Electrocardiosignal generation method, cardiovascular disease diagnosis device, equipment and medium
Through the ECG signal generation method, multi-lead ECG signals are generated based on the ECG signal generation model, which solves the problem of daily wearable devices measuring a single lead signal, and realizes comprehensive ECG signal acquisition and efficient monitoring of cardiovascular diseases .
Patent Information
- Application Number
- CN202410964495.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-18
AI Technical Summary
In the prior art, daily wearable electrocardiogram signal measurement devices can only measure the electrocardiogram signal of a single lead, which is difficult to fully reflect the health status of the heart, limiting the long-term monitoring and diagnosis of cardiovascular diseases.
Through the electrocardiogram signal generation method, a multi-lead electrocardiogram signal is generated based on the electrocardiogram collected under a lead method. The electrocardiogram signal generation model is used to predict and splice signals under multiple lead methods to determine the optimal lead Method, suitable for smart wearable devices, improving signal accuracy and comprehensiveness.
It realizes the acquisition of comprehensive ECG signals without wearing complex patches, improves the accuracy and practicality of ECG signals, and facilitates efficient and reliable monitoring and diagnosis of cardiovascular diseases.
Smart Images

Figure CN118948292A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and in particular to an electrocardiogram signal generation method, a cardiovascular disease diagnosis device, equipment and medium. Background Art
[0002] Cardiovascular disease is one of the leading causes of death worldwide. Electrocardiograms are often used for early diagnosis of cardiovascular disease and play a vital role in reducing patient mortality. Medical workers can diagnose cardiovascular disease by analyzing typical changes in electrocardiograms, such as abnormalities in characteristic waves such as P waves, QRS complexes, and ST segments.
[0003] At present, the electrocardiogram commonly used in medicine contains multi-lead electrocardiogram signals, which can more comprehensively reflect the health status of the heart and help doctors make accurate clinical diagnoses of patients. However, it is difficult to collect standard multi-lead electrocardiogram signals. For example, collecting 12-lead electrocardiogram signals requires configuring more patches, which restricts the patient's free movement. It is only suitable for clinical medical diagnosis and is not conducive to long-term cardiovascular disease monitoring. In order to achieve long-term monitoring of cardiovascular diseases, daily wearable electrocardiogram signal measurement devices (such as smart bracelets, patches, etc.) have been developed and used. These daily wearable devices reduce the impact on the patient's normal life while monitoring the electrocardiogram signals in real time. However, daily wearable electrocardiogram signal measurement devices have an obvious limitation, that is, most of them can only measure the electrocardiogram signals of a single lead, which is less practical and difficult to comprehensively reflect the health status of the heart.
[0004] In summary, the problems existing in related technologies need to be solved urgently. Summary of the invention
[0005] The purpose of this application is to solve one of the technical problems existing in the related art to at least a certain extent.
[0006] To this end, an object of embodiments of the present application is to provide an electrocardiogram signal generation method, a cardiovascular disease diagnosis device, equipment and medium.
[0007] In order to achieve the above technical objectives, the technical solutions adopted by the embodiments of the present application include:
[0008] On the one hand, an embodiment of the present application provides a method for generating an electrocardiogram signal, the method comprising:
[0009] Acquire multi-lead ECG signals corresponding to multiple sample persons; wherein the multi-lead ECG signals include first ECG signals measured in various lead modes, and there are multiple types of lead modes;
[0010] Inputting a first ECG signal corresponding to a first lead mode into an ECG signal generation model corresponding to the first lead mode, and obtaining a first prediction signal through prediction by the ECG signal generation model; the first prediction signal is used to represent the ECG signal corresponding to the sample person predicted by the ECG signal generation model in a lead mode other than the first lead mode; wherein the first lead mode is any lead mode;
[0011] Determining a loss value for training according to the first prediction signal and the multi-lead ECG signal;
[0012] According to the loss value, updating the parameters of the electrocardiogram signal generation model to obtain the electrocardiogram signal generation model trained in the first lead mode;
[0013] Evaluate the effects of the ECG signal generation models trained under various lead modes, determine the performance indicators corresponding to the trained ECG signal generation models, determine the optimal lead mode according to the performance indicators, and determine the ECG signal generation model trained under the optimal lead mode as the target model;
[0014] Acquiring a second electrocardiogram signal measured by the target person in the optimal lead mode;
[0015] Inputting the second ECG signal into the target model, and obtaining a second prediction signal through prediction by the target model; the second prediction signal is used to represent the ECG signal corresponding to the target person predicted by the target model under a lead mode other than the optimal lead mode;
[0016] The second electrocardiogram signal and the second prediction signal are concatenated to obtain a generated target electrocardiogram signal.
[0017] In addition, the electrocardiogram signal generation method according to the above embodiment of the present application may also have the following additional technical features:
[0018] Furthermore, in one embodiment of the present application, the ECG signal generation model is constructed by a convolutional autoencoder.
[0019] Further, in one embodiment of the present application, the convolutional autoencoder includes an encoder and a decoder;
[0020] The encoder comprises a convolutional layer, a first batch normalization layer, and a first activation function;
[0021] The decoder includes a deconvolution layer, a second batch normalization layer, and a second activation function.
[0022] Furthermore, in one embodiment of the present application, after obtaining the multi-lead ECG signals corresponding to the multiple sample persons, the method further includes:
[0023] Removing baseline drift noise in the multi-lead ECG signal by wavelet decomposition and reconstruction method;
[0024] The power frequency interference and myoelectric interference in the multi-lead electrocardiogram signal are removed by wavelet threshold method.
[0025] Further, in one embodiment of the present application, the effect evaluation of the ECG signal generation model trained under each lead mode, determining the performance index corresponding to the trained ECG signal generation model, and determining the optimal lead mode according to the performance index includes:
[0026] Determine the index values of the first prediction signal and the multi-lead ECG signal predicted in each lead mode by at least one index algorithm selected from the group consisting of mean square error, mean absolute error, and Pearson correlation coefficient;
[0027] The lead mode corresponding to the smallest mean square error and the smallest mean absolute error is determined as the optimal lead mode, or the lead mode corresponding to the largest Pearson correlation coefficient is determined as the optimal lead mode.
