CN106214145B - Electrocardiogram classification method based on deep learning algorithm - Google Patents
Electrocardiogram classification method based on deep learning algorithm Download PDFInfo
- Publication number
- CN106214145B CN106214145B CN201610572216.4A CN201610572216A CN106214145B CN 106214145 B CN106214145 B CN 106214145B CN 201610572216 A CN201610572216 A CN 201610572216A CN 106214145 B CN106214145 B CN 106214145B
- Authority
- CN
- China
- Prior art keywords
- electrocardiogram
- waveform data
- waveform
- information
- pqrst
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
-
- 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
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- 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
- A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- 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
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- 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
- A61B5/7271—Specific aspects of physiological measurement analysis
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Power Engineering (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
the invention discloses an electrocardiogram classification method based on a deep learning algorithm, which comprises the following steps: a. acquiring original electrocardiogram waveform data with the measuring time of more than 8 seconds and electrocardiogram additional information, and acquiring electrocardiogram rhythm information and representative PQRST waveform data according to the original electrocardiogram waveform data; b. training a neural network of the deep learning algorithm, arranging electrocardiogram rhythm information, PQRST waveform data and electrocardiogram additional information into one-dimensional data, and then carrying out waveform classification through the trained deep learning algorithm to obtain an electrocardiogram classification result. The deep learning method is introduced into the field of electrocardiogram classification, the characteristics of electrocardiogram classification are reasonably combined, the deep learning method is trained through the steps, and the deep learning method is used for waveform classification, so that the information quality of providing electrocardiogram classification auxiliary information for doctors can be greatly improved.
Description
Technical Field
The invention relates to an electrocardiogram classification method, in particular to an electrocardiogram classification method based on a deep learning algorithm.
Background
The electrocardiogram waveform data acquisition and the electrocardiogram classification result are important auxiliary means and reference information for doctors to diagnose heart diseases, usually, the electrocardiogram waveform data acquisition and classification are carried out in hospitals or physical examination centers, and have the defects of inconvenient detection, low detection frequency and the like. In recent years, with the popularization of networks and mobile smart phones, the introduction of portable electrocardiographic monitors and home personal electrocardiographic wave monitors has become possible. The classification algorithm of the monitors in the market is based on the traditional electrocardiogram measurement classification method, so that the monitors are easy to be classified by mistake when the single waveform measurement characteristic is not obvious, the clinical reliability and the accuracy are low, and the actual requirement of providing auxiliary diagnosis information for doctors cannot be met.
Disclosure of Invention
The invention provides a novel electrocardiogram classification method based on a deep learning algorithm, aiming at the defects that the traditional electrocardiogram measurement classification method in the prior art is easy to generate error classification when the single waveform measurement characteristic is not obvious, has low clinical reliability and accuracy, and can not meet the actual requirement of providing auxiliary diagnosis information for doctors.
in order to solve the technical problems, the invention is realized by the following technical scheme:
an electrocardiogram classification method based on a deep learning algorithm comprises the following steps:
a. acquiring original electrocardiogram waveform data and electrocardiogram additional information with the measuring time of more than 8 seconds, extracting electrocardiogram rhythm information and representative PQRST waveform according to the original electrocardiogram waveform data, and acquiring electrocardiogram rhythm information and representative PQRST waveform data;
b. training a neural network of the deep learning algorithm, arranging the electrocardiogram rhythm information, the PQRST waveform data and the electrocardiogram additional information obtained in the step a into one-dimensional data, and then carrying out waveform classification through the trained deep learning algorithm to obtain an electrocardiogram classification result.
The deep learning algorithm is a machine learning method in the field of artificial intelligence, and comprises a multi-layer perceptron with multiple hidden layers, and forms more abstract high-layer representation attribute categories or features by combining low-layer features so as to find distributed feature representation of data. The deep learning method is introduced into the field of electrocardiogram classification, the characteristics of electrocardiogram classification are reasonably combined, the deep learning method is trained through the steps, and the deep learning method is used for waveform classification, so that the information quality of providing electrocardiogram classification auxiliary information for doctors can be greatly improved.
wherein the original electrocardiogram waveform data of more than 8 seconds has enough waveforms, so that the extracted electrocardiogram rhythm information and the PQRST waveform data are more accurate. The extraction of the representative PQRST waveform can effectively reduce the influence of waveform change caused by non-classified elements such as human body movement and electrode instability, and meanwhile, the data volume of the representative PQRST waveform is much less than that of the waveform data in the original electrocardiogram waveform data, so that the waveform data is more stable, the training amount of a later deep learning algorithm can be greatly reduced, the calculation efficiency of the deep learning algorithm is improved, and the quality of electrocardiogram auxiliary classification information provided for doctors is improved. Extraction of the electrocardiographic rhythm information can be used to improve the accuracy of the associated electrocardiographic classification information. The electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information are arranged into one-dimensional data and then are subjected to waveform classification through a deep learning algorithm, so that the deep learning algorithm can analyze the relevance among the information, and the accuracy of final electrocardiogram classification is improved.
Preferably, in the method for classifying an electrocardiogram based on a deep learning algorithm, in the step a, raw electrocardiogram waveform data having a measurement time of 8 seconds or more is acquired, and then the raw electrocardiogram waveform data is denoised.
The baseline drift noise, the electromyographic interference, the power frequency interference and the like of the original electrocardiogram waveform data can be removed, so that the accuracy of the final electrocardiogram classification result is further improved.
Preferably, the above-mentioned electrocardiogram classification method based on the deep learning algorithm includes the following steps:
a11. Removing baseline drift noise by adopting a high-pass filter;
a12. confirming whether the noise is too high or not based on the standard deviation of the PQ section signal and a threshold value method;
a13. When the noise is too high, a low-pass Butterworth filter is used for removing noise interference.
by the steps, baseline drift noise, myoelectricity interference, power frequency interference and the like in the original electrocardiogram waveform data can be effectively removed, so that the accuracy of the final electrocardiogram classification auxiliary information is further improved.
