Electrocardiogram generation system and method based on deep learning algorithm

Abstract

The present application relates to a kind of electrocardiogram generation system and method based on deep learning algorithm.According to the present application, electrocardiogram generation system based on deep learning algorithm includes: data input unit, receives 12-lead electrocardiogram measured by multiple patients;Data extraction unit extracts learning data from the input 12-lead electrocardiogram;Learning unit learns the characteristics of electrocardiogram by inputting the extracted learning data into multiple learning models;Electrocardiogram generation unit receives more than one reference electrocardiogram from the measured object, generates virtual electrocardiogram by inputting the input reference electrocardiogram into multiple learning models that have completed learning;And control unit, the reference electrocardiogram and the generated virtual electrocardiogram are synchronized with each other, and the waveforms of synchronous reference electrocardiogram and virtual electrocardiogram are output.

Description

Technical Field

[0001] This invention relates to an electrocardiogram (ECG) generation system and method based on a deep learning algorithm, and more specifically, to an ECG generation system and method that utilizes a deep learning algorithm to generate multiple ECGs from one or more leads. Background Technology

[0002] The standard 12-lead electrocardiogram used in hospitals collects and synthesizes information from all 12 leads to diagnose diseases by attaching six electrodes to the anterior surface of the chest and three electrodes to each of the four limbs (or four electrodes if grounding is included).

[0003] A 12-lead electrocardiogram records the heart's electrical potential in twelve electrical directions with the heart as the center. This allows for the assessment of the heart's condition in each direction, thus enabling accurate readings of heart diseases localized to a specific area.

[0004] The significance of measuring cardiac potential in multiple directions is that it allows us to understand the characteristics of the heart from various angles. Therefore, it is recommended to measure a standard 12-lead electrocardiogram in the medical process for the diagnosis of heart diseases (such as myocardial infarction).

[0005] However, to take a 12-lead electrocardiogram, the chest needs to be exposed so that chest electrodes can be attached. It is difficult for ordinary people to attach nine electrodes (three on the limbs and six on the chest) in the correct positions, making it difficult to perform measurements at home or in daily life. Furthermore, because it is difficult to attach ten electrodes and move them, it is difficult to use for real-time monitoring.

[0006] Therefore, devices that can measure one-lead or two-lead electrocardiograms have recently been developed so that they can be used in everyday life.

[0007] First, one-lead ECG devices using two electrodes include watch-type ECG devices (Apple Watch or Galaxy Watch). In a watch-type ECG device, the back of the watch contacts the left wrist, and the fingers of the right hand contact the crown of the watch, thereby bringing the left arm electrode into contact with the right arm electrode, and using the potential difference between the two electrodes to measure the lead I ECG.

[0008] Furthermore, the watch-style electrocardiograph measures lead I by wearing it on the left arm and touching the crown with the right hand. With the watch resting on the abdomen, lead II is measured by touching the crown with the right hand, and lead III is measured by touching the crown with the left hand while the watch is resting on the abdomen. Then, with the left hand touching the crown, the back of the watch is placed against the V1-V6 electrodes to measure the electrocardiogram of leads V1-V6.

[0009] The above method has poor usability because it requires the user to accurately contact the V1-V6 leads of the electrocardiogram. In the case of V1-V6 leads, unlike the standard chest lead electrocardiogram which requires showing the potential difference between the virtual center point and the V electrode even if the corresponding position is contacted, the problem is that the standard chest lead cannot show the potential difference between the right arm electrode and the V electrode.

[0010] Furthermore, by holding the electrodes with both hands and touching the electrodes on the back of the device to the legs or ankles, and with the electrodes in contact with the left arm, right arm, and left leg respectively, it is possible to measure two or more leads of ECG. However, the above method is not suitable for long-term monitoring (1 hour, 24 hours, 7 days, etc.), and its disadvantage is that the device itself needs to have three electrodes.

[0011] The background technology of this invention is disclosed in Korean Patent No. 10-2180135 (published on November 17, 2020). Summary of the Invention

[0012] Technical issues

[0013] Therefore, according to the present invention, an electrocardiogram (ECG) generation system and method are provided that utilize deep learning algorithms to generate multiple ECGs from one or more leads.