[0028] Further, in one embodiment of the present application, the effect evaluation of the ECG signal generation model trained under each lead mode, determining the performance index corresponding to the trained ECG signal generation model, and determining the optimal lead mode according to the performance index includes:
[0029] Obtaining cardiovascular disease label information corresponding to each of the sample persons; the cardiovascular disease label information is used to characterize the true category of the cardiovascular disease possessed by the sample person;
[0030] According to the first prediction signal output by the trained electrocardiogram signal generation model in each lead mode, the cardiovascular disease of the sample person is predicted by the cardiovascular disease prediction model to obtain a first prediction result;
[0031] Determining the prediction accuracy corresponding to each of the lead modes according to the first prediction result and the cardiovascular disease label information;
[0032] The corresponding lead mode with the highest prediction accuracy is determined as the optimal lead mode.
[0033] On the other hand, an embodiment of the present application provides an electrocardiogram signal generating device, the device comprising:
[0034] A first acquisition unit is used to acquire multi-lead ECG signals corresponding to multiple sample persons; wherein the multi-lead ECG signals include first ECG signals measured in various lead modes, and there are multiple types of lead modes;
[0035] A first prediction unit is used to input a first ECG signal corresponding to a first lead mode into an ECG signal generation model corresponding to the first lead mode, and obtain a first prediction signal through prediction by the ECG signal generation model; the first prediction signal is used to represent the ECG signal corresponding to the sample person predicted by the ECG signal generation model in a lead mode other than the first lead mode; wherein the first lead mode is any lead mode;
[0036] a processing unit, configured to determine a loss value of training according to the first prediction signal and the multi-lead ECG signal;
[0037] An updating unit, configured to update the parameters of the electrocardiogram signal generation model according to the loss value, so as to obtain the electrocardiogram signal generation model trained in the first lead mode;
[0038] An evaluation unit is used to evaluate the effects of the ECG signal generation models trained under various lead modes, determine the performance indicators corresponding to the trained ECG signal generation models, determine the optimal lead mode according to the performance indicators, and determine the ECG signal generation model trained under the optimal lead mode as the target model;
[0039] A second acquisition unit, used for acquiring a second electrocardiogram signal measured by the target person in the optimal lead mode;
[0040] A second prediction unit is used to input the second ECG signal into the target model, and obtain a second prediction signal through prediction by the target model; the second prediction signal is used to represent the ECG signal corresponding to the target person predicted by the target model in a lead mode other than the optimal lead mode;
[0041] The integration unit is used to splice the second ECG signal and the second prediction signal to obtain a generated target ECG signal.
[0042] On the other hand, an embodiment of the present application provides a cardiovascular disease diagnosis device, comprising:
[0043] An input unit, used for inputting the target ECG signal generated by the aforementioned ECG signal generation method into a cardiovascular disease prediction model;
[0044] The third prediction unit is used to predict the cardiovascular disease of the target person through the cardiovascular disease prediction model to obtain a second prediction result.
[0045] On the other hand, an embodiment of the present application provides an electronic device, including:
[0046] at least one processor;
[0047] at least one memory for storing at least one program;
[0048] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned electrocardiographic signal generating method.
[0049] On the other hand, an embodiment of the present application further provides a computer-readable storage medium, in which a program executable by a processor is stored. When the program executable by the processor is executed by the processor, it is used to implement the above-mentioned electrocardiogram signal generation method.
[0050] The advantages and benefits of the present application will be partially given in the following description, and partially become apparent from the following description, or be understood through the practice of the present application:
[0051] The ECG signal generation method, cardiovascular disease diagnosis device, equipment and medium disclosed in the embodiments of the present application obtain multi-lead ECG signals corresponding to multiple sample persons; wherein the multi-lead ECG signals include first ECG signals measured under various lead modes, and there are multiple types of lead modes; the first ECG signal corresponding to the first lead mode is input into an ECG signal generation model corresponding to the first lead mode, and a first prediction signal is obtained by predicting the ECG signal generation model; the first prediction signal is used to characterize the ECG signal generation model predicting the corresponding ECG signal of the sample person under lead modes other than the first lead mode; wherein the first lead mode is any lead mode; according to the first prediction signal and the multi-lead ECG signals, a loss value of training is determined; according to the loss value, the ECG signal generation model is trained. The model is updated with parameters to obtain an ECG signal generation model trained under the first lead mode; the effect of the ECG signal generation model trained under each lead mode is evaluated to determine the performance index corresponding to the trained ECG signal generation model, and the optimal lead mode is determined according to the performance index, and the ECG signal generation model trained under the optimal lead mode is determined as the target model; a second ECG signal measured by the target person under the optimal lead mode is obtained; the second ECG signal is input into the target model, and a second prediction signal is obtained by predicting the target model; the second prediction signal is used to characterize the ECG signal corresponding to the target person predicted by the target model under lead modes other than the optimal lead mode; the second ECG signal and the second prediction signal are spliced to obtain a generated target ECG signal. This method can generate multi-lead ECG signals based on ECG signals collected in one lead mode, without the need for users to wear complex patches to collect ECG signals. It is suitable for application scenarios of smart wearable devices and can obtain more comprehensive ECG signals with better practicality. Moreover, this method can evaluate and select the optimal lead mode suitable as the basis for signal generation, improve the accuracy of the generated ECG signals, and facilitate efficient and reliable monitoring of the user's health status based on the generated ECG signals. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following introduction is made to the drawings of the embodiments of the present application or the related technical solutions in the prior art. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present application. For technical personnel in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0053] Figure 1 A schematic diagram of an implementation environment of an electrocardiogram signal generation method provided in an embodiment of the present application;
[0054] Figure 2 A schematic diagram of a flow chart of an electrocardiogram signal generation method provided in an embodiment of the present application;
[0055] Figure 3 A schematic diagram of the structure of an electrocardiogram signal generation model provided in an embodiment of the present application;
[0056] Figure 4 A schematic diagram of the structure of an encoder provided in an embodiment of the present application;
[0057] Figure 5 A schematic diagram of the structure of a decoder provided in an embodiment of the present application;
[0058] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0059] The present application is further described below in conjunction with the accompanying drawings and specific embodiments. The described embodiments should not be regarded as limiting the present application, and all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.
