WO2020052640A1 - Deep learning algorithm-based electrocardiogram feature extraction method, apparatus, system, device, and classification method - Google Patents
Deep learning algorithm-based electrocardiogram feature extraction method, apparatus, system, device, and classification method Download PDFInfo
- Publication number
- WO2020052640A1 WO2020052640A1 PCT/CN2019/105624 CN2019105624W WO2020052640A1 WO 2020052640 A1 WO2020052640 A1 WO 2020052640A1 CN 2019105624 W CN2019105624 W CN 2019105624W WO 2020052640 A1 WO2020052640 A1 WO 2020052640A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- ecg
- model
- feature extraction
- deep learning
- learning algorithm
- Prior art date
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 70
- 238000013135 deep learning Methods 0.000 title claims abstract description 33
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 32
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 34
- 230000000747 cardiac effect Effects 0.000 claims abstract description 17
- 238000004590 computer program Methods 0.000 claims description 22
- 210000002569 neuron Anatomy 0.000 claims description 17
- 238000011176 pooling Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 7
- 238000013145 classification model Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 3
- 230000007958 sleep Effects 0.000 claims description 2
- 238000013461 design Methods 0.000 abstract description 3
- 238000012360 testing method Methods 0.000 description 14
- 230000033764 rhythmic process Effects 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 7
- 206010003119 arrhythmia Diseases 0.000 description 7
- 230000006793 arrhythmia Effects 0.000 description 7
- 238000001514 detection method Methods 0.000 description 5
- 230000002861 ventricular Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 206010006578 Bundle-Branch Block Diseases 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000001746 atrial effect Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000002565 electrocardiography Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 206010003658 Atrial Fibrillation Diseases 0.000 description 2
- 206010003662 Atrial flutter Diseases 0.000 description 2
- 206010003674 Atrioventricular block first degree Diseases 0.000 description 2
- 206010006582 Bundle branch block right Diseases 0.000 description 2
- 208000000418 Premature Cardiac Complexes Diseases 0.000 description 2
- 208000004301 Sinus Arrhythmia Diseases 0.000 description 2
- 208000007888 Sinus Tachycardia Diseases 0.000 description 2
- 206010040738 Sinus arrest Diseases 0.000 description 2
- 206010040741 Sinus bradycardia Diseases 0.000 description 2
- 208000003734 Supraventricular Tachycardia Diseases 0.000 description 2
- 208000009729 Ventricular Premature Complexes Diseases 0.000 description 2
- 206010047281 Ventricular arrhythmia Diseases 0.000 description 2
- 230000005059 dormancy Effects 0.000 description 2
- 230000009977 dual effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 201000002934 first-degree atrioventricular block Diseases 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 201000007916 right bundle branch block Diseases 0.000 description 2
- 201000002932 second-degree atrioventricular block Diseases 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 201000002931 third-degree atrioventricular block Diseases 0.000 description 2
- 206010047302 ventricular tachycardia Diseases 0.000 description 2
- 206010003130 Arrhythmia supraventricular Diseases 0.000 description 1
- 206010006580 Bundle branch block left Diseases 0.000 description 1
- 206010015856 Extrasystoles Diseases 0.000 description 1
- 206010042600 Supraventricular arrhythmias Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000002763 arrhythmic effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 208000019622 heart disease Diseases 0.000 description 1
- 201000001715 left bundle branch hemiblock Diseases 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000004962 physiological condition Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7246—Details of waveform analysis using correlation, e.g. template matching or determination of similarity
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
Definitions
- the present invention relates to the technical field of electrocardiogram signal processing, and in particular, to an electrocardiogram feature extraction method, device, system, device, and classification method based on a deep learning algorithm.
- ECG detection is the most important method for detecting and diagnosing heart diseases.
- the human electrocardiogram (ECG) is the comprehensive performance of the heart's electrical activity on the surface of the body. By extracting the characteristic information of the ECG, you can get the heart parts. Physiological condition.
- the current arrhythmia analysis mainly uses the waveform analysis method and the template matching method.
- the waveform analysis method first obtains characteristic waveform parameters, such as characteristic waveform amplitude, time length, rise / fall time, waveform interval, etc. These waveform parameters are compared with the judgment threshold obtained based on clinical experience, and the analysis result of arrhythmia can be obtained.
- the template matching method mainly calculates the average RR interval and the average R wave shape of the subject's ECG signal as templates, and compares the RR interval and R wave shape of each heartbeat of the subject with the template. If the difference exceeds a certain range, arrhythmia is considered to have occurred.
- the waveform analysis method uses the empirical threshold of the solidified characteristic waveform parameters as the basis for judgment. It is relatively simple and intuitive, with fewer characteristic values, limited classification types, and ECG morphology is very sensitive to noise. Various transform domains and statistical methods are useful for arrhythmia types. The definition is confusing, and the classification results and effects are also different.
