Method, device and equipment for evaluating electrocardiogram signal quality

Abstract

The embodiment of the present application discloses a method, device and equipment for evaluating the quality of an ECG signal, which relates to the field of ECG technology, and includes: obtaining an ECG signal to be evaluated, wherein the ECG signal includes multiple characteristic wave signals; extracting a parameter vector sequence of the ECG signal, wherein the parameter vector sequence includes multiple parameter vectors for characterizing the characteristic wave signals, and each characteristic wave signal corresponds to a parameter vector; extracting the quality features of the ECG signal according to the parameter vector sequence; evaluating the quality of the ECG signal according to the quality features, and obtaining the quality evaluation level of the ECG signal. The above method can solve the technical problems in the related art that the application scope of ECG signal quality evaluation is small and the quality evaluation results are not intuitive enough.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of electrocardiography technology, and in particular to an electrocardiography signal quality assessment method, device and equipment. Background Art

[0002] As society ages, the incidence and harmfulness of cardiovascular diseases continue to rise. Electrocardiogram (ECG) has been widely used as an important means of detecting cardiovascular status. Electrocardiogram (ECG) technology can be used in medical devices such as ECG machines, ECG monitors, and Holter monitors to monitor and analyze ECG signals.

[0003] With the development of intelligent technology, more and more wearable devices (such as smart watches) have also applied ECG technology to monitor the human body's ECG signals at any time. However, compared with medical devices, wearable devices are limited by factors such as power consumption, volume, cost, and complex usage scenarios, making the signal quality of the ECG signals they collect less stable. At this time, in order to ensure the accuracy of subsequent ECG signal analysis, it is necessary to first evaluate the quality of the collected ECG signals.

[0004] In the related art, the quality of ECG signals can be evaluated through frequency analysis (such as using fast Fourier transform, Welch method power spectrum calculation and other frequency analysis methods), time domain analysis (such as using wavelet analysis, empirical mode decomposition and other time domain analysis methods), nonlinear analysis (such as using Lyapunov exponents, entropy and other nonlinear analysis) and other methods. However, the above frequency analysis, time domain analysis and nonlinear analysis all reflect some statistical characteristics of the entire ECG signal, but ignore the specific waveform of the ECG signal in the time domain. Therefore, they cannot intuitively reflect the quality evaluation results of the ECG signal. In the related art, the quality of ECG signals can also be evaluated based on the matching of heartbeat position. However, the above method can only be applied to tasks related to heartbeat detection and cannot be applied to other tasks (such as tasks such as T wave amplitude measurement). Summary of the invention

[0005] The embodiments of the present application provide an electrocardiogram (ECG) signal quality assessment method, device, and apparatus to solve the technical problems in related arts that the application scope of ECG signal quality assessment is small and the quality assessment results are not intuitive enough.

[0006] In a first aspect, an embodiment of the present application provides a method for evaluating the quality of an electrocardiogram signal, comprising:

[0007] Acquiring an electrocardiogram signal to be evaluated, wherein the electrocardiogram signal includes a plurality of characteristic wave signals;

[0008] Extracting a parameter vector sequence of the electrocardiogram signal, wherein the parameter vector sequence includes a plurality of parameter vectors for characterizing the characteristic wave signal, and each characteristic wave signal corresponds to a parameter vector;

[0009] Extracting quality features of the electrocardiogram signal according to the parameter vector sequence;

[0010] The quality of the electrocardiogram signal is evaluated according to the quality feature, and a quality evaluation grade of the electrocardiogram signal is obtained.

[0011] In a second aspect, an embodiment of the present application further provides an electrocardiogram signal quality assessment device, comprising:

[0012] A signal acquisition module, used to acquire an electrocardiogram signal to be evaluated, wherein the electrocardiogram signal includes a plurality of characteristic wave signals;

[0013] A parameter extraction module, used for extracting a parameter vector sequence of the electrocardiogram signal, wherein the parameter vector sequence includes a plurality of parameter vectors for characterizing the characteristic wave signal, and each characteristic wave signal corresponds to a parameter vector;

[0014] A feature extraction module, used for extracting the quality features of the electrocardiogram signal according to the parameter vector sequence;

[0015] The quality assessment module is used to perform quality assessment on the ECG signal according to the quality feature and obtain the quality assessment grade of the ECG signal.

[0016] In a third aspect, the present application also provides an electrocardiogram signal quality assessment device, including:

[0017] one or more processors;

[0018] A memory for storing one or more programs;

[0019] When the one or more programs are executed by the one or more processors, the one or more processors implement the electrocardiogram signal quality assessment method as described in the first aspect.

[0020] The above-mentioned ECG signal quality assessment method, device and equipment obtain an ECG signal to be assessed that contains multiple characteristic wave signals, then extract a parameter vector sequence of the ECG signal to be assessed, each parameter vector in the parameter vector sequence corresponds to a characteristic wave signal, then extract the quality features of the ECG signal according to the parameter vector sequence, and perform quality assessment on the ECG signal according to the quality features to obtain a technical solution for the quality assessment level of the ECG signal, which solves the technical problems of the small application scope of ECG signal quality assessment in the related technology and the insufficient intuitiveness of the quality assessment results. By extracting the parameter vector of each special wave signal and extracting the quality features according to the parameter vector, it is possible to combine the specific waveform of the ECG signal in the time domain (i.e., the waveform of each characteristic wave signal) to obtain the quality features related to the waveform, and then obtain the quality assessment results that intuitively reflect the ECG signal. In addition, different quality assessment levels can be set in combination with actual needs, and the quality assessment level of the ECG signal can be obtained according to the quality features, which can meet the needs of different ECG signal analysis tasks, making the application scope of ECG signal quality assessment larger, more comprehensive and more detailed. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flowchart of an electrocardiogram signal quality assessment method provided in one embodiment of the present application;

[0022] Figure 2 A first schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0023] Figure 3 A second schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0024] Figure 4 A third schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0025] Figure 5 A fourth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0026] Figure 6 A fifth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0027] Figure 7 A flowchart of another method for evaluating the quality of an electrocardiogram signal provided in one embodiment of the present application;

[0028] Figure 8 A sixth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0029] Fig. 9 A seventh schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0030] Fig.10An eighth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0031] Fig.11 A ninth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0032] Fig.12 A tenth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0033] Fig.13 An eleventh schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0034] Fig.14 A twelfth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0035] Fig.15 A thirteenth schematic diagram of an electrocardiogram signal provided in accordance with an embodiment of the present application;

[0036] Fig.16 A fourteenth schematic diagram of an electrocardiogram signal provided in accordance with an embodiment of the present application;

[0037] Fig.17 A fifteenth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application;

[0038] Fig.18 A schematic diagram of the structure of an electrocardiogram signal quality assessment device provided in one embodiment of the present application;

[0039] Fig.19 A schematic diagram of the structure of an electrocardiogram signal quality assessment device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0040] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only the parts related to the present application, rather than all structures, are shown in the accompanying drawings.

[0041] The ECG signal quality assessment method provided in the embodiment of the present application can be performed by an ECG signal quality assessment device, which can be implemented by software and / or hardware, and the ECG signal quality assessment device can be composed of two or more physical entities, or can be composed of one physical entity, which is not limited in the embodiment. For example, the ECG signal quality assessment device can be a smart watch with an ECG, a portable ECG monitor, and other devices.

[0042] For example, Figure 1 A flowchart of a method for evaluating the quality of an electrocardiogram signal provided by an embodiment of the present application. Figure 1 , the ECG signal quality assessment method specifically includes:

[0043] Step 110: Acquire an electrocardiogram signal to be evaluated, where the electrocardiogram signal includes a plurality of characteristic wave signals.

[0044] Before and after the heart beats, the myocardium is excited, and during the process of myocardial excitement, a weak bioelectric current is generated. At this time, each cardiac cycle of the heart is accompanied by bioelectric changes. This bioelectric change can be transmitted to various parts of the body surface. Because the tissues of various parts of the body are different and the distances from the heart are different, the potentials of the electrocardiogram in different parts of the body are also different. For a normal heart, the direction, frequency, and intensity of this bioelectric change are regular. If the electrical signals of different parts of the body surface are detected by electrodes, amplified by amplifiers, and recorded by recorders, an electrocardiogram (ECG) can be obtained.

[0045] Since each beat of the heart is regular, the waveform in the electrocardiogram is also regular. The waveform in the electrocardiogram that can fully represent one cardiac cycle of the heart is called a heartbeat signal. A typical heartbeat signal consists of a P wave, a QRS complex and a T wave, wherein the P wave represents the electrical activity of atrial polarization, the QRS complex and the T wave represent the electrical activity of ventricular polarization and repolarization, respectively, and the first downward wave in the QRS complex is the Q wave, the upward wave is the R wave, and the next downward wave is the S wave. At this time, the P wave, Q wave, R wave, S wave and T wave can be considered as characteristic wave signals. Sometimes, the heartbeat signal also includes a U wave, which is a wide and low wave that appears 0.02s to 0.04s after the T wave. Generally speaking, an electrocardiogram records more than 100,000 heartbeat signals. In one embodiment, an electrocardiogram containing multiple heartbeats can be intercepted from the electrocardiogram, and the electrocardiogram contains multiple characteristic wave signals. For ECG signals, the horizontal axis is the time axis, which is measured in time, and the vertical axis is the strength of the ECG signal, which is represented by voltage. At this time, each point in the ECG signal has a corresponding time and intensity to represent the acquisition time and the collected voltage of the point. In clinical practice, the straight segment (TS segment) between the T wave and the S wave in the ECG signal is usually extended to obtain the time axis.