[0060] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.
[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.
[0062] 1) Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making. Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level technology and software-level technology. Basic artificial intelligence technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.
[0063] 2) Machine Learning (ML) is a multi-disciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning (deep learning) usually includes artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.
[0064] Cardiovascular disease is one of the leading causes of death worldwide. Electrocardiograms are often used for early diagnosis of cardiovascular disease and play a vital role in reducing patient mortality. Medical workers can diagnose cardiovascular disease by analyzing typical changes in electrocardiograms, such as abnormalities in characteristic waves such as P waves, QRS complexes, and ST segments.
[0065] At present, the electrocardiogram commonly used in medicine contains multi-lead electrocardiogram signals, which can more comprehensively reflect the health status of the heart and help doctors make accurate clinical diagnoses of patients. However, it is difficult to collect standard multi-lead electrocardiogram signals. For example, collecting 12-lead electrocardiogram signals requires configuring more patches, which restricts the patient's free movement. It is only suitable for clinical medical diagnosis and is not conducive to long-term cardiovascular disease monitoring. In order to achieve long-term monitoring of cardiovascular diseases, daily wearable electrocardiogram signal measurement devices (such as smart bracelets, patches, etc.) have been developed and used. These daily wearable devices reduce the impact on the patient's normal life while monitoring the electrocardiogram signals in real time. However, daily wearable electrocardiogram signal measurement devices have an obvious limitation, that is, most of them can only measure the electrocardiogram signals of a single lead, which is less practical and difficult to comprehensively reflect the health status of the heart.
[0066] In view of this, an embodiment of the present application provides an ECG signal generation method, which can generate a multi-lead ECG signal based on an ECG signal collected in a one-lead mode. The method does not require the user to wear a complex patch to collect the ECG signal. The method is suitable for application scenarios of smart wearable devices, and can obtain a more comprehensive ECG signal with better practicality. Moreover, the method can evaluate and select the optimal lead mode suitable as the basis for signal generation, which can improve the accuracy of the generated ECG signal and facilitate efficient and reliable monitoring of the user's health status based on the generated ECG signal.
[0067] Please refer to Figure 1 , Figure 1 The schematic diagram of the implementation environment of an electrocardiogram signal generation method provided in the embodiment of the present application is shown. In the implementation environment, the main hardware and software entities involved include a terminal device 110 and a background server 120. The terminal device 110 and the background server 120 are connected in communication.
[0068] Specifically, the ECG signal generation method provided in the embodiments of the present application can be executed solely on the terminal device 110 side, or solely on the background server 120 side, or based on data interaction between the terminal device 110 and the background server 120.
[0069] The terminal device 110 of the above embodiment may include a mobile phone, a computer, a smart wearable device, a PDA device, an intelligent voice interaction device, a vehicle terminal, etc., but is not limited thereto. The backend server 120 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.
[0070] The terminal device 110 and the backend server 120 may establish a communication connection via a wireless network or a wired network. The wireless network or wired network uses standard communication technology and / or protocols, and the network may be set to the Internet or any other network, such as but not limited to a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or any combination of a virtual private network.
[0071] Of course, it is understandable that Figure 1 The implementation environment in the present application is only some optional application scenarios of the ECG signal generation method provided in the embodiment of the present application, and the actual application is not fixed. Figure 1 The hardware and software environment shown.
[0072] Below, in combination with the introduction of the aforementioned implementation environment, an electrocardiogram signal generation method provided in an embodiment of the present application is introduced and explained.
[0073] Please refer to Figure 2 , Figure 2 is a schematic diagram of an electrocardiogram signal generation method provided in an embodiment of the present application, and the electrocardiogram signal generation method includes but is not limited to:
[0074] Step 210: obtaining multi-lead ECG signals corresponding to multiple sample persons; wherein the multi-lead ECG signals include first ECG signals measured in various lead modes, and there are multiple types of lead modes;
[0075] Step 220: inputting a first ECG signal corresponding to a first lead mode into an ECG signal generation model corresponding to the first lead mode, and obtaining a first prediction signal through prediction by the ECG signal generation model; the first prediction signal is used to represent the ECG signal corresponding to the lead mode other than the first lead mode predicted by the ECG signal generation model for the sample person; wherein the first lead mode is any lead mode;
[0076] Step 230: determining a training loss value according to the first prediction signal and the multi-lead ECG signal;
[0077] Step 240: updating the parameters of the ECG signal generation model according to the loss value to obtain the ECG signal generation model trained in the first lead mode;
[0078] Step 250: Evaluate the effects of the ECG signal generation models trained under various lead modes, determine the performance indicators corresponding to the trained ECG signal generation models, determine the optimal lead mode according to the performance indicators, and determine the ECG signal generation model trained under the optimal lead mode as the target model;
[0079] Step 260, obtaining a second electrocardiogram signal of the target person measured in the optimal lead mode;
[0080] Step 270: input the second ECG signal into the target model, and obtain a second prediction signal through prediction by the target model; the second prediction signal is used to represent the ECG signal corresponding to the target person predicted by the target model in a lead mode other than the optimal lead mode;
[0081] Step 280: concatenate the second ECG signal and the second prediction signal to obtain a generated target ECG signal.
[0082] In an embodiment of the present application, a method for generating an ECG signal is provided. The method can automatically generate a multi-lead ECG signal based on an ECG signal generation model. In actual use, it is only necessary to collect the ECG signal of the user in one lead mode. The user does not need to wear a complex patch to collect the ECG signal. The method is suitable for application scenarios of smart wearable devices.
