A method and system for identifying I-lead electrocardiogram data
By building and setting a neural network model in I-lead ECG devices, the recognition problem of multi-lead devices without network is solved, and the autonomous recognition and convenient operation of I-lead ECG data is realized, which improves the accuracy of identification rate and efficiency.
Patent Information
- Application Number
- CN202110268180.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-03-11
- Publication Date
- 2024-12-27
- Estimated Expiration
- 2041-03-11
AI Technical Summary
In the prior art, multi-lead electrocardiogram devices cannot perform real-time identification of electrocardiogram data when there is no network connection, and the operation steps for the elderly are cumbersome and inconvenient.
Using the I-lead ECG data recognition method, by constructing the I-lead ECG data identification neural network model and setting it in the I-lead ECG device, the device can independently collect and identify the ECG data and avoid sending it to the cloud server. It is processed and extracted features through ResECG11 deep convolutional neural network, quantized and compressed and burned into an AI chip to achieve offline recognition.
It realizes independent identification of ECG data in the absence of network, reduces operation steps, improves convenience, and improves identification accuracy and efficiency.
Smart Images
Figure CN113040781B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of electrocardiogram data recognition, and in particular to a method and system for recognizing lead I electrocardiogram data. Background Art
[0002] Cardiovascular disease is one of the most difficult diseases to cure at present. In order to effectively prevent cardiovascular disease, ECG data is generally monitored and analyzed in real time, so as to prevent it by identifying whether the ECG data is abnormal. In the existing technology, multi-lead ECG equipment is generally used to collect ECG data, and the ECG data is sent to a cloud server, and the artificial intelligence algorithm running in the cloud server is used to interpret it, thereby realizing the recognition of ECG data.
[0003] However, in order to collect ECG data through multi-lead ECG equipment and send the ECG data to a cloud server for operation, the multi-lead ECG device needs to be connected to the Internet or a mobile phone for use. When there is no Internet, the data cannot be sent to the cloud server for calculation and recognition. Moreover, connecting a multi-lead ECG device through a mobile phone often requires connecting the terminal device in the mobile application, applying for an account and other cumbersome steps. For the elderly, it is not so intuitive and convenient to use. Summary of the invention
[0004] In view of this, the present invention proposes a method and system for identifying lead I ECG data, which can solve the defects of existing ECG data identification, such as being unable to calculate and identify without a network and being troublesome to use.
[0005] The technical solution of the present invention is achieved in this way:
[0006] A method for identifying I-lead ECG data, the method being based on an I-lead ECG device and comprising the following steps:
[0007] Step S1, constructing an I-lead ECG data recognition neural network model, wherein the input end of the I-lead ECG data recognition neural network model is the I-lead ECG data, and the output end of the I-lead ECG data recognition neural network model is a judgment result of whether the I-lead ECG data is abnormal;
[0008] Step S2, setting the constructed I-lead ECG data recognition neural network model in the I-lead ECG device;
[0009] Step S3, the I-lead ECG device collects the I-lead ECG data to be identified, and identifies it through the I-lead ECG data identification neural network model;
[0010] Step S4, the I-lead ECG data recognition neural network model outputs a judgment result of the I-lead ECG data to be recognized, and the I-lead ECG device displays according to the judgment result.
[0011] As a further optional solution of the I-lead ECG data recognition method, the construction of the I-lead ECG data recognition neural network model in step S1 specifically includes the following steps:
[0012] Step S11, collecting I-lead ECG data and its corresponding ECG judgment result;
[0013] Step S12, preprocessing the I-lead ECG data to obtain a training data matrix;
[0014] Step S13, inputting the training data matrix and the ECG judgment result into a deep convolutional neural network model for training to obtain an I-lead ECG data recognition neural network model.
[0015] As a further optional solution of the I-lead ECG data recognition method, the deep convolutional neural network model adopts the ResECG11 deep convolutional neural network.
[0016] As a further optional solution of the I-lead ECG data recognition method, the ResECG11 deep convolutional neural network includes a convolution layer, a normalization layer, an orthogonal linear unit layer, an identity mapping module and a fully connected layer.
[0017] As a further optional solution of the I-lead ECG data recognition method, step S12 includes the following steps:
[0018] Step S121, intercepting the I-lead ECG data to obtain intercepted I-lead ECG data;
[0019] Step S122, processing the intercepted I lead ECG data through a linear interpolation algorithm to obtain a training data matrix.