[0046] It can be understood that the ECG signal quality assessment device collects electrical signals (analog signals) related to myocardial excitation on the human body surface, wherein the acquisition means embodiment is not limited. Afterwards, the collected signals are subjected to impedance matching, filtering, amplification and other processing by analog circuits. Then the electrical signals are converted into digital signals by an analog-to-digital converter and stored by a memory. Generally speaking, the electrical signals collected from the human body surface are low-amplitude, low-frequency, weak physiological electrical signals measured by electrodes on the body surface, and their amplitude is usually between 0.1mv-5mv, and the energy is concentrated in the range of 0.5Hz-40Hz. They are very susceptible to the influence of the environment and various interferences (such as myoelectric interference, baseline drift, and power frequency interference, etc.). In order to avoid the influence of interference, the electrical signals need to be preprocessed. In one embodiment, the preprocessing may include removing low-frequency noise and power frequency noise, resampling, cutting into multiple electrical signals, and normalizing. After the ECG signal is intercepted from the electrocardiogram, the ECG signal is preprocessed to obtain the ECG signal to be evaluated.

[0047] Step 120: extract a parameter vector sequence of the ECG signal, where the parameter vector sequence includes a plurality of parameter vectors for characterizing the characteristic wave signals, and each characteristic wave signal corresponds to a parameter vector.

[0048] The parameter vector sequence is composed of multiple parameter vectors, each parameter vector corresponds to a characteristic wave signal in the ECG signal or a signal similar to the time domain waveform of the characteristic wave. The parameter vector represents the parameters of different dimensions of the characteristic wave signal, which is related to the waveform of the characteristic wave signal in the time domain. After the parameter vectors are arranged in time order, a parameter vector sequence representing the ECG signal can be obtained, that is, the parameter vector sequence can be considered as a parametric representation of the ECG signal.

[0049] The parameter type in the parameter vector can be selected according to the actual situation, such as the parameter vector includes the amplitude of the characteristic wave signal, the offset relative to the time axis, the position information and other parameter types. In one embodiment, when the ECG signal quality is evaluated, the parameter vector includes: one or more of the key points of the characteristic wave signal, the matching degree between the characteristic wave signal and the characteristic wave signal template, the width of the characteristic wave signal template, the position information of the characteristic wave signal, the amplitude of the characteristic wave signal and the offset of the characteristic wave signal relative to the time axis. In one embodiment, multiple key points can be set for each characteristic wave signal, and two adjacent key points have the same spacing on the horizontal axis. The number of key points can be set according to the actual situation, and the embodiment does not limit this. In one embodiment, a spline interpolation calculation is performed on multiple key points corresponding to each characteristic wave signal, and a standardized characteristic wave signal template with an amplitude of 1, a mean of 0, and within a certain width range can be obtained. After that, the matching degree between the characteristic wave signal template and the characteristic wave signal can be calculated, wherein the matching degree is used to represent the error between the characteristic wave signal template and the characteristic wave signal or the similarity of the waveforms of the two. The greater the matching degree, the smaller the error or the greater the similarity. Among them, the width of the characteristic wave signal template is set according to the clinical prior knowledge in the field of electrocardiographic analysis. For example, the width of a normal P wave is less than 0.1s, so the width of the corresponding characteristic wave signal template should be less than 0.1s. For another example, the width of a normal QRS complex is 0.06s-0.12s, so the width of the corresponding characteristic wave signal template should be between 0.06s-0.12s. In one embodiment, the position information of the characteristic wave signal refers to the position information of a certain point on the characteristic wave signal (such as a point with a maximum value of the ordinate) on the time axis. In one embodiment, the amplitude of the characteristic wave signal refers to the difference between the ordinates of the highest point and the lowest point of the characteristic wave signal, that is, the difference between the maximum voltage value and the minimum voltage value on the characteristic wave signal. In one embodiment, the offset of the characteristic wave signal relative to the time axis refers to the ordinate of the position where the mean of the characteristic wave signal is located.

[0050] In one embodiment, a parameter vector sequence can be obtained by a determined parameterized model. The parameterized model is used to parameterize the ECG signal to obtain the parameters in each parameter vector. The parameterized model can be obtained by training a large number of ECG signal samples. The ECG signal samples are ECG signals used to train the parameterized model. When determining the parameterized model, the model parameters of the parameterized model include relative position information of multiple key points used to reconstruct the ECG signal. Optionally, the relative position information of the key points can be optimized when the parameterized model is trained by a large number of ECG signal samples. The relative position information of the key points can determine the shape of the characteristic wave signal when reconstructing the characteristic wave signal. Exemplarily, the ECG signal is input into the parameterized model for processing, and parameters of different dimensions representing the characteristic wave signal can be obtained. The parameters are organized into the form of vectors to obtain the parameter vector of the characteristic wave signal, and then the parameter vector sequence of the ECG signal is composed of the parameter vectors.

[0051] Step 130: extracting quality features of the ECG signal according to the parameter vector sequence.

[0052] Exemplarily, the quality feature is a feature related to the quality assessment level, which can reflect the degree of interference of the ECG signal, the regularity of the waveform, etc. In one embodiment, the quality feature includes at least one of the characteristic wave signal category entropy, the ECG reconstruction signal signal-to-noise ratio, the characteristic wave signal peak value of the ECG signal to the signal energy ratio, the mean of the amplitude kurtosis, and the mean of the bias kurtosis.

[0053] Among them, the characteristic wave signal category entropy can intuitively reflect the disorder degree of the characteristic wave signal in the ECG signal, and it has a good recognition effect on the artifact noise, wherein the artifact noise refers to other noise signals in the ECG signal except the electrical signal generated by the heart beat. The greater the artifact noise interference, the higher the disorder degree of each characteristic wave signal in the ECG signal, and the smaller the artifact noise interference, the more ordered the characteristic wave signals in the ECG signal. Optionally, according to the key points of the characteristic wave signal in the vector parameters of each characteristic wave signal and the amplitude of the characteristic wave signal, the characteristic wave signals with smaller amplitude differences and the same key points are classified into one category, and then the characteristic wave signal category entropy can be obtained based on the probability of occurrence of each type of characteristic wave signal in the ECG signal.

[0054] The signal-to-noise ratio of the ECG reconstruction signal can directly reflect the degree of interference of white noise and high-frequency noise in the ECG signal on the waveform of the characteristic wave signal. The ECG reconstruction signal refers to the signal obtained by reconstructing the ECG signal according to the relative position information of the key points. Optionally, the signal-to-noise ratio of the ECG reconstruction signal can be calculated based on the ECG reconstruction signal and the noise signal between the ECG signal.

[0055] The ratio of the peak value of the characteristic wave signal of the ECG signal to the signal energy can intuitively reflect whether the characteristic wave signal is easy to distinguish in the ECG signal. Optionally, the amplitude of the characteristic wave signal and the width of the characteristic wave signal template in the parameter vector corresponding to the characteristic wave signal are obtained, and the ratio between the two parameters is calculated. Then, among the ratios corresponding to each characteristic wave signal, the characteristic wave signal corresponding to the maximum ratio is selected and the amplitude of the characteristic wave signal is obtained. The amplitude can be used as the signal peak value. Then, the average energy of the ECG signal is calculated, and the ratio of the characteristic wave signal peak value to the signal energy can be obtained according to the signal peak value and the average energy.

[0056] The mean of the amplitude kurtosis can intuitively reflect the tightness of the amplitude distribution of the characteristic wave signals belonging to the same category in the ECG signal. This quality feature has a good recognition effect on artifact noise. Optionally, the characteristic wave signals are classified according to the parameter vectors, wherein the classification basis of the characteristic wave signals can be set according to the actual situation. For example, if the key points of the two parameter vectors are the same, the characteristic wave signals corresponding to the two parameter vectors belong to one category. After classifying the characteristic wave signals, the number of characteristic wave signals of each category is obtained, and then the corresponding amplitude average value of each category is calculated according to the number of characteristic wave signals of each category and the amplitude of each characteristic wave signal, and then the corresponding amplitude fourth-order center distance and amplitude variance of each category are obtained. Then, the ratio of the fourth-order center distance of the amplitude to the square of the amplitude variance under the same category is taken as the amplitude kurtosis under the category, and then the average value of the amplitude kurtosis of each category is taken as the mean value of the amplitude kurtosis.

[0057] The mean of the bias kurtosis can directly reflect the tightness of the bias distribution of the characteristic wave signals belonging to the same category in the ECG signal, and has a good recognition effect on the baseline drift noise. Among them, the change or disturbance that causes the amplitude bias of the characteristic wave signal to change can be called baseline drift. Optionally, the characteristic wave signal is classified according to the parameter vector, wherein the classification basis of the characteristic wave signal can be set according to the actual situation. For example, if the key points of the two parameter vectors are the same, then the characteristic wave signals corresponding to the two parameter vectors belong to one category. Optionally, after the characteristic wave signal is classified, the number of characteristic wave signals of each category is obtained, and then the corresponding bias average value of each category is calculated according to the corresponding number of each category and the bias of the characteristic wave signal under each category relative to the time axis, and then the corresponding bias fourth-order center distance and bias variance of each category are obtained. Then, the ratio of the bias fourth-order center distance to the bias variance square under the same category is used as the variance kurtosis under the category, and then the average value of the variance kurtosis of each category is used as the mean value of the variance kurtosis.