[0083] In the embodiment of the present application, before generating ECG signals, it is necessary to first train and determine the ECG signal generation model to be used. In the embodiment of the present application, the ECG signal generation model finally put into use is recorded as the target model. The process of determining the target model will be introduced and explained below.
[0084] Specifically, when determining the target model, you can first obtain multi-lead ECG signals corresponding to multiple sample persons. Here, the sample persons are the persons whose relevant data are used when training and evaluating the ECG signal generation model. These persons can be patients in the hospital who have undergone ECG testing, etc. The number of sample persons can be multiple, and this application does not limit the specific number. In the embodiment of the present application, the multi-lead ECG signals corresponding to the sample persons are obtained. In order to facilitate subsequent processing, the number of leads and the specific lead method for obtaining the multi-lead ECG signals of each sample person can be the same. For example, in some embodiments, a standard 12-lead ECG signal can be obtained as a multi-lead ECG signal. Here, the standard 12-lead ECG signal captures the heart's electrical signals by placing electrodes on different parts of the body. There are 12 ways to set the leads, such as standard lead I: left forelimb to right forelimb; standard lead II: right forelimb to left lower limb; standard lead III: left lower limb to right forelimb; pressurized limb lead aVR: right upper limb (positive pole) to left upper limb and left lower limb (negative pole); pressurized limb lead aVL: left upper limb (positive pole) to right upper limb and left lower limb (negative pole); augmented limb lead aVF: left lower limb (positive pole) to right upper limb and left upper limb (negative pole); chest lead V1: at the fourth intercostal space on the right edge of the sternum; chest lead V2: at the fourth intercostal space on the left edge of the sternum; chest lead V3: between V2 and V4 (approximately at the midline of the sternum); chest lead V4: at the intersection of the left midclavicular line and the fifth intercostal space; chest lead V5: at the left anterior axillary line at the same level as V4; chest lead V6: at the left midaxillary line at the same level as V4.
[0085] In the embodiment of the present application, the ECG signals measured in each lead mode in the multi-lead ECG signal are recorded as the first ECG signal. It is understandable that in the embodiment of the present application, the multi-lead ECG signal may also only include the first ECG signals in some lead modes in the standard 12-lead ECG signal, and the present application does not limit the specific number and category of lead modes.
[0086] Of course, it should be noted that when obtaining multi-lead ECG signals of sample persons, relevant laws and regulations must be complied with and the privacy of the sample persons must not be disclosed.
[0087] In the embodiment of the present application, after obtaining the first ECG signal measured by each lead mode of the sample person, the ECG signal can be divided according to the lead mode, and multiple ECG signal generation models can be trained. Here, the ECG signal generation model can be used to generate ECG signals, and its input can be the ECG signal measured in a certain lead mode, and the output is the predicted ECG signal corresponding to other multiple lead modes.
[0088] Specifically, in the embodiment of the present application, for each lead mode, in the process of training the ECG signal generation model corresponding to it, it can be recorded as the first lead mode. The first ECG signal corresponding to the first lead mode is input into the ECG signal generation model corresponding to the first lead mode, and a prediction signal can be predicted by the ECG signal generation model. In the embodiment of the present application, it is recorded as the first prediction signal. The first prediction signal is the ECG signal corresponding to the lead mode other than the first lead mode predicted by the ECG signal generation model. For example, assuming that the current multi-lead ECG signal is a 12-lead signal, a lead mode is first determined as the first lead mode, and the first ECG signal under the first lead mode can be used as input to predict the ECG signals under the other 11 lead modes through the ECG signal generation model. Then, the prediction effect of the ECG signal generation model can be determined based on the predicted first prediction signal and the original multi-lead ECG signal, that is, the loss value between the first prediction signal and the multi-lead ECG signal can be calculated by the relevant loss function. This application does not limit the specific type of loss function used. According to the loss value, the parameters of the ECG signal generation model can be updated to obtain the ECG signal generation model trained in the first lead mode.
[0089] For example, in some scenarios, the Adam optimizer can be used to train the model with the mean squared error (MSE) as the loss function. The initial learning rate is set to 0.01, and the early stopping strategy is adopted. If the loss function value of the validation set does not improve for 3 consecutive epochs, the learning rate is reduced to one tenth of the previous value; if the loss function of the validation set does not improve for 8 consecutive epochs, the model training is stopped. The batch size (batch_size) of the model is 128, and the maximum number of iterations is 100.
[0090] It is understandable that in the embodiments of the present application, each lead mode is taken as the first lead mode in turn, and the ECG signal generation model corresponding to each lead mode can be trained. After obtaining the ECG signal generation models trained under each lead mode, they can be evaluated for their effects, and the performance indicators corresponding to each model can be determined. According to the performance indicators, the optimal lead mode can be selected from each lead mode, and the ECG signal generation model trained under the optimal lead mode can be determined as the target model. For example, in some embodiments, it can be evaluated which ECG signal generation model has the highest accuracy and determined as the target model. This will be described in detail in the subsequent embodiments and will not be repeated here.
[0091] In an embodiment of the present application, after determining the optimal lead mode and the target model, the application of ECG signal generation can be implemented according to the optimal lead mode and the target model. Specifically, in an embodiment of the present application, the person who needs to actually generate the ECG signal can be recorded as the target person, and the ECG signal measured by the target person under the optimal lead mode can be obtained and recorded as the second ECG signal. Then, the second ECG signal can be input into the target model, and the target model can be used to predict the corresponding ECG signal of the target person under lead modes other than the optimal lead mode, and these signals are recorded as the second predicted signal. By splicing the second ECG signal and the second predicted signal, the target ECG signal under multiple lead modes can be obtained.
