A self-supervised method and system for evaluating the signal-to-noise ratio of electrocardiogram

Abstract

The present invention belongs to the technical field of electrocardiogram (ECG) signal processing. Aiming at the noise in the ECG signal frequency band, a self-supervised ECG signal-to-noise ratio (SNR) evaluation method and system are provided. The original ECG data is added with composite noise to obtain composite ECG data. Then, the upper and lower envelopes of the original ECG data and the composite noise are calculated, and the amplitude difference of the envelopes is calculated. Subsequently, the SNR of the composite waveform is calculated through the SNR formula. The composite ECG data is input into the network, and after obtaining the result, the loss is calculated. The network parameters are updated through an optimization algorithm until the result reaches the expectation. Whether the noise segment in the electrocardiogram will affect the result of the subsequent automatic ECG analysis algorithm is judged by the magnitude of the SNR. This method utilizes the self-supervised feature to overcome the one-sidedness of the previous models in identifying noise segments, can identify complex noises in the real world to a greater extent, and the proposed method has a low data cost and remarkable effect on identifying noises with the same distribution as the ECG signal.

Description

Technical Field

[0001] The present invention belongs to the technical field of electrocardiogram (ECG) signal processing, and particularly relates to a self-supervised method and system for evaluating the signal-to-noise ratio of ECG signals. Background Art

[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] ECG activity is one of the most important vital signs and is the main observation index for many diseases. In modern society, the incidence of cardiovascular diseases is high. Due to the convenience of ECG examination, high reusability of results, reliability, wide disease coverage, low cost, etc., ECG examination is widely used in clinical diagnosis. Since its application in clinical practice in 1903, electrocardiology has made great contributions to the medical discipline and the cause of human health.

[0004] In ECG diagnosis, noise will bring great interference to doctors' diagnosis. The sources of noise are complex. Some noise can be filtered by signal processing methods, and these noises have little impact on the diagnosis results. However, some noises cannot be removed, such as noise with a frequency in the ECG frequency band. If the signal-to-noise ratio of the ECG signal can be evaluated, the collaborative analysis efficiency of the ECG automatic analysis algorithm and the doctor will be greatly improved.

[0005] The traditional methods for evaluating the signal-to-noise ratio have the following defects:

[0006] 1. Traditional methods need to filter the noise, that is, separate it, to eliminate the influence of noise on the signal. For example, a low-pass filter can filter high-frequency noise, and a notch filter can filter power-frequency noise, etc.

[0007] 2. Traditional methods can only filter "out-of-band" noise, that is, noise outside the ECG signal frequency band; and for "in-band" noise, that is, noise with a frequency range in the ECG signal frequency domain and an amplitude similar to that of the sub-waveform of the ECG signal, it cannot be separated, and its influence on the ECG signal cannot be eliminated by effective means.

[0008] 3. Most deep learning-based evaluation methods rely on data annotation, and the data cost is high. Summary of the Invention

[0009] In order to solve at least one of the technical problems in the above background art, the present invention provides a self-supervised method and system for evaluating the signal-to-noise ratio of ECG signals, which does not need to separate "in-band" noise, but effectively evaluates the signal-to-noise ratio of ECG data containing noise in the same frequency band as the ECG signal. Based on this evaluation result, the degree to which the noise affects the waveform of the original ECG signal and whether it affects the results of subsequent ECG automatic analysis algorithms can be obtained.

[0010] To achieve the above object, the present invention adopts the following technical solutions:

[0011] The first aspect of the present invention provides a self-supervised electrocardiogram signal-to-noise ratio evaluation method, including the following steps:

[0012] Obtain the electrocardiogram data to be evaluated;

[0013] Based on the electrocardiogram data to be evaluated and the trained signal-to-noise ratio prediction model, obtain the noise level of the electrocardiogram data to be evaluated;

[0014] Among them, when training the signal-to-noise ratio prediction model, the obtained electrocardiogram data samples and the composite waveforms obtained by further superimposing the composite noise obtained by superimposing segment noise, recombination noise and baseline noise are used as training samples and test samples;

[0015] Combining the result output by the signal-to-noise ratio prediction model and the signal-to-noise ratio calculation loss value calculated using the signal-to-noise ratio formula, using an optimization method, update the network parameters of the signal-to-noise ratio prediction model according to this loss value until the loss value reaches the set threshold.