[0058] It should be noted that in practical applications, other quality features can be extracted in combination with actual needs, such as calculating the category entropy of characteristic wave signals of different widths, the reconstruction error of characteristic wave signals of different widths, etc., and the embodiment does not limit this.

[0059] Step 140: Perform quality assessment on the ECG signal according to the quality feature, and obtain a quality assessment grade of the ECG signal.

[0060] The quality assessment level can be set according to the requirements of subsequent analysis tasks of the ECG signal. Different quality assessment levels can be applicable to different analysis tasks. In one embodiment, the quality assessment level includes at least two of: a morphological analysis level, a rhythm analysis level, a beat localization level, a reconfigurable denoising level, and an unavailable level. Among them, the morphological analysis level means that the ECG signal can be used for morphological analysis, that is, the nature of the ECG signal morphology can be analyzed. For example, Figure 2 A first schematic diagram of an electrocardiogram signal provided by an embodiment of the present application, Figure 2 The ECG signal shown is an ECG signal that meets the morphological analysis level and has a relatively regular waveform. Figure 2 When the two are similar, it can be considered that the quality assessment level of the ECG signal to be evaluated is the morphologically analyzable level. It is understandable that when symptoms such as myocardial infarction and myocardial ischemia occur, the morphology of the ECG signal will change significantly. When the ECG signal meets the morphologically analyzable level, by analyzing the morphology of the ECG signal, it is possible to determine whether symptoms such as myocardial infarction and myocardial ischemia have occurred. The rhythm analyzable level means that the ECG signal can be used for rhythm analysis, where rhythm refers to the origin site that controls the electrical activity of the heart. For example, the heart rhythm formed by the excitement of ectopic pacemakers is called ectopic rhythm. When premature beats occur, the ectopic pacemakers will send out premature impulses that cause heart beats. At this time, by analyzing the rhythm of the ECG signal, it is possible to determine whether premature beats have occurred. For example, Figure 3 A second schematic diagram of an electrocardiogram signal provided by an embodiment of the present application, Figure 3 The ECG signal shown is an ECG signal that meets the rhythm analysis level and has a clear rhythm. Figure 3 When the quality assessment level is similar, the rhythm analysis level is available. The beat location level means that the ECG signal can be used for beat location. Among them, a beat is composed of a continuous P wave, a QRS complex and a T wave. Beat location refers to identifying the position of the QRS complex in the ECG signal (i.e., the horizontal coordinate in the ECG signal). For example, Figure 4 A third schematic diagram of an electrocardiogram signal provided in an embodiment of the present application, Figure 4 The ECG signal shown is an ECG signal that meets the level of beat location, and its QRS complex is relatively clear. Figure 4 When the ECG signals are similar, the quality assessment level is the heart beat localization level. The reconstructable denoising level means that the ECG signal can be reconstructed and the noise can be removed, that is, the ECG signal can be used for subsequent analysis after reconstruction and denoising. For example, Figure 5 A fourth schematic diagram of an electrocardiogram signal provided in an embodiment of the present application, Figure 5 The ECG signal shown is an ECG signal that meets the reconstructible denoising level. Its waveform is relatively irregular. Through reconstruction and denoising, a relatively regular ECG waveform can be obtained. Figure 5 When the quality evaluation level is similar, the reconstructible denoising level is used. The unusable level means that the ECG signal cannot be used for subsequent analysis. In this case, the ECG signal can be deleted or discarded. For example, Figure 6 A fifth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application, Figure 6 The ECG signal shown is an ECG signal that meets the unusable level. Its waveform is very irregular and cannot be used. Figure 6 When the quality is similar, the quality assessment level is unusable.

[0061] Exemplarily, the quality features are analyzed to determine the quality assessment level to which the ECG signal belongs based on the quality features. In one embodiment, the quality features are identified using a machine learning method. Exemplarily, the ECG signal is classified according to the quality features using a machine learning classifier, and then the quality assessment level of the ECG feature is determined based on the classification result, wherein each quality assessment level can be used as a category. Among them, the machine learning classifier can be set according to the actual situation, for example, a machine learning classifier such as a support vector machine (SVM), a decision tree (Decision Tree) or a naive Bayesian model (NBM) is used for classification to achieve the evaluation of the quality assessment level of the ECG signal. Exemplarily, after obtaining the quality assessment level of the ECG signal, the ECG signal can be subsequently processed according to the quality assessment level.

[0062] In the above, by obtaining an ECG signal to be evaluated containing multiple characteristic wave signals, a parameter vector sequence of the ECG signal to be evaluated is extracted, each parameter vector in the parameter vector sequence corresponds to a characteristic wave signal, the quality feature of the ECG signal is extracted according to the parameter vector sequence, the quality of the ECG signal is evaluated according to the quality feature, and the technical solution of the quality evaluation level of the ECG signal is obtained, which solves the technical problems that the application scope of the ECG signal quality evaluation in the related technology is small and the quality evaluation result is not intuitive enough. By extracting the parameter vector of each special wave signal and extracting the quality feature according to the parameter vector, it is possible to combine the specific waveform of the ECG signal in the time domain (that is, the waveform of each characteristic wave signal) to obtain the quality feature related to the waveform, and then obtain the quality evaluation result that intuitively reflects the ECG signal. In addition, different quality evaluation levels can be set according to actual needs, and the quality evaluation level of the ECG signal can be obtained according to the quality feature, which can meet the needs of different ECG signal analysis tasks, making the application scope of ECG signal quality evaluation larger, more comprehensive and more detailed.

[0063] Figure 7 A flowchart of another method for evaluating the quality of an electrocardiogram signal provided by an embodiment of the present application. Figure 7 , the ECG signal quality assessment method includes:

[0064] Step 210: Acquire an electrocardiogram signal to be evaluated, where the electrocardiogram signal includes a plurality of characteristic wave signals.

[0065] Step 220: Determine a parameterized model, where model parameters of the parameterized model include relative position information of multiple key points used to reconstruct the electrocardiogram signal.

[0066] Exemplarily, the parameterized model is used to parameterize the ECG signal to obtain parameters for characterizing different dimensions of the ECG signal. The model parameters of the parameterized model include relative position information of multiple key points used to reconstruct the ECG signal, and the relative position information of the key points can be optimized by training the parameterized model with a large number of ECG signal samples. For example, 5 key points are set for each characteristic wave signal, and two adjacent key points have the same spacing on the horizontal axis. The relative position information of multiple key points can be determined by the position of each key point on the vertical axis.

[0067] In one embodiment, the training process of the parameterized model may be: obtaining an ECG signal sample, wherein the ECG signal sample includes a plurality of characteristic wave signal samples. The ECG signal sample is an ECG signal used to train the parameterized model, and the acquisition method thereof is not limited in the embodiment. A parameterized model is determined, wherein the model parameters include the relative position information of a plurality of key points used to reconstruct the ECG signal sample. When the model is just established, the relative position information of the key points is randomly generated, that is, the value of the ordinate of the key points is randomly generated. In one embodiment, the parameterized model includes a template generation module, a matching degree calculation module, and an amplitude and bias calculation module. Among them, the template generation module is used to perform spline interpolation calculation on the key points to generate a characteristic wave signal template; the matching degree calculation module is used to calculate the matching degree between the characteristic wave signal template and the characteristic wave signal sample; the amplitude and bias calculation module is used to calculate the amplitude of the characteristic wave signal sample and the bias relative to the time axis. Afterwards, the ECG signal sample is input into the parameterized model for processing to obtain a parameter vector sequence of the ECG signal sample. Afterwards, the ECG signal sample is reconstructed according to the parameter vector sequence to obtain a reconstructed signal sample. Exemplarily, each parameter vector in the parameter vector sequence is reconstructed to obtain a reconstructed characteristic wave signal sample, and the reconstructed characteristic wave signal sample is spliced ​​in the order of time to obtain a reconstructed signal sample. Afterwards, the error between the ECG signal sample and the reconstructed signal sample is calculated. Optionally, several points are selected from the ECG signal sample, and the mean square error between the ordinate values ​​of these points and the ordinate values ​​of the points corresponding to the positions on the reconstructed signal sample is calculated, and the mean square error is used as the error between the ECG signal sample and the reconstructed signal sample. It is understandable. The root mean square error and the mean absolute error of the ECG signal sample and the reconstructed signal sample can also be used as the error between the two, and the embodiment is not limited here. Afterwards, the relative position information of the key points in the parameterized model is updated according to the error. Exemplarily, the error is compared with the preset error. When it is greater than the preset error, the relative position information of the key points in the parameterized model is updated, and the ECG signal samples are continuously obtained to train the parameterized model until the calculated error is less than or equal to the preset error, and the model training is determined to be completed. Among them, the preset error can be set according to the actual situation. It can be understood that during training, the vertical coordinate value of the key point is used as the parameter to be optimized, the calculated error is used as the objective function, and then the gradient descent method is used to find the vertical coordinate value of the key point that makes the error less than or equal to the preset error.