[0092] It can be understood that the ECG signal generation method provided in the embodiment of the present application can generate multi-lead ECG signals based on ECG signals collected in one lead mode, without the need for users to wear complex patches to collect ECG signals. It is suitable for application scenarios of smart wearable devices, and can obtain more comprehensive ECG signals with better practicality. Moreover, the method can evaluate and select the optimal lead mode suitable as the basis for signal generation, which can improve the accuracy of the generated ECG signals and facilitate efficient and reliable monitoring of the user's health status based on the generated ECG signals.
[0093] Specifically, in some embodiments, the ECG signal generation model is constructed by a convolutional autoencoder.
[0094] In the embodiment of the present application, a convolutional autoencoder can be used to build an ECG signal generation model. Figure 3 , Figure 3 A schematic diagram of the structure of a convolutional autoencoder provided in an embodiment of the present application is shown. In the embodiment of the present application, the convolutional autoencoder is a neural network structure, which may include an encoder and a decoder. The encoder may compress the input signal into a feature representation of a latent space, and then the decoder may be used to reconstruct the input signal from the feature representation of the latent space. The decoder may also be used to generate a signal of the same type as the input signal based on the feature representation in the latent space. For applications in the embodiment of the present application, an ECG signal measured in a single lead mode may be used as input, and a convolutional autoencoder may be used to generate ECG signals in other multiple lead modes. Figure 1 As shown, in an embodiment of the present application, a convolutional autoencoder can first generate the remaining 11-lead ECG signals based on the input single-lead ECG signal, and then the input real single-lead ECG signal and the generated 11-lead ECG signal are spliced to combine into a standard 12-lead ECG signal.
[0095] Reference Figure 4 and Figure 5 ,in Figure 4is a schematic diagram of the structure of an encoder provided in an embodiment of the present application, Figure 5 It is a structural diagram of a decoder provided in an embodiment of the present application. In an embodiment of the present application, the encoder in the convolutional autoencoder may include a convolutional layer (Conv), a batch normalization layer (BN) and an activation function, and the batch normalization layer and the activation function are recorded as the first batch normalization layer and the first activation function. Specifically, the convolutional layer (Conv) can be used to learn the characteristic information of the ECG signal. Compared with the fully connected layer, the convolutional layer has fewer parameters and is easier to extract the local invariant features in the ECG signal. In an embodiment of the present application, the convolutional layer is used to learn the characteristic information in the original single-lead ECG signal and compress it into a feature representation of the latent space. Since the ECG signal data is one-dimensional data, the convolutional layer is also designed as a one-dimensional convolution. The number of output channels of the five convolutional layers is 16, 32, 64, 128 and 256 respectively, and the convolution kernel size is set to 3, and the step size is 2.
[0096] The first batch normalization layer (BN) can normalize each batch of data so that its mean is close to 0 and its variance is close to 1. This helps to solve the problem of internal covariate shift, makes the network training more stable, and improves the generalization ability of the network. The first activation function can select the LeakyReLU function, which is a nonlinear function that can improve the expressiveness of the model so that the model can better fit nonlinear data. At the same time, compared with the ReLU activation function, the LeakyReLU function sets a non-zero slope for the negative part, avoiding the problem of neuron death. In an embodiment of the present application, the slope set by the LeakyReLU function for the negative part is 0.2.
[0097] In the embodiment of the present application, the decoder includes a deconvolution layer (ConvTranspose), a batch normalization layer (BN) and an activation function, wherein the batch normalization layer and the activation function are recorded as the second batch normalization layer and the second activation function. Specifically, the deconvolution layer (ConvTranspose) is a way of upsampling. Contrary to the convolution layer, it can increase the dimension of the feature information and is used to upsample and generate a new ECG signal from the feature information in the latent space. In the embodiment of the present application, the deconvolution layer can be used to generate the remaining 11-lead ECG signals according to the feature representation of the input single-lead ECG signal in the latent space. Since the ECG signal data is one-dimensional data, the deconvolution layer is also designed as a one-dimensional deconvolution. The number of output channels of the five deconvolution layers is 128, 64, 32, 16 and 11, respectively, and the convolution kernel size is set to 3, with a step size of 2. The second activation function can still use the LeakyReLU function. The parameter settings and functions of the second batch normalization layer (BN) and the second activation function are the same as those of the encoder, and are not described here.
[0098] Specifically, in some embodiments, after obtaining the multi-lead ECG signals corresponding to the multiple sample persons, the method further includes:
[0099] Removing baseline drift noise in the multi-lead ECG signal by wavelet decomposition and reconstruction method;
[0100] The power frequency interference and myoelectric interference in the multi-lead electrocardiogram signal are removed by wavelet threshold method.
[0101] It should be pointed out that the original ECG signal will be interfered by various noises during the acquisition process, and these noises will have an adverse effect on the subsequent model training and the evaluation of the model effect. Therefore, in the embodiment of the present application, after the ECG signals are acquired, they can be denoised.
[0102] Specifically, common noise interferences in the original ECG signal include baseline drift noise, power frequency interference, and myoelectric interference. Baseline drift noise is generally generated by the breathing or movement of the person being measured. It is a low-frequency signal with a frequency range of about 0.05 to 1 Hz. Wavelet decomposition and reconstruction can be used to remove baseline drift noise. Taking the sampling frequency of 100 Hz as an example, the ECG signal can be decomposed into 6 levels using the db6 wavelet basis function. The baseline drift noise is mainly located in the frequency band corresponding to the 6th layer after wavelet decomposition, so the approximate coefficients and detail coefficients corresponding to the 6th layer can be directly forced to zero, thereby filtering out the baseline drift noise in the ECG signal.
[0103] Power frequency interference is caused by the influence of power lines and surrounding electrical equipment. It is a high-frequency noise with a frequency range of 50 to 60 Hz. Myoelectric interference is caused by the patient's muscle contraction and relaxation. It is a high-frequency noise with a frequency range of 5 to 500 Hz. For these two types of noise, the wavelet threshold method can be used to remove them. The ECG signal is decomposed at level 1 using the db6 wavelet basis function, and the threshold is calculated using the unbiased risk estimation method. The threshold is used to perform soft threshold denoising on the detail coefficients of each scale obtained after decomposition, and the denoised ECG signal can be obtained.