Preferably, in the method for classifying an electrocardiogram based on a deep learning algorithm, in step a, the electrocardiographic rhythm information includes an average ventricular rate, an average RR interval, a difference between a longest RR interval and a shortest RR interval, a standard deviation of RR intervals, consistency P-wave information, PR intervals and an average value of beats under the sinus rhythm, a detection result of a pre-shock wave in an R-wave, QT intervals and QTc intervals and an average value of beats under the sinus rhythm, a sinoatrial QRS average wave width, a sinoatrial P wave width and an average wave width, extra-systolic information, an extra-systolic type, an extra-systolic shape, a detection result of an F-wave of atrial flutter and an F-wave of atrial fibrillation, and a detection result of an asynchronous P-wave.
the above information has a great influence on the final electrocardiogram classification result, and the accuracy of the final electrocardiogram classification can be further improved.
Preferably, in the method for classifying an electrocardiogram based on a deep learning algorithm, the step a of extracting the representative PQRST waveform includes the steps of:
a21. Detecting original electrocardiogram waveform data by a first-order differentiation method and a threshold value method to obtain characteristic points of P waves, QRS waves and T waves;
a22. Clustering and classifying all PQRST waves in original electrocardiogram waveform data, taking the type with the largest number of PQRST waves as a representative PQRST waveform according to a classification result, if the type with the largest number is more than 2, selecting the type with the largest average amplitude of R waves as the representative PQRST waveform, and finally calculating the average waveform of the PQRST waves of each heartbeat as the representative PQRST waveform by using a superposition average method.
Through the steps, the characteristic points of P waves, QRS waves and T waves in the original electrocardiogram waveform data can be effectively extracted, all PQRST waves in the original electrocardiogram waveform data are clustered and classified, PQRST waveforms interfered by noise artifact in the original electrocardiogram waveform data and QRST waveforms related to rhythms can be effectively removed, and the obtained representative PQRST waveforms can be guaranteed to transmit more accurate effective information to carry out electrocardiogram classification.
preferably, in the method for classifying electrocardiograms based on a deep learning algorithm, in the step a, the additional information of electrocardiograms includes sex, height, chest circumference, weight, fat rate and race.
The information is related to the electrocardiogram classification standard, has a large influence on the final classification result, and the accuracy of the electrocardiogram classification result can be improved by considering the information.
preferably, in the method for classifying an electrocardiogram based on a deep learning algorithm, the original electrocardiogram waveform data is single-lead data.
The single-lead data is generally applicable to portable electrocardiogram detection instruments, so that the invention has wider application range.
preferably, in the method for classifying electrocardiograms based on the deep learning algorithm, the original electrocardiogram waveform data in the step a is multi-lead data, the electrocardiogram rhythm information is formed by serially connecting electrocardiogram rhythm information of each lead into one-dimensional data, and the representative PQRST waveform data is formed by serially connecting representative PQRST waveform data of each lead into one-dimensional data.
The multi-lead original electrocardiogram waveform data has more sufficient information and can improve the accuracy of related electrocardiogram classification auxiliary information, and when the electrocardiogram rhythm information formed by serially connecting the electrocardiogram rhythm information of each lead and the PQRST waveform data formed by serially connecting the PQRST waveform data of each lead are subjected to waveform classification by a deep learning algorithm, the correlation among the leads can be effectively summarized after sufficient training, so that the accuracy of the final electrocardiogram classification auxiliary information can be further improved.
Preferably, in the method for classifying an electrocardiogram based on a deep learning algorithm, in the step b, the deep learning algorithm is a convolutional neural network, an iterative neural network, or a deep neural network.
The three neural networks have higher accuracy, and the accuracy of the final electrocardiogram classification auxiliary information can be ensured.
The invention has the following beneficial effects:
1. the invention reasonably combines the deep learning algorithm in the field of artificial intelligence with the traditional electrocardiogram classification algorithm, and can greatly improve the accuracy of the final electrocardiogram classification auxiliary information. The invention utilizes the proven effective information data in the traditional electrocardiogram classification method, and simultaneously utilizes the super-strong learning ability, automatic feature extraction, automatic feature distribution relation classification and other superior abilities of the deep learning algorithm to make up the defects of insufficient feature extraction and insufficient feature correlation classification in the traditional electrocardiogram classification method.
2. the invention can more effectively provide auxiliary information of electrocardiogram classification required by doctors in early treatment. The traditional electrocardiogram classification algorithm is a static algorithm and does not have self-learning ability, but the invention combines the traditional electrocardiogram classification algorithm with the deep learning algorithm, thereby improving the accuracy and robustness of electrocardiogram classification on one hand, improving the understanding of various electrocardiogram classifications on the other hand, and providing auxiliary information for doctors to explain the mechanisms of various heart diseases.
drawings
FIG. 1 is a flowchart of an electrocardiogram classification method based on a deep learning algorithm according to the present invention.
Detailed Description
The invention will be described in further detail below with reference to the accompanying figure 1 and the detailed description, but they are not intended to limit the invention:
Example 1
A deep learning algorithm-based electrocardiogram classification method is shown in a flow chart of fig. 1, and specifically comprises the following steps:
a, (1) acquiring single-lead electrocardiogram waveform data and electrocardiogram additional information, and intercepting data with the length of 10 seconds from the single-lead electrocardiogram waveform data to serve as original electrocardiogram waveform data, wherein the single-lead electrocardiogram waveform data and the electrocardiogram additional information can be acquired through an existing database such as a Coulter Europe style electrocardiogram waveform database (CSE) or other ways, and the electrocardiogram additional information comprises gender, height, chest circumference, weight, fat rate and race.
(2) According to the requirement, the original electrocardiogram waveform data obtained in the step (1) can be subjected to denoising treatment, wherein the denoising treatment comprises the following steps:
a11. Removing baseline drift noise by adopting a high-pass filter;
a12. Confirming whether the noise is too high or not based on the standard deviation of the PQ section signal and a threshold value method;
a13. when the noise is too high, a low-pass Butterworth filter is used for removing noise interference.