[0014] Technical solution

[0015] To address this technical problem, according to an embodiment of the present invention, an electrocardiogram (ECG) generation system based on a deep learning algorithm includes: a data input unit that receives 12-lead ECGs measured by multiple patients and patient information corresponding to the 12-lead ECGs; a data extraction unit that classifies and stores the input 12-lead ECGs according to the patient information and extracts learning data from the stored 12-lead ECGs; a learning unit that learns ECG features by inputting the extracted learning data into one or more learning models; an ECG generation unit that receives one or more reference ECGs from the measurement object and generates a virtual ECG by inputting the input reference ECGs into one or more learning models that have completed learning; and a control unit that synchronizes the reference ECGs and the generated virtual ECGs and outputs the waveforms of the synchronized reference ECGs and virtual ECGs.

[0016] The patient information may include at least one of the following: gender, age, whether the patient has heart disease, and the measured electrocardiogram potential vector.

[0017] The learning unit can input n leads of ECG, which are part of the 6-lead ECG of the limbs and the 6-lead ECG of the chest, into the first learning model to enable the first learning model to learn, so as to identify the potential vector of the input ECG and generate 12-n ECGs.

[0018] The aforementioned learning unit can construct 12 second learning models for 6-lead electrocardiograms of the limbs and 6-lead electrocardiograms of the chest. When the 12-n electrocardiograms generated by the constructed first learning model are input into the second learning model, the second learning model learns to transform the input electrocardiograms into patterns with corresponding potential vectors and outputs them as n virtual electrocardiograms.

[0019] The first and second learning models described above can be constructed by using a hybrid adversarial generative network and autoencoder approach.

[0020] The aforementioned electrocardiogram (ECG) generation unit can generate multiple remaining virtual ECGs by inputting an input reference ECG into a first learning model and extracting the potential vector of the reference ECG. The generated virtual ECGs can then be input into a second learning model learned through the extracted potential vector to regenerate virtual ECGs with the same leads as the reference ECG.

[0021] The control unit matches and synchronizes the reference electrocardiogram with the multiple virtual electrocardiograms, and outputs the synchronized multiple electrocardiograms through the monitor.

[0022] Furthermore, according to embodiments of the present invention, an electrocardiogram (ECG) generation method using an ECG generation system includes the following steps: the ECG generation system based on a deep learning algorithm receives 12-lead ECGs measured by multiple patients and patient information corresponding to the 12-lead ECGs; classifies and stores the input 12-lead ECGs according to the patient information, and extracts learning data from the stored 12-lead ECGs; inputs the extracted learning data into one or more learning models to enable them to learn the features of the ECGs; receives one or more reference ECGs from the measurement object, and generates a virtual ECG by inputting the input reference ECGs into one or more learning models that have completed learning; and synchronizes the reference ECGs and the generated virtual ECGs to each other, and outputs the waveforms of the synchronized reference ECGs and the virtual ECGs.

[0023] The effects of the invention

[0024] Therefore, according to the present invention, multiple synchronized electrocardiograms can be generated from two electrocardiograms measured at different points using deep learning algorithms, thus enabling accurate readings of heart disease with the same precision as a 12-lead electrocardiogram.

[0025] Furthermore, according to the present invention, since two-lead electrocardiograms are used, measurements can be taken at home or in daily life, even while on the move. Therefore, real-time monitoring is possible. Synchronous electrocardiograms are generated using electrocardiogram information measured at different time points, allowing medical personnel to read synchronous lead electrocardiogram information that matches the corresponding fluctuations and make more accurate diagnoses based on this. Attached Figure Description

[0026] Figure 1 This is a structural diagram illustrating the electrocardiogram (ECG) generation system of an embodiment of the present invention.

[0027] Figure 2 This is a flowchart illustrating the electrocardiogram (ECG) generation method using an ECG generation system according to an embodiment of the present invention.

[0028] Figure 3 An illustrative diagram illustrating the measurement method of a standard 12-lead electrocardiogram.

[0029] Figure 4 For illustrative purposes Figure 2 An example diagram of step S240 is shown.