[0068] Step 230: Input the ECG signal into the parameterized model for processing to obtain a parameter vector sequence of the ECG signal.

[0069] In one embodiment, the parameter vector sequence includes the parameter vectors of each characteristic wave signal in the ECG signal, that is, each characteristic wave signal in the ECG signal corresponds to a parameter vector. The parameter vector includes: the key point of the characteristic wave signal, the matching degree between the characteristic wave signal and the characteristic wave signal template, the width of the characteristic wave signal template, the position information of the characteristic wave signal, the amplitude of the characteristic wave signal and the offset of the characteristic wave signal relative to the time axis. In one embodiment, the total number of parameter vectors contained in the parameter vector sequence is recorded as N seq .

[0070] Exemplarily, a plurality of key points corresponding to each characteristic wave signal are subjected to spline interpolation calculation to obtain a standardized characteristic wave signal template with an amplitude of 1, a mean of 0, and a certain width. Among them, interpolation is to interpolate a continuous function on the basis of discrete data so that this continuous curve passes through all discrete points (i.e., multiple key points), and the approximate value of the function at other points can also be estimated. Spline interpolation is to determine a function between every two points (i.e., key points), and this function is a spline. Different functions have different splines. The width mentioned here is the duration of the characteristic wave signal. Optionally, the width is set according to clinical prior knowledge in the field of electrocardiographic analysis. For example, a plurality of key points of the characteristic wave signal are subjected to cubic spline interpolation to generate a spline wave signal with an amplitude of 1 and a mean of 0. After that, the spline wave signal is scaled to a preset width to obtain a standardized characteristic wave signal template.

[0071] After obtaining the characteristic wave signal template, the matching degree between the characteristic wave signal template and the characteristic wave signal can be calculated. Exemplarily, the correlation coefficient between the characteristic wave signal template and the characteristic wave signal is calculated as a matching degree sequence. Among them, the correlation coefficient is used to characterize the degree of correlation between the characteristic wave signal template and the characteristic wave signal. Optionally, the Pearson correlation coefficient is used to characterize the degree of correlation between the characteristic wave signal template and the characteristic wave signal to obtain a matching degree sequence, or the matching degree sequence is obtained by using Cosine similarity, the square of the Euclidean distance, etc. Afterwards, the maximum value is screened out from each matching degree in the above matching degree sequence. Exemplarily, a sliding window is used to slide over the matching degree sequence, and the width of the sliding window is at least greater than the width of the two matching degrees. For example, the width of the sliding window is equal to the width of the standardized characteristic wave signal template. When the sliding window slides over a certain area on the matching degree sequence, the local maximum value in the area is obtained, and non-maximum suppression is performed at the same time to reduce the non-maximum value in the sliding window (such as reduced to 0.1). After sliding across the entire area, the largest local maximum value among the local maximum values ​​obtained is used as the matching degree between the characteristic wave signal template and the characteristic wave signal.

[0072] Exemplarily, the position information of the characteristic wave signal on the time axis may be the position information of a certain point on the characteristic wave signal on the time axis, and the position information of the characteristic wave signal on the time axis can be determined based on the position information of the point. In one embodiment, in the aforementioned steps, the sliding window records the position of the local maximum during the sliding process, and at this time, the position of the largest local maximum can be used as the position information of the characteristic wave signal.

[0073] Exemplarily, the amplitude of the characteristic wave signal and the offset of the characteristic wave signal relative to the time axis can be calculated based on the least squares method. For example, assuming that the fitting function of the characteristic wave signal is f(x)=ax+b, taking several points on the characteristic wave signal, the horizontal coordinate values ​​of these points are xi, taking the points at the corresponding positions of the electrocardiogram signal to be evaluated, the vertical coordinate values ​​of these points are yi, and the function value of the corresponding fitting function is f(xi). The least squares method is used to calculate the sum of the squares of the differences between the two, and the calculation formula is: n is the total number of points on the characteristic wave signal. The parameters of the fitting function f(x) when the above formula reaches the minimum value are determined by iterative calculation, that is, the amplitude a of the characteristic wave signal and the offset b of the characteristic wave signal relative to the time axis are obtained.

[0074] Exemplarily, the relative position information of the key points corresponding to the characteristic wave signal obtained in the above steps, the matching degree between the characteristic wave signal and the characteristic wave signal template, the amplitude of the characteristic wave signal, the offset of the characteristic wave signal relative to the time axis, the width of the characteristic wave signal template and the position information of the characteristic wave signal are sorted into parameter vectors. Among them, the relative position information of the key points is the ordinate of each key point (because the spacing between two adjacent key points on the horizontal axis is equal). Then, the parameter vectors are arranged in the order of the position information of the characteristic wave signal to obtain the parameter vector sequence of the electrocardiogram signal.

[0075] Step 240: extracting quality features of the ECG signal according to the parameter vector sequence.

[0076] The quality features include characteristic wave signal category entropy, ECG reconstruction signal signal-to-noise ratio, ECG signal characteristic wave signal peak to signal energy ratio, amplitude kurtosis mean and bias kurtosis mean.

[0077] In one embodiment, the quality feature includes characteristic wave signal category entropy. In this case, step 240 includes: step 241-step 245:

[0078] Step 241: Filter out the parameter vector of the key characteristic wave signal from the parameter vector sequence.

[0079] The key characteristic wave signal refers to the characteristic wave signal that needs to be paid special attention to when evaluating the quality level, and the waveform characteristics of the key characteristic wave signal are relatively obvious. The key characteristic wave signal can be set according to the actual situation. For example, the key characteristic wave signal refers to a characteristic wave signal with a signal width between 0.06s and 0.20s and an absolute amplitude greater than 0.5mV. In one embodiment, a characteristic wave signal with a signal width between 0.06s and 0.20s and an absolute amplitude greater than 0.5mV is selected from the electrocardiogram signal as the key characteristic wave signal. Generally speaking, there are multiple key characteristic wave signals. Afterwards, in the parameter vector sequence, the parameter vector corresponding to the key characteristic wave signal is selected.

[0080] Step 242: cluster the selected parameter vectors according to a preset first clustering condition, where the first clustering condition is that the key points of the parameter vectors are the same and the difference in amplitude is less than a preset threshold.

[0081] The parameter vectors corresponding to each key characteristic wave signal are clustered to determine the parameter vectors belonging to the same category, thereby realizing the classification of the key characteristic wave signals. In one embodiment, the parameter vectors are clustered according to the first clustering condition. Among them, the first clustering condition can be set according to the actual situation. In one embodiment, the first clustering condition is that the key points of the parameter vectors are the same and the difference in amplitude is less than a preset threshold. Take the parameter vectors of any two key characteristic wave signals as an example for description. Exemplarily, the parameter vectors of any two key characteristic wave signals are obtained, and then the key points and amplitudes in the parameter vectors are obtained. It is judged whether the key points of the two parameter vectors are the same, and it is determined whether the difference in amplitude of the two parameter vectors is less than a preset threshold. Exemplarily, if the horizontal coordinates and vertical coordinates of the key points in the two parameter vectors are the same, it is determined that the key points in the two parameter vectors are the same. For example, if there are 5 key points, and the horizontal coordinates and vertical coordinates of the 5 key points in the two parameter vectors are the same, it is determined that the key points in the two parameter vectors are the same, otherwise, it is determined that the key points in the two parameter vectors are not the same. Exemplarily, the difference in amplitude of the characteristic wave signals in the two parameter vectors is calculated, and optionally, the absolute value of the difference is taken. Afterwards, the difference in amplitude is compared with a preset threshold, wherein the preset threshold can be set according to actual conditions. In one embodiment, the preset threshold is 1 mV. If the key points in the two parameter vectors are the same and the difference in amplitude is less than the preset threshold, the two parameter vectors are classified into one category. According to the above method, the parameter vectors with the same key points and the difference in amplitude less than the preset threshold can be found and classified into one category, thereby realizing clustering of the parameter vectors. After clustering, K (K≥2) categories can be obtained, and the numbers of the K categories are: 1, 2, ..., K.

[0082] Step 243: Count the first quantity of characteristic wave signals in each category.

[0083] Exemplarily, the number of parameter vectors contained in each category is counted, and the number is used as the number of characteristic wave signals (key characteristic wave signals) in the category. In one embodiment, the number counted in step 243 is recorded as the first number. At this time, each category has a corresponding first number.

[0084] Step 244: Calculate the occurrence probability of the characteristic wave signal in the corresponding category according to the first quantity.

[0085] In one embodiment, the probability of occurrence of a characteristic wave signal under a certain category refers to the probability of occurrence of the characteristic wave signal under this category in the ECG signal, and each category corresponds to an occurrence probability. Exemplarily, the calculation formula for the occurrence probability of the kth (1≤k≤K) category is: p k =n k / N seq , where the first number corresponding to the kth category is n k , the total number of parameter vectors contained in the parameter vector sequence is N seq , that is, the number of characteristic wave signals in the ECG signal is N seq , the probability of occurrence of the kth category is p k .

[0086] Step 245: Obtain the characteristic wave signal category entropy according to the occurrence probability corresponding to each category.

