes of ventricular repolarization. Analysis of the T loop's shape, size, and orientation can assess the synchrony of cardiac repolarization and the direction of the cardiac electrical axis. Furthermore, the morphology and orientation of the T loop may be associated with both the physiological state and pathological changes of the heart, thus possessing important application value in clinical diagnosis and research.

[0096] The three-dimensional space ring diagram of the electrocardiogram vector ring and its two-dimensional projection surface diagram are as follows: Figure 4 shown; among them, Figure 4 (a) Figure 4 (b) and Figure 4 (c) Schematic diagram of the amplitudes of the three leads X, Y, and Z of the ECG vector loop; Figure 4 (d) shows the three-dimensional spatial loop diagram and two-dimensional projection of the vector cardiogram (ECG) loop. The vector cardiogram (ECG) loop consists of the QRS loop and the T loop. Both loops are constructed by analyzing the potential variations of the ECG signal in three dimensions. They reflect the electrical activity during ventricular depolarization and repolarization, respectively. By comparing and analyzing the characteristics of these two loops, we can gain a more comprehensive understanding of the electrophysiological characteristics and functional status of the heart, providing important reference information for the diagnosis and treatment of heart diseases.

[0097] The maximum QT vector angle is calculated using the following formula:

[0098] .

[0099] in, Indicates the maximum vector angle of QT; Represents the cosine value of the maximum QT vector angle; V1 and V2 represent the first vector composed of the sampling points in the first time period between the starting point of the QRS complex and the end point of the T wave and the second vector composed of the sampling points in the second time period, respectively; and The QT maximum vector angle represents the first and second vector moduli, respectively. It measures the direction of QT interval change or cardiac electrical activity. The QT interval represents the duration of a complete cardiac electrical cycle (from ventricular depolarization to repolarization), while the maximum vector angle refers to the angle of change in the QT interval vector at different time points, describing the directional shift in cardiac electrical activity from depolarization to repolarization. Changes in this angle can, to a certain extent, reveal abnormal electrophysiological changes in the heart, indicating the occurrence of myocardial infarction (MI).

[0100] The maximum vector length of the projected surface loops of the QRS loop and T loop on the two-dimensional plane was calculated using the Euclidean distance formula.

[0101] Use the inverse tangent function to calculate the angle of the projected surface loop, the angle, and the ratio of the 12 quadrant areas.

[0102] Calculate the maximum distance between two points in the projection of the QRS loop and T loop on a two-dimensional plane. Use arctan2 to calculate angles in different quadrants.

[0103] .

[0104] in, The projection plane of QRS loop and T loop on the two-dimensional plane is Point and The maximum distance between points, that is, the Euclidean distance between two points. and The projection plane of QRS loop and T loop on the two-dimensional plane is The horizontal and vertical coordinates of the point, and The projection plane of QRS loop and T loop on the two-dimensional plane is The horizontal and vertical coordinates of the point, The projection plane of QRS loop and T loop on the two-dimensional plane is Point and The angle between the points, arctan2 is the inverse tangent function.

[0105] The 12-quadrant area ratio is a feature derived from clinical observations of the offset and shape changes of VCG (vector cardiogram) projections. The projected area is calculated using the Gaussian area formula, ensuring that each vertex is connected to the next. The area of the irregular shape can be calculated by taking the area of the triangles and then adding them up and dividing by 2.

[0106] Calculate the 12-quadrant area ratio using the following formula:

[0107] .

[0108] .

[0109] in, Indicates the The ratio of the area of each quadrant to the total area of the projection surface; =1,2,3.....,12; Indicates the quadrant area; Represents the total area of the projection surface; J represents the total number of sampling points in the projection surface; x0 and y0 represent the horizontal and vertical coordinates of the starting sampling point in the projection surface respectively; j and y j Respectively represent the horizontal and vertical coordinates of the j-th sampling point in the projection surface; x j+1 and y j+1 They represent the horizontal and vertical coordinates of the j+1th sampling point in the projection surface respectively.

[0110] The three-dimensional spatial features of step 204 specifically include: the length, maximum vector and angle of the QRS loop and T loop; the angle between the maximum vector of the QRS loop and the T loop; the octant value, octant time ratio, octant vector variance, and octant maximum vector length of the QRS loop and T loop.