(3) Calculating the discrimination points of the PQRST waveform according to the original electrocardiogram waveform data, thereby extracting electrocardiogram rhythm information according to the discrimination points of the PQRST waveform to obtain the electrocardiogram rhythm information, wherein the electrocardiogram rhythm information comprises an average ventricular rate, an average RR interval, the difference between the longest RR interval and the shortest RR interval, the standard deviation of the RR interval, consistency P wave information, PR interval and average value of each heart beat under sinus rhythm, detection result of pre-shock wave in R wave, QT interval and QTc interval and average value of each heart beat under sinus rhythm, sinus rhythm QRS average wave width, sinus rhythm P wave width and average wave width, extra-systole information, extra-systole type, extra-systole shape, detection result of F wave of atrial flutter and F wave of atrial fibrillation, and detection result of non-contemporaneous P wave, and the representative PQRST waveform is extracted through the following steps:
a21. detecting original electrocardiogram waveform data by a first-order differentiation method and a threshold value method to obtain characteristic points of P waves, QRS waves and T waves;
a22. Clustering and classifying all PQRST waves in original electrocardiogram waveform data, taking the type with the largest number of PQRST waves as a representative PQRST waveform according to a classification result, if the type with the largest number is more than 2, selecting the type with the largest average amplitude of R waves as a representative PQRST waveform class, and finally calculating the average waveform of the PQRST waves of each heartbeat as the representative PQRST waveform by using a superposition average method.
(4) In order to train the convolutional neural network, it is further required to acquire training data, which can be acquired from other corresponding physical examination results, or alternatively from an existing database, for example, from the european common body electrocardiogram waveform database (CSE), wherein the training data includes other corresponding single-lead electrocardiogram waveform data and electrocardiogram additional information, and the acquisition step of the training data is as follows: and (3) processing each information in the European-Council electrocardiogram waveform database (CSE) according to the steps (1) to (3) to acquire each electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the European-Council electrocardiogram waveform database (CSE).
And b, (1) setting the number of nodes of an input layer, a hidden layer and an output layer of the convolutional neural network, and randomly setting the weight among the nodes of adjacent layers.
(2) Arranging the electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the training data obtained in the step (4) in the step a into one-dimensional data, inputting the one-dimensional data from the input end of the convolutional neural network, and inputting the corresponding real electrocardiogram classification result to the result end of the convolutional neural network to train the convolutional neural network.
(3) After the training of the convolutional neural network is finished, the electrocardiogram additional information in the step (1) of the step a, the PQRST waveform data represented in the step (3) of the step a and the electrocardiogram rhythm information are arranged into one-dimensional data and then input to the input end of the convolutional neural network, namely, the waveform classification can be carried out through the convolutional neural network to obtain an electrocardiogram classification result, wherein the arrangement modes of the electrocardiogram additional information, the PQRST waveform data represented in the step (3) of the step a can be selected according to actual conditions.
Example 2
a deep learning algorithm-based electrocardiogram classification method is shown in a flow chart of fig. 1, and specifically comprises the following steps:
a, (1) acquiring single-lead electrocardiogram waveform data and electrocardiogram additional information, and intercepting data with the length of 8 seconds according to the single-lead electrocardiogram waveform data to serve as original electrocardiogram waveform data, wherein the single-lead electrocardiogram waveform data and the electrocardiogram additional information can be acquired through an existing database such as a Coulter Europe style electrocardiogram waveform database (CSE) or other ways, and the electrocardiogram additional information comprises gender, height, chest circumference, weight, fat rate and race.
(2) According to the requirement, the original electrocardiogram waveform data obtained in the step (1) can be subjected to denoising treatment, wherein the denoising treatment comprises the following steps:
a11. Removing baseline drift noise by adopting a high-pass filter;
a12. Confirming whether the noise is too high or not based on the standard deviation of the PQ section signal and a threshold value method;
a13. When the noise is too high, a low-pass Butterworth filter is used for removing noise interference.
(3) Calculating the discrimination points of the PQRST waveform according to the original electrocardiogram waveform data, thereby extracting electrocardiogram rhythm information according to the discrimination points of the PQRST waveform to obtain the electrocardiogram rhythm information, wherein the electrocardiogram rhythm information comprises an average ventricular rate, an average RR interval, the difference between the longest RR interval and the shortest RR interval, the standard deviation of the RR interval, consistency P wave information, PR interval and average value of each heart beat under sinus rhythm, detection result of pre-shock wave in R wave, QT interval and QTc interval and average value of each heart beat under sinus rhythm, sinus rhythm QRS average wave width, sinus rhythm P wave width and average wave width, extra-systole information, extra-systole type, extra-systole shape, detection result of F wave of atrial flutter and F wave of atrial fibrillation, and detection result of non-contemporaneous P wave, and the representative PQRST waveform is extracted through the following steps:
a21. detecting original electrocardiogram waveform data by a first-order differentiation method and a threshold value method to obtain characteristic points of P waves, QRS waves and T waves;
a22. Clustering and classifying all PQRST waves in original electrocardiogram waveform data, taking the type with the largest number of PQRST waves as a representative PQRST waveform according to a classification result, if the type with the largest number is more than 2, selecting the type with the largest average amplitude of R waves as a representative PQRST waveform class, and finally calculating the average waveform of the PQRST waves of each heartbeat as the representative PQRST waveform by using a superposition average method.
(4) in order to train the iterative neural network, it is further required to acquire training data, which may be acquired from an existing database, for example, from the european common body electrocardiogram waveform database (CSE), wherein the training data includes other corresponding single-lead electrocardiogram waveform data and electrocardiogram additional information, and the acquisition step of the training data is as follows: and (3) processing each information in the European-Council electrocardiogram waveform database (CSE) according to the steps (1) to (3) to acquire each electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the European-Council electrocardiogram waveform database (CSE).
and b, (1) setting the number of nodes of an input layer, a hidden layer and an output layer of the iterative neural network, and randomly setting the weight among the nodes of adjacent layers.