[0104] Specifically, in some embodiments, the effect evaluation of the ECG signal generation model trained under each lead mode, determining the performance index corresponding to the trained ECG signal generation model, and determining the optimal lead mode according to the performance index includes:
[0105] Determine the index values of the first prediction signal and the multi-lead ECG signal predicted in each lead mode by at least one index algorithm selected from the group consisting of mean square error, mean absolute error, and Pearson correlation coefficient;
[0106] The lead mode corresponding to the smallest mean square error and the smallest mean absolute error is determined as the optimal lead mode, or the lead mode corresponding to the largest Pearson correlation coefficient is determined as the optimal lead mode.
[0107] In an embodiment of the present application, when evaluating the effect of the ECG signal generation model, in some scenarios, indicators such as mean square error (MSE), mean absolute error (MAE) and Pearson Correlation Coefficient can be used for evaluation. The smaller the mean square error and the mean absolute error and the larger the Pearson correlation coefficient, the better the effect of the model. Specifically, please refer to Table 1 below. In an embodiment of the present application, taking the task of generating a complete 12-lead ECG signal from a single-lead ECG signal as an example, the corresponding ECG signal generation models under different lead modes are trained, and their performance indicators are evaluated. Table 1 shows some of the obtained indicator values.
[0108] Table 1
[0109]
[0110]
[0111]
[0112] It can be observed from Table 1 that when the ECG signals of 12 different lead modes are used as input, the average value of the mean square error of the remaining 11 lead ECG signals generated by the model is less than 0.02, the average value of the mean absolute error is less than 0.06, and the average value of the Pearson correlation coefficient is more than 0.7 compared with the real signal. In particular, when the V3 lead is used as input, the average value of the mean square error and the average value of the mean absolute error of the remaining 11 lead ECG signals generated by the model are the smallest compared with the real signal, which are 0.011 and 0.050 respectively; when the III lead is used as input, the average value of the mean square error and the average value of the mean absolute error of the remaining 11 lead ECG signals generated by the model are the largest compared with the real signal, which are 0.018 and 0.059 respectively. When lead II is used as input, the average value of the Pearson correlation coefficient of the remaining 11-lead ECG signals generated by the model is the largest compared with the true signal, which is 0.799; when lead V1 is used as input, the average value of the Pearson correlation coefficient of the remaining 11-lead ECG signals generated by the model is the smallest compared with the true signal, which is 0.712. In some embodiments, the lead mode with the smallest mean square error and mean absolute error can be determined as the optimal lead mode. For example, lead V3 can be determined as the optimal lead mode, and the ECG signal generation model trained under lead V3 can be determined as the target model. Of course, in some embodiments, the lead mode with the largest corresponding Pearson correlation coefficient can also be determined as the optimal lead mode, for example, lead II can be determined as the optimal lead mode, and the ECG signal generation model trained under lead II can be determined as the target model. In other embodiments, different indicator values can also be weighted to determine the optimal lead mode, and this application does not limit this.
[0113] Specifically, in some embodiments, the effect evaluation of the ECG signal generation model trained under each lead mode, determining the performance index corresponding to the trained ECG signal generation model, and determining the optimal lead mode according to the performance index includes:
[0114] Obtaining cardiovascular disease label information corresponding to each of the sample persons; the cardiovascular disease label information is used to characterize the true category of the cardiovascular disease possessed by the sample person;
[0115] According to the first prediction signal output by the trained electrocardiogram signal generation model in each lead mode, the cardiovascular disease of the sample person is predicted by the cardiovascular disease prediction model to obtain a first prediction result;
[0116] Determining the prediction accuracy corresponding to each of the lead modes according to the first prediction result and the cardiovascular disease label information;
[0117] The corresponding lead mode with the highest prediction accuracy is determined as the optimal lead mode.
[0118] In the embodiment of the present application, generally speaking, the generated ECG signal can be subsequently used in a cardiovascular disease prediction model. In order to improve the practicality of the target ECG signal generated in the embodiment of the present application, the ECG signal generation model can also be evaluated in combination with subsequent classification prediction tasks.
[0119] Specifically, in an embodiment of the present application, a cardiovascular disease prediction model can be used to predict the cardiovascular disease that a person has, and the input of the cardiovascular disease prediction model is an electrocardiogram signal, and the output is a predicted category of cardiovascular disease. In an embodiment of the present application, the cardiovascular disease label information corresponding to the sample person can be obtained, which can be used to characterize the true category of cardiovascular disease that the sample person has. Exemplarily, for example, the true category of cardiovascular disease can include health, myocardial infarction, ST-T segment changes, myocardial hypertrophy, conduction disorder, etc., and this application is not limited to this. After training to obtain each electrocardiogram signal generation model, the cardiovascular disease that the sample person has can be predicted by the cardiovascular disease prediction model according to the first prediction signal output by the trained electrocardiogram signal generation model and the input of the electrocardiogram signal generation model (the first electrocardiogram corresponding to the first lead mode), and a prediction result is obtained, which is recorded as the first prediction result. The first prediction result can characterize the category of cardiovascular disease that the sample person has predicted by the cardiovascular disease prediction model. In an embodiment of the present application, the prediction accuracy corresponding to each lead mode can be determined according to the first prediction result and the cardiovascular disease label information.
[0120] In order to verify the practicality of the ECG signal generation model in the diagnosis of cardiovascular diseases, a disease diagnosis classification experiment was carried out on the PTB database using the residual network as a classifier. The PTB database is a diagnostic database related to myocardial infarction, which is a common cardiovascular disease. In the embodiment of the present application, when performing classification, the selected classification performance evaluation indicators can be accuracy (Acc) and F1 score (F1), which are defined as follows:
[0121]
[0122] Among them, TP, TN, FP, and FN represent the number of true positives, true negatives, false positives, and false negatives, respectively.