[0111] The dynamics of ECG signals can reflect dynamic changes in cardiac electrical activity, such as changes in the signal conduction pathway or changes in the length of the conduction pathway. These changes may manifest as adjustments in the circuit morphology or shifts in the spatial distribution of electrical activity. By calculating the cumulative length from each point in the circuit to the next, the circuit length can be quantified to reflect the overall cardiac electrical activity. The formula is as follows:

[0112] .

[0113] Where I represents the total number of ECG vectors on the loop; 、 and Respectively represent The coordinate components of the electrocardiogram vector on the coordinate axes X, Y and Z; Indicates the maximum length of the loop; Indicates the The modulus of the electrocardiogram vector.

[0114] The eight equal regions of space, defined by the ECG vector loop, are constructed by dividing the spatial projection of cardiac electrical activity into eight regions to facilitate analysis of the direction and intensity distribution of cardiac electrical activity. Analysis of these regions through the ECG vector loop facilitates more detailed spatial localization and diagnosis of cardiac electrical activity. Dividing space into eight equal regions allows for clearer clinical assessment of cardiac electrical activity. Extracting features such as the mean, time scale, vector variance, and maximum vector length within each octant allows for understanding changes in cardiac electrical activity and assessing cardiac health.

[0115] Octant Mean: Each octant mean reflects the intensity and distribution of electrical activity in a specific direction or region of the VEC loop. It represents the central tendency or representative value for each region and helps describe the distribution of data within that region. Significant changes in the mean will alter the VEC loop's shape, providing a clearer picture of cardiac physiological status.

[0116] .

[0117] in, Indicates the The mean of a segment is the average value of all ECG vector values in the area, which is used to measure the central trend of electrical activity in the area. The first A separate area, ={1,2,....,8}, The value range of is 1 to 8, corresponding to 8 different regions or parts. Each part represents a specific area of cardiac electrical activity in space. Indicates the The total number of sampling points in a split, or The number of electrocardiographic vectors contained in a body. Indicates the The first ECG vector data points. Specifically, The range is 1 to , that is, to accumulate and sum all the data points in each segment, Indicates the The first The vector cardiogram value of the vector cardiogram data point.

[0118] Octant time ratio: The ratio measures the proportion of each area in the entire time period and can reflect the temporal distribution characteristics.

[0119] .

[0120] in, Indicates the The duration of a division is usually the duration of electrical activity in that division. Represents the total time, which is the sum of the time for the entire ECG vector loop data acquisition. Indicates the The time ratio of each split, a specific time period Total time It reflects the ratio of The proportion of a part or region in the total time.

[0121] Octant vector variance: Variance is used to measure the fluctuation of data points relative to the mean, reflecting the stability or uncertainty of each octant.

[0122] .

[0123] in, Indicates the The vector variance of the split, Indicates the The mean amplitude of each segment is the average amplitude of all ECG vector values in the area, which is used to measure the central trend of electrical activity in the area.

[0124] Octant maximum vector length: The maximum vector length represents the maximum extension of the data point in space and can reflect the spatial distribution range of the data point.

[0125] .

[0126] in, Indicates the The maximum length of the central electric vector of a split body, which is used to measure the maximum strength of the electric vector in the heart of the area. The calculation is the The first The magnitude of the ECG vector of the ECG vector data point.

[0127] When processing high-dimensional data, feature selection is a key step to improve classification accuracy and computational efficiency. This application uses a random forest method based on dynamic average threshold setting for feature selection. This method has strong feature evaluation capabilities and can effectively identify the most informative features from a large number of features. Random forest is an integrated learning method that performs classification or regression tasks by constructing multiple decision trees. Each decision tree only uses part of the features of the data during training, which helps to reduce overfitting and improve the generalization ability of the model. During the classification process, random forest calculates the importance of each feature, which is determined by analyzing the contribution of each feature to the model decision process. Specifically, random forest evaluates the relative importance of each feature by calculating the average split gain of each feature in all trees.

[0128] In another exemplary embodiment of the present application, the calculation formula for the preset importance score threshold is:

[0129] .