[0016] The second aspect of the present invention provides a self-supervised electrocardiogram signal-to-noise ratio evaluation system, including:

[0017] A data acquisition module, which is used to obtain the electrocardiogram data to be evaluated;

[0018] An electrocardiogram signal-to-noise ratio evaluation module, which is used to obtain the noise level of the electrocardiogram data to be evaluated based on the electrocardiogram data to be evaluated and the trained signal-to-noise ratio prediction model;

[0019] Among them, when training the signal-to-noise ratio prediction model, the obtained electrocardiogram data samples and the composite waveforms obtained by further superimposing the composite noise obtained by superimposing segment noise, recombination noise and baseline noise are used as training samples and test samples;

[0020] Combining the result output by the signal-to-noise ratio prediction model and the signal-to-noise ratio calculation loss value calculated using the signal-to-noise ratio formula, using an optimization method, update the network parameters of the signal-to-noise ratio prediction model according to this loss value until the loss value reaches the set threshold.

[0021] The third aspect of the present invention provides a computer-readable storage medium.

[0022] A computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps in a self-supervised electrocardiogram signal-to-noise ratio evaluation method as described in the first aspect above.

[0023] The fourth aspect of the present invention provides a computer device.

[0024] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in a self-supervised electrocardiogram signal-to-noise ratio evaluation method as described in the first aspect above.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] 1. Aiming at the defect that the traditional signal-to-noise ratio evaluation method needs to filter noise to eliminate the influence of noise on the signal, the present invention does not need to separate the "in-band" noise, but evaluates the level of the "in-band" noise by using a neural network to evaluate the signal-to-noise ratio of the electrocardiogram signal.

[0027] 2. Aiming at the defect that the traditional signal-to-noise ratio evaluation method can only filter the "out-of-band" noise and cannot separate the "in-band" noise, the present invention effectively evaluates the signal-to-noise ratio of electrocardiogram data containing noise in the same frequency band as the electrocardiogram signal, and uses the composite waveform obtained by further superimposing the acquired electrocardiogram data samples and the composite noise obtained by superimposing segment noise, recombinant noise, and baseline noise as training samples and test samples.

[0028] 3. Aiming at the defect that the existing evaluation method based on deep learning depends on data annotation and has a high data cost, the composite noise generated by the present invention has a wide coverage range of frequency bands and amplitudes, can simulate most of the "in-band" noises in reality, and can identify complex noises in the real world to a greater extent through this method. This method has a low data cost and has a significant effect on identifying noises with the same distribution as the electrocardiogram signal.

[0029] Advantages of additional aspects of the present invention will be partially given in the following description, partially become apparent from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0031] Figure 1 It is the overall flowchart of the self-supervised electrocardiogram signal-to-noise ratio evaluation method in the first embodiment of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0033] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0034] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0035] Embodiment 1

[0036] Refer to Figure 1 , this embodiment provides a self-supervised electrocardiogram signal-to-noise ratio evaluation method, including the following steps:

[0037] Step 1: Obtain the electrocardiogram data to be evaluated;

[0038] Step 2: Preprocess the electrocardiogram data to be evaluated to obtain the preprocessed electrocardiogram data to be evaluated;

[0039] Step 3: Based on the preprocessed electrocardiogram data to be evaluated and the trained signal-to-noise ratio prediction model, obtain the noise level of the electrocardiogram data to be evaluated;

[0040] Among them, when training the signal-to-noise ratio prediction model, the obtained electrocardiogram data samples and the composite waveforms obtained by further superimposing the composite noise obtained by superimposing segment noise, recombinant noise, and baseline noise are used as training samples and test samples;

[0041] Combining the result output by the signal-to-noise ratio prediction model and the loss value calculated using the signal-to-noise ratio formula, using an optimization algorithm, update the network parameters of the signal-to-noise ratio prediction model according to this loss value until the loss value reaches a set threshold.

[0042] The noise level of the electrocardiogram data to be evaluated includes:

[0043] Three levels, namely 0, 1, and 2 levels. Among them, level 0 indicates that the noise is small and does not affect the shape of the electrocardiogram sub-waveform; level 1 indicates that the noise affects the P, T, U waves, and F waves (atrial flutter waves), but is not sufficient to affect the QRS complex signal; level 2 indicates that the noise is extremely chaotic and has affected the shape of all sub-waveforms.

[0044] In step 1, the electrocardiogram data Predict_Wave to be evaluated is a single 12-lead electrocardiogram data, with a dimension of (1, 12, len), indicating that the number of electrocardiograms is 1, the leads of the electrocardiogram are 12, and len represents the length of the electrocardiogram data.

[0045] In step 2, the preprocessing includes data format conversion and data truncation.

[0046] Among them, the data format is converted as follows: converting data of dictionary type into data of array type.