[0087] Entropy is a term in statistical physics and information theory, which can also be recorded as information entropy or Shannon entropy. The characteristic wave signal category entropy can be recorded as entropy class , the characteristic wave signal category entropy is determined by the probability of occurrence, which can reflect the disorder degree of the characteristic wave signal appearing in the ECG signal, and has a good recognition effect on the artifact noise. The larger the characteristic wave signal category entropy, the higher the disorder degree of the ECG signal, the larger the artifact noise in the ECG signal, and the more irregular the waveform of the ECG signal. The smaller the characteristic wave signal category entropy, the lower the disorder degree of the ECG signal, the smaller the artifact noise in the ECG signal, and the more regular the waveform of the ECG signal. For example, Figure 8 A sixth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application, Fig. 9 The seventh schematic diagram of an electrocardiogram signal provided by an embodiment of the present application. Figure 8 The ECG signal contains a lot of artifact noise, the waveform is relatively irregular, and the characteristic wave signal category entropy corresponding to the ECG signal is large. Fig. 9 The ECG signal contains less artifact noise, the waveform is more regular, and the characteristic wave signal category entropy corresponding to the ECG signal is smaller.

[0088] In one embodiment, the calculation formula of the characteristic wave signal category entropy is: Among them, entropy classrepresents the characteristic wave signal category entropy, K is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the first clustering condition, k represents the kth category, 1≤k≤K, p k Represents the probability of occurrence of the kth category.

[0089] In one embodiment, the quality feature includes the signal-to-noise ratio of the ECG reconstruction signal. In this case, step 240 includes: steps 246 to 248:

[0090] Step 246: reconstruct the ECG signal according to the parameter vector sequence to obtain an ECG reconstructed signal.

[0091] Exemplarily, the waveform of the characteristic wave signal can be determined according to the key points, and then the characteristic wave signal can be adjusted according to the parameters in the parameter vector, such as adjusting the amplitude of the characteristic wave signal, the offset of the characteristic wave signal relative to the time axis, and the position information on the time axis, so that the characteristic wave signal can be reconstructed to obtain a reconstructed ECG signal. In the embodiment, the reconstructed ECG signal is recorded as an ECG reconstruction signal.

[0092] Step 247: Calculate the noise signal between the ECG signal and the ECG reconstruction signal.

[0093] The noise signal can reflect the difference between the ECG signal and the ECG reconstruction signal. In one embodiment, the calculation formula of the noise signal is: sig noise =sig-sig rec Among them, sig represents the ECG signal to be evaluated, sig rec Represents the ECG reconstruction signal, sig noise Represents the noise signal. From the above formula, we can know that the noise signal can be obtained by subtracting the ECG reconstruction signal from the ECG signal.

[0094] Step 248: Calculate the signal-to-noise ratio of the ECG reconstruction signal according to the noise signal and the ECG reconstruction signal.

[0095] The signal-to-noise ratio of the ECG reconstruction signal can reflect the ratio of the ECG reconstruction signal to the noise signal. The signal-to-noise ratio of the ECG reconstruction signal can be recorded as SNR rec . The signal-to-noise ratio of the ECG reconstruction signal can be calculated by the noise signal and the ECG reconstruction signal. The signal-to-noise ratio of the ECG reconstruction signal can intuitively reflect the degree of interference of white noise and high-frequency noise in the ECG signal on the characteristic wave signal. The signal-to-noise ratio of the ECG reconstruction signal can evaluate the performance impact of white noise and high-frequency noise on tasks such as T wave amplitude measurement in the ECG signal. Among them, white noise refers to random noise with the same energy density at the frequency, and high-frequency noise refers to noise with a main frequency higher than the set frequency. In one embodiment, the lower the signal-to-noise ratio of the ECG reconstruction signal, the greater the interference of white noise and high-frequency noise in the ECG signal. The higher the signal-to-noise ratio of the ECG reconstruction signal, the smaller the interference of white noise and high-frequency noise in the ECG signal. For example, Fig.10 An eighth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application, Fig.11 The ninth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application. Fig.10 , the ECG signal contains a lot of white noise, that is, there is white noise interference, so the ECG signal waveform has more burrs, and the ECG reconstruction signal corresponding to the ECG signal has a small signal-to-noise ratio. Fig.11 The ECG signal contains less white noise, the ECG signal is relatively clean, and the ECG reconstruction signal corresponding to the ECG signal has a relatively large signal-to-noise ratio.

[0096] In one embodiment, the calculation formula of the signal-to-noise ratio of the ECG reconstruction signal is: Among them, SNR rec Represents the signal-to-noise ratio of the ECG reconstruction signal, sig rec Represents the ECG reconstruction signal, sig noise Represents the noise signal. Among them, sig rec 2 Represents the square of the ordinate of each point in the ECG reconstruction signal, sig noise 2 Represents the square of the ordinate of each point in the noise signal.

[0097] In one embodiment, the quality feature includes the peak value of the characteristic wave signal of the electrocardiogram signal and the signal energy ratio. In this case, step 240 includes: steps 249 to 2412:

[0098] Step 249, calculating a first ratio corresponding to each parameter vector in the parameter vector sequence, where the first ratio is a ratio of the amplitude of the characteristic wave signal in the parameter vector to the width of the characteristic wave signal template.

[0099] In one embodiment, the amplitude of the characteristic wave signal and the width of the characteristic wave signal template recorded in each parameter vector in the parameter vector sequence are obtained, and then the ratio of the amplitude to the width in the same parameter vector is calculated, and the ratio is recorded as the first ratio. At this time, each parameter vector corresponds to a first ratio.

[0100] Step 2410: Find the largest first ratio among all first ratios, and take the amplitude corresponding to the largest first ratio as the signal peak value.

[0101] The larger the first ratio is, the larger the amplitude of the ECG signal corresponding to the parameter vector is relative to the width of the characteristic wave signal template. In one embodiment, among the first ratios, the first ratio with the largest ratio is selected. Then, the parameter vector corresponding to the first ratio with the largest ratio is obtained, and the amplitude in the parameter vector is determined as the signal peak value of the ECG signal to be evaluated, where the peak value can be understood as the maximum instantaneous value.

[0102] Step 2411: Calculate the average energy of the ECG signal.

[0103] For example, the average energy is calculated as: avg =∑sig 2 / N sig Among them, Energy avg represents the average energy of the ECG signal to be evaluated, sig represents the ECG signal to be evaluated, N sig Indicates the number of sampling points contained in the ECG signal to be evaluated. 2 Represents the square of the ordinate of each point in the ECG signal to be evaluated.

[0104] Step 2412: Obtain the ratio of the peak value of the characteristic wave signal of the ECG signal to the signal energy according to the signal peak value and the average energy.

[0105] The ratio of the characteristic wave signal peak to the signal energy can intuitively reflect whether the characteristic wave signal is easy to distinguish in the ECG signal. This feature can evaluate the performance of the ECG signal for the task of heartbeat positioning. The ratio of the characteristic wave signal peak to the signal energy can be denoted as PSR. The ratio of the characteristic wave signal peak to the signal energy can reflect the ratio of the signal peak to the average energy. In one embodiment, the higher the ratio of the characteristic wave signal peak to the signal energy, the more regular and easier to distinguish the characteristic wave signal in the ECG signal, and the cleaner the ECG signal. The lower the ratio of the characteristic wave signal peak to the signal energy, the more irregular and difficult to distinguish the ECG signal, and the greater the interference of the ECG signal. For example, Fig.12 The tenth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application is Fig.13 The eleventh schematic diagram of an electrocardiogram signal provided by an embodiment of the present application. Fig.12 The ECG signal has a large interference, its signal waveform is relatively messy and difficult to distinguish, and the peak value and signal energy of the characteristic wave signal are relatively small. Fig.13 The ECG signal is relatively regular, the characteristic wave signals are obvious and easy to distinguish, the ECG signal is relatively clean, and the characteristic wave signal peak and signal energy are relatively large.

[0106] In one embodiment, the calculation formula of the characteristic wave signal peak value and signal energy ratio is: Among them, PSR represents the ratio of the peak value of the characteristic wave signal of the ECG signal to the signal energy, w peak Indicates the signal peak value of the ECG signal to be evaluated, Energy avg Represents the average energy of the ECG signal to be evaluated.

[0107] In one embodiment, the quality feature includes the mean of the amplitude kurtosis. In this case, step 240 includes: step 2413-step 2418:

[0108] Step 2413: Filter out the parameter vector of the key characteristic wave signal from the parameter vector sequence.

[0109] This step may refer to step 241 and will not be described in detail here.

[0110] Step 2414: cluster the selected parameter vectors according to a preset second clustering condition, where the second clustering condition is that the key points of the two parameter vectors are the same.