[0130] .

[0131] .

[0132] Among them, K represents the preset importance score threshold; represents the mean importance score; k represents an adjustable parameter used to control the strictness of the threshold; represents the standard deviation of importance scores; represents the importance score of the i-th dynamic feature, and m represents the number of dynamic features.

[0133] The different dimensional features screened by the random forest algorithm are concatenated in sequence in a specific order to form a new high-dimensional feature vector. This concatenation order is scientifically arranged based on the contribution of each feature to the overall ECG signal feature expression and the inherent logical relationship between the features of each dimension. Features that play a key role in reflecting the change law of the ECG vector and can significantly improve the richness of feature expression are given priority in concatenation. At the same time, the complementarity between features of different dimensions is fully considered to ensure that the concatenated feature vector can retain the effective information from each dimension to the greatest extent, and enhance the comprehensive expression ability of the multi-dimensional dynamic information of the ECG signal. The formula is as follows:

[0134] .

[0135] in, For comprehensive feature information, The first key dynamic feature, the second key dynamic feature and the key dynamic features.

[0136] The purpose of feature extraction and fusion based on multi-dimensional dynamic information of ECG signals is to improve the deficiencies of existing methods in expressing cardiac electrical activity and the problem of information redundancy. By extracting features that express the dynamic information of cardiac electrical activity from three dimensions: multi-lead characteristic waves, plane projections, and three-dimensional ECG vector distribution, the dynamic changes and spatial distribution characteristics of cardiac electrical activity are fully captured. Combined with feature screening based on statistical distribution and serial fusion technology, it can effectively remove redundant information and retain the most discriminative features, thereby constructing high-dimensional feature vectors and improving the ability to express cardiac electrical activity. The organic combination of multi-dimensional features not only makes up for the limitations of traditional methods in feature extraction, but also provides more comprehensive and accurate technical support for in-depth analysis and research on cardiac electrical activity.

[0137] The present application also provides an application scenario, which applies the above-mentioned method for processing ECG signals. Specifically: the method for processing ECG signals provided in this embodiment can be applied in a remote medical monitoring scenario. The remote medical monitoring scenario includes a data acquisition link, a data processing link, and a data analysis link; the data enters the data processing link from the data acquisition link, obtains corresponding feature information through ECG signal processing, and enters the data analysis link. The method for processing ECG signals provided in this embodiment belongs to the feature extraction sub-link in the data processing link. Specifically in the feature extraction sub-link, the method provides high-quality feature data for subsequent data analysis and clinical diagnosis by accurately identifying and extracting the position information of the QRS wave, T wave, and ST wave band in the ECG signal, and calculating and fusing key dynamic features.

[0138] Based on the same inventive concept, embodiments of the present application also provide an ECG signal processing system for implementing the ECG signal processing method described above. The solution provided by this system is similar to the solution described in the method described above. Therefore, the specific limitations of one or more ECG signal processing system embodiments provided below can be found in the limitations of the ECG signal processing method described above and will not be further elaborated here.

[0139] In an exemplary embodiment, Figure 7 As shown, a system for processing an electrocardiogram signal is provided, comprising:

[0140] The signal acquisition module 401 is used to acquire an ECG signal based on the Frank lead system.

[0141] The beat cutting and characteristic wave detection module 402 is used to perform beat cutting on the ECG signal to obtain multiple beats, and obtain the position information of the QRS wave, T wave and ST wave segment of each beat through the Pan-Tompkins algorithm; each beat contains a preset number of sampling points.

[0142] The feature extraction module 403 is used to extract the dynamic features of each heartbeat based on the position information of the QRS wave, T wave and ST wave segment of each heartbeat; the dynamic features include multi-lead characteristic wave features, plane projection features and three-dimensional space features.

[0143] The feature screening module 404 is configured to calculate the importance score of each dynamic feature relative to all dynamic features of the ECG signal using a random forest algorithm, and select dynamic features corresponding to values greater than a preset importance score threshold as key dynamic features.

[0144] The feature fusion module 405 is used to perform serial fusion of the key dynamic features based on the importance scores corresponding to the key dynamic features to obtain comprehensive feature information of the ECG signal.