[0047] The data is truncated as follows: limiting overly long electrocardiogram data within a certain range.

[0048] In step 3, the training of the signal-to-noise ratio prediction model specifically includes:

[0049] Step 301: Obtain electrocardiogram data samples, and after preprocessing, obtain the preprocessed electrocardiogram data samples Raw_wave, with a dimension of (b, 12, len), where b represents the number of electrocardiograms;

[0050] Step 302: Use the electrocardiogram data sample Raw_Wave obtained after preprocessing and the composite waveform Wave obtained by further superimposing the composite noise Final_Noise obtained by superimposing segment noise, recombination noise, and baseline noise as training samples and test samples;

[0051] In step 302, the composite noise is randomly superimposed according to the proportions of segment noise, recombination noise, and baseline noise.

[0052] Among them, in step 302, the segment noise is two different types of segment noise, specifically:

[0053] The first type of segment noise is denoted as noise1. Noise1 is composed of 12 noises with different frequencies and different positions where the noise appears in the electrocardiogram data, with a dimension of (b, 12, len). The second type of segment noise is denoted as noise2. Noise2 is composed of 12 noises with the same frequency, the same position where the noise appears in the electrocardiogram data, and different amplitudes, with a dimension of (b, 12, len).

[0054] Among them, the first type of segment noise is obtained from a random number matrix P with a dimension of (b, 12, len) through a Chebyshev type-I filter, that is, the noise noise1 in the electrocardiogram signal frequency band;

[0055] For the noise noise1 in the electrocardiogram signal frequency band, randomly generate regions where the noise appears according to leads. The regions do not overlap, and the noise in the remaining regions tends to be 0. The amplitude of the regional noise gradually increases and decreases according to a trapezoidal change trend, and at the beginning and end, there is a smooth transition.

[0056] Among them, the second type of segment noise is obtained from a random number matrix Q with a dimension of (b, 1, len) through a Chebyshev type-I filter, that is, the noise noise2 in the electrocardiogram signal frequency band.

[0057] After being filtered by the Chebyshev filter, its frequency is in the electrocardiogram signal frequency band, and the amplitude range is large enough to cover amplitude changes in various real situations.

[0058] In the same way, the noise noise2 in the ECG signal frequency band is randomly generated according to the lead, the areas where the noise appears do not overlap, and the noise in the remaining areas is 0, so as to obtain the second type of regional segment noise, so that noise2 becomes a noise that appears regionally within the ECG length range and increases and decreases in amplitude in a trapezoidal trend.

[0059] The first type of regional segment noise is superimposed, and the second type of regional segment noise becomes noise: 12 random numbers are generated to make noise2 become 12 noises with the same frequency and the same noise position, but different amplitudes. Then noise1 and noise2 are superimposed to become noise, with a dimension of (b,12,len).

[0060] The reconstructed noise is represented as noise3. Noise3 is obtained by separating the noise frequency and phase by Fourier transforming the data and reconstructing the data by inverse Fourier transforming. The dimension of the noise3 is (b, 12, len).

[0061] The generation process of the recombinant noise is as follows:

[0062] (a) A clean segment of the ECG and a noisy segment of the ECG of the same patient are captured. After multiple experimental comparisons and analyses, some noisy segments of different patients are selected as processing data.

[0063] (b) The clean ECG data and the noisy fragments are processed by Fourier transform to obtain the frequency and phase of the corresponding signal, and the noise frequency is obtained by subtracting the frequency of the clean ECG data from the frequency of the noisy fragment ECG data.

[0064] (c) During the inverse Fourier transform, the phase of the reconstructed noise is a randomly generated phase; the generated phase and the separated frequency are inversely transformed through Fourier transform to obtain the reconstructed noise noise3, with dimension (b, 12, len).

[0065] The noise with overlapping frequency band and amplitude range and ECG signal is separated by Fourier transform and inverse transform, and then added to the composite noise.

[0066] The baseline noise is two different baseline noises, wherein the first type of baseline noise is represented by baseline1, baseline1 is 12 noises with different frequencies, and the second type of baseline noise is represented by baseline2, baseline2 is 12 noises with the same frequency and different amplitudes, and the dimensions of the two baseline noises are both (b, 12, len).

[0067] The generation process of baseline noise includes:

[0068] (a) Generate the first type of baseline, baseline1: Obtain baseline1 from a random number matrix R with dimensions (b, 12, len) through a Chebyshev type-I filter, where b represents the number of electrocardiograms, 12 represents the 12 leads of the electrocardiogram, and len represents the length of the electrocardiogram data.