[0111] The parameter vectors corresponding to each key characteristic wave signal are clustered to determine the parameter vectors belonging to the same category, thereby realizing the classification of the key characteristic wave signals. In one embodiment, the parameter vectors are clustered according to the second clustering condition. Among them, the second clustering condition can be set according to the actual situation. In one embodiment, the second clustering condition is that the key points of the parameter vectors are the same. Take the parameter vectors of any two key characteristic wave signals as an example for description. Exemplarily, the parameter vectors of any two key characteristic wave signals are obtained, and then it is determined whether the key points in the two parameter vectors are the same. If the horizontal coordinates and the vertical coordinates of the key points in the two parameter vectors are the same, it is determined that the key points in the two parameter vectors are the same, otherwise, it is determined that the key points in the two parameter vectors are not the same. If the key points in the two parameter vectors are the same, the two parameter vectors are classified into one category. According to the above method, the parameter vectors with the same key points can be found and classified into one category, thereby realizing the clustering of the parameter vectors. After clustering, M (M≥2) categories can be obtained, and the numbers of the M categories are: 1, 2, ..., M.

[0112] Step 2415: Count the second number of characteristic wave signals in each category.

[0113] Exemplarily, the number of parameter vectors contained in each category is counted, and the number is used as the number of characteristic wave signals (key characteristic wave signals) in the category. In one embodiment, the number counted in step 2415 is recorded as the second number. At this time, each category has a corresponding second number.

[0114] Step 2416: Calculate the average amplitude value of the characteristic wave signal in each category according to the second quantity and the amplitude value of the characteristic wave signal in the parameter vector, and calculate the fourth-order center distance and amplitude variance of the characteristic wave signal in each category according to the amplitude average value.

[0115] Exemplarily, the amplitude of the characteristic wave signal in the parameter vector under each category is obtained, and the average value of the amplitude of the characteristic wave signal under each category is calculated in combination with the second quantity corresponding to each category. In one embodiment, the sum of the amplitudes of the characteristic wave signals under the same category is calculated, and then the quotient of the sum and the second quantity is calculated, and the quotient is used as the amplitude average value of the characteristic wave signal under the category. At this time, each category corresponds to an amplitude average value.

[0116] The k-order central moment is a parameter commonly used in probability theory or statistics. For example, the fourth-order central distance refers to the central distance with k being 4. The amplitude fourth-order central distance refers to the fourth-order central distance with the amplitude of the characteristic wave signal as a variable, and each category corresponds to an amplitude fourth-order central distance. In one embodiment, the amplitude fourth-order central distance is calculated as follows: Among them, E w,m represents the fourth-order center distance of the amplitude of the mth category, 1≤m≤M, M is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the second clustering condition, is the average amplitude value corresponding to the mth category, w m,i is the amplitude of the i-th characteristic wave signal under the m-th category, 1≤i≤n m , n m is the second quantity corresponding to the m-th category.

[0117] Exemplarily, the amplitude variance can also be recorded as the amplitude second-order center distance, which refers to the second-order center distance with the amplitude of the characteristic wave signal as a variable. Each category corresponds to an amplitude variance. In one embodiment, the calculation formula of the amplitude variance is: Among them, D w,m represents the amplitude variance of the mth category, 1≤m≤M, M is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the second clustering condition, is the average amplitude value corresponding to the mth category, w m,i is the amplitude of the i-th characteristic wave signal under the m-th category, 1≤i≤n m , n m is the second quantity corresponding to the m-th category.

[0118] Step 2417: Calculate the amplitude kurtosis of the characteristic wave signal under the same category according to the fourth-order center distance of the amplitude and the amplitude variance under the same category.

[0119] Kurtosis can characterize the characteristic number of the peak height of the probability density distribution curve at the mean value. Intuitively, kurtosis reflects the sharpness of the peak. In one embodiment, the amplitude kurtosis can reflect the sharpness of the amplitude of the characteristic wave signal under the same category. Exemplarily, the ratio of the fourth-order center distance of the amplitude of a certain category to the amplitude variance is taken as the amplitude kurtosis under the category. In one embodiment, the calculation formula of the amplitude kurtosis is: Among them, kurtosis w,m represents the amplitude kurtosis of the mth category, 1≤m≤M, M is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the second clustering condition, E w,m represents the fourth-order center distance of the magnitude of the mth category, D w,m represents the amplitude variance of the mth category. At this time, each category corresponds to an amplitude kurtosis.

[0120] Step 2418: Calculate the first average value of the amplitude kurtosis corresponding to each category, and use the first average value as the mean value of the amplitude kurtosis.

[0121] Exemplarily, after determining the amplitude kurtosis of the characteristic wave signal under each category, the average value of each amplitude kurtosis can be calculated, and the average value can be used to reflect the average amplitude sharpness of the ECG signal. In one embodiment, the average value of the amplitude kurtosis is recorded as the first average value. The first average value can intuitively reflect the compactness of the amplitude distribution of the characteristic wave signals belonging to the same category in the ECG signal, and has a good recognition effect on artifact noise. The mean of the first average value can be recorded as kurtosis w In one embodiment, the lower the first average value, the more regular and clean the ECG signal is, and the higher the first average value, the greater the interference of the ECG signal and the less regular it is. For example, Fig.14 The twelfth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application is Fig.15 The thirteenth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application. Fig.14 The ECG signal has a large interference, its signal waveform is relatively messy, and the first average value (the mean of the amplitude kurtosis) is relatively large. Fig.15 The ECG signal is relatively regular and clean, and the first average value (the mean of the amplitude kurtosis) is relatively small.

[0122] The calculation formula of the first average value is: Among them, kurtosis w Indicates the first mean, kurtosis w,m represents the amplitude kurtosis of the mth category, 1≤m≤M, and M is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the second clustering condition. After calculating the first average value, the first average value is used as the mean value of the amplitude kurtosis.

[0123] In one embodiment, the quality feature includes the mean of the bias kurtosis. In this case, step 240 includes: step 2419-step 2424:

[0124] Step 2419: Filter out the parameter vector of the key characteristic wave signal from the parameter vector sequence.

[0125] This step may refer to step 241 and will not be described in detail here.

[0126] Step 2420: cluster the selected parameter vectors according to a preset third clustering condition, where the third clustering condition is that the key points of the parameter vectors are the same.

[0127] The parameter vectors corresponding to each key characteristic wave signal are clustered to determine the parameter vectors belonging to the same category, thereby realizing the classification of the key characteristic wave signals. In one embodiment, the parameter vectors are clustered according to the third clustering condition. Among them, the third clustering condition can be set according to the actual situation. In one embodiment, the third clustering condition is that the key points of the parameter vectors are the same. At this time, the third clustering condition is the same as the second clustering condition. The technical means for clustering the parameter vectors based on the third clustering condition can refer to the clustering of the parameter vectors based on the second clustering condition, and the embodiment will not go into details. After clustering, M (M≥2) categories can be obtained, and the numbers of the M categories are: 1, 2, ..., M.

[0128] Step 2421: Count the third quantity of characteristic wave signals in each category.

[0129] Exemplarily, the number of parameter vectors contained in each category is counted, and the number is used as the number of characteristic wave signals (key characteristic wave signals) in the category. In one embodiment, the number counted in step 2421 is recorded as the third number. At this time, each category has a corresponding third number. When the third clustering condition is the same as the second clustering condition, the second number in each category is the same as the third number. It can be understood that when two quality features, the mean of the bias kurtosis and the mean of the amplitude kurtosis, exist at the same time, clustering can be performed only once and the number of characteristic wave signals in each category can be counted once.

[0130] Step 2422: Calculate the biased average value of the characteristic wave signal under each category according to the bias of the characteristic wave signal relative to the time axis in the third quantity and parameter vector, and calculate the biased fourth-order center distance and biased variance of the characteristic wave signal under each type according to the biased average value.

[0131] Exemplarily, the offset of the characteristic wave signal in the parameter vector under each category relative to the time axis is obtained, and the average value of the offset of the characteristic wave signal under each category relative to the time axis is calculated in combination with the third quantity corresponding to each category. In one embodiment, the sum of the offsets of each characteristic wave signal under the same category relative to the time axis is calculated, and then the quotient of the sum and the third quantity is calculated, and the quotient is used as the offset average value of the characteristic wave signal under the category. At this time, each category corresponds to a bias average value.

[0132] The offset fourth-order center distance refers to the fourth-order center distance with the offset of the characteristic wave signal relative to the time axis as a variable, and each category corresponds to a offset fourth-order center distance. In one embodiment, the offset fourth-order center distance is calculated as follows: Among them, E b,m represents the offset fourth-order center distance of the mth category, 1≤m≤M, M is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the third clustering condition, is the average bias value corresponding to the mth category, b m,i is the offset of the i-th characteristic wave signal under the m-th category relative to the time axis, 1≤i≤n m , n m is the third quantity corresponding to the mth category.

[0133] Exemplarily, the bias variance can also be recorded as the bias second-order center distance, which refers to the second-order center distance with the bias of the characteristic wave signal relative to the time axis as a variable, and each category corresponds to a bias variance. In one embodiment, the calculation formula of the bias variance is: Among them, D b,m represents the bias variance of the mth category, 1≤m≤M, M is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the third clustering condition, is the average bias value corresponding to the mth category, b m,i is the bias of the i-th characteristic wave signal under the m-th category, 1≤i≤n m , n m is the third quantity corresponding to the mth category.

[0134] Step 2423, calculate the bias kurtosis of the characteristic wave signal under the same category according to the bias fourth-order center distance and bias variance under the same category.