[0030] Figure 5 For illustrative purposes Figure 2 An example diagram of step S260 is shown.

[0031] Best practice

[0032] The electrocardiogram (ECG) generation system 100 of this invention includes a data input unit 110, a data extraction unit 120, a learning unit 130, an ECG generation unit 140, and a control unit 150.

[0033] First, the data input unit 110 receives 12-lead electrocardiograms measured from multiple patients and information about the patients corresponding to the 12-lead electrocardiograms.

[0034] The 12-lead electrocardiogram includes a 6-lead electrocardiogram of the limbs and a 6-lead electrocardiogram of the chest. The patient's information includes at least one of the following: gender, age, whether the patient has heart disease, and the measured electrocardiogram potential vector.

[0035] The data extraction unit 120 classifies the 12-lead electrocardiograms using the input patient information and stores them in a database. Furthermore, the data extraction unit 120 generates learning data by randomly extracting multiple 12-lead electrocardiograms stored in the database.

[0036] Next, the learning unit 130 uses the learning data to enable the constructed learning model to learn. More specifically, the learning unit 130 constructs a first learning model and a second learning model. The first learning model generates 12-n lead virtual electrocardiograms from n lead reference electrocardiograms by extracting the potential vector information of the input lead electrocardiograms. The second learning model generates n lead virtual electrocardiograms from the 12-n lead virtual electrocardiograms.

[0037] Furthermore, the learning unit 130 enables the first learning model or the second learning model to learn by inputting the generated learning data into the constructed first learning model or the second learning model.

[0038] The electrocardiogram (ECG) generation unit 140 acquires one or more reference ECGs from the measurement subject. Furthermore, the ECG generation unit 140 generates a first virtual ECG by inputting the acquired reference ECGs into a first learning model that has completed learning, and based on the potential vectors of the reference ECGs. The ECG generation unit 140 inputs the generated first virtual ECG into a second learning model that has completed learning, causing the second learning model to output a second virtual ECG.

[0039] Finally, the control unit 150 synchronizes the reference electrocardiogram with the virtual electrocardiogram and outputs multiple synchronized electrocardiograms from the reference electrocardiogram and the virtual electrocardiogram through the monitoring device. Detailed Implementation

[0040] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this process, for clarity and convenience, the thickness of the lines or the dimensions of the components shown in the drawings may be enlarged.

[0041] Furthermore, the terms used below, defined in consideration of their function in this invention, may vary depending on the intentions or practices of the user or operator. Therefore, the definitions of these terms should be based on the entirety of this specification.

[0042] The following uses Figure 1 The electrocardiogram generation system based on deep learning algorithm according to the embodiments of the present invention will be described in more detail.

[0043] Figure 1 This is a structural diagram illustrating the electrocardiogram (ECG) generation system of an embodiment of the present invention.

[0044] like Figure 1 As shown, the electrocardiogram (ECG) generation system 100 of this embodiment includes a data input unit 110, a data extraction unit 120, a learning unit 130, an ECG generation unit 140, and a control unit 150.

[0045] First, the data input unit 110 receives 12-lead electrocardiograms measured from multiple patients and information about the patients corresponding to the 12-lead electrocardiograms.

[0046] The 12-lead electrocardiogram includes a 6-lead electrocardiogram of the limbs and a 6-lead electrocardiogram of the chest. The patient's information includes at least one of the following: gender, age, whether the patient has heart disease, and the measured electrocardiogram potential vector.

[0047] The data extraction unit 120 classifies the 12-lead electrocardiograms using the input patient information and stores them in a database. Furthermore, the data extraction unit 120 generates learning data by randomly extracting multiple 12-lead electrocardiograms stored in the database.

[0048] Next, the learning unit 130 uses the learning data to enable the constructed learning model to learn. More specifically, the learning unit 130 constructs a first learning model and a second learning model. The first learning model generates 12-n lead virtual electrocardiograms from n lead reference electrocardiograms by extracting the potential vector information of the input lead electrocardiograms. The second learning model generates n lead virtual electrocardiograms from the 12-n lead virtual electrocardiograms.