[0069] (b) Generate the second type of baseline, baseline2: Obtain baseline2 from a random number matrix S with dimensions (b, 1, len) through a Chebyshev type-I filter, where b represents the number of electrocardiograms, 1 represents generating only one piece of noise, and len represents the length of the electrocardiogram data.

[0070] (c) Synthesize the final baseline noise, baseline: Superimpose baseline2 and baseline1 to obtain baseline with dimensions (b, 12, len).

[0071] Step 303: Calculate the signal-to-noise ratio SNR of the composite waveform Wave according to the signal-to-noise ratio calculation formula to obtain the first signal-to-noise ratio;

[0072] Among them, in step 303, the signal-to-noise ratio calculation formula is:

[0073]

[0074] Among them, V S , V N respectively represent the "root mean square values" of the signal and noise voltages.

[0075] By calculating the amplitude difference between Raw_Wave and Final_Noise, the signal-to-noise ratio SNR of the composite waveform Wave is then obtained.

[0076] The calculation process of the signal-to-noise ratio of the composite electrocardiogram data is as follows:

[0077] (a) Calculate the amplitude difference of the original electrocardiogram data Raw_Wave: Obtain the upper and lower envelopes of the original electrocardiogram data through morphological filtering, and subtract the upper and lower envelopes to obtain the amplitude difference of the original electrocardiogram data Raw_Wave.

[0078] (b) Calculate the amplitude difference of the composite noise Final_Noise: Obtain the upper and lower envelopes of the noise Final_noise through morphological filtering, and subtract the upper and lower envelopes to obtain the amplitude difference of the noise Final_Noise.

[0079] (c) Calculate the signal-to-noise ratio according to the signal-to-noise ratio formula using the amplitude difference of the original electrocardiogram data Raw_Wave and the amplitude difference of the noise Final_Noise.

[0080] Step 304: Input the composite waveform Wave into the signal-to-noise ratio prediction neural network, and record the signal-to-noise ratio prediction result as the second signal-to-noise ratio;

[0081] Among them, in step 304, the signal-to-noise ratio prediction neural network can be a convolutional neural network, LSTM, recurrent neural network, etc., but is not limited to the above models.

[0082] Step 305: Input the first signal-to-noise ratio and the second signal-to-noise ratio into the loss function to obtain the loss value loss. Adopt an optimization algorithm to update the network parameters of the neural network according to the loss value loss until loss reaches the set threshold and the training ends.

[0083] It should be noted that the threshold set in this embodiment can be set according to the requirements for the loss value loss, and no specific description is given in this embodiment.

[0084] In this embodiment, the loss function is the mean square loss function.

[0085] In this embodiment, the optimization algorithm adopts the AdamW optimization algorithm.

[0086] Embodiment 2

[0087] This embodiment provides a self-supervised electrocardiogram signal-to-noise ratio evaluation system, including:

[0088] A data acquisition module, which is used to acquire the electrocardiogram data to be evaluated;

[0089] An electrocardiogram signal-to-noise ratio evaluation module, which is used to obtain the noise level of the electrocardiogram data to be evaluated based on the electrocardiogram data to be evaluated and the trained signal-to-noise ratio prediction model;

[0090] Among them, when training the signal-to-noise ratio prediction model, the acquired electrocardiogram data sample and the composite waveform obtained by further superimposing the composite noise obtained by superimposing segment noise, recombination noise and baseline noise are used as training samples and test samples;

[0091] Combine the result output by the signal-to-noise ratio prediction model and the loss value calculated by using the signal-to-noise ratio formula. Adopt an optimization method to update the network parameters of the signal-to-noise ratio prediction model according to this loss value until the loss value reaches the set threshold.

[0092] Embodiment 3

[0093] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the steps in a self-supervised electrocardiogram signal-to-noise ratio evaluation method as described in Embodiment 1 above.

[0094] Embodiment 4

[0095] This embodiment provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps in a self-supervised heart telemetry signal-to-noise ratio evaluation method as described in Embodiment 1 above.

[0096] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories and optical memories, etc.) containing computer-usable program code.