[0123] For the convenience of evaluation, this application sets up two situations and compares them. Situation (1): directly using a single-lead ECG signal for disease diagnosis and classification; Situation (2): first using the pre-trained ECG signal generation model to generate a complete 12-lead ECG signal based on the single-lead ECG signal, and then using the complete ECG signal for diagnosis and classification. During the training process, the convolutional autoencoder parameters are frozen, and only the parameters of the residual network classifier are trained. The results are shown in Table 2.
[0124] Table 2
[0125]
[0126] As can be seen from Table 2, except for Lead aVF and Lead V1, the performance of the remaining 10 leads after the pre-trained ECG signal generation model generates a complete 12-lead signal, and then diagnoses and classifies the disease, and its performance is improved compared with the direct use of single-lead ECG signals for disease diagnosis and classification. Among them, when Lead I is used as the input signal, the application of the ECG signal generation model has the most obvious improvement in the performance of disease diagnosis and classification, with the accuracy and F1 score increased by 9.4% and 10.64% respectively; when Lead V6 is used as the input signal, the model has the best performance in disease diagnosis and classification, with the accuracy and F1 score being 43.04% and 37.37% respectively when the convolutional autoencoder is not used (case 1), and the accuracy is 48.21% and 42.88% after the convolutional autoencoder is used (case 2). It shows that the application of the ECG signal generation model can improve the accuracy of the overall model for myocardial infarction diagnosis, and has certain practical value. Therefore, the V6 lead can be determined as the optimal lead mode.
[0127] After comprehensive analysis, when diagnosing myocardial infarction, a typical cardiovascular disease, the single-lead ECG signal of lead V6 can be selected as the acquisition lead. By using the ECG signal generation model, a complete 12-lead signal can be generated based on the V6 lead signal, which can further improve the accuracy of the model in diagnosing myocardial infarction.
[0128] In an embodiment of the present application, for the target person, after obtaining the corresponding target ECG signal, it can also be input into a cardiovascular disease prediction model, and the cardiovascular disease prediction model is used to predict the cardiovascular disease of the target person to obtain a prediction result, which is recorded as the second prediction result.
[0129] In an embodiment of the present application, there is also provided an electrocardiogram signal generating device, comprising:
[0130] A first acquisition unit is used to acquire multi-lead ECG signals corresponding to multiple sample persons; wherein the multi-lead ECG signals include first ECG signals measured in various lead modes, and there are multiple types of lead modes;
[0131] A first prediction unit is used to input a first ECG signal corresponding to a first lead mode into an ECG signal generation model corresponding to the first lead mode, and obtain a first prediction signal through prediction by the ECG signal generation model; the first prediction signal is used to represent the ECG signal corresponding to the sample person predicted by the ECG signal generation model in a lead mode other than the first lead mode; wherein the first lead mode is any lead mode;
[0132] a processing unit, configured to determine a loss value of training according to the first prediction signal and the multi-lead ECG signal;
[0133] An updating unit, configured to update the parameters of the electrocardiogram signal generation model according to the loss value, so as to obtain the electrocardiogram signal generation model trained in the first lead mode;
[0134] An evaluation unit is used to evaluate the effects of the ECG signal generation models trained under various lead modes, determine the performance indicators corresponding to the trained ECG signal generation models, determine the optimal lead mode according to the performance indicators, and determine the ECG signal generation model trained under the optimal lead mode as the target model;
[0135] A second acquisition unit, used for acquiring a second electrocardiogram signal measured by the target person in the optimal lead mode;
[0136] A second prediction unit is used to input the second ECG signal into the target model, and obtain a second prediction signal through prediction by the target model; the second prediction signal is used to represent the ECG signal corresponding to the target person predicted by the target model in a lead mode other than the optimal lead mode;
[0137] The integration unit is used to splice the second ECG signal and the second prediction signal to obtain a generated target ECG signal.
[0138] In an embodiment of the present application, a cardiovascular disease diagnosis device is also provided, comprising:
[0139] An input unit, used for inputting the target ECG signal generated by the aforementioned ECG signal generation method into a cardiovascular disease prediction model;
[0140] The third prediction unit is used to predict the cardiovascular disease of the target person through the cardiovascular disease prediction model to obtain a second prediction result.
[0141] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0142] Reference Figure 6 , an embodiment of the present application provides an electronic device, including:
[0143] at least one processor 610;
[0144] At least one memory 620, used to store at least one program;
[0145] When at least one program is executed by at least one processor 610, at least one processor 610 implements the above-mentioned electrocardiogram signal generating method.
[0146] Similarly, the contents of the above method embodiments are all applicable to the electronic device embodiments. The functions specifically implemented by the electronic device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0147] The embodiment of the present application further provides a computer-readable storage medium, in which a program executable by the processor 610 is stored. The program executable by the processor 610 is used to execute the above-mentioned electrocardiogram signal generation method when executed by the processor 610.
[0148] Similarly, the contents of the above method embodiments are all applicable to the computer-readable storage medium embodiments. The functions specifically implemented by the computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0149] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the application is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. The optional embodiment is expected, wherein the order of various operations is changed and the sub-operation described as a part of a larger operation is performed independently.
[0150] In addition, although the present application is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present application. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional techniques of the engineer. Therefore, those skilled in the art can implement the present application set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present application, which is determined by the full scope of the attached claims and their equivalents.
[0151] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.
[0152] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.
[0153] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.
[0154] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0155] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0156] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the claims and their equivalents.
[0157] The above is a specific description of the preferred implementation of the present application, but the present application is not limited to the embodiments. Technical personnel familiar with the field can make various equivalent modifications or substitutions without violating the spirit of the present application. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present application.