[0145] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store electrocardiogram signal processing data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for processing electrocardiogram signals is implemented.

[0146] Those skilled in the art will understand that Figure 8 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the above-mentioned method embodiments when executing the computer program.

[0147] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0148] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0149] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0150] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0151] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0152] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0153] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for processing an electrocardiogram signal, characterized in that: The electrocardiogram signal processing method includes: Acquire ECG signals based on the Frank lead system; The electrocardiogram signal is subjected to beat segmentation to obtain multiple beats, and the position information of the QRS wave, T wave, and ST wave of each beat is obtained by using the Pan-Tompkins algorithm, specifically including: each beat contains a preset number of sampling points; the beat signal of each beat is subjected to median filtering and bandpass filtering to obtain a denoised beat signal of each beat; After performing differentiation and square operations on the denoised heartbeat signal of each heartbeat, a squared heartbeat signal is obtained; The position corresponding to the sampling point in the squared heartbeat signal that is greater than a preset R wave peak amplitude threshold and is within a preset time threshold range is determined as the R wave peak position; The first local minimum before the R wave peak is determined as the Q wave starting point; the first local minimum after the R wave peak is determined as the S wave starting point; the first local minimum after the first local maximum after the S wave starting point is determined as the S wave end point; Determine the minimum and maximum T wave durations. The formulas for calculating the minimum and maximum T wave durations are: ； ； in, Indicates the minimum duration of the T wave; Indicates the actual total duration of the QRS complex; Indicates the maximum upper limit of the QRS complex duration; Indicates the minimum duration of the initial T wave; Indicates the maximum duration of the T wave; Indicates the maximum duration of the initial T wave; The S wave endpoint is used as the starting point of a first preset sliding window, and the position corresponding to the sampling point within the first preset sliding window that is greater than a first preset T wave amplitude threshold and a first preset T wave slope threshold is used as the starting point of the T wave; the size of the first preset sliding window is twice the maximum duration of the T wave; The starting point of the T wave is used as the midpoint of a second preset sliding window, and the position corresponding to the sampling point within the second preset sliding window that is greater than a second preset T wave amplitude threshold and a second preset T wave slope threshold is used as the first screening position of the T wave; the size of the second preset sliding window is twice the minimum duration of the T wave; The position obtained by adding the QRS wave end position to the minimum T wave duration is used as the second screening position of the T wave; Select the position corresponding to the sampling point with the largest time between the first screening position and the second screening position as the end position of the T wave; The end point of the S wave and the starting point of the T wave are used as the starting point and end point of the ST segment respectively; Based on the position information of the QRS wave, T wave and ST wave of each heartbeat, dynamic features are extracted respectively; the dynamic features include multi-lead characteristic wave features, plane projection features and three-dimensional space features; The importance score of each dynamic feature relative to all dynamic features of the ECG signal is calculated using the random forest algorithm, and the dynamic features corresponding to those with a score greater than a preset importance threshold are regarded as key dynamic features; Based on the importance scores corresponding to the key dynamic features, the key dynamic features are fused in series to obtain the comprehensive feature information of the ECG signal.

2. The method for processing an electrocardiogram signal according to claim 1, wherein: The first preset T wave amplitude threshold is: ； in, represents the first preset T wave amplitude threshold; N represents the total number of sampling points in the first preset sliding window; represents the signal amplitude of the nth sampling point in the first preset sliding window; The first preset T wave slope threshold is: ； in, represents a first preset T wave slope threshold; Indicates the slope of the nth sampling point in the first preset sliding window.

3. The method for processing an electrocardiogram signal according to claim 1, wherein: The multi-lead characteristic wave features specifically include: the signal amplitude mean of all sampling points in the ST band, the T wave peak value, and the sample entropy from the QRS wave to the T wave.

4. The method for processing an electrocardiogram signal according to claim 1, wherein: The plane projection features specifically include: the maximum QT vector angle, the maximum vector length, angle, and 12-quadrant area ratio of the projection surface loops of the QRS loop and T loop on a two-dimensional plane; the QRS loop is formed by the potential values of the QRS wave of each heartbeat on the X, Y, and Z leads based on the Frank lead system; the T loop is formed by the potential values of the T wave of each heartbeat on the X, Y, and Z leads based on the Frank lead system.