- the template matching method can only make an effective judgment when the R-wave morphology of the subject is significantly different from that of the template, and cannot effectively judge the arrhythmia waveforms that are not significantly different.
- the human body is a non-linear complex system.
- the shape of individual ECG signals will change over time, and physical health conditions also have a large impact on ECG signals.
- ECG signal data is very large and complex, and more adequate feature information extraction is required. .
- the invention provides a method, a device, a system, a device, and a classification method for extracting ECG features based on a deep learning algorithm, which improves the accuracy and diversity of ECG signal feature extraction.
- a method for extracting ECG features based on a deep learning algorithm includes the following steps: randomly extracting a continuous ECG signal from a 12-lead ECG to be processed, the ECG signal including at least two cardiac cycles; and the ECG signal to be intercepted
- the feature extraction model is input in the form of pictures, and the features of the ECG signal are extracted.
- the feature extraction model is obtained by training based on a ResNet model, an Inception model, or an Inception-ResNet model.
- the following also provides a number of optional methods, but it is not intended as an additional limitation on the above-mentioned overall scheme. It is only a further addition or preference. Without technical or logical contradiction, each of the alternative methods can be performed separately for the above-mentioned overall scheme Combination can also be a combination of multiple optional ways.
- the features extracted by the feature extraction model should be understood broadly, including both the information directly extracted from the electrocardiogram and the information obtained by progressively processing the directly extracted information, and further processing including classification and other processing methods.
- the feature extraction model includes a convolution layer, a pooling layer, a dropout layer, and a fully-connected layer, wherein the convolution layer, the pooling layer, and the dropout layer all adopt a Relu activation function or a variant thereof, and are fully connected
- the layer uses the softmax function.
- the first convolution layer in the convolution layer includes 6 convolution channels, and the convolution kernel of each convolution channel is ⁇ 1, 1, 2 ⁇ , ⁇ 1, 2, 1 ⁇ , ⁇ 2,1,1 ⁇ .
- the size of the pooling window of the pooling layer is (2, 2).
- the fully connected layer includes three layers, the number of neurons in the first layer is 448, the number of neurons in the second layer is 112, and the number of neurons in the third layer is 28.
- the dormancy ratio of the neurons in the dropout layer is 0.5.
- An ECG feature extraction device based on a deep learning algorithm includes: a preprocessing module for randomly intercepting a continuous ECG signal from a 12-lead ECG to be processed, the ECG signal including at least two cardiac cycles; extraction The module is used for inputting the intercepted electrocardiogram signal into a feature extraction model in the form of a picture to extract the characteristics of the electrocardiogram signal; the feature extraction model is obtained by training based on a ResNet model, an Inception model, or an Inception-ResNet model.
- a computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the processor implements the following steps: Randomly intercepts a continuous ECG signal in a 12-lead ECG to be processed.
- the ECG signal includes at least two cardiac cycles; the intercepted ECG signal is input into a feature extraction model in the form of a picture to extract the characteristics of the ECG signal; the feature extraction model is based on a ResNet model, or an Inception model, or an Inception-ResNet model training get.
- An ECG feature extraction system based on a deep learning algorithm includes a terminal and a server, the server includes a memory and a processor, the memory stores a computer program, and the server obtains an electrocardiogram from the terminal; the processor executes all When describing a computer program, the following steps are implemented:
- the intercepted electrocardiogram signal is input into a feature extraction model in the form of a picture, and the features of the electrocardiogram signal are extracted; the feature extraction model is obtained by training based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- a computer-readable storage medium stores a computer program in the computer-readable storage medium.
- the computer program is executed by a computer processor, the following steps are implemented:
- the intercepted electrocardiogram signal is input into a feature extraction model in the form of a picture, and the features of the electrocardiogram signal are extracted; the feature extraction model is obtained by training based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- a method for classifying an ECG signal based on a deep neural network includes the following steps: Randomly intercept a segment of a continuous ECG signal from a 12-lead ECG to be processed, the ECG signal including at least two cardiac cycles; and dividing the intercepted ECG signal into The picture form is input into a classification model to obtain a classification result; the classification model is obtained by training based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- the ECG feature extraction method, device, system, equipment and classification method based on the deep learning algorithm provided by the present invention reduce the incompleteness caused by artificial design features and improve the accuracy of ECG feature extraction and classification based on the deep learning algorithm. And diversity.
- FIG. 1 is a flowchart of an ECG feature extraction method based on a deep learning algorithm in an embodiment
- FIG. 2 is a schematic diagram of an intercepted electrocardiogram signal
- FIG. 3 is a schematic diagram of a computer device in one embodiment
- FIG. 4 is a flowchart of an ECG feature extraction method based on a deep learning algorithm in one embodiment.