Claims
1. A method for generating an electrocardiogram signal, characterized in that: The method comprises: Acquire multi-lead ECG signals corresponding to multiple sample persons; wherein the multi-lead ECG signals include first ECG signals measured in various lead modes, and there are multiple types of lead modes; Inputting a first ECG signal corresponding to a first lead mode into an ECG signal generation model corresponding to the first lead mode, and obtaining a first prediction signal through prediction by the ECG signal generation model; the first prediction signal is used to represent the ECG signal corresponding to the sample person predicted by the ECG signal generation model in a lead mode other than the first lead mode; wherein the first lead mode is any lead mode; Determining a training loss value according to the first prediction signal and the multi-lead ECG signal; According to the loss value, updating the parameters of the electrocardiogram signal generation model to obtain the electrocardiogram signal generation model trained in the first lead mode; Evaluate the effects of the ECG signal generation models trained under various lead modes, determine the performance indicators corresponding to the trained ECG signal generation models, determine the optimal lead mode according to the performance indicators, and determine the ECG signal generation model trained under the optimal lead mode as the target model; Acquiring a second electrocardiogram signal measured by the target person in the optimal lead mode; Inputting the second ECG signal into the target model, and obtaining a second prediction signal through prediction by the target model; the second prediction signal is used to represent the ECG signal corresponding to the target person predicted by the target model under a lead mode other than the optimal lead mode; The second electrocardiogram signal and the second prediction signal are concatenated to obtain a generated target electrocardiogram signal.
2. The method for generating an electrocardiogram signal according to claim 1, characterized in that: The ECG signal generation model is built through a convolutional autoencoder.
3. The method for generating an electrocardiogram signal according to claim 2, characterized in that: The convolutional autoencoder includes an encoder and a decoder; The encoder comprises a convolutional layer, a first batch normalization layer, and a first activation function; The decoder includes a deconvolution layer, a second batch normalization layer, and a second activation function.
4. The method for generating an electrocardiogram signal according to claim 1, characterized in that: After obtaining the multi-lead ECG signals corresponding to the multiple sample persons, the method further includes: Removing baseline drift noise in the multi-lead ECG signal by wavelet decomposition and reconstruction method; The power frequency interference and myoelectric interference in the multi-lead electrocardiogram signal are removed by wavelet threshold method.
5. The method for generating an electrocardiogram signal according to claim 1, characterized in that: The effect evaluation of the ECG signal generation model trained under each lead mode is performed, the performance index corresponding to the trained ECG signal generation model is determined, and the optimal lead mode is determined according to the performance index, including: Determine the index values of the first prediction signal and the multi-lead ECG signal predicted in each lead mode by at least one index algorithm selected from the group consisting of mean square error, mean absolute error, and Pearson correlation coefficient; The lead mode corresponding to the smallest mean square error and the smallest mean absolute error is determined as the optimal lead mode, or the lead mode corresponding to the largest Pearson correlation coefficient is determined as the optimal lead mode.
6. The method for generating an electrocardiogram signal according to claim 1, characterized in that: The effect evaluation of the ECG signal generation model trained under each lead mode is performed, the performance index corresponding to the trained ECG signal generation model is determined, and the optimal lead mode is determined according to the performance index, including: Obtaining cardiovascular disease label information corresponding to each of the sample persons; the cardiovascular disease label information is used to characterize the true category of the cardiovascular disease possessed by the sample person; According to the first prediction signal output by the trained electrocardiogram signal generation model in each lead mode, the cardiovascular disease of the sample person is predicted by the cardiovascular disease prediction model to obtain a first prediction result; Determining the prediction accuracy corresponding to each of the lead modes according to the first prediction result and the cardiovascular disease label information; The corresponding lead mode with the highest prediction accuracy is determined as the optimal lead mode.
7. An electrocardiogram signal generating device, characterized in that: The device comprises: A first acquisition unit is used to acquire multi-lead ECG signals corresponding to multiple sample persons; wherein the multi-lead ECG signals include first ECG signals measured in various lead modes, and there are multiple types of lead modes; A first prediction unit is used to input a first ECG signal corresponding to a first lead mode into an ECG signal generation model corresponding to the first lead mode, and obtain a first prediction signal through prediction by the ECG signal generation model; the first prediction signal is used to represent the ECG signal corresponding to the sample person predicted by the ECG signal generation model in a lead mode other than the first lead mode; wherein the first lead mode is any lead mode; a processing unit, configured to determine a loss value of training according to the first prediction signal and the multi-lead ECG signal; An updating unit, configured to update the parameters of the electrocardiogram signal generation model according to the loss value, so as to obtain the electrocardiogram signal generation model trained in the first lead mode; An evaluation unit is used to evaluate the effects of the ECG signal generation models trained under various lead modes, determine the performance indicators corresponding to the trained ECG signal generation models, determine the optimal lead mode according to the performance indicators, and determine the ECG signal generation model trained under the optimal lead mode as the target model; A second acquisition unit, used for acquiring a second electrocardiogram signal measured by the target person in the optimal lead mode; A second prediction unit is used to input the second ECG signal into the target model, and obtain a second prediction signal through prediction by the target model; the second prediction signal is used to represent the ECG signal corresponding to the target person predicted by the target model in a lead mode other than the optimal lead mode; The integration unit is used to splice the second ECG signal and the second prediction signal to obtain a generated target ECG signal.
8. A cardiovascular disease diagnosis device, characterized in that: The device comprises: An input unit, used for inputting a target ECG signal generated by the ECG signal generation method according to any one of claims 1 to 6 into a cardiovascular disease prediction model; The third prediction unit is used to predict the cardiovascular disease of the target person through the cardiovascular disease prediction model to obtain a second prediction result.
9. An electronic device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the electrocardiogram signal generating method according to any one of claims 1 to 6.
10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to implement an electrocardiogram signal generation method according to any one of claims 1 to 6 when executed by the processor.
Citation Information
Patent Citations
Method and device for processing multi-lead synchronized electrocardiosignal
CN101467879A
Lead selection method for emotional electroencephalogram based on independent component analysis
CN108937968A
Electrocardio lead expansion method and device
CN110507314A
Depression detection system based on multi-lead EEG optimal lead selection
CN110876626A
Cascaded binary classifier for identifying rhythms in a single-lead electrocardiogram (ECG) signal
US20190082988A1