[0049] Furthermore, the learning unit 130 enables the first learning model or the second learning model to learn by inputting the generated learning data into the constructed first learning model or the second learning model.

[0050] The electrocardiogram (ECG) generation unit 140 acquires one or more reference ECGs from the measurement subject. Furthermore, the ECG generation unit 140 generates a first virtual ECG by inputting the acquired reference ECGs into a first learning model that has completed learning, and based on the potential vectors of the reference ECGs. The ECG generation unit 140 inputs the generated first virtual ECG into a second learning model that has completed learning, causing the second learning model to output a second virtual ECG.

[0051] Finally, the control unit 150 synchronizes the reference electrocardiogram with the virtual electrocardiogram and outputs multiple synchronized electrocardiograms from the reference electrocardiogram and the virtual electrocardiogram through the monitoring device.

[0052] The following uses Figures 2 to 5 The electrocardiogram (ECG) generation method using the ECG generation system 100 according to an embodiment of the present invention will be described in more detail.

[0053] Figure 2 This is a flowchart illustrating the electrocardiogram (ECG) generation method using an ECG generation system according to an embodiment of the present invention.

[0054] like Figure 2As shown, the electrocardiogram (ECG) generation method using the ECG generation system in this embodiment of the invention is divided into a step of enabling the learning model to learn and a step of generating an ECG using the learned model that has completed learning.

[0055] In the step of enabling the learning model to learn, firstly, the electrocardiogram generation system 100 receives lead electrocardiograms measured by multiple patients and information about the patients (step S210).

[0056] First, a 12-lead electrocardiogram shows cardiac potentials recorded in 12 electrical directions centered on the heart.

[0057] The standard method for viewing a leaded electrocardiogram (ECG) involves first attaching four electrodes (limb leads) to the patient's arms and legs. In this case, the ECG measuring device uses the electrodes on the arms and left leg to measure the electrical potential, while the electrode on the right leg acts as a grounding electrode.

[0058] Figure 3 An illustrative diagram illustrating the measurement method of a standard 12-lead electrocardiogram.

[0059] like Figure 3 As shown, the electrocardiogram (ECG) measuring device generates a lead I ECG by subtracting the potential of the right arm from the potential of the left arm, a lead II ECG by subtracting the potential of the right arm from the potential of the left leg, and a lead III ECG by subtracting the potential of the left arm from the potential of the left leg.

[0060] The electrocardiogram (ECG) measuring device calculates the potential of the virtual center point by averaging the potentials of the electrodes in both arms and the left leg. Furthermore, the device generates the ECG for lead aVL by subtracting the potential of the virtual center point from the potential in the left arm, and generates the ECG for lead aVR by subtracting the potential of the virtual center point from the potential in the right arm. Similarly, the device generates the ECG for lead aVF by subtracting the potential of the virtual center point from the potential in the left leg electrode. As described above, the ECG measuring device uses three (or four if grounded) limb electrodes to generate a total of six leads (six rows) of ECG.

[0061] Next, the electrocardiogram (ECG) measuring device generates a 6-lead chest ECG using the potential difference between the virtual center point determined in the limb leads and the six electrodes attached to the chest. Specifically, the six electrodes, V1, V2, V3, V4, V5, and V6, are attached to the front of the chest at a designated location, extending to the left side.

[0062] Furthermore, the electrocardiogram measuring device generates a V1 lead electrocardiogram by subtracting the measured potential from the V1 electrode and the potential of the virtual center point obtained by averaging the potentials of the aforementioned limb electrodes.

[0063] As described above, the electrocardiogram (ECG) measuring device generates a 12-lead ECG using electrodes attached to multiple patients. Furthermore, the generated 12-lead ECG is transmitted to the ECG generation system 100.

[0064] In this case, the electrocardiogram generation system 100 additionally receives the generated 12-lead electrocardiogram and the corresponding patient information.

[0065] The patient's information includes at least one of the following: gender, age, whether they have heart disease, and the measured electrocardiogram potential vector.

[0066] When step S210 is completed, the ECG generation system 100 uses the collected 12-lead ECG and patient information to extract learning data (step S220).