[0020] As a further optional solution of the I-lead ECG data recognition method, step S2 includes the following steps:
[0021] Step S21, quantizing and compressing the I-lead ECG data recognition neural network model;
[0022] Step S22, burning the quantized and compressed I-lead ECG data recognition neural network model into the AI chip of the I-lead ECG device.
[0023] As a further optional solution of the I-lead ECG data recognition method, step S3 includes the following steps:
[0024] Step S31, the I-lead ECG device collects the I-lead ECG data to be identified;
[0025] Step S32, preprocessing the I-lead ECG data to be identified to obtain a data matrix to be identified;
[0026] Step S33, inputting the data matrix to be identified into the I-lead ECG data identification neural network model for identification.
[0027] As a further optional solution of the I-lead ECG data identification method, step S32 includes the following steps:
[0028] Step S321, intercepting the I-lead ECG data to be identified to obtain intercepted I-lead ECG data to be identified;
[0029] Step S322, processing the intercepted I-lead ECG data to be identified by a linear interpolation algorithm to obtain a data matrix to be identified.
[0030] As a further optional solution of the I-lead ECG data recognition method, step S4 includes the following steps:
[0031] Step S41, the I-lead ECG data recognition neural network model outputs a judgment result of the I-lead ECG data to be recognized, wherein the judgment result includes whether the I-lead ECG data to be recognized is normal or whether the I-lead ECG data to be recognized is abnormal;
[0032] Step S42: If the judgment result is that the I-lead ECG data to be identified is normal, the I-lead ECG device displays the data as normal; if the judgment result is that the I-lead ECG data to be identified is abnormal, the I-lead ECG device displays the data as abnormal and prompts the user to seek medical treatment in time.
[0033] A I-lead electrocardiogram data recognition system, the system using any one of the above recognition methods.
[0034] The beneficial effects of the present invention are as follows: by constructing an I-lead ECG data recognition neural network model and setting the I-lead ECG data recognition neural network model in an I-lead ECG device, the I-lead ECG device can autonomously recognize the collected I-lead ECG data, avoiding sending it to a cloud server for processing, thereby avoiding the situation in the prior art that the data cannot be sent to a cloud server for calculation and recognition without a network. In addition, by adopting an I-lead ECG device, the effect of all-weather use can be achieved, thereby reducing the steps of operation for the elderly and improving convenience. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0036] Figure 1 The figure is a flow chart of a method for identifying lead I ECG data according to the present invention. DETAILED DESCRIPTION
[0037] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0038] refer to Figure 1 , a method for identifying I-lead ECG data, the method is based on an I-lead ECG device, and comprises the following steps:
[0039] Step S1, constructing an I-lead ECG data recognition neural network model, wherein the input end of the I-lead ECG data recognition neural network model is the I-lead ECG data, and the output end of the I-lead ECG data recognition neural network model is a judgment result of whether the I-lead ECG data is abnormal;
[0040] Step S2, setting the constructed I-lead ECG data recognition neural network model in the I-lead ECG device;
[0041] Step S3, the I-lead ECG device collects the I-lead ECG data to be identified, and identifies it through the I-lead ECG data identification neural network model;
[0042] Step S4, the I-lead ECG data recognition neural network model outputs a judgment result of the I-lead ECG data to be recognized, and the I-lead ECG device displays according to the judgment result.
[0043] In this embodiment, by constructing an I-lead ECG data recognition neural network model and setting the I-lead ECG data recognition neural network model in an I-lead ECG device, the I-lead ECG device can autonomously identify the collected I-lead ECG data, avoiding sending it to a cloud server for processing, thereby avoiding the situation in the prior art that the data cannot be sent to a cloud server for calculation and recognition without a network. In addition, by adopting an I-lead ECG device, the effect of all-weather use can be achieved, thereby reducing the operation steps of the elderly and improving convenience.
[0044] Preferably, the step S1 of constructing a neural network model for identifying lead I ECG data specifically includes the following steps:
[0045] Step S11, collecting I-lead ECG data and its corresponding ECG judgment result;
[0046] Step S12, preprocessing the I-lead ECG data to obtain a training data matrix;
[0047] Step S13, inputting the training data matrix and the ECG judgment result into a deep convolutional neural network model for training to obtain an I-lead ECG data recognition neural network model.