(2) And (b) arranging the electrocardiogram rhythm information, the PQRST waveform data and the electrocardiogram additional information in the training data obtained in the step (4) in the step (a) into one-dimensional data, inputting the one-dimensional data from the input end of the iterative neural network, and inputting the corresponding real classification result to the result end of the iterative neural network to train the iterative neural network.
(3) After the training of the iterative neural network is finished, the electrocardiogram additional information in the step (1) of the step a, the PQRST waveform data represented in the step (3) of the step a and the electrocardiogram rhythm information are arranged into one-dimensional data and then input to the input end of the iterative neural network, namely, the waveform classification can be carried out through the iterative neural network to obtain an electrocardiogram classification result, wherein the arrangement modes of the electrocardiogram additional information, the PQRST waveform data represented in the step (3) of the step a can be selected according to actual conditions.
Example 3
A deep learning algorithm-based electrocardiogram classification method is shown in a flow chart of fig. 1, and specifically comprises the following steps:
a, (1) acquiring single-lead electrocardiogram waveform data and electrocardiogram additional information, and intercepting data with the length of 16 seconds from the single-lead electrocardiogram waveform data to serve as original electrocardiogram waveform data, wherein the single-lead electrocardiogram waveform data and the electrocardiogram additional information can be acquired through an existing database such as a Coulter Europe style electrocardiogram waveform database (CSE) or other ways, and the electrocardiogram additional information comprises gender, height, chest circumference, weight, fat rate and race.
(2) According to the requirement, the original electrocardiogram waveform data obtained in the step (1) can be subjected to denoising treatment, wherein the denoising treatment comprises the following steps:
a11. Removing baseline drift noise by adopting a high-pass filter;
a12. confirming whether the noise is too high or not based on the standard deviation of the PQ section signal and a threshold value method;
a13. When the noise is too high, a low-pass Butterworth filter is used for removing noise interference.
(3) Calculating the discrimination points of the PQRST waveform according to the original electrocardiogram waveform data, thereby extracting electrocardiogram rhythm information according to the discrimination points of the PQRST waveform to obtain the electrocardiogram rhythm information, wherein the electrocardiogram rhythm information comprises an average ventricular rate, an average RR interval, the difference between the longest RR interval and the shortest RR interval, the standard deviation of the RR interval, consistency P wave information, PR interval and average value of each heart beat under sinus rhythm, detection result of pre-shock wave in R wave, QT interval and QTc interval and average value of each heart beat under sinus rhythm, sinus rhythm QRS average wave width, sinus rhythm P wave width and average wave width, extra-systole information, extra-systole type, extra-systole shape, detection result of F wave of atrial flutter and F wave of atrial fibrillation, and detection result of non-contemporaneous P wave, and the representative PQRST waveform is extracted through the following steps:
a21. Detecting original electrocardiogram waveform data by a first-order differentiation method and a threshold value method to obtain characteristic points of P waves, QRS waves and T waves;
a22. clustering and classifying all PQRST waves in original electrocardiogram waveform data, taking the type with the largest number of PQRST waves as a representative PQRST waveform according to a classification result, if the type with the largest number is more than 2, selecting the type with the largest average amplitude of R waves as a representative PQRST waveform class, and finally calculating the average waveform of the PQRST waves of each heartbeat as the representative PQRST waveform by using a superposition average method.
(4) in order to train the deep neural network, it is further required to acquire training data, which can be acquired from other corresponding physical examination results, or alternatively from an existing database, for example, from the european common body electrocardiogram waveform database (CSE), wherein the training data includes other corresponding single-lead electrocardiogram waveform data and electrocardiogram additional information, and the acquisition step of the training data is as follows: and (3) processing each information in the European-Council electrocardiogram waveform database (CSE) according to the steps (1) to (3) to acquire each electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the European-Council electrocardiogram waveform database (CSE).
And b, (1) setting the number of nodes of an input layer, a hidden layer and an output layer of the deep neural network, and randomly setting the weight among the nodes of adjacent layers.
(2) arranging the electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the training data obtained in the step (4) in the step a into one-dimensional data, inputting the one-dimensional data from the input end of the deep neural network, and inputting the corresponding real electrocardiogram classification result into the result end of the deep neural network to train the deep neural network.
(3) After the deep neural network is trained, arranging the electrocardiogram additional information in the step (1) of the step a, the PQRST waveform data represented in the step (3) of the step a and the electrocardiogram rhythm information into one-dimensional data and inputting the one-dimensional data into an input end of the deep neural network, namely performing waveform classification through the deep neural network to obtain an electrocardiogram classification result, wherein the arrangement modes of the electrocardiogram additional information, the PQRST waveform data represented in the step (3) of the step a can be selected according to actual conditions.
Example 4
A deep learning algorithm-based electrocardiogram classification method is shown in a flow chart of fig. 1, and specifically comprises the following steps:
a, (1) acquiring multi-lead electrocardiogram waveform data and electrocardiogram additional information, and intercepting data with the length of 10 seconds according to the multi-lead electrocardiogram waveform data to be used as original electrocardiogram waveform data, wherein the multi-lead electrocardiogram waveform data and the electrocardiogram additional information can be acquired through an existing database such as a Coulter Electrocardiogram waveform database (CSE) or other ways, and the multi-lead electrocardiogram waveform data can be multi-lead electrocardiogram waveform data such as twelve-lead electrocardiogram waveform data, three-lead electrocardiogram waveform data, six-lead electrocardiogram waveform data, eighteen-lead electrocardiogram waveform data and the like. The additional information of electrocardiogram comprises sex, height, chest circumference, weight, fat rate and race.
(2) According to the requirement, the original electrocardiogram waveform data obtained in the step (1) can be subjected to denoising treatment, wherein the denoising treatment comprises the following steps:
a11. removing baseline drift noise by adopting a high-pass filter;
a12. confirming whether the noise is too high or not based on the standard deviation of the PQ section signal and a threshold value method;
a13. When the noise is too high, a low-pass Butterworth filter is used for removing noise interference.