5. The method for processing an electrocardiogram signal according to claim 1, wherein: The three-dimensional spatial features specifically include: the length, maximum vector and angle of the QRS loop and T loop; the angle between the maximum vector of the QRS loop and the T loop; the octant value, octant time ratio, octant vector variance, and octant maximum vector length of the QRS loop and T loop.

6. The method for processing an electrocardiogram signal according to claim 1, wherein: The calculation formula for the preset importance score threshold is: ； ； ； Where K represents the preset importance score threshold; represents the mean importance score; k represents an adjustable parameter; represents the standard deviation of importance scores; represents the importance score of the i-th dynamic feature, and m represents the number of dynamic features.

7. The method for processing an electrocardiogram signal according to claim 1, wherein: Perform beat segmentation on the ECG signal to obtain multiple beats. Previously, this also included: The discrete wavelet transform method is used to denoise the ECG signal.

8. A system for processing electrocardiogram signals, characterized in that: The electrocardiogram signal processing system applies the electrocardiogram signal processing method according to any one of claims 1 to 7, and the electrocardiogram signal processing system includes: Signal acquisition module, used to acquire ECG signals based on the Frank lead system; The heartbeat segmentation and characteristic wave detection module is used to segment the ECG signal to obtain multiple heartbeats and obtain the position information of the QRS wave, T wave and ST wave segment of each heartbeat through the Pan-Tompkins algorithm. Specifically, each heartbeat contains a preset number of sampling points; the heartbeat signal of each heartbeat is subjected to median filtering and bandpass filtering to obtain a denoised heartbeat signal of each heartbeat; After performing differentiation and square operations on the denoised heartbeat signal of each heartbeat, a squared heartbeat signal is obtained; The position corresponding to the sampling point in the squared heartbeat signal that is greater than a preset R wave peak amplitude threshold and is within a preset time threshold range is determined as the R wave peak position; The first local minimum before the R wave peak is determined as the Q wave starting point; the first local minimum after the R wave peak is determined as the S wave starting point; the first local minimum after the first local maximum after the S wave starting point is determined as the S wave end point; Determine the minimum and maximum T wave durations. The formulas for calculating the minimum and maximum T wave durations are: ； ； in, Indicates the minimum duration of the T wave; Indicates the actual total duration of the QRS complex; Indicates the maximum upper limit of the QRS complex duration; Indicates the minimum duration of the initial T wave; Indicates the maximum duration of the T wave; Indicates the maximum duration of the initial T wave; The S wave endpoint is used as the starting point of a first preset sliding window, and the position corresponding to the sampling point within the first preset sliding window that is greater than a first preset T wave amplitude threshold and a first preset T wave slope threshold is used as the starting point of the T wave; the size of the first preset sliding window is twice the maximum duration of the T wave; The starting point of the T wave is used as the midpoint of a second preset sliding window, and the position corresponding to the sampling point within the second preset sliding window that is greater than a second preset T wave amplitude threshold and a second preset T wave slope threshold is used as the first screening position of the T wave; the size of the second preset sliding window is twice the minimum duration of the T wave; The position obtained by adding the QRS wave end position to the minimum T wave duration is used as the second screening position of the T wave; Select the position corresponding to the sampling point with the largest time between the first screening position and the second screening position as the end position of the T wave; The end point of the S wave and the starting point of the T wave are used as the starting point and end point of the ST segment respectively; A feature extraction module is used to extract the dynamic characteristics of each heartbeat based on the position information of the QRS wave, T wave and ST wave segment of each heartbeat; the dynamic characteristics include multi-lead characteristic wave characteristics, plane projection characteristics and three-dimensional space characteristics; A feature screening module is used to calculate the importance score of each dynamic feature relative to all dynamic features of the ECG signal using a random forest algorithm, and to select dynamic features corresponding to values greater than a preset importance score threshold as key dynamic features; The feature fusion module is used to fuse the key dynamic features in series based on the importance scores corresponding to the key dynamic features to obtain the comprehensive feature information of the ECG signal.

Images