[0097] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in one Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0098] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the specified functions in one Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0099] 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 generate a computer-implemented process. Therefore, the instructions executed on the computer or other programmable device provide steps for implementing the specified functions in one Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0100] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above various methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0101] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A self-supervised heart telemetry signal-to-noise ratio evaluation method, characterized in that, It includes the following steps: Obtain the electrocardiogram data to be evaluated; Based on the electrocardiogram data to be evaluated and the trained signal-to-noise ratio prediction model, obtain the noise level of the electrocardiogram data to be evaluated; Among them, when training the signal-to-noise ratio prediction model, the composite waveform obtained by further superimposing the obtained electrocardiogram data samples and the composite noise obtained by superimposing segment noise, recombination noise, and baseline noise is used as the training samples and test samples; Combining the result output by the signal-to-noise ratio prediction model and the signal-to-noise ratio calculation loss value calculated using the signal-to-noise ratio formula, using an optimization algorithm, update the network parameters of the signal-to-noise ratio prediction model according to this loss value until the loss value reaches the set threshold; The segment noise includes the first type of segment noise and the second type of segment noise. The first type of segment noise is 12 noises with different frequencies and different occurrence positions in the electrocardiogram data; the second type of segment noise is 12 noises with the same frequency, the same occurrence position in the electrocardiogram data, but different amplitudes.

2. The self-supervised heart telemetry signal-to-noise ratio evaluation method according to claim 1, characterized in that, The composite noise is randomly superimposed according to the ratio of segment noise, recombination noise, and baseline noise.

3. The self-supervised heart telemetry signal-to-noise ratio evaluation method according to claim 1, characterized in that, Both the first type of segment noise and the second type of segment noise are obtained through a Chebyshev type I filter based on a random number matrix. After filtering, the regions where the noise appears are randomly generated according to the leads, and the regions do not overlap, and the noise in the remaining regions tends to be 0.

4. The self-supervised heart telemetry signal-to-noise ratio evaluation method according to claim 1, characterized in that, The generation process of the recombination noise is as follows: Intercept a clean electrocardiogram segment and a noisy electrocardiogram segment from the same electrocardiogram; Process the clean electrocardiogram and the noisy electrocardiogram through Fourier transform to obtain the frequencies and phases of the corresponding signals. Subtract the frequency of the clean electrocardiogram from the frequency of the noisy electrocardiogram to obtain the noise frequency; Randomly generate a phase, and through the inverse Fourier transform of this phase and the noise frequency, obtain the recombination noise.

5. The self-supervised heart telemetry signal-to-noise ratio evaluation method according to claim 1, characterized in that, The process of calculating the signal-to-noise ratio using the signal-to-noise ratio formula includes: Based on the electrocardiogram data, obtain the first pair of upper and lower envelopes, and based on the difference between the first pair of upper and lower envelopes, obtain the amplitude difference of the electrocardiogram data; Based on the composite noise, obtain the second pair of upper and lower envelopes, and based on the difference between the second pair of upper and lower envelopes, obtain the amplitude difference of the composite noise; Substitute the amplitude difference of the electrocardiogram data and the amplitude difference of the composite noise into the signal-to-noise ratio calculation formula to obtain the signal-to-noise ratio of the composite waveform.

6. The self-supervised heart telemetry signal-to-noise ratio evaluation method according to claim 1, characterized in that, The loss function for calculating the loss value uses the mean square loss function, and the optimization algorithm uses the AdamW optimization algorithm.

7. A self-supervised heart telemetry signal-to-noise ratio evaluation system, characterized in that, It includes: A data acquisition module, which is used to obtain the electrocardiogram data to be evaluated; An electrocardiogram signal-to-noise ratio evaluation module, which is used to obtain the noise level of the electrocardiogram data to be evaluated based on the electrocardiogram data to be evaluated and the trained signal-to-noise ratio prediction model; Among them, when training the signal-to-noise ratio prediction model, the composite waveform obtained by further superimposing the obtained electrocardiogram data samples and the composite noise obtained by superimposing segment noise, recombination noise, and baseline noise is used as the training samples and test samples; Combining the result output by the signal-to-noise ratio prediction model and the signal-to-noise ratio calculation loss value calculated using the signal-to-noise ratio formula, using an optimization algorithm, update the network parameters of the signal-to-noise ratio prediction model according to this loss value until the loss value reaches the set threshold; The segment noise includes a first type of segment noise and a second type of segment noise. The first type of segment noise consists of 12 noises with different frequencies and different occurrence positions in the electrocardiogram data. The second type of segment noise consists of 12 noises with the same frequency, the same occurrence position in the electrocardiogram data, but different amplitudes.

8. A computer-readable storage medium, on which a computer program is stored, characterized in that, When the program is executed by the processor, it implements the steps in a self-supervised electrocardiogram signal-to-noise ratio evaluation method as described in any one of claims 1-6.

9. A computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in a self-supervised electrocardiogram signal-to-noise ratio evaluation method as described in any one of claims 1-6.

Images