[0048] In this embodiment, by preprocessing the I-lead ECG data, the data collected each time can obtain input data with the same time length and the same sampling rate, so as to obtain an accurate training data matrix based on the input data, and by inputting the accurate training matrix data into the deep convolutional neural network model for training, the recognition accuracy of the I-lead ECG data recognition neural network model can be improved; it should be noted that the same time length is 30s, the same sampling rate is 320Hz, and the shape of the training data matrix is [1, 9600].
[0049] Preferably, the deep convolutional neural network model adopts the ResECG11 deep convolutional neural network.
[0050] In this embodiment, by adopting the ResECG11 deep convolutional neural network, the characteristics of the I-lead ECG data can be accurately extracted, thereby improving the training effect and further improving the recognition effect of the constructed I-lead ECG data recognition neural network model.
[0051] Preferably, the ResECG11 deep convolutional neural network includes a convolutional layer, a normalization layer, an orthogonal linear unit layer, an identity mapping module and a fully connected layer.
[0052] In this embodiment, the I-lead ECG data is subjected to feature abstraction extraction through the convolution layer, and after multi-channel abstraction, it is subjected to the normalization layer. Normalization is equivalent to redistributing the obtained data in the standard normal distribution, and is also the process of optimizing the training effect. The purpose is to reduce the model collapse during training, so as to obtain a better model, and to better realize the process of gradually improving the expressiveness in the residual network. The orthogonal linear unit layer is a process of connecting each residual block and weighting the output of the residual block of the previous level, so that important information can be more consistently transmitted backward, and the total number of weights can be reduced to obtain better results. The final fully connected layer outputs the features linearly to obtain the final state value, which ranges between 0 and 1.
[0053] Preferably, the step S12 comprises the following steps:
[0054] Step S121, intercepting the I-lead ECG data to obtain intercepted I-lead ECG data;
[0055] Step S122, processing the intercepted I lead ECG data through a linear interpolation algorithm to obtain a training data matrix.
[0056] In this embodiment, the collected I-lead ECG data is intercepted so that the input is intercepted as 30 seconds of I-lead ECG data. If the ECG data collection process is less than 30 seconds, it will be judged as an invalid collection process, and the user will be prompted to re-measure and collect. The intercepted I-lead ECG data is processed into I-lead ECG data with a sampling rate of 320 Hz through a linear interpolation algorithm, thereby forming a training data matrix.
[0057] Preferably, step S2 comprises the following steps:
[0058] Step S21, quantizing and compressing the I-lead ECG data recognition neural network model;
[0059] Step S22, burning the quantized and compressed I-lead ECG data recognition neural network model into the AI chip of the I-lead ECG device.
[0060] In this embodiment, in order to be able to use it in the chip system, the file size of the I-lead ECG data recognition neural network model should not be too large. By truncating each weight storage space from 32 bits to 8 bits and adjusting the output result through statistical methods, a neural network model with 1 / 4 the size of the original model can be obtained; through the burning software provided by the chip manufacturer, the neural network is burned into the chip system and connected to the data input of the ECG measurement module, the local ECG recognition function can be realized.
[0061] Preferably, step S3 comprises the following steps:
[0062] Step S31, the I-lead ECG device collects the I-lead ECG data to be identified;
[0063] Step S32, preprocessing the I-lead ECG data to be identified to obtain a data matrix to be identified;
[0064] Step S33, inputting the data matrix to be identified into the I-lead ECG data identification neural network model for identification.
[0065] In this embodiment, by preprocessing the I-lead ECG data to be identified, the workload of the I-lead ECG data identification neural network model can be reduced, thereby improving the recognition accuracy and efficiency of the I-lead ECG data identification neural network model.
[0066] Preferably, the step S32 comprises the following steps:
[0067] Step S321, intercepting the I-lead ECG data to be identified to obtain intercepted I-lead ECG data to be identified;
[0068] Step S322, processing the intercepted I-lead ECG data to be identified by a linear interpolation algorithm to obtain a data matrix to be identified.
[0069] Preferably, step S4 comprises the following steps:
[0070] Step S41, the I-lead ECG data recognition neural network model outputs a judgment result of the I-lead ECG data to be recognized, wherein the judgment result includes whether the I-lead ECG data to be recognized is normal or whether the I-lead ECG data to be recognized is abnormal;
[0071] Step S42: If the judgment result is that the I-lead ECG data to be identified is normal, the I-lead ECG device displays the data as normal; if the judgment result is that the I-lead ECG data to be identified is abnormal, the I-lead ECG device displays the data as abnormal and prompts the user to seek medical treatment in time.