(3) Calculating the discrimination points of the PQRST waveform according to the original electrocardiogram waveform data, thereby extracting electrocardiogram rhythm information according to the discrimination points of the PQRST waveform to obtain the electrocardiogram rhythm information, wherein the electrocardiogram rhythm information comprises an average ventricular rate, an average RR interval, the difference between the longest RR interval and the shortest RR interval, the standard deviation of the RR interval, consistency P wave information, PR interval and average value of each heart beat under sinus rhythm, detection result of pre-shock wave in R wave, QT interval and QTc interval and average value of each heart beat under sinus rhythm, sinus rhythm QRS average wave width, sinus rhythm P wave width and average wave width, extra-systole information, extra-systole type, extra-systole shape, detection result of F wave of atrial flutter and F wave of atrial fibrillation, and detection result of non-contemporaneous P wave, and the representative PQRST waveform is extracted through the following steps:
a21. Detecting original electrocardiogram waveform data by a first-order differentiation method and a threshold value method to obtain characteristic points of P waves, QRS waves and T waves;
a22. Clustering and classifying all PQRST waves in original electrocardiogram waveform data, taking the type with the largest number of PQRST waves as a representative PQRST waveform according to a classification result, if the type with the largest number is more than 2, selecting the type with the largest average amplitude of R waves as a representative PQRST waveform class, and finally calculating the average waveform of the PQRST waves of each heartbeat as the representative PQRST waveform by using a superposition average method.
(4) In order to train the convolutional neural network, it is further required to acquire training data, which can be acquired from an existing database, for example, from the european community electrocardiogram waveform database (CSE), wherein the training data includes other corresponding multi-lead electrocardiogram waveform data and additional electrocardiogram information, and the acquisition steps of the training data are as follows: and (3) processing each information in the European-Council electrocardiogram waveform database (CSE) according to the steps (1) to (3) to acquire each electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the European-Council electrocardiogram waveform database (CSE).
and b, (1) setting the number of nodes of an input layer, a hidden layer and an output layer of the convolutional neural network, and randomly setting the weight among the nodes of adjacent layers.
(2) Arranging the electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the training data obtained in the step (4) in the step a into one-dimensional data, inputting the one-dimensional data from the input end of the convolutional neural network, and inputting the corresponding real electrocardiogram classification result to the result end of the convolutional neural network to train the convolutional neural network.
(3) After the training of the convolutional neural network is finished, the electrocardiogram additional information in the step (1) of the step a, the PQRST waveform data represented in the step (3) of the step a and the electrocardiogram rhythm information are arranged into one-dimensional data and then input to the input end of the convolutional neural network, namely, the waveform classification can be carried out through the convolutional neural network to obtain an electrocardiogram classification result, wherein the arrangement modes of the electrocardiogram additional information, the PQRST waveform data represented in the step (3) of the step a can be selected according to actual conditions.
example 5
a deep learning algorithm-based electrocardiogram classification method is shown in a flow chart of fig. 1, and specifically comprises the following steps:
a, (1) acquiring multi-lead electrocardiogram waveform data and electrocardiogram additional information, and intercepting data with the length of 8 seconds according to the multi-lead electrocardiogram waveform data to be used as original electrocardiogram waveform data, wherein the multi-lead electrocardiogram waveform data and the electrocardiogram additional information can be acquired through an existing database such as a Coulter Electrocardiogram waveform database (CSE) or other ways, and the multi-lead electrocardiogram waveform data can be multi-lead electrocardiogram waveform data such as twelve-lead electrocardiogram waveform data, three-lead electrocardiogram waveform data, six-lead electrocardiogram waveform data, eighteen-lead electrocardiogram waveform data and the like. The additional information of electrocardiogram comprises sex, height, chest circumference, weight, fat rate and race.
(2) according to the requirement, the original electrocardiogram waveform data obtained in the step (1) can be subjected to denoising treatment, wherein the denoising treatment comprises the following steps:
a11. Removing baseline drift noise by adopting a high-pass filter;
a12. Confirming whether the noise is too high or not based on the standard deviation of the PQ section signal and a threshold value method;
a13. when the noise is too high, a low-pass Butterworth filter is used for removing noise interference.
(3) calculating the discrimination points of the PQRST waveform according to the original electrocardiogram waveform data, thereby extracting electrocardiogram rhythm information according to the discrimination points of the PQRST waveform to obtain the electrocardiogram rhythm information, wherein the electrocardiogram rhythm information comprises an average ventricular rate, an average RR interval, the difference between the longest RR interval and the shortest RR interval, the standard deviation of the RR interval, consistency P wave information, PR interval and average value of each heart beat under sinus rhythm, detection result of pre-shock wave in R wave, QT interval and QTc interval and average value of each heart beat under sinus rhythm, sinus rhythm QRS average wave width, sinus rhythm P wave width and average wave width, extra-systole information, extra-systole type, extra-systole shape, detection result of F wave of atrial flutter and F wave of atrial fibrillation, and detection result of non-contemporaneous P wave, and the representative PQRST waveform is extracted through the following steps:
a21. detecting original electrocardiogram waveform data by a first-order differentiation method and a threshold value method to obtain characteristic points of P waves, QRS waves and T waves;
a22. clustering and classifying all PQRST waves in original electrocardiogram waveform data, taking the type with the largest number of PQRST waves as a representative PQRST waveform according to a classification result, if the type with the largest number is more than 2, selecting the type with the largest average amplitude of R waves as a representative PQRST waveform class, and finally calculating the average waveform of the PQRST waves of each heartbeat as the representative PQRST waveform by using a superposition average method.