- a method for extracting ECG features based on a deep learning algorithm includes the following steps:
- step S1 a continuous ECG signal is randomly intercepted from the 12-lead ECG to be processed, and the ECG signal includes at least two cardiac cycles.
- the ECG signal (ie, the electrocardiogram signal) can be a unipolar ECG signal or a bipolar ECG signal.
- an individual's ECG can be recorded by one or more electrodes placed on the surface of the human body.
- the 12-lead ECG can provide spatial information about ECG activity, and each of the 12 leads represents a specific location in the space where the heart's ECG activity is measured.
- lead I right arm to left arm
- lead II right arm to left leg
- lead III left arm to left leg
- leads V1, V2 , V3, V4, V5 and V6 unipolar chest leads.
- two electrodes need to be in contact with the two skin surfaces of the limbs, such as the right and left arms.
- an electrode is required to make contact with the skin surface of the chest.
- the term "contact” herein may refer to the direct contact between the electrode and the bare skin, or the indirect contact between the electrode and the bare skin with a conductive material (such as a conductive patch or conductive clothing).
- the ECG data originally collected was discrete data.
- One-dimensional ECG signals were obtained by Bspline interpolation.
- the one-dimensional ECG signals were pre-processed to generate noise reduction signals, and then the ECG signal feature extraction was performed.
- the noise sources of the ECG signal include the background of the electromyographic signal, baseline drift, power frequency interference, and motion artifacts.
- Preprocessing can use existing techniques, for example, wavelet transform or filter to remove baseline drift and power frequency from the ECG signal. Interference and high-frequency noise.
- step S2 the intercepted electrocardiogram signal is input into a feature extraction model in the form of a picture, and the features of the electrocardiogram signal are extracted; the feature extraction model is obtained by training based on a ResNet model, an Inception model, or an Inception-ResNet model.
- the ECG signal is input to the feature extraction model in the form of a picture as shown in FIG. 2, and the picture size of the ECG signal is 1200 ⁇ 600 pixels.
- the process of extracting ECG features usually requires artificial design features, which are prone to miss detection and misjudgment.
- the feature extraction model is based on the ResNet model, or Inception model, or Inception-ResNet model training, can be automatically and effectively extracted from a large number of electrocardiograms to fully describe the characteristics of the electrocardiogram, including low-dimensional heartbeat features, improve the integrity and diversity of ECG signal feature extraction, robust Stronger performance and better detection accuracy.
- the feature extraction model is retrained to ensure that the feature extraction model has a better extraction accuracy rate.
- the feature extraction model includes a convolution layer, a pooling layer, a dropout layer, and a fully connected layer, wherein the convolution layer, the pooling layer, and the dropout layer all adopt a Relu activation function or a variant thereof, a fully connected layer Using the softmax function.
- the first convolution layer in the convolution layer includes 6 convolution channels, and the convolution kernel of each convolution channel is ⁇ 1, 1, 2 ⁇ , ⁇ 1, 2, 1 ⁇ , ⁇ 2,1,1 ⁇ .
- the first convolution layer is used to enhance boundary information of a picture.
- the pooling window size of the pooling layer is (2, 2).
- the pooling layer is used for dimensionality reduction, thereby reducing operation parameters and improving calculation speed.
- the fully connected layer includes three layers, the number of neurons in the first layer is 448, the number of neurons in the second layer is 112, and the number of neurons in the third layer is 28.
- the fully connected layer is used to integrate multi-channel data into a one-dimensional feature vector. According to the characteristics of the multilayer neural network, the abstract one-dimensional feature vector is classified at the fully connected layer.
- the number of neurons in the third layer of the fully connected layer determines the number of extracted ECG features. In this embodiment, 28 types of ECG features can be extracted.
- the proportion of dormant neurons in the dropout layer is 0.5.
- a dropout layer is added between every two fully connected layers to avoid overfitting.
- the neurons that are trained each time will temporarily sleep according to the set ratio.
- the dormancy ratio of the neurons is set to 0.5 according to experience, that is, each time an ECG sample enters the model, half of the neurons in the fully connected layer will not participate in the training of the sample.
- the feature extraction model is trained in deep learning frameworks TensorFlow and Keras.
- the training process of the feature extraction model includes:
- Step a construct a training database and a test database.
- the data of the training database and test database can come from hospitals and physical examination centers.
- the total number of ECGs in the training database and test database is not less than 100,000, of which 70% are normal ECGs, and the rest are abnormal ECGs.
- Each ECG has passed at least two ECG experts Manually label abnormal features independently. There is no overlap between the ECG data in the training database and the test database, and the amount of ECG data in the training database is greater than the amount of ECG data in the test database.
- the ECG data in the training database and the test database are from different individuals, including males between the ages of 10 and 99 and females between the ages of 10 and 99. The proportion of males and females is half.
- the training and test databases Randomly selected from the data set of the electrocardiograph, including various representative waveform and artifact record numbers, various uncommon but clinically significant data, and some complex ventricular, nodular, and supraventricular arrhythmias And conduction abnormalities.
- Each ECG signal data in the training database and test database includes the ECG signal format, sampling frequency, length, and related information about the patient, such as the place of collection, the patient's condition, and medication, etc.
- the signal analysis results mainly include information such as heartbeat, rhythm and signal quality.
- the annotations of each ECG signal data in the training database and test database contain the basic information of ECG signals and the results of signal analysis by ECG diagnostic experts, such as: signal duration, heart rate, signal quality, location of arrhythmic beats, Number and arrhythmia characteristics.
- the abnormal characteristics of electrical signals in the training database and test data center determine the features that the feature extraction model can extract.
- ECG signals include, but are not limited to, the following types of heartbeats: normal, pacing, sinus tachycardia, sinus bradycardia, sinus arrhythmia, sinus arrest, escape beat, and escape beat rhythm (Further divided into atrial rhythm, borderline rhythm, ventricular rhythm, other escape beats and escape beat rhythms) Atrial premature beat, supraventricular dual rhythm, premature ventricular beat, fast ventricular rate, slow ventricular rate, supraventricular Tachycardia, ventricular tachycardia, atrial flutter, atrial fibrillation, first degree atrioventricular block, second degree atrioventricular block, third degree atrioventricular block, indoor block (further divided into left bundle branch Conduction block, right bundle branch block, left anterior branch block, other indoor blocks), noise, others.
- heartbeats normal, pacing, sinus tachycardia, sinus bradycardia, sinus arrhythmia, sinus arrest, escape beat, and escape beat rhythm (Further divided into atrial rhythm, borderline rhythm
- Step b preprocessing each ECG in the training database and the test database.
- a continuous ECG signal is randomly intercepted, and the ECG signal includes at least two cardiac cycles.
- the electrocardiogram signal is converted into a picture format, which is the same as the method for converting the central electrocardiogram signal into a picture format in step S2.
- the ECG signals converted into pictures in the training database are used as input for the ResNet model (or Inception model or Inception-ResNet model) for training.
- the trained models are tested using the ECG in the test database.
- the test results are divided into four types: True positive (TP), true negative (TN), false positive (FP), false negative (FN).
- the four results are used to evaluate the feature extraction results.
- Sen Sensitivity
- Spe Specificity
- PPV Positive rate
- Acc Accuracy
- Acc (TP + TN) / (TP + FP + FN + TN).
- Sen Spe Acc PPV 99% 98.6% 98.8% 98.1%
- steps in the flowchart of FIG. 1 are sequentially displayed according to the directions of the arrows, these steps are not necessarily performed sequentially in the order indicated by the arrows. Unless explicitly stated in this document, the execution of these steps is not strictly limited, and these steps can be performed in other orders. Moreover, at least a part of the steps in FIG. 1 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. The execution of these sub-steps or stages The sequence is not necessarily performed sequentially, but may be performed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.
- an ECG feature extraction device based on a deep learning algorithm includes a preprocessing module for randomly intercepting a continuous ECG signal from a 12-lead ECG to be processed, where the ECG signal includes at least Two cardiac cycles; an extraction module, which is used to input the intercepted ECG signals into a feature extraction model in the form of pictures to extract the characteristics of the ECG signal; the feature extraction model is based on a ResNet model, or an Inception model, or an Inception-ResNet model. .
- a preprocessing module for randomly intercepting a continuous ECG signal from a 12-lead ECG to be processed, where the ECG signal includes at least Two cardiac cycles
- an extraction module which is used to input the intercepted ECG signals into a feature extraction model in the form of pictures to extract the characteristics of the ECG signal
- the feature extraction model is based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- Each module in the above-mentioned ECG feature extraction device based on a deep learning algorithm may be implemented in whole or in part by software, hardware, and a combination thereof.
- the above-mentioned modules may be embedded in the hardware form or independent of the processor in the computer device, or may be stored in the memory of the computer device in the form of software, so that the processor calls and performs the operations corresponding to the above modules.
- the ECG feature extraction device based on the deep learning algorithm provided in this embodiment may be configured at a remote end, and the ECG signal may be obtained through a remote terminal connected to the device.
- the device of this embodiment may be configured at the terminal (for example, a computer used by a user). Or medical detection equipment) to obtain the ECG signal directly through the ECG acquisition device.
- a computer device is provided.
- the computer device may be a terminal, and its internal structure diagram may be as shown in FIG. 3.
- the computer equipment includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus.
- the processor of the computer device is used to provide computing and control capabilities.
- the memory of the computer device includes a non-volatile storage medium and an internal memory.
- the non-volatile storage medium stores an operating system and a computer program.
- the internal memory provides an environment for running an operating system and computer programs in a non-volatile storage medium.
- the network interface of the computer device is used to communicate with an external terminal through a network connection.
- the computer program is executed by a processor to implement an ECG feature extraction method based on a deep learning algorithm.
- the display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen.
- the input device of the computer device may be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the computer device casing. , Or an external keyboard, trackpad, or mouse.
- FIG. 3 is only a block diagram of a part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer equipment to which the scheme of the present application is applied.
- the specific computer equipment may be Include more or fewer parts than shown in the figure, or combine certain parts, or have a different arrangement of parts.
- a computer device including a memory and a processor.
- the memory stores a computer program
- the processor implements the following steps when the computer program executes:
- the intercepted electrocardiogram signal is input into a feature extraction model in the form of a picture, and the features of the electrocardiogram signal are extracted; the feature extraction model is obtained by training based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- an ECG feature extraction system based on a deep learning algorithm which includes a terminal and a server.
- the server includes a memory and a processor.
- the memory stores a computer program, and the server obtains the information from the terminal. ECG; when the processor executes the computer program, the following steps are implemented:
- the intercepted electrocardiogram signal is input into a feature extraction model in the form of a picture, and the features of the electrocardiogram signal are extracted; the feature extraction model is obtained by training based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- the terminal may be, but is not limited to, a computer, a laptop, a smart phone, a tablet, and a portable wearable device.
- the server may be a remote background server or a server in a cloud platform.
- An independent server or It is implemented by a server cluster composed of multiple servers.
- the ECG is transmitted between the terminal and the server through a communication network.
- the communication network can be, but is not limited to, 3G, 4G, 5G, and wifi.
- information related to the ECG can be transmitted from the terminal to the server.
- the related information of the ECG includes, but is not limited to, user information and detection time.
- the server receives the ECG data and transmits the extracted feature results to the terminal through the communication network.
- the server stores the received electrocardiogram data and feature extraction results.
- a computer-readable storage medium stores a computer program, and when the computer program is executed by a computer processor, the following steps are implemented:
- the intercepted electrocardiogram signal is input into a feature extraction model in the form of a picture, and the features of the electrocardiogram signal are extracted; the feature extraction model is obtained by training based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
- Volatile memory can include random access memory (RAM) or external cache memory.
- RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Synchlink DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
- SRAM static RAM
- DRAM dynamic RAM
- SDRAM synchronous DRAM
- DDRSDRAM dual data rate SDRAM
- ESDRAM enhanced SDRAM
- SLDRAM synchronous chain Synchlink DRAM
- Rambus direct RAM
- DRAM direct memory bus dynamic RAM
- RDRAM memory bus dynamic RAM
- an ECG signal classification method based on a deep neural network including the following steps:
- the intercepted electrocardiogram signals are input into a classification model in the form of pictures to obtain a classification result; the classification model is obtained by training based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- the classification categories of ECG signals include, but are not limited to: normal, pacing, sinus tachycardia, sinus bradycardia, sinus arrhythmia, sinus arrest, escape beat, and escape beat rhythm (further divided into atrial Heart rhythm, border rhythm, ventricular rhythm, other escape beats and escape beat rhythms) Atrial premature beats, supraventricular duplex, premature ventricular beats, fast ventricular rate, slow ventricular rate, supraventricular tachycardia, ventricular Tachycardia, atrial flutter, atrial fibrillation, first degree atrioventricular block, second degree atrioventricular block, third degree atrioventricular block, indoor block (further divided into left bundle branch block, right bundle Branch block, left anterior branch block, other indoor blocks), noise, other.
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Cardiology (AREA)
- Physics & Mathematics (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Veterinary Medicine (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
Description
SenSen | SpeSpe | AccAcc | PPVPPV |
99%99% | 98.6%98.6% | 98.8%98.8% | 98.1%98.1% |
Claims (10)
- 一种基于深度学习算法的心电特征提取方法,其特征在于,包括如下步骤:A method for extracting ECG features based on a deep learning algorithm is characterized in that it includes the following steps:在待处理的十二导联心电图中随机截取一段连续的心电图信号,该心电图信号至少包含两个心动周期;Randomly intercept a continuous ECG signal from the 12-lead ECG to be processed, and the ECG signal includes at least two cardiac cycles;将截取的心电图信号以图片形式输入特征提取模型中,提取得到心电图信号特征;所述特征提取模型基于ResNet模型、或Inception模型、或 Inception-ResNet模型训练得到。The intercepted electrocardiogram signal is input into a feature extraction model in the form of a picture, and the features of the electrocardiogram signal are extracted; the feature extraction model is obtained by training based on a ResNet model, or an Inception model, or an Inception-ResNet model.
- 如权利要求1所述的基于深度学习算法的心电特征提取方法,其特征在于,所述特征提取模型包括卷积层、池化层、dropout层和全连接层,其中,卷积层、池化层、dropout层均采用Relu激活函数或其变体,全连接层采用softmax函数。The ECG feature extraction method based on deep learning algorithm according to claim 1, wherein the feature extraction model comprises a convolution layer, a pooling layer, a dropout layer, and a fully connected layer, wherein the convolution layer, the pool The chemical layer and the dropout layer both use the Relu activation function or a variant thereof, and the fully connected layer uses the softmax function.
- 如权利要求2所述的基于深度学习算法的心电特征提取方法,其特征在于,所述卷积层中第一卷积层包括6个卷积通道,每个卷积通道的卷积核均为{{1,1,2},{1,2,1},{2,1,1}}。The ECG feature extraction method based on deep learning algorithm according to claim 2, wherein the first convolution layer in the convolution layer includes 6 convolution channels, and the convolution kernel of each convolution channel is For {{1,1,2}, {1,2,1}, {2,1,1}}.
- 如权利要求2所述的基于深度学习算法的心电特征提取方法,其特征在于,所述池化层的池化窗口大小为(2,2)。The method for extracting ECG features based on a deep learning algorithm according to claim 2, wherein the size of the pooling window of the pooling layer is (2, 2).
- 如权利要求2所述的基于深度学习算法的心电特征提取方法,其特征在于,所述全连接层包含三层,第一层的神经元个数为448,第二层的神经元个数为112,第三层的神经元个数为28。The method for extracting ECG features based on a deep learning algorithm according to claim 2, wherein the fully connected layer comprises three layers, the number of neurons in the first layer is 448, and the number of neurons in the second layer Is 112, and the number of neurons in the third layer is 28.
- 如权利要求2所述的基于深度学习算法的心电特征提取方法,其特征在于,所述dropout层中神经元的休眠比例为0.5。The method for extracting ECG features based on a deep learning algorithm according to claim 2, wherein the sleep ratio of the neurons in the dropout layer is 0.5.
- 一种基于深度学习算法的心电特征提取装置,其特征在于,包括:预处理模块,用于在待处理的十二导联心电图中随机截取一段连续的心电图信号,该心电图信号至少包含两个心动周期; 提取模块,用于将截取的心电图信号以图片形式输入特征提取模型中,提取得到心电图信号特征;所述特征提取模型基于ResNet模型、或Inception模型、或Inception-ResNet模型训练得到。A device for extracting ECG features based on a deep learning algorithm, comprising: a pre-processing module for randomly intercepting a continuous ECG signal from a 12-lead ECG to be processed, the ECG signal including at least two Cardiac cycle An extraction module is used to input the intercepted electrocardiogram signal into a feature extraction model in the form of a picture to extract the characteristics of the electrocardiogram signal; the feature extraction model is obtained by training based on a ResNet model, an Inception model, or an Inception-ResNet model.
- 一种计算机设备,包括存储器和处理器,所述存储器存储有计算机程序,其特征在于,所述处理器执行所述计算机程序时实现权利要求1 至6中任一项所述的基于深度学习算法的心电特征提取方法。A computer device includes a memory and a processor, and the memory stores a computer program, wherein the processor implements the deep learning algorithm according to any one of claims 1 to 6 when executing the computer program. ECG feature extraction method.
- 一种基于深度学习算法的心电特征提取系统,包括终端以及服务器,所述服务器包括存储器和处理器,所述存储器内存储有计算机程序, 其特征在于,所述服务器从终端获取心电图;所述处理器执行所述计算机程序时,实现如权利要求1~6任一项所述的基于深度学习算法的心电特征提取方法。An ECG feature extraction system based on a deep learning algorithm includes a terminal and a server. The server includes a memory and a processor. The memory stores a computer program. It is characterized in that the server obtains an electrocardiogram from a terminal; when the processor executes the computer program, the method for extracting an electrocardiogram feature based on a deep learning algorithm according to any one of claims 1 to 6 is implemented.
- 一种基于深度学习算法的心电图信号分类方法,其特征在于,包括如下步骤:An ECG signal classification method based on a deep learning algorithm is characterized in that it includes the following steps:在待处理的十二导联心电图中随机截取一段连续的心电图信号,该心电图信号至少包含两个心动周期;Randomly intercept a continuous ECG signal from the 12-lead ECG to be processed, and the ECG signal includes at least two cardiac cycles;将截取的心电图信号以图片形式输入分类模型中,得到分类结果;所述分类模型基于ResNet模型、或Inception模型、或Inception-ResNet模型训练得到。The intercepted electrocardiogram signals are input into a classification model in the form of pictures to obtain a classification result; the classification model is obtained by training based on a ResNet model, an Inception model, or an Inception-ResNet model.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811075592.8A CN110897629A (en) | 2018-09-14 | 2018-09-14 | Deep learning algorithm-based electrocardiogram feature extraction method, device, system, equipment and classification method |
CN201811075592.8 | 2018-09-14 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020052640A1 true WO2020052640A1 (en) | 2020-03-19 |
Family
ID=69776772
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/105624 WO2020052640A1 (en) | 2018-09-14 | 2019-09-12 | Deep learning algorithm-based electrocardiogram feature extraction method, apparatus, system, device, and classification method |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN110897629A (en) |
WO (1) | WO2020052640A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114652322A (en) * | 2022-03-24 | 2022-06-24 | 中南大学 | Electrocardiosignal classification method and system based on multi-domain feature learning |
CN118285809A (en) * | 2024-03-21 | 2024-07-05 | 济南大学 | Multi-feature auxiliary diagnosis method based on double-lead ECG signal, terminal and readable storage medium |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
RU2767157C2 (en) * | 2020-07-21 | 2022-03-16 | Общество с ограниченной ответственностью "СберМедИИ" (ООО "СберМедИИ") | Method and system for automatic ecg analysis |
CN112568908A (en) * | 2020-12-14 | 2021-03-30 | 上海数创医疗科技有限公司 | Electrocardiogram waveform positioning and classifying model device adopting multi-scale visual field depth learning |
CN113647959B (en) * | 2021-07-27 | 2024-01-12 | 东软集团股份有限公司 | Waveform identification method, device and equipment for electrocardiographic waveform signals |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160135706A1 (en) * | 2014-11-14 | 2016-05-19 | Zoll Medical Corporation | Medical Premonitory Event Estimation |
CN105748063A (en) * | 2016-04-25 | 2016-07-13 | 山东大学齐鲁医院 | Intelligent arrhythmia diagnosis method based on multiple-lead and convolutional neural network |
CN107714023A (en) * | 2017-11-27 | 2018-02-23 | 乐普(北京)医疗器械股份有限公司 | Static ecg analysis method and apparatus based on artificial intelligence self study |
CN107837082A (en) * | 2017-11-27 | 2018-03-27 | 乐普(北京)医疗器械股份有限公司 | Electrocardiogram automatic analysis method and device based on artificial intelligence self study |
CN107951485A (en) * | 2017-11-27 | 2018-04-24 | 乐普(北京)医疗器械股份有限公司 | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study |
CN107981858A (en) * | 2017-11-27 | 2018-05-04 | 乐普(北京)医疗器械股份有限公司 | Electrocardiogram heartbeat automatic recognition classification method based on artificial intelligence |
CN108113666A (en) * | 2017-12-19 | 2018-06-05 | 中国科学院深圳先进技术研究院 | Recognition methods, device and the equipment of atrial fibrillation signal |
CN108309263A (en) * | 2018-02-24 | 2018-07-24 | 乐普(北京)医疗器械股份有限公司 | Multi-parameter monitoring data analysing method and multi-parameter monitoring system |
CN108478209A (en) * | 2018-02-24 | 2018-09-04 | 乐普(北京)医疗器械股份有限公司 | Ecg information dynamic monitor method and dynamic monitor system |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170156592A1 (en) * | 2015-12-02 | 2017-06-08 | Mediatek Inc. | Healthcare systems and monitoring method for physiological signals |
CN108175402A (en) * | 2017-12-26 | 2018-06-19 | 智慧康源(厦门)科技有限公司 | The intelligent identification Method of electrocardiogram (ECG) data based on residual error network |
CN108464827A (en) * | 2018-03-08 | 2018-08-31 | 四川大学 | It is a kind of it is Weakly supervised under electrocardio image-recognizing method |
-
2018
- 2018-09-14 CN CN201811075592.8A patent/CN110897629A/en active Pending
-
2019
- 2019-09-12 WO PCT/CN2019/105624 patent/WO2020052640A1/en active Application Filing
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160135706A1 (en) * | 2014-11-14 | 2016-05-19 | Zoll Medical Corporation | Medical Premonitory Event Estimation |
CN105748063A (en) * | 2016-04-25 | 2016-07-13 | 山东大学齐鲁医院 | Intelligent arrhythmia diagnosis method based on multiple-lead and convolutional neural network |
CN107714023A (en) * | 2017-11-27 | 2018-02-23 | 乐普(北京)医疗器械股份有限公司 | Static ecg analysis method and apparatus based on artificial intelligence self study |
CN107837082A (en) * | 2017-11-27 | 2018-03-27 | 乐普(北京)医疗器械股份有限公司 | Electrocardiogram automatic analysis method and device based on artificial intelligence self study |
CN107951485A (en) * | 2017-11-27 | 2018-04-24 | 乐普(北京)医疗器械股份有限公司 | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study |
CN107981858A (en) * | 2017-11-27 | 2018-05-04 | 乐普(北京)医疗器械股份有限公司 | Electrocardiogram heartbeat automatic recognition classification method based on artificial intelligence |
CN108113666A (en) * | 2017-12-19 | 2018-06-05 | 中国科学院深圳先进技术研究院 | Recognition methods, device and the equipment of atrial fibrillation signal |
CN108309263A (en) * | 2018-02-24 | 2018-07-24 | 乐普(北京)医疗器械股份有限公司 | Multi-parameter monitoring data analysing method and multi-parameter monitoring system |
CN108478209A (en) * | 2018-02-24 | 2018-09-04 | 乐普(北京)医疗器械股份有限公司 | Ecg information dynamic monitor method and dynamic monitor system |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114652322A (en) * | 2022-03-24 | 2022-06-24 | 中南大学 | Electrocardiosignal classification method and system based on multi-domain feature learning |
CN118285809A (en) * | 2024-03-21 | 2024-07-05 | 济南大学 | Multi-feature auxiliary diagnosis method based on double-lead ECG signal, terminal and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN110897629A (en) | 2020-03-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Merdjanovska et al. | Comprehensive survey of computational ECG analysis: Databases, methods and applications | |
US20220015711A1 (en) | System and method for automated analysis and detection of cardiac arrhythmias from electrocardiograms | |
US9872652B2 (en) | Method and apparatus for heart rate monitoring using an electrocardiogram sensor | |
WO2020052640A1 (en) | Deep learning algorithm-based electrocardiogram feature extraction method, apparatus, system, device, and classification method | |
CN106214145B (en) | Electrocardiogram classification method based on deep learning algorithm | |
Lai et al. | Non-standardized patch-based ECG lead together with deep learning based algorithm for automatic screening of atrial fibrillation | |
Yang | Multiscale recurrence quantification analysis of spatial cardiac vectorcardiogram signals | |
JP7429371B2 (en) | Method and system for quantifying and removing asynchronous noise in biophysical signals | |
Demirel et al. | Energy-efficient real-time heart monitoring on edge–fog–cloud internet of medical things | |
CN109497986A (en) | Electrocardiogram intelligent analysis method, device, computer equipment and system based on deep neural network | |
Nandagopal et al. | Newly constructed real time ECG monitoring system using labview | |
Singhal et al. | A systematic review on artificial intelligence-based techniques for diagnosis of cardiovascular arrhythmia diseases: challenges and opportunities | |
US9549681B2 (en) | Matrix-based patient signal analysis | |
Prashar et al. | Novel cardiac arrhythmia processing using machine learning techniques | |
Khatun et al. | Detection of myocardial infarction and arrhythmia from single-lead ECG data using bagging trees classifier | |
Mathur et al. | Analysis of CNN and feed-forward ANN model for the evaluation of ECG signal | |
Vizcaya et al. | Standard ECG lead I prospective estimation study from far-field bipolar leads on the left upper arm: A neural network approach | |
Gavriel et al. | Smartphone as an ultra-low cost medical tricorder for real-time cardiological measurements via ballistocardiography | |
CN111528833B (en) | Rapid identification and processing method and system for electrocardiosignals | |
CN110897628A (en) | Deep neural network-based electrocardiogram signal feature extraction method, device, system, equipment and classification method | |
Rajoub | Machine learning in biomedical signal processing with ECG applications | |
Marhoon et al. | A new real-time resource-efficient algorithm for ECG denoising, feature extraction and classification-based wearable sensor network | |
Rashkovska et al. | Clustering of heartbeats from ECG recordings obtained with wireless body sensors | |
Oktivasari et al. | A simple real-time system for detection of normal and myocardial ischemia in the ST segment and t wave ECG signal | |
CN110897626A (en) | Deep neural network-based electrocardiogram analysis method, device, computer equipment and system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19859597 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19859597 Country of ref document: EP Kind code of ref document: A1 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19859597 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 04.02.2022) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19859597 Country of ref document: EP Kind code of ref document: A1 |