[0067] To reiterate, the data extraction unit 120 categorizes the collected 12-lead electrocardiograms (ECGs) based on patient information and stores them in a database. Furthermore, the data extraction unit 120 generates learning data by randomly extracting data from multiple stored 12-lead ECGs.

[0068] Next, the learning unit 130 uses the generated learning data to enable the first learning model and the second learning model to learn respectively (step S230).

[0069] First, the learning unit 130 inputs learning data consisting of a 12-lead electrocardiogram (ECG) into a first learning model and a second learning model, enabling the two models to learn the characteristics of the ECG. To reiterate, in a lead ECG, the direction of current varies depending on the potential vector, and the pattern of the ECG is affected by the patient's age and gender. Specifically, the heart muscle decreases with age, thus the amplitude of the ECG tends to decrease. In women, the breast tissue causes the electrode position to drop or the distance between the heart and the electrodes to increase, thus distorting the shape of the ECG.

[0070] Furthermore, in cases of chronic lung disease (chronic obstructive pulmonary disease), as lung volume increases, the heart stands vertically between the two lungs, so in three-dimensional space, the electrical current of the heart changes in the vertical direction.

[0071] Therefore, the learning unit 130 inputs the electrocardiograms (ECGs) of the input leads, which are separated and stored based on the patient's information, namely, the patient's age, gender, health status, and potential vectors, into the first and second learning models. Then, the first and second learning models extract the potential vector information of the input ECGs by learning the features of the input ECGs. However, learning can also be performed by receiving ECGs without distinguishing based on patient information, and this method can also enable more than one learning model to learn.

[0072] Furthermore, the learning unit 130 constructs 12 second learning models corresponding to the 12-lead electrocardiogram. Moreover, the learning unit 130 trains each second learning model to learn based on the characteristics of the electrocardiogram.

[0073] To reiterate, the first learning model learns the correlation between the input ECG and the 12-lead ECG based on the deep learning algorithm to extract features from the input ECG, namely, at least one of age, gender, and potential vector. Thus, it generates 12-n lead virtual ECGs from a reference ECG of n leads. Conversely, the second learning model generates n lead virtual ECGs from the 12-n lead virtual ECGs. For example, assuming the generation of a V1 lead ECG, the learning unit 130 learns a pattern in the first learning model to generate virtual ECGs of leads I, II, III, aVL, aVR, aVF, V2, V3, V4, V5, and V6 from the V1 lead ECG, and the second learning model learns a pattern to generate a virtual V1 lead ECG from the virtual ECGs of leads I, II, III, aVL, aVR, aVF, V2, V3, V4, V5, and V6. Then, the second learning model generates a virtual electrocardiogram by converting the input electrocardiogram into a V1 lead style.

[0074] In the first learning model, in order to determine which lead the input electrocardiogram (ECG) belongs to, the ECG of any lead is... Figure 1 It learns from the input. Thus, when the learned data is used, it can generate a 12-lead ECG even if the input does not provide any ECG of the leads.

[0075] In this case, the first learning model and the second learning model are based on a deep learning algorithm consisting of an autoencoder or a generative adversarial network. The deep learning algorithm can be implemented using one of the autoencoders or generative adversarial networks, or it can be implemented by mixing autoencoders and generative adversarial networks.

[0076] When the learning model completes its learning through steps S210 and S230, the electrocardiogram generation system 100 of this embodiment of the invention uses the learned learning model to generate an electrocardiogram.

[0077] First, the electrocardiogram generation system 100 receives a reference electrocardiogram for measurement via electrodes attached to the body of the subject being measured (step S240).

[0078] The reference electrocardiogram does not contain information related to the potential vector.

[0079] Next, the ECG generation unit 140 generates multiple virtual ECGs by inputting the input reference ECGs into the first learning model and the second learning model (step S250).

[0080] First, the electrocardiogram generation unit 140 extracts features from the reference electrocardiogram by inputting the reference electrocardiogram into the first learning model.

[0081] The features include at least one of the measurement object's age, gender, and potential vector.

[0082] Thus, a first virtual electrocardiogram is generated from the reference electrocardiogram.

[0083] Next, the ECG generation unit 140 inputs the generated first virtual ECG into the second learning model. Then, the second learning model generates a second virtual ECG based on the input first virtual ECG.

[0084] Figure 4 For illustrative purposes Figure 2 An example diagram of step S240 is shown.

[0085] like Figure 4 As shown, the first electrocardiogram (ECG) represents a reference ECG. In this case, the first learning model assumes the features of the reference ECG to be L1. Then, in the ECG generation unit 140, the first learning model generates a virtual ECG for the remaining 11 leads based on the L1 ECG. The generated 11-lead virtual ECG is then input into the second learning model to generate a virtual L1 lead ECG. Each learning model includes a discriminator, which determines whether the ECG generated by the ECG generator is correctly generated and improves the accuracy of the generated ECG through feedback.

[0086] When step S250 is completed, the control unit 150 outputs a reference electrocardiogram and 11 virtual electrocardiograms through the monitoring device (step S260).

[0087] Figure 5 For illustrative purposes Figure 2 An example diagram of step S260 is shown.

[0088] The control unit 150 outputs a reference electrocardiogram and 11 virtual electrocardiograms. Furthermore, as... Figure 5 As shown, the control unit 150 determines whether the health of the measured subject is abnormal by synchronizing the output reference electrocardiogram and 11 virtual electrocardiograms. In this case, a virtual electrocardiogram with the same leads as the reference electrocardiogram is also generated in the second learning model, so it can be identified as 12 virtual electrocardiograms.

[0089] As described above, the electrocardiogram (ECG) outputs different slopes, amplitudes, etc., based on the age, gender, and presence of health abnormalities of the measurement subject. Therefore, the control unit 150 outputs 12 ECGs synchronized with the reference ECG and 11 virtual ECGs via a monitoring device.

[0090] As described above, according to the present invention, multiple synchronized electrocardiograms can be generated from two electrocardiograms measured at different points using deep learning algorithms, thus enabling accurate readings of heart disease with the same precision as a 12-lead electrocardiogram.

[0091] Furthermore, the electrocardiogram (ECG) generation system of the present invention utilizes two-lead ECGs, thus enabling measurements to be taken at home or in daily life, even while on the move. This allows for real-time monitoring, generating synchronized ECGs using ECG information measured at different time points. This allows medical personnel to read synchronized lead ECG information that matches the corresponding beats and make more accurate diagnoses based on this information.

[0092] Although the present invention has been described with reference to the embodiments shown in the accompanying drawings, these are merely illustrative, and those skilled in the art will understand that various modifications and other equivalent embodiments are possible. Therefore, the true scope of protection of the present invention needs to be determined by the technical concept covered by the appended claims.

[0093] Explanation of reference numerals in the attached figures

[0094] 100: Electrocardiogram Generation System

[0095] 110: Data Input Department

[0096] 120: Data Extraction Department

[0097] 130: Study Department

[0098] 140: Electrocardiogram Generation Department

[0099] 150: Control Department

[0100] Industrial availability

[0101] This invention can generate multiple synchronized electrocardiograms from two electrocardiograms measured at different points using a deep learning algorithm. Therefore, it can accurately read heart disease with the same precision as a 12-lead electrocardiogram and has industrial applicability in various deep learning algorithm-based electrocardiogram generation systems.

Claims

1. An electrocardiogram (ECG) generation system, based on a deep learning algorithm, characterized in that, include: The data input unit receives 12-lead electrocardiograms measured from multiple patients and patient information including the potential vectors of the measured 12-lead electrocardiograms; The data extraction unit categorizes and stores the input 12-lead electrocardiograms based on the aforementioned patient information, and extracts learning data from the stored 12-lead electrocardiograms. The learning unit learns the features of electrocardiograms by inputting the extracted learning data into multiple learning models, including a first learning model and a second learning model. The first learning model receives n-lead electrocardiograms to generate 12-n-lead virtual electrocardiograms, and the second learning model receives the 12-n-lead virtual electrocardiograms generated by the first learning model to generate n-lead virtual electrocardiograms. An electrocardiogram (ECG) generation unit receives one or more reference ECGs from a measurement object, inputs the input reference ECGs into the first learning model among multiple learned learning models, determines the potential vector of the reference ECGs, and generates a virtual ECG; and The control unit synchronizes the aforementioned reference electrocardiogram with the generated 12-n lead virtual electrocardiogram, and outputs the waveforms of the synchronized reference electrocardiogram and the 12-n lead virtual electrocardiogram.

2. The electrocardiogram generation system according to claim 1, characterized in that, The patient information also includes gender, age, and whether they have at least one of the following: heart disease.

3. The electrocardiogram generation system according to claim 2, characterized in that, The aforementioned learning unit inputs n leads of electrocardiograms, which are part of the 6-lead electrocardiograms of the limbs and the 6-lead electrocardiogram of the chest, into the first learning model to enable the first learning model to learn and determine the potential vector of the input electrocardiogram.

4. The electrocardiogram generation system according to claim 3, characterized in that, The aforementioned study department is configured as follows: When the 12-n lead virtual electrocardiograms generated by the first learning model are input into the second learning model, the input electrocardiograms are transformed into a style with corresponding potential vectors and output as n virtual electrocardiograms.

5. The electrocardiogram generation system according to claim 1, characterized in that, The first and second learning models described above can be constructed by using adversarial generative networks and autoencoders separately or by combining them.

6. The electrocardiogram generation system according to claim 1, characterized in that, The control unit uses at least one of the amplitude, slope, and electrode position of the output waveforms of the reference electrocardiogram and the virtual electrocardiogram to determine whether there is an abnormality in the health of the measured object.

7. A method for generating an electrocardiogram (ECG), utilizing an ECG generation system, characterized in that, Includes the following steps: Receive 12-lead electrocardiograms measured from multiple patients and patient information including the potential vectors of the aforementioned 12-lead electrocardiograms; Based on the above patient information, classify and store the input 12-lead electrocardiograms, and extract learning data from the stored 12-lead electrocardiograms; The extracted learning data is input into multiple learning models, including a first learning model and a second learning model, to learn the features of the electrocardiogram. The first learning model receives n leads of electrocardiograms to generate 12-n leads of virtual electrocardiograms, and the second learning model receives the 12-n leads of virtual electrocardiograms generated by the first learning model to generate n virtual electrocardiograms. Receive one or more reference electrocardiograms from the measurement object, input the input reference electrocardiograms into the first learning model among multiple learning models that have completed learning, determine the potential vector of the reference electrocardiograms, and generate a virtual electrocardiogram; and The reference electrocardiogram (ECG) is synchronized with the generated 12-n lead virtual ECG, and the waveforms of the synchronized reference ECG, the 12-n lead virtual ECG, and whether the health of the measured object is abnormal are output.

8. The electrocardiogram generation method according to claim 7, characterized in that, The patient information also includes gender, age, and whether they have at least one of the following: heart disease.

9. The electrocardiogram generation method according to claim 7, characterized in that, In the step of learning the features of the above-mentioned electrocardiogram, the first learning model is trained by inputting n leads of electrocardiograms, which are part of the 6-lead electrocardiograms of the limbs and the 6-lead electrocardiograms of the chest, into the first learning model to determine the potential vector of the input electrocardiogram and generate a first virtual electrocardiogram.

10. The electrocardiogram generation method according to claim 9, characterized in that, In the step of learning the features of the above-mentioned electrocardiogram, when the above-mentioned first virtual electrocardiogram is input into the constructed second learning model, the above-mentioned second learning model learns to transform the input above-mentioned lead electrocardiogram into a style with corresponding potential vectors and outputs it as a second virtual electrocardiogram.

11. The electrocardiogram generation method according to claim 7, characterized in that, The first and second learning models described above can be constructed by using adversarial generative networks and autoencoders separately or by combining them.

12. The electrocardiogram generation method according to claim 7, characterized in that, In the step of outputting whether the above-mentioned health is abnormal, at least one of the amplitude, slope, and electrode position of the waveform of the above-mentioned reference electrocardiogram and virtual electrocardiogram is used to determine whether the health of the measured object is abnormal.

Images