(4) In order to train the iterative neural network, it is further required to acquire training data, which can be acquired from other corresponding physical examination results, or alternatively from an existing database, for example, from the european common body electrocardiogram waveform database (CSE), wherein the training data includes other corresponding multi-lead electrocardiogram waveform data and additional electrocardiogram information, and the acquisition step of the training data is as follows: and (3) processing each information in the European-Council electrocardiogram waveform database (CSE) according to the steps (1) to (3) to acquire each electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the European-Council electrocardiogram waveform database (CSE).
And b, (1) setting the number of nodes of an input layer, a hidden layer and an output layer of the iterative neural network, and randomly setting the weight among the nodes of adjacent layers.
(2) And (b) arranging the electrocardiogram rhythm information, the PQRST waveform data and the electrocardiogram additional information in the training data obtained in the step (4) in the step (a) into one-dimensional data, inputting the one-dimensional data from the input end of the iterative neural network, and inputting the corresponding real electrocardiogram classification result to the result end of the iterative neural network to train the iterative neural network.
(3) after the training of the iterative neural network is finished, the electrocardiogram additional information in the step (1) of the step a, the PQRST waveform data represented in the step (3) of the step a and the electrocardiogram rhythm information are arranged into one-dimensional data and then input to the input end of the iterative neural network, namely, the waveform classification can be carried out through the iterative neural network to obtain an electrocardiogram classification result, wherein the arrangement modes of the electrocardiogram additional information, the PQRST waveform data represented in the step (3) of the step a can be selected according to actual conditions.
Example 6
A deep learning algorithm-based electrocardiogram classification method is shown in a flow chart of fig. 1, and specifically comprises the following steps:
a, (1) acquiring multi-lead electrocardiogram waveform data and electrocardiogram additional information, and intercepting data with the length of 16 seconds according to the multi-lead electrocardiogram waveform data to be used as original electrocardiogram waveform data, wherein the multi-lead electrocardiogram waveform data and the electrocardiogram additional information can be acquired through an existing database such as a Coulter Electrocardiogram waveform database (CSE) or other ways, and the multi-lead electrocardiogram waveform data can be multi-lead electrocardiogram waveform data such as twelve-lead electrocardiogram waveform data, three-lead electrocardiogram waveform data, six-lead electrocardiogram waveform data, eighteen-lead electrocardiogram waveform data and the like. The additional information of electrocardiogram comprises sex, height, chest circumference, weight, fat rate and race.
(2) According to the requirement, the original electrocardiogram waveform data obtained in the step (1) can be subjected to denoising treatment, wherein the denoising treatment comprises the following steps:
a11. removing baseline drift noise by adopting a high-pass filter;
a12. Confirming whether the noise is too high or not based on the standard deviation of the PQ section signal and a threshold value method;
a13. When the noise is too high, a low-pass Butterworth filter is used for removing noise interference.
(3) Calculating the discrimination points of the PQRST waveform according to the original electrocardiogram waveform data, thereby extracting electrocardiogram rhythm information according to the discrimination points of the PQRST waveform to obtain the electrocardiogram rhythm information, wherein the electrocardiogram rhythm information comprises an average ventricular rate, an average RR interval, the difference between the longest RR interval and the shortest RR interval, the standard deviation of the RR interval, consistency P wave information, PR interval and average value of each heart beat under sinus rhythm, detection result of pre-shock wave in R wave, QT interval and QTc interval and average value of each heart beat under sinus rhythm, sinus rhythm QRS average wave width, sinus rhythm P wave width and average wave width, extra-systole information, extra-systole type, extra-systole shape, detection result of F wave of atrial flutter and F wave of atrial fibrillation, and detection result of non-contemporaneous P wave, and the representative PQRST waveform is extracted through the following steps:
a21. detecting original electrocardiogram waveform data by a first-order differentiation method and a threshold value method to obtain characteristic points of P waves, QRS waves and T waves;
a22. Clustering and classifying all PQRST waves in original electrocardiogram waveform data, taking the type with the largest number of PQRST waves as a representative PQRST waveform according to a classification result, if the type with the largest number is more than 2, selecting the type with the largest average amplitude of R waves as a representative PQRST waveform class, and finally calculating the average waveform of the PQRST waves of each heartbeat as the representative PQRST waveform by using a superposition average method.
(4) In order to train the deep neural network, it is also necessary to acquire training data, which can be acquired from other corresponding physical examination results, or alternatively from an existing database, for example, from the european common body electrocardiogram waveform database (CSE), wherein the training data includes other corresponding multi-lead electrocardiogram waveform data and additional electrocardiogram information, and the acquisition step of the training data is as follows: and (3) processing each information in the European-Council electrocardiogram waveform database (CSE) according to the steps (1) to (3) to acquire each electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the European-Council electrocardiogram waveform database (CSE).
and b, (1) setting the number of nodes of an input layer, a hidden layer and an output layer of the deep neural network, and randomly setting the weight among the nodes of adjacent layers.
(2) arranging the electrocardiogram rhythm information, the representative PQRST waveform data and the electrocardiogram additional information in the training data obtained in the step (4) in the step a into one-dimensional data, inputting the one-dimensional data from the input end of the deep neural network, and inputting the corresponding real electrocardiogram classification result into the result end of the deep neural network to train the deep neural network.
(3) After the deep neural network is trained, arranging the electrocardiogram additional information in the step (1) of the step a, the PQRST waveform data represented in the step (3) of the step a and the electrocardiogram rhythm information into one-dimensional data and inputting the one-dimensional data into an input end of the deep neural network, namely performing waveform classification through the deep neural network to obtain an electrocardiogram classification result, wherein the arrangement modes of the electrocardiogram additional information, the PQRST waveform data represented in the step (3) of the step a can be selected according to actual conditions.
Example 7
this example compares the classification results of the electrocardiograms of examples 1 to 6 with those of the electrocardiograms of the conventional measurement methods in terms of sensitivity and specificity, and the results are shown in the following table:
as can be seen from the above table, the sensitivity and specificity of the classification result of electrocardiogram obtained by the present invention are improved by about 10% compared with the classification result of electrocardiogram obtained by the traditional measurement method, and basically kept at about 97%, which can well meet the actual requirement of providing the classification information of electrocardiogram required for auxiliary diagnosis for doctors.