[0135] In one embodiment, the bias kurtosis can reflect the sharpness of the bias of the characteristic wave signal relative to the time axis under the same category. Exemplarily, the ratio of the bias fourth-order center distance and the bias variance of a certain category is used as the bias kurtosis under the category. In one embodiment, the calculation formula of the bias kurtosis is: Among them, kurtosis b,m represents the bias kurtosis of the mth category, 1≤m≤M, M is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the third clustering condition, E b,m Denotes the offset fourth-order center distance of the mth category, D b,m represents the bias variance of the mth category. At this time, each category corresponds to a bias kurtosis.

[0136] Step 2424: Calculate the second average value of the bias kurtosis corresponding to each category, and use the second average value as the mean value of the bias kurtosis.

[0137] The mean of the bias kurtosis can directly reflect the tightness of the bias distribution of characteristic wave signals belonging to the same category in the ECG signal, and has a good recognition effect on baseline drift noise. The mean of the bias kurtosis can be recorded as kurtosis b. Exemplarily, after determining the bias kurtosis of the characteristic wave signal under each category, the average value of each bias kurtosis can be calculated, and the average value is used to reflect the average of the bias of each characteristic wave signal in the ECG signal relative to the time axis. In one embodiment, the average value of the bias kurtosis is recorded as the second average value. In one embodiment, the lower the second average value, the smaller the baseline drift noise in the ECG signal, and the more regular and clean the ECG signal. The higher the second average value, the greater the baseline drift noise in the ECG signal, and the more irregular the ECG signal. For example, Fig.16 A fourteenth schematic diagram of an electrocardiogram signal provided in one embodiment of the present application, Fig.17 The fifteenth schematic diagram of an electrocardiogram signal provided by an embodiment of the present application. Fig.16 The ECG signal has a large baseline drift noise, its signal waveform is relatively messy, and the second average value (the mean of the bias kurtosis) is relatively large. Fig.17 The baseline drift noise of the ECG signal is small, the ECG signal is relatively regular and clean, and the second average value (the mean of the bias kurtosis) is relatively small.

[0138] The calculation formula of the second average value is: Among them, kurtosis b represents the second mean, kurtosis b,m represents the bias kurtosis of the mth category, 1≤m≤M, and M is the total number of categories obtained after clustering the parameter vector of the key characteristic wave signal according to the third clustering condition. After calculating the second average value, the second average value is used as the mean value of the bias kurtosis.

[0139] Step 250: Perform quality assessment on the ECG signal according to the quality feature, and obtain a quality assessment grade of the ECG signal.

[0140] Among them, the quality assessment levels include: morphological analysis level, rhythm analysis level, heart beat localization level, reconfigurable denoising level and unusable level.

[0141] In the above, by obtaining an ECG signal to be evaluated containing multiple characteristic wave signals, a parameter vector sequence of the ECG signal to be evaluated is obtained using a parameterized model, each parameter vector in the parameter vector sequence corresponds to a characteristic wave signal, the quality features of the ECG signal are extracted according to the parameter vector sequence, the quality of the ECG signal is evaluated according to the quality features, and the technical solution of the quality evaluation level of the ECG signal is obtained, which solves the technical problems that the application scope of the ECG signal quality evaluation in the related technology is small and the quality evaluation results are not intuitive enough. By extracting the parameter vector of each special wave signal through the parameterized model, and then directly using these parameter vectors to extract the quality features, it is possible to combine the specific waveform of the ECG signal in the time domain (that is, the waveform of each characteristic wave signal) to obtain the quality features related to the waveform, and then obtain the quality evaluation results that intuitively reflect the ECG signal. And. Different quality evaluation levels can be set in combination with actual needs, and corresponding quality features can be set, which can meet the needs of different ECG signal analysis tasks, making the application scope of ECG signal quality evaluation larger, more comprehensive and more detailed, and the calculation method of the quality features is simple and easy to apply.

[0142] Fig.18 A schematic diagram of the structure of an electrocardiogram signal quality assessment device provided by an embodiment of the present application. Fig.18 The electrocardiogram signal quality assessment device includes: a signal acquisition module 301, a parameter extraction module 302, a feature extraction module 303, and a quality assessment module 304.

[0143] Among them, the signal acquisition module 301 is used to obtain the ECG signal to be evaluated, and the ECG signal includes multiple characteristic wave signals; the parameter extraction module 302 is used to extract the parameter vector sequence of the ECG signal, and the parameter vector sequence includes multiple parameter vectors for characterizing the characteristic wave signals, and each characteristic wave signal corresponds to a parameter vector; the feature extraction module 303 is used to extract the quality characteristics of the ECG signal according to the parameter vector sequence; the quality assessment module 304 is used to perform quality assessment on the ECG signal according to the quality characteristics and obtain the quality assessment level of the ECG signal.

[0144] In one embodiment, the parameter vector includes: the key points of the characteristic wave signal, the matching degree between the characteristic wave signal and the characteristic wave signal template, the width of the characteristic wave signal template, the position information of the characteristic wave signal, the amplitude of the characteristic wave signal and the offset of the characteristic wave signal relative to the time axis.

[0145] In one embodiment, the parameter extraction module 302 includes: a model determination unit, used to determine a parameterized model, the model parameters of the parameterized model include relative position information of multiple key points used to reconstruct the ECG signal; a signal processing unit, used to input the ECG signal into the parameterized model for processing to obtain a parameter vector sequence of the ECG signal, the parameter vector sequence includes multiple parameter vectors for characterizing characteristic wave signals, and each characteristic wave signal corresponds to a parameter vector.

[0146] In one embodiment, the quality feature includes characteristic wave signal category entropy, and the feature extraction module 303 includes: a first screening unit, used to screen out parameter vectors of key characteristic wave signals from a parameter vector sequence; a first clustering unit, used to cluster the screened parameter vectors according to a preset first clustering condition, and the first clustering condition is that the key points of the parameter vectors are the same and the difference in amplitude is less than a preset threshold; a first statistical unit, used to count the first number of characteristic wave signals under each category; a probability calculation unit, used to calculate the occurrence probability of the characteristic wave signal under the corresponding category according to the first number; an entropy calculation unit, used to obtain the characteristic wave signal category entropy according to the occurrence probability corresponding to each category.

[0147] In one embodiment, the quality feature includes the signal-to-noise ratio of the ECG reconstruction signal, and the feature extraction module 303 includes: a signal reconstruction unit, used to reconstruct the ECG signal according to the parameter vector sequence to obtain the ECG reconstruction signal; a noise calculation unit, used to calculate the noise signal between the ECG signal and the ECG reconstruction signal; and a signal-to-noise ratio calculation unit, used to calculate the signal-to-noise ratio of the ECG reconstruction signal based on the noise signal and the ECG reconstruction signal.

[0148] In one embodiment, the quality feature includes the ratio of the peak value of the characteristic wave signal of the ECG signal to the signal energy, and the feature extraction module 303 includes: a ratio calculation unit, which is used to calculate the first ratio corresponding to each parameter vector in the parameter vector sequence, and the first ratio is the ratio of the amplitude of the characteristic wave signal in the parameter vector to the width of the characteristic wave signal template; a peak search unit, which is used to find the largest first ratio among all the first ratios, and take the amplitude corresponding to the largest first ratio as the signal peak; an energy calculation unit, which is used to calculate the average energy of the ECG signal; and an energy ratio calculation unit, which is used to obtain the ratio of the peak value of the characteristic wave signal of the ECG signal to the signal energy based on the signal peak value and the average energy.

[0149] In one embodiment, the quality feature includes the mean of the amplitude kurtosis, and the feature extraction module 303 includes: a second screening unit, used to screen out the parameter vector of the key characteristic wave signal from the parameter vector sequence; a second clustering unit, used to cluster the screened parameter vector according to a preset second clustering condition, and the second clustering condition is that the key points of the parameter vector are the same; a second statistical unit, used to count the second number of characteristic wave signals under each category; a first center distance calculation unit, used to calculate the amplitude average value of the characteristic wave signal under each category according to the second number and the amplitude of the characteristic wave signal in the parameter vector, and calculate the fourth-order center distance of the amplitude and the amplitude variance of the characteristic wave signal under each category according to the amplitude average value; a first kurtosis calculation unit, used to calculate the amplitude kurtosis of the characteristic wave signal under the same category according to the fourth-order center distance of the amplitude and the amplitude variance under the same category; a first mean calculation unit, used to calculate the first average value of the amplitude kurtosis corresponding to each category, and use the first average value as the mean value of the amplitude kurtosis.

[0150] In one embodiment, the quality feature includes the mean of the bias kurtosis, and the feature extraction module 303 includes: a third screening unit, used to screen out the parameter vector of the key characteristic wave signal from the parameter vector sequence; a third clustering unit, used to cluster the screened parameter vector according to a preset third clustering condition, and the third clustering condition is that the key points of the parameter vector are the same; a third statistical unit, used to count the third number of characteristic wave signals under each category; a second center distance calculation unit, used to calculate the biased average value of the characteristic wave signal under each category according to the third number and the bias of the characteristic wave signal in the parameter vector relative to the time axis, and calculate the biased fourth-order center distance and bias variance of the characteristic wave signal under each category according to the biased average value; a second kurtosis calculation unit, used to calculate the bias kurtosis of the characteristic wave signal under the same category according to the biased fourth-order center distance and bias variance under the same category; a second mean calculation unit, used to calculate the second average value of the bias kurtosis corresponding to each category, and use the second average value as the mean of the bias kurtosis.