[0072] In this embodiment, the identified I-lead ECG data abnormalities include ventricular premature beats, atrial premature beats, atrial fibrillation, atrial flutter, complete left bundle branch block, complete right bundle branch block, sinus bradycardia and sinus tachycardia. If the judgment result is that the I-lead ECG data to be identified is abnormal, the I-lead ECG device displays the data abnormality and prompts the user to seek medical treatment in time, otherwise the I-lead ECG device displays the data normal.
[0073] A I-lead electrocardiogram data recognition system, the system using any one of the above recognition methods.
[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for identifying I-lead electrocardiogram data, characterized in that: The method is based on an I-lead ECG device and includes the following steps: Step S1, constructing an I-lead ECG data recognition neural network model, wherein the input end of the I-lead ECG data recognition neural network model is the I-lead ECG data, and the output end of the I-lead ECG data recognition neural network model is a judgment result of whether the I-lead ECG data is abnormal; Step S2, setting the constructed I-lead ECG data recognition neural network model in the I-lead ECG device; Step S3, the I-lead ECG device collects the I-lead ECG data to be identified, and identifies it through the I-lead ECG data identification neural network model; Step S4, the I-lead ECG data recognition neural network model outputs the judgment result of the I-lead ECG data to be identified, and the I-lead ECG device displays the judgment result according to the result; The step S1 of constructing the I-lead ECG data recognition neural network model specifically includes the following steps: Step S11, collecting I-lead ECG data and its corresponding ECG judgment result; Step S12, preprocessing the I-lead ECG data to obtain a training data matrix; Step S13, inputting the training data matrix and the ECG judgment result into a deep convolutional neural network model for training to obtain an I-lead ECG data recognition neural network model; The deep convolutional neural network model adopts the ResECG11 deep convolutional neural network; The ResECG11 deep convolutional neural network includes a convolutional layer, a normalization layer, an orthogonal linear unit layer, an identity mapping module and a fully connected layer; The step S2 comprises the following steps: Step S21, quantizing and compressing the I-lead ECG data recognition neural network model, wherein the quantization and compression is to truncate each weight storage space from 32 bits to 8 bits, and adjust the output result by statistical means to obtain an I-lead ECG data recognition neural network model with a size of 1 / 4 of the original model; Step S22, burning the quantized and compressed I-lead ECG data recognition neural network model into the AI chip of the I-lead ECG device.
2. A method for identifying lead I ECG data according to claim 1, characterized in that: The step S12 comprises the following steps: Step S121, intercepting the I-lead ECG data to obtain intercepted I-lead ECG data; Step S122, processing the intercepted I lead ECG data through a linear interpolation algorithm to obtain a training data matrix.
3. A method for identifying lead I ECG data according to claim 2, characterized in that: The step S3 comprises the following steps: Step S31, the I-lead ECG device collects the I-lead ECG data to be identified; Step S32, preprocessing the I-lead ECG data to be identified to obtain a data matrix to be identified; Step S33, inputting the data matrix to be identified into the I-lead ECG data identification neural network model for identification.
4. A method for identifying lead I ECG data according to claim 3, characterized in that: The step S32 comprises the following steps: Step S321, intercepting the I-lead ECG data to be identified to obtain intercepted I-lead ECG data to be identified; Step S322, processing the intercepted I-lead ECG data to be identified by a linear interpolation algorithm to obtain a data matrix to be identified.
5. A method for identifying lead I ECG data according to claim 4, characterized in that: The step S4 comprises the following steps: Step S41, the I-lead ECG data recognition neural network model outputs a judgment result of the I-lead ECG data to be recognized, wherein the judgment result includes whether the I-lead ECG data to be recognized is normal or whether the I-lead ECG data to be recognized is abnormal; Step S42: If the judgment result is that the I-lead ECG data to be identified is normal, the I-lead ECG device displays the data as normal; if the judgment result is that the I-lead ECG data to be identified is abnormal, the I-lead ECG device displays the data as abnormal and prompts the user to seek medical treatment in time.
6. A I-lead ECG data recognition system, characterized in that: The system uses the I-lead ECG data recognition method according to any one of claims 1-5.