In summary, the above-mentioned embodiments are only preferred embodiments of the present invention, and all equivalent changes and modifications made within the scope of the claims of the present invention should be covered by the present invention.
Claims (5)
1. An electrocardiogram classification method based on a deep learning algorithm is characterized by comprising the following steps:
a. obtaining original electrocardiogram waveform data and electrocardiogram additional information with the measuring time of more than 8 seconds, and carrying out denoising processing on the original electrocardiogram waveform data, wherein the denoising processing comprises the following steps:
a11. Removing baseline drift noise by adopting a high-pass filter;
a12. confirming whether the noise is too high or not based on the standard deviation of the PQ section signal and a threshold value method;
a13. When the noise is too high, a low-pass Butterworth filter is used for removing noise interference;
Extracting electrocardiogram rhythm information and representative PQRST waveform according to the original electrocardiogram waveform data to obtain electrocardiogram rhythm information and representative PQRST waveform data;
In the step a, the extraction of the representative PQRST waveform comprises the following steps:
a21. Detecting original electrocardiogram waveform data by a first-order differentiation method and a threshold value method to obtain characteristic points of P waves, QRS waves and T waves;
a22. clustering and classifying all PQRST waves in original electrocardiogram waveform data, taking the type with the largest number of PQRST waves as a representative PQRST waveform according to a classification result, if the type with the largest number is more than 2, selecting the type with the largest average amplitude of R waves as the representative PQRST waveform, and finally calculating the average waveform of the PQRST waves of each heartbeat as the representative PQRST waveform by using a superposition average method;
b. training a neural network of a deep learning algorithm, wherein the deep learning algorithm is a convolutional neural network or an iterative neural network or a deep neural network, and after the electrocardiogram rhythm information, the PQRST waveform data and the electrocardiogram additional information obtained in the step a are arranged into one-dimensional data, carrying out waveform classification through the trained deep learning algorithm to obtain an electrocardiogram classification result.
2. The electrocardiogram classification method based on the deep learning algorithm as claimed in claim 1, wherein: in the step a, the electrocardiographic rhythm information includes an average ventricular heart rate, an average RR interval, a difference between a longest RR interval and a shortest RR interval, a standard deviation of RR intervals, consistency P-wave information, PR intervals and mean values of beats under sinus rhythm, pre-shock detection results in R-waves, QT intervals and QTc intervals and mean values of beats under sinus rhythm, a sinus rhythm QRS mean width, sinus rhythm P-wave width and mean wave width, extra-systole information, extra-systole types, extra-systole shapes, detection results of F-waves of atrial flutter and F-waves of atrial fibrillation, and detection results of non-contemporaneous P-waves.
3. The electrocardiogram classification method based on the deep learning algorithm as claimed in claim 1, wherein: in the step a, the electrocardiogram additional information comprises sex, height, chest circumference, weight, fat rate and race.
4. The electrocardiogram classification method based on the deep learning algorithm as claimed in claim 1, wherein: the original electrocardiogram waveform data is single-lead data.
5. The electrocardiogram classification method based on the deep learning algorithm as claimed in claim 1, wherein: the original electrocardiogram waveform data in the step a is multi-lead data, the electrocardiogram rhythm information is formed by connecting electrocardiogram rhythm information of each lead in series into one-dimensional data, and the representative PQRST waveform data is formed by connecting the representative PQRST waveform data of each lead in series into one-dimensional data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610572216.4A CN106214145B (en) | 2016-07-20 | 2016-07-20 | Electrocardiogram classification method based on deep learning algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610572216.4A CN106214145B (en) | 2016-07-20 | 2016-07-20 | Electrocardiogram classification method based on deep learning algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106214145A CN106214145A (en) | 2016-12-14 |
| CN106214145B true CN106214145B (en) | 2019-12-10 |
Family
ID=57530993
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610572216.4A Active CN106214145B (en) | 2016-07-20 | 2016-07-20 | Electrocardiogram classification method based on deep learning algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106214145B (en) |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778685A (en) * | 2017-01-12 | 2017-05-31 | 司马大大（北京）智能系统有限公司 | Electrocardiogram image-recognizing method, device and service terminal |
| CN108509823A (en) * | 2017-02-24 | 2018-09-07 | 深圳市理邦精密仪器股份有限公司 | The detection method and device of QRS complex |
| CN108511055B (en) * | 2017-02-27 | 2021-10-12 | 中国科学院苏州纳米技术与纳米仿生研究所 | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules |
| CN107203692B (en) * | 2017-05-09 | 2020-05-05 | 哈尔滨工业大学（威海） | Electrocardio data digital signal processing method based on deep convolutional neural network |
| CN107256393B (en) * | 2017-06-05 | 2020-04-24 | 四川大学 | Feature extraction and state recognition of one-dimensional physiological signals based on deep learning |
| CN107981858B (en) | 2017-11-27 | 2020-12-01 | 上海优加利健康管理有限公司 | Automatic electrocardiogram heart beat identification and classification method based on artificial intelligence |
| CN107837082B (en) * | 2017-11-27 | 2020-04-24 | 乐普（北京）医疗器械股份有限公司 | Automatic electrocardiogram analysis method and device based on artificial intelligence self-learning |
| CN107714023B (en) * | 2017-11-27 | 2020-09-01 | 上海优加利健康管理有限公司 | Static electrocardiogram analysis method and device based on artificial intelligence self-learning |