[0151] In one embodiment, the quality assessment levels include: a morphology analyzable level, a rhythm analyzable level, a heart beat localization level, a reconfigurable denoising level, and an unusable level.

[0152] The ECG signal quality assessment device provided above can be used to execute the ECG signal quality assessment method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0153] It is worth noting that in the embodiment of the above-mentioned ECG signal quality assessment device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application.

[0154] Fig.19 A schematic diagram of the structure of an electrocardiogram signal quality assessment device provided in one embodiment of the present application. Fig.19 As shown, the ECG signal quality assessment device includes a processor 40, a memory 41, an input device 42, and an output device 43; the number of the processor 40 in the ECG signal quality assessment device can be one or more. Fig.19 A processor 40 is taken as an example. The processor 40, the memory 41, the input device 42, and the output device 43 in the electrocardiogram signal quality assessment device can be connected via a bus or other means. Fig.19 The example of connecting through bus is taken in the following.

[0155] The memory 41, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the ECG signal quality assessment method in the embodiment of the present application (for example, the signal acquisition module 301, the parameter extraction module 302, the feature extraction module 303, and the quality assessment module 304 in the ECG signal quality assessment device). The processor 40 executes various functional applications and data processing of the ECG signal quality assessment device by running the software programs, instructions and modules stored in the memory 41, that is, realizes the above-mentioned ECG signal quality assessment method.

[0156] The memory 41 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the electrocardiogram signal quality assessment device, etc. In addition, the memory 41 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 41 may further include a memory remotely arranged relative to the processor 40, and these remote memories may be connected to the electrocardiogram signal quality assessment device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0157] The input device 42 can be used to receive input digital or character information, and generate key signal input related to user settings and function control of the ECG signal quality assessment device. The output device 43 can include display devices such as a display screen. The ECG signal quality assessment device can also include a collection device, which can collect ECG signals.

[0158] The above-mentioned ECG signal quality assessment device includes an ECG signal quality assessment apparatus, which can be used to execute any ECG signal quality assessment method and has corresponding functions and beneficial effects.

[0159] Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products.

[0160] Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0161] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0162] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0163] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0164] Note that the above are only preferred embodiments of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A method for evaluating the quality of an electrocardiogram signal, characterized in that: include: Acquiring an electrocardiogram signal to be evaluated, wherein the electrocardiogram signal includes a plurality of characteristic wave signals; Extracting a parameter vector sequence of the electrocardiogram signal, wherein the parameter vector sequence includes a plurality of parameter vectors for characterizing the characteristic wave signal, each characteristic wave signal corresponds to a parameter vector, and the parameter vector includes: a key point of the characteristic wave signal, a matching degree between the characteristic wave signal and a characteristic wave signal template, a width of the characteristic wave signal template, position information of the characteristic wave signal, an amplitude of the characteristic wave signal, and an offset of the characteristic wave signal relative to a time axis; Extracting quality features of the ECG signal according to the parameter vector sequence, the quality features comprising at least one of characteristic wave signal category entropy, ECG reconstruction signal signal-to-noise ratio, characteristic wave signal peak to signal energy ratio of the ECG signal, mean value of amplitude kurtosis, and mean value of bias kurtosis; The quality of the ECG signal is evaluated according to the quality characteristics, and a quality evaluation level of the ECG signal is obtained, wherein the quality evaluation level includes: a morphological analysis level, a rhythm analysis level, a beat localization level, a reconfigurable denoising level and an unusable level.

2. The electrocardiogram signal quality assessment method according to claim 1, characterized in that: The parameter vector sequence of extracting the electrocardiogram signal comprises: Determining a parameterized model, wherein model parameters of the parameterized model include relative position information of a plurality of key points used to reconstruct the electrocardiogram signal; The electrocardiogram signal is input into the parameterized model for processing to obtain a parameter vector sequence of the electrocardiogram signal.

3. The electrocardiogram signal quality assessment method according to claim 1, characterized in that: The quality features include characteristic wave signal category entropy, The extracting the quality feature of the electrocardiogram signal according to the parameter vector sequence comprises: Filtering out parameter vectors of key characteristic wave signals from the parameter vector sequence; Clustering the selected parameter vectors according to a preset first clustering condition, wherein the first clustering condition is that the key points of the parameter vectors are the same and the difference in amplitude is less than a preset threshold; Count the first number of characteristic wave signals under each category; Calculating the occurrence probability of characteristic wave signals under the corresponding category according to the first quantity; The characteristic wave signal category entropy is obtained according to the occurrence probability corresponding to each category.

4. The electrocardiogram signal quality assessment method according to claim 1, characterized in that: The quality features include the signal-to-noise ratio of the ECG reconstruction signal, The extracting the quality feature of the electrocardiogram signal according to the parameter vector sequence comprises: Reconstructing the ECG signal according to the parameter vector sequence to obtain an ECG reconstruction signal; Calculating a noise signal between the ECG signal and the ECG reconstruction signal; The signal-to-noise ratio of the ECG reconstruction signal is calculated according to the noise signal and the ECG reconstruction signal.

5. The method for evaluating the quality of an electrocardiogram signal according to claim 1, wherein: The quality feature includes the peak value of the characteristic wave signal of the electrocardiogram signal and the signal energy ratio, The extracting the quality feature of the electrocardiogram signal according to the parameter vector sequence comprises: Calculating a first ratio corresponding to each parameter vector in the parameter vector sequence, wherein the first ratio is a ratio of the amplitude of the characteristic wave signal in the parameter vector to the width of the characteristic wave signal template; Finding the largest first ratio among all the first ratios, and taking the amplitude corresponding to the largest first ratio as the signal peak value; Calculating the average energy of the electrocardiogram signal; The ratio of the characteristic wave signal peak value to the signal energy of the electrocardiogram signal is obtained according to the signal peak value and the average energy.

6. The method for evaluating the quality of an electrocardiogram signal according to claim 1, characterized in that: The quality characteristics include the mean of the amplitude kurtosis, The extracting the quality feature of the electrocardiogram signal according to the parameter vector sequence comprises: Filtering out parameter vectors of key characteristic wave signals from the parameter vector sequence; Clustering the selected parameter vectors according to a preset second clustering condition, where the second clustering condition is that the key points of the parameter vectors are the same; Counting the second quantity of characteristic wave signals under each category; Calculating the amplitude average of the characteristic wave signal in each category according to the second quantity and the amplitude of the characteristic wave signal in the parameter vector, and calculating the amplitude fourth-order center distance and amplitude variance of the characteristic wave signal in each category according to the amplitude average; Calculating the amplitude kurtosis of the characteristic wave signal under the same category according to the amplitude fourth-order center distance and the amplitude variance under the same category; A first average value of the amplitude kurtosis corresponding to each category is calculated, and the first average value is used as the mean value of the amplitude kurtosis.

7. The method for evaluating the quality of an electrocardiogram signal according to claim 1, characterized in that: The quality characteristics include the mean of the bias kurtosis, The extracting the quality feature of the electrocardiogram signal according to the parameter vector sequence comprises: Filtering out parameter vectors of key characteristic wave signals from the parameter vector sequence; Clustering the selected parameter vectors according to a preset third clustering condition, wherein the third clustering condition is that the key points of the parameter vectors are the same; Count the third quantity of characteristic wave signals under each category; Calculating the biased average value of the characteristic wave signal under each category according to the third quantity and the bias of the characteristic wave signal in the parameter vector relative to the time axis, and calculating the biased fourth-order center distance and biased variance of the characteristic wave signal under each type according to the biased average value; Calculating the bias kurtosis of the characteristic wave signal under the same category according to the bias fourth-order center distance and the bias variance under the same category; A second average value of the bias kurtosis corresponding to each category is calculated, and the second average value is used as the mean value of the bias kurtosis.

8. An electrocardiogram signal quality assessment device, characterized in that: include: A signal acquisition module, used to acquire an electrocardiogram signal to be evaluated, wherein the electrocardiogram signal includes a plurality of characteristic wave signals; A parameter extraction module, used for extracting a parameter vector sequence of the electrocardiogram signal, wherein the parameter vector sequence includes a plurality of parameter vectors for characterizing the characteristic wave signal, each characteristic wave signal corresponds to a parameter vector, and the parameter vector includes: a key point of the characteristic wave signal, a matching degree between the characteristic wave signal and a characteristic wave signal template, a width of the characteristic wave signal template, a position information of the characteristic wave signal, an amplitude of the characteristic wave signal, and an offset of the characteristic wave signal relative to a time axis; A feature extraction module, used for extracting quality features of the ECG signal according to the parameter vector sequence, wherein the quality features include at least one of characteristic wave signal category entropy, ECG reconstruction signal signal-to-noise ratio, characteristic wave signal peak value to signal energy ratio of the ECG signal, mean value of amplitude kurtosis and mean value of bias kurtosis; The quality assessment module is used to perform quality assessment on the ECG signal according to the quality characteristics and obtain the quality assessment level of the ECG signal, wherein the quality assessment level includes: morphological analysis level, rhythm analysis level, beat localization level, reconfigurable denoising level and unusable level.

9. An electrocardiogram signal quality assessment device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the electrocardiogram signal quality assessment method as described in any one of claims 1-7.

Images