| CN108399369B (en) * | 2018-02-02 | 2021-10-19 | 东南大学 | Electrocardio beat classification method based on distributed computation and deep learning |
| CN108836302B (en) * | 2018-03-19 | 2021-06-04 | 武汉海星通技术股份有限公司 | Intelligent electrocardiogram analysis method and system based on deep neural network |
| CN108926341A (en) * | 2018-04-20 | 2018-12-04 | 平安科技（深圳）有限公司 | Detection method, device, computer equipment and the storage medium of ECG signal |
| CN108960050A (en) * | 2018-05-25 | 2018-12-07 | 东软集团股份有限公司 | Classification model training method, ECG data classifying method, device and equipment |
| CN109077721B (en) * | 2018-07-20 | 2021-03-23 | 广州视源电子科技股份有限公司 | Atrial fibrillation detection apparatus and storage medium |
| CN109480825B (en) * | 2018-12-13 | 2021-08-06 | 武汉中旗生物医疗电子有限公司 | Electrocardio data processing method and device |
| CN109620210B (en) * | 2019-01-28 | 2019-10-25 | 山东科技大学 | A kind of electrocardiosignal classification method of the CNN based on from coding mode in conjunction with GRU |
| CN109770861B (en) * | 2019-03-29 | 2022-04-12 | 广州视源电子科技股份有限公司 | Method, device, equipment and storage medium for training and detecting cardioelectric rhythm model |
| CN110750770B (en) * | 2019-08-18 | 2023-10-03 | 浙江好络维医疗技术有限公司 | Electrocardiogram-based method for unlocking electronic equipment |
| CN111110224A (en) * | 2020-01-17 | 2020-05-08 | 武汉中旗生物医疗电子有限公司 | Electrocardiogram classification method and device based on multi-angle feature extraction |
| CN111297350B (en) * | 2020-02-27 | 2021-08-31 | 福州大学 | Three-heart beat multi-model comprehensive decision-making electrocardiogram feature classification method integrating source end influence |
| CN111557660A (en) * | 2020-06-08 | 2020-08-21 | 东北大学 | Arrhythmia identification method under sub-population deep learning framework |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101969842A (en) * | 2008-01-14 | 2011-02-09 | 皇家飞利浦电子股份有限公司 | Atrial fibrillation monitoring |
| CN103038772A (en) * | 2010-03-15 | 2013-04-10 | 新加坡保健服务集团有限公司 | Method of predicting the survivability of a patient |
| CN104873186A (en) * | 2015-04-17 | 2015-09-02 | 中国科学院苏州生物医学工程技术研究所 | Wearable artery detection device and data processing method thereof |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI374727B (en) * | 2008-11-19 | 2012-10-21 | Univ Nat Yang Ming | Chip for sensing a physiological signal and sensing method thereof |
| CN101766484B (en) * | 2010-01-18 | 2011-09-07 | 董军 | Method and equipment for identification and classification of electrocardiogram |
| CN102028460B (en) * | 2011-01-04 | 2014-01-08 | 复旦大学 | Ventricular fibrillation signal sequence automatic-detection system |
| CN102908135B (en) * | 2012-10-08 | 2015-06-10 | 中国科学院深圳先进技术研究院 | ECG diagnosis system and operating method of ECG diagnosis system |
-
2016
- 2016-07-20 CN CN201610572216.4A patent/CN106214145B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101969842A (en) * | 2008-01-14 | 2011-02-09 | 皇家飞利浦电子股份有限公司 | Atrial fibrillation monitoring |
| CN103038772A (en) * | 2010-03-15 | 2013-04-10 | 新加坡保健服务集团有限公司 | Method of predicting the survivability of a patient |
| CN104873186A (en) * | 2015-04-17 | 2015-09-02 | 中国科学院苏州生物医学工程技术研究所 | Wearable artery detection device and data processing method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106214145A (en) | 2016-12-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106214145B (en) | Electrocardiogram classification method based on deep learning algorithm | |
| CN107358196B (en) | Heart beat type classification method and device and electrocardiograph | |
| EP3692904B1 (en) | Method and device for self-learning dynamic electrocardiography analysis employing artificial intelligence | |
| CN109171712B (en) | Atrial fibrillation identification method, atrial fibrillation identification device, atrial fibrillation identification equipment and computer readable storage medium | |
| CN106108889B (en) | Electrocardiogram classification method based on deep learning algorithm | |
| JP7429371B2 (en) | Method and system for quantifying and removing asynchronous noise in biophysical signals | |
| CN106214123B (en) | A kind of electrocardiogram compressive classification method based on deep learning algorithm | |
| Oweis et al. | QRS detection and heart rate variability analysis: A survey | |
| Wen et al. | A feasible feature extraction method for atrial fibrillation detection from BCG | |
| WO2012106729A1 (en) | System and method for evaluating an electrophysiological signal | |
| CN109288515B (en) | Periodicity monitoring method and device based on premature beat signal in wearable electrocardiosignal | |
| CN111265210A (en) | Atrial fibrillation prediction device and equipment based on deep learning | |
| Turnip et al. | An application of modified filter algorithm fetal electrocardiogram signals with various subjects | |
| Lee et al. | A real-time abnormal beat detection method using a template cluster for the ECG diagnosis of IoT devices | |
| CN111832537A (en) | Abnormal electrocardiosignal identification method and abnormal electrocardiosignal identification device | |
| Farrokhi et al. | Reliable peak detection and feature extraction for wireless electrocardiograms | |
| CN109394206B (en) | Real-time monitoring method and device based on premature beat signal in wearable electrocardiosignal | |
| Lim et al. | Enhanced wavelet transformation for feature extraction in highly variated ECG signal | |
| CN111839505A (en) | Atrial fibrillation classification method based on cardiovascular system electrical-mechanical activity information | |
| Sanamdikar et al. | Extraction of different features of ECG signal for detection of cardiac arrhythmias by using wavelet transformation Db 6 | |
| Ghosal et al. | Ecg beat quality assessment using self organizing map | |
| CN117338309A (en) | Electrocardiosignal approximation threshold calculation method, identification method and storage medium | |
| Oktivasari et al. | A simple real-time system for detection of normal and myocardial ischemia in the ST segment and t wave ECG signal | |
| Revanth et al. | Detection of QRS Complexes in ECG Signal | |
| CN113349753A (en) | Arrhythmia detection method based on portable dynamic electrocardiogram monitor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |