CN109893118A - A kind of electrocardiosignal classification diagnosis method based on deep learning - Google Patents
A kind of electrocardiosignal classification diagnosis method based on deep learning Download PDFInfo
- Publication number
- CN109893118A CN109893118A CN201910164534.0A CN201910164534A CN109893118A CN 109893118 A CN109893118 A CN 109893118A CN 201910164534 A CN201910164534 A CN 201910164534A CN 109893118 A CN109893118 A CN 109893118A
- Authority
- CN
- China
- Prior art keywords
- electrocardiosignal
- wave
- characteristic wave
- signal
- feature
- 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.)
- Pending
Links
Abstract
The electrocardiosignal classification diagnosis method based on deep learning that the invention proposes a kind of.The present invention is acquired to obtain original electro-cardiologic signals using electrocardiogram acquisition equipment to the electrocardiosignal between left upper extremity and right upper extremity；Original electro-cardiologic signals are pre-processed to electrocardiosignal after being denoised, extract the characteristic wave of electrocardiosignal after denoising；Wavelet coefficient is constructed transformation matrix according to certain rule by the wavelet transformation that the characteristic wave of electrocardiosignal is carried out to the western small echo of the more shellfishes of quadravalence；The corresponding time-frequency figure of the characteristic wave that transformation matrix is considered as electrocardiosignal is passed to deep learning module, the disease that the person that obtains ecg signal acquiring may suffer from.The present invention can analyze and highlight the feature of signal on frequency domain, realize tentative diagnosis of the electrocardiosignal to a variety of heart diseases, improve the accuracy rate of diagnosis.
Description
Technical field
The present invention relates to medical signals process fields, and in particular to a kind of electrocardiosignal classification diagnosis based on deep learning
Method.
Background technique
Heart is one of most important organ of the mankind, and the power of flowing can be provided to blood and blood is transported to body
Each part of body.The relevant disease incidence of heart is rapid, consequence is serious, has become the No.1 of threat human life in recent years
Killer.
Arrhythmia cordis is the Etiological of cardiovascular disease, refers specifically to the irregular transformation of heart rate, including atrial fibrillation,
Ventricular extra-systolic, ventricular fibrillation and heart rate are too fast, and lasting arrhythmia cordis may cause to influence for a long time on the health of people, therefore to the heart
Rate is regularly monitored particularly significant for preventing and managing cardiovascular disease.Body fluid around heart is electrically conductive, every
In a cardiac cycle, the bioelectricity of heart changes, these bioelectricity are referred to as electrocardio, Single Cardiac Cell variation
Summation can reflect on body surface, make between each point of body surface that there are potential differences.Electrocardiogram (ECG) is located on human body
Electrode is most strong one of the diagnostic tool of analysis of cardiac activity, the research to electrocardiosignal for the record of cardiac electrical activity
And the inspection of heart has important reference role.It can detecte that whether there is or not cardiac arrhythmia, myocardial damages etc. to ask by electrocardiosignal
It inscribes and analyzes it and identify.By ECG signal sampling heart, whether there is or not diseases, just by normal or abnormal electrocardiosignal
Really classification, can help to check heart disease, this is of great significance.But still there are many problems to be solved, such as the heart at present
Electric signal is influenced to will lead to by outside noise, and result reliability is not high, and accuracy is lower.
In the detection of arrhythmia cordis, the form and feature of electrocardiosignal play a crucial role detection, the heart
Electric signal has the wave of different amplitudes and shape, respectively becomes P, Q, R, S, T wave, different ventricles and artery generate these waves
Physical activity is also different.Therefore, most of medical experts can consider the following of electrocardiosignal when carrying out arrhythmia cordis diagnosis
Feature: between the time interval, QT wave between the relative position of P, Q, R, S, T wave, size, form and signature waveform when
Between be spaced etc. other features.
According to before about the detection work of detection arrhythmia cordis it is found that passing through (i.e. adjacent characteristic wave between detection R-R wave
Shape) variation obtain the statistical parameters such as mean value, variance, standard deviation, can tentatively judge whether signal is to make an uproar according to statistical parameter
Sound, therefore we are for statistical analysis with regard to the signature waveform of electrocardiosignal herein, are pre-processed to be provided with electrocardiosignal.
The present more existing patent (including invention granted patent and Invention Announce patent) about ECG's data compression
It is as follows:
Publication No. is that the patent of invention of 106805965 A of CN proposes the side that electrocardiosignal classification is carried out with deep learning
Method obtains training set and test set data to electrocardiosignal segmentation, and instructs with data of the deep learning to training set
Disaggregated model is got, is classified with the disaggregated model to electrocardiosignal, can be classified extremely to different types of heart rate；
The patent of invention that publication No. is 109063552 A of CN propose a kind of classification of multi-lead electrocardiosignal method and
System extracts the signal characteristic of multi-lead electrocardiosignal by multiple branch circuit convolution residual error neural network, is based on Softmax function
The flexibility and adaptability that classification improves electrocardiosignal classification are carried out to the fused multi-lead electrocardiosignal；
Publication No. is that the patent of invention of 107495959 A of CN proposes a kind of electrocardio based on one-dimensional convolutional neural networks
Modulation recognition method, this method obtain several signature waveform candidate segments on the basis of the R crest value point detected, establish towards the heart
The one-dimensional convolutional neural networks model of electric signal classifies to these signature waveform candidate segments, and this method avoid electrocardiosignals
The problem of the necessary precise positioning of characteristic point.
As can be seen that the patent of invention about electrocardio processing existing at present is all based on greatly at one-dimensional electrocardiosignal
Reason, and existing handling implement is more suitable for the processing of time-frequency figure；On the other hand, existing electrocardio processing method is all based on greatly letter
Number processing, but signal processing calculates cumbersome, and is limited by signal quality, cannot accurately extract the spy in electrocardiosignal
Sign；On the other hand, deep learning has the ability of very strong data processing and classification, and on ECG's data compression using compared with
It is few, advantage can not be played.
Therefore, find it is a kind of convert time-frequency figure for one-dimensional electrocardiosignal, and signal processing and deep learning are combined,
Accurately differentiate according to electrocardiosignal and the method for Diagnosing Cardiac disease has become a urgent problem to be solved.In recent years, when
Frequency-domain transform has been widely used on ECG's data compression, and ECG's data compression and deep learning are combined into electrocardiosignal
The network of one trend of processing, RNN and some shallow-layers also achieves good results in this respect.Innovative point of the invention exists
In one-dimensional electrocardiosignal is changed into time-frequency figure using wavelet transformation, use for reference and using the two dimensional image that currently have been relatively mature
Processing method and deep neural network classify to signal, realize the Combined Treatment of signal processing and deep learning, and then sentence
Break signal gathered person whether illness and suffer from which kind of disease, to obtain diagnostic result.For conventional method, mention
The high utilization rate of electrocardiosignal and the accuracy rate of heart disease diagnosis.
Summary of the invention
In order to overcome the disadvantage that accuracy rate is lower on traditional ECG's data compression, the invention proposes one kind to be based on depth
The electrocardiosignal classification diagnosis method of habit.
The technical scheme is that a kind of electrocardiosignal classification diagnosis method based on deep learning, specifically includes following
Step:
Step 1: the electrocardiosignal between left upper extremity and right upper extremity being acquired to obtain using electrocardiogram acquisition equipment original
Electrocardiosignal；
Step 2: original electro-cardiologic signals being pre-processed to electrocardiosignal after being denoised, extract electrocardiosignal after denoising
Characteristic wave；
Step 3: the characteristic wave of electrocardiosignal being carried out to the wavelet transformation of the western small echo of the more shellfishes of quadravalence, by wavelet coefficient according to one
Set pattern rule building transformation matrix；
Step 4: the corresponding time-frequency figure of the characteristic wave that transformation matrix is considered as electrocardiosignal is passed to deep learning module, obtains
The disease that may be suffered to ecg signal acquiring person.
Preferably, original electro-cardiologic signals described in step 1 are sorig=[a1,a2,…,aN], wherein N is the sample of acquisition
This number；
Preferably, original electro-cardiologic signals described in step 2 are pre-processed electrocardiosignal after being denoised are as follows:
By original electro-cardiologic signals sorig=[a1,a2,…,aN], wherein N is that the sample number of acquisition is input to 6 rank Butterworths
Low-pass filter filters out the high-frequency noise of original electro-cardiologic signals, electrocardiosignal s after being denoiseddenoise=[b1,b2,…,bN]；
The characteristic wave s of electrocardiosignal after extraction denoising described in step 2featureAre as follows:
Using the algorithm of qrs wave group in the quick detection electrocardiosignal of proposition, s is extracteddenoise=[b1,b2,…,bN] in
R wave and calculate n-th section of electrocardiosignal sdenoiseThe total quantity K of middle R waven, n is the serial number of R wave electrocardiosignal, n ∈ [1, M], M
For the quantity of R wave band；
Within t minutes electrocardiosignal sampling time, according to relevant medical knowledge, every section of electrocardiosignal sdenoiseMiddle R wave number
Measure KnIt should be within the scope of 50t≤Kn≤100t, to electrocardiosignal progress Preliminary detection, K after extracting characteristic waven> 100t
Or KnThe R wave of < 50t is regarded as noise signal or excessive electrocardiosignal affected by noise, replaces n-th section of heart with full 0 sequence
Electric signal sdenoise；
According to n-th section of electrocardiosignal sdenoiseThe total quantity K of middle R wavenTo electrocardiosignal sdenoiseExtract characteristic wave
sfeature, it is as follows to extract characteristic wave process:
If n-th section of electrocardiosignal sdenoiseFor non-zero sequence, then within the scope of without departing from signal length, selection is located at n-th section
Electrocardiosignal sdenoiseThe R peak-to-peak value point at middle part, and centered on the point, respectively take a certain number of signaling points to make from left and right respectively
For the characteristic wave of the electrocardiosignal, in order to more preferably embody the feature of signal and reduce calculation amount, the characteristic wave length L of selection includes
Four complete heartbeat waveforms under normal cardiac rate；
If n-th section of electrocardiosignal sdenoiseFor 0 sequence, then use 0 sequence of length L as the characteristic wave of this kind of signal；
Preferably, the feature s of electrocardiosignal described in step 3featureCarry out the wavelet transformation of the western small echo of the more shellfishes of quadravalence
Are as follows:
Claim spy to wave sfeatureWavelet analysis is carried out to signal with mallet algorithm, using the western small echo of the more shellfishes of quadravalence to feature
Wave carries out the wavelet transformation that scale is f/60-0.6f, and wherein f is sample frequency, extracts the wavelet coefficient of signal, is mutually strained
Wavelet coefficient λ after changingd, d ∈ [1, D]；
Wavelet coefficient is stored in a matrix according to certain rule described in step 3 are as follows:
Characteristic wave sfeatureCoefficient lambda after the wavelet transformation of different scale1, λ2..., λD, it is put into matrix in sequence
In every a line, each characteristic wave s is obtainedfeatureCorresponding transformation matrix A=[λ1, λ2..., λD]；
Preferably, deep learning module is passed to described in step 4, the disease that the person that obtains ecg signal acquiring may suffer from
Disease:
Transformation matrix is considered as characteristic wave sfeatureCorresponding time-frequency figure；
Each characteristic wave sfeatureCorresponding transformation matrix A=[λ1, λ2..., λD] ResNet-34 model is used,
Original electro-cardiologic signals s can be obtained using the networkorig=[a1,a2,…,aN] the disease that may suffer from of picker and may
Property size, a possibility that every kind of disease illness, is calculated by softmax function, when possibility is greater than given threshold
Think signal acquisition person with this kind of disease.
The present invention help to obtain accurate medical diagnosis on disease, and final output signal picker can with every heart disease
It can property.Carry out the tentative diagnosis of heart disease in a manner of non-intruding in daily life and hospital applying, the present invention can be square
Just heart disease is detected in time and early warning, mitigate doctor work load and patient check expense burden.This hair
The method of the electrocardiosignal Diagnosing Cardiac disease based on deep learning of bright proposition can be very good to extract the frequency domain of electrocardiosignal
Feature, and signal processing, image procossing and medicine are combined, it improves and the accurate of heart disease detection is carried out by electrocardiosignal
Rate.
Detailed description of the invention
Fig. 1: for the method for the present invention flow chart；
Fig. 2: for the general technical block diagram of the embodiment of the present invention.
Specific embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, with reference to the accompanying drawings and embodiments to this hair
It is bright to be described in further detail, it should be understood that implementation example described herein is merely to illustrate and explain the present invention, not
For limiting the present invention.
Fig. 1 is flow chart of the method for the present invention, and Fig. 2 is the general frame of technical solution of the present invention, and the present invention is based on depth
Spend the method for the electrocardiosignal Diagnosing Cardiac disease of study.Removing outputs and inputs part, and include following three modules: data are pre-
Processing module, one-dimensional signal are converted to time-frequency module and medical diagnosis on disease module.The major function of data preprocessing module is benefit
Collected original electro-cardiologic signals are filtered with signal processing, image procossing and medical knowledge and characteristic wave extracts, are calculated
The quantity of characteristic wave in the signal acquisition time, and intercept one section on this basis and correspond to the signal of acquisition time as the electrocardio
The characteristic wave of signal；The major function that one-dimensional signal is converted to time-frequency module is that the small echo of different scale is carried out to characteristic wave
Transformation, transformation coefficient is stored in transition matrix, and transition matrix is converted to time-frequency figure according to certain rules, passes through this
The electrocardiosignal of time domain can be transformed to Frequency domain by a module, meanwhile, it can solve traditional heart using the algorithm of image procossing
The complex problem of Electric signal processing algorithm, improves the accuracy rate of Modulation recognition；The major function of medical diagnosis on disease module is benefit
Classified with the ResNet-34 network of transformation to the image after conversion, electrocardiosignal corresponding to the image is judged with this
Whether picker may suffer from certain heart disease.Whether whole system final output signal picker health and may suffer from
The information such as heart disease.
Below with reference to Fig. 1 to Fig. 2, a specific embodiment of the invention, the technical side of the specific embodiment of the invention are introduced
Case is a kind of electrocardiosignal classification diagnosis method based on deep learning, specifically includes the following steps:
Step 1: the electrocardiosignal between left upper extremity and right upper extremity being acquired to obtain using electrocardiogram acquisition equipment original
Electrocardiosignal；
Original electro-cardiologic signals described in step 1 are sorig=[a1,a2,…,aN], wherein N is the sample number of acquisition, is selected
ECG signal data collection contains the work of the ecg signal data under people's rest state of 121818 all ages and classes difference heart states
For original electro-cardiologic signals, each data include the .mat file comprising electrocardiogram and a .hea text comprising shape information
Part marks the heart disease information suffered from file containing each signal acquisition person；
Step 2: original electro-cardiologic signals being pre-processed to electrocardiosignal after being denoised, extract electrocardiosignal after denoising
Characteristic wave；
Original electro-cardiologic signals described in step 2 are pre-processed denoised after electrocardiosignal are as follows:
By original electro-cardiologic signals sorig=[a1,a2,…,aN], wherein N is that the sample number of acquisition is input to 6 rank Butterworths
Low-pass filter filters out the high-frequency noise of original electro-cardiologic signals, electrocardiosignal s after being denoiseddenoise=[b1,b2,…,bN]；
The characteristic wave s of electrocardiosignal after extraction denoising described in step 2featureAre as follows:
Using the algorithm of qrs wave group in the quick detection electrocardiosignal of proposition, s is extracteddenoise=[b1,b2,…,bN] in
R wave and calculate n-th section of electrocardiosignal sdenoiseThe total quantity K of middle R waven, n is the serial number of R wave electrocardiosignal, n ∈ [1, M], M
For the quantity of R wave band；
Within t minutes electrocardiosignal sampling time, according to relevant medical knowledge, every section of electrocardiosignal sdenoiseMiddle R wave number
Measure KnIt should be within the scope of 50t≤Kn≤100t, to electrocardiosignal progress Preliminary detection, K after extracting characteristic waven> 100t
Or KnThe R wave of < 50t is regarded as noise signal or excessive electrocardiosignal affected by noise, replaces n-th section of heart with full 0 sequence
Electric signal sdenoise；
According to n-th section of electrocardiosignal sdenoiseThe total quantity K of middle R wavenTo electrocardiosignal sdenoiseExtract characteristic wave
sfeature, it is as follows to extract characteristic wave process:
If n-th section of electrocardiosignal sdenoiseFor non-zero sequence, then within the scope of without departing from signal length, selection is located at n-th section
Electrocardiosignal sdenoiseThe R peak-to-peak value point at middle part, and centered on the point, respectively take a certain number of signaling points to make from left and right respectively
For the characteristic wave of the electrocardiosignal, in order to more preferably embody the feature of signal and reduce calculation amount, the characteristic wave length L of selection includes
Four complete heartbeat waveforms under normal cardiac rate；
If n-th section of electrocardiosignal sdenoiseFor 0 sequence, then use 0 sequence of length L as the characteristic wave of this kind of signal；
Step 3: the characteristic wave of electrocardiosignal being carried out to the wavelet transformation of the western small echo of the more shellfishes of quadravalence, by wavelet coefficient according to one
Set pattern rule building transformation matrix；
The feature s of electrocardiosignal described in step 3featureCarry out the wavelet transformation of the western small echo of the more shellfishes of quadravalence are as follows:
Claim spy to wave sfeatureWavelet analysis is carried out to signal with mallet algorithm, using the western small echo of the more shellfishes of quadravalence to feature
Wave carries out the wavelet transformation that scale is f/60-0.6f, and wherein f is sample frequency, extracts the wavelet coefficient of signal, is mutually strained
Wavelet coefficient λ after changingd, d ∈ [1, D]；
Wavelet coefficient is stored in a matrix according to certain rule described in step 3 are as follows:
Characteristic wave sfeatureCoefficient lambda after the wavelet transformation of different scale1, λ2..., λD, it is put into matrix in sequence
In every a line, each characteristic wave s is obtainedfeatureCorresponding transformation matrix A=[λ1, λ2..., λD]；
Step 4: the corresponding time-frequency figure of the characteristic wave that transformation matrix is considered as electrocardiosignal is passed to deep learning module, obtains
The disease that may be suffered to ecg signal acquiring person.
Deep learning module is passed to described in step 4, the disease that the person that obtains ecg signal acquiring may suffer from:
Transformation matrix is considered as characteristic wave sfeatureCorresponding time-frequency figure；
Obtain the image data collection of corresponding (0.6f-f/60) the * 4f of each signal；
Each characteristic wave sfeatureCorresponding transformation matrix A=[λ1, λ2..., λD] using the residual error nerve of modification
Network ResNet-34 model can obtain original electro-cardiologic signals s using the networkorig=[a1,a2,…,aN] picker, can
The size of the disease and possibility that can suffer from；A possibility that every kind of disease illness, is calculated by softmax function, works as possibility
It is believed that signal acquisition person suffers from this kind of disease when greater than given threshold.
It should be understood that the above-mentioned description for preferred embodiment is more detailed, can not therefore be considered to this
The limitation of invention patent protection range, those skilled in the art under the inspiration of the present invention, are not departing from power of the present invention
Benefit requires to make replacement or deformation under protected ambit, fall within the scope of protection of the present invention, this hair
It is bright range is claimed to be determined by the appended claims.
Claims (5)
1. a kind of electrocardiosignal classification diagnosis method based on deep learning characterized by comprising
Step 1: the electrocardiosignal between left upper extremity and right upper extremity being acquired to obtain original electrocardiographicdigital using electrocardiogram acquisition equipment
Signal；
Step 2: original electro-cardiologic signals being pre-processed to electrocardiosignal after being denoised, extract the feature of electrocardiosignal after denoising
Wave；
Step 3: the characteristic wave of electrocardiosignal being carried out to the wavelet transformation of the western small echo of the more shellfishes of quadravalence, by wavelet coefficient according to a set pattern
Rule building transformation matrix；
Step 4: the corresponding time-frequency figure of the characteristic wave that transformation matrix is considered as electrocardiosignal is passed to deep learning module, obtains the heart
The disease that electrical signal collection person may suffer from.
2. the electrocardiosignal classification diagnosis method according to claim 1 based on deep learning, it is characterised in that: step 1
Described in original electro-cardiologic signals be sorig=[a1,a2,…,aN], wherein N is the sample number of acquisition.
3. the electrocardiosignal classification diagnosis method according to claim 1 based on deep learning, it is characterised in that: step 2
Described in original electro-cardiologic signals pre-processed denoised after electrocardiosignal are as follows:
By original electro-cardiologic signals sorig=[a1,a2,…,aN], wherein N is that the sample number of acquisition is input to 6 rank Butterworth low passes
Filter filters out the high-frequency noise of original electro-cardiologic signals, electrocardiosignal s after being denoiseddenois=e[b1,b2,…,bN]；
The characteristic wave s of electrocardiosignal after extraction denoising described in step 2featureAre as follows:
Using the algorithm of qrs wave group in the quick detection electrocardiosignal of proposition, s is extracteddenoise=[b1,b2,…,bN] in R wave
And calculate n-th section of electrocardiosignal sdenoiseThe total quantity K of middle R waven, n is the serial number of R wave electrocardiosignal, and n ∈ [1, M], M are R wave
The quantity of section；
Within t minutes electrocardiosignal sampling time, according to relevant medical knowledge, every section of electrocardiosignal sdenoiseMiddle R wave number amount KnIt answers
It should be within the scope of 50t≤Kn≤100t, to electrocardiosignal progress Preliminary detection, K after extracting characteristic waven> 100t or Kn<
The R wave of 50t is regarded as noise signal or excessive electrocardiosignal affected by noise, replaces n-th section of electrocardiosignal with full 0 sequence
sdenoise；
According to n-th section of electrocardiosignal sdenoiseThe total quantity K of middle R wavenTo electrocardiosignal sdenoiseExtract characteristic wave sfeature, extract
Characteristic wave process is as follows:
If n-th section of electrocardiosignal sdenoiseFor non-zero sequence, then within the scope of without departing from signal length, selection is located at n-th section of electrocardio
Signal sdenoiseThe R peak-to-peak value point at middle part, and centered on the point, respectively taking a certain number of signaling points to be used as from left and right respectively should
The characteristic wave of electrocardiosignal, in order to more preferably embody the feature of signal and reduce calculation amount, the characteristic wave length L of selection includes normal
Four complete heartbeat waveforms under heart rate；
If n-th section of electrocardiosignal sdenoiseFor 0 sequence, then use 0 sequence of length L as the characteristic wave of this kind of signal.
4. the electrocardiosignal classification diagnosis method according to claim 1 based on deep learning, it is characterised in that: step 3
Described in electrocardiosignal feature sfeatureCarry out the wavelet transformation of the western small echo of the more shellfishes of quadravalence are as follows:
Claim spy to wave sfeatureWavelet analysis is carried out to signal with mallet algorithm, using the western small echo of the more shellfishes of quadravalence to characteristic wave into
Row scale is the wavelet transformation of f/60-0.6f, and wherein f is sample frequency, the wavelet coefficient of signal is extracted, after accordingly being converted
Wavelet coefficient λd, d ∈ [1, D]；
Wavelet coefficient is stored in a matrix according to certain rule described in step 3 are as follows:
Characteristic wave sfeatureCoefficient lambda after the wavelet transformation of different scale1, λ2..., λD, it is put into every a line of matrix in sequence
In, obtain each characteristic wave sfeatureCorresponding transformation matrix A=[λ1, λ2..., λD]。
5. the electrocardiosignal classification diagnosis method according to claim 1 based on deep learning, it is characterised in that: step 4
Described in be passed to deep learning module, the disease that the person that obtains ecg signal acquiring may suffer from:
Transformation matrix is considered as characteristic wave sfeatureCorresponding time-frequency figure；
Each characteristic wave sfeatureCorresponding transformation matrix A=[λ1, λ2..., λD] ResNet-34 model is used, utilize this
Network can obtain original electro-cardiologic signals sorig=[a1,a2,…,aN] the disease that may suffer from of picker and possibility it is big
Small, a possibility that every kind of disease illness, is calculated by softmax function, when possibility is greater than given threshold it is believed that believing
Number picker suffers from this kind of disease.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910164534.0A CN109893118A (en) | 2019-03-05 | 2019-03-05 | A kind of electrocardiosignal classification diagnosis method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910164534.0A CN109893118A (en) | 2019-03-05 | 2019-03-05 | A kind of electrocardiosignal classification diagnosis method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109893118A true CN109893118A (en) | 2019-06-18 |
Family
ID=66946458
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910164534.0A Pending CN109893118A (en) | 2019-03-05 | 2019-03-05 | A kind of electrocardiosignal classification diagnosis method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109893118A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598549A (en) * | 2019-08-07 | 2019-12-20 | 王满 | Convolutional neural network information processing system based on cardiac function monitoring and training method |
| CN110916645A (en) * | 2019-12-10 | 2020-03-27 | 电子科技大学 | QRS wave identification method combining wavelet transformation and image segmentation network |
| CN111368627A (en) * | 2019-11-20 | 2020-07-03 | 山东大学 | Heart sound classification method and system based on CNN combined with improved frequency wavelet slice transformation |
| CN111449644A (en) * | 2020-03-19 | 2020-07-28 | 复旦大学 | Bioelectricity signal classification method based on time-frequency transformation and data enhancement technology |
| CN113095302A (en) * | 2021-05-21 | 2021-07-09 | 中国人民解放军总医院 | Depth model for arrhythmia classification, method and apparatus using the same |
| CN113876332A (en) * | 2021-10-27 | 2022-01-04 | 深圳大学 | Electrocardiosignal monitoring device and method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1951320A (en) * | 2006-11-16 | 2007-04-25 | 上海交通大学 | Abnormal electrocardiogram recognition method based on ultra-complete characteristics |
| CN107203692A (en) * | 2017-05-09 | 2017-09-26 | 哈尔滨工业大学（威海） | The implementation method of atrial fibrillation detection based on depth convolutional neural networks |
| EP3284403A1 (en) * | 2016-08-16 | 2018-02-21 | American Reliance, Inc. | Methods and systems for disease analysis based on transformations of diagnostic signals |
| CN109077715A (en) * | 2018-09-03 | 2018-12-25 | 北京工业大学 | A kind of electrocardiosignal automatic classification method based on single lead |
| CN109124620A (en) * | 2018-06-07 | 2019-01-04 | 深圳市太空科技南方研究院 | A kind of atrial fibrillation detection method, device and equipment |
-
2019
- 2019-03-05 CN CN201910164534.0A patent/CN109893118A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1951320A (en) * | 2006-11-16 | 2007-04-25 | 上海交通大学 | Abnormal electrocardiogram recognition method based on ultra-complete characteristics |
| EP3284403A1 (en) * | 2016-08-16 | 2018-02-21 | American Reliance, Inc. | Methods and systems for disease analysis based on transformations of diagnostic signals |
| CN107203692A (en) * | 2017-05-09 | 2017-09-26 | 哈尔滨工业大学（威海） | The implementation method of atrial fibrillation detection based on depth convolutional neural networks |
| CN109124620A (en) * | 2018-06-07 | 2019-01-04 | 深圳市太空科技南方研究院 | A kind of atrial fibrillation detection method, device and equipment |
| CN109077715A (en) * | 2018-09-03 | 2018-12-25 | 北京工业大学 | A kind of electrocardiosignal automatic classification method based on single lead |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598549A (en) * | 2019-08-07 | 2019-12-20 | 王满 | Convolutional neural network information processing system based on cardiac function monitoring and training method |
| CN111368627A (en) * | 2019-11-20 | 2020-07-03 | 山东大学 | Heart sound classification method and system based on CNN combined with improved frequency wavelet slice transformation |
| CN111368627B (en) * | 2019-11-20 | 2023-08-22 | 山东大学 | Method and system for classifying heart sounds by combining CNN (computer numerical network) with improved frequency wavelet slice transformation |
| CN110916645A (en) * | 2019-12-10 | 2020-03-27 | 电子科技大学 | QRS wave identification method combining wavelet transformation and image segmentation network |
| CN111449644A (en) * | 2020-03-19 | 2020-07-28 | 复旦大学 | Bioelectricity signal classification method based on time-frequency transformation and data enhancement technology |
| CN113095302A (en) * | 2021-05-21 | 2021-07-09 | 中国人民解放军总医院 | Depth model for arrhythmia classification, method and apparatus using the same |
| CN113095302B (en) * | 2021-05-21 | 2023-06-23 | 中国人民解放军总医院 | Depth model for arrhythmia classification, method and device using same |
| CN113876332A (en) * | 2021-10-27 | 2022-01-04 | 深圳大学 | Electrocardiosignal monitoring device and method |
| CN113876332B (en) * | 2021-10-27 | 2023-12-08 | 深圳大学 | Electrocardiosignal monitoring device and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110876626B (en) | Depression detection system based on optimal lead selection of multi-lead electroencephalogram | |
| CN109893118A (en) | A kind of electrocardiosignal classification diagnosis method based on deep learning | |
| WO2019100566A1 (en) | Artificial intelligence self-learning-based static electrocardiography analysis method and apparatus | |
| WO2019100565A1 (en) | Method and device for self-learning dynamic electrocardiography analysis employing artificial intelligence | |
| JP7429371B2 (en) | Method and system for quantifying and removing asynchronous noise in biophysical signals | |
| CN107184198A (en) | A kind of electrocardiosignal classifying identification method | |
| CN107837082A (en) | Electrocardiogram automatic analysis method and device based on artificial intelligence self study | |
| He et al. | Recognition of ECG patterns using artificial neural network | |
| Srivastava et al. | DWT-based feature extraction from ECG signal | |
| Bsoul et al. | Detection of P, QRS, and T components of ECG using wavelet transformation | |
| CN109044348A (en) | atrial fibrillation detection device and storage medium | |
| Talatov et al. | Algorithmic and software analysis and processing of ECG signals | |
| Mahesh et al. | ECG arrhythmia classification based on logistic model tree | |
| Zhang et al. | Deep learning-based signal quality assessment for wearable ECGs | |
| Zhao et al. | PVC recognition for wearable ECGs using modified frequency slice wavelet transform and convolutional neural network | |
| Kusumandari et al. | Application of convolutional neural network classifier for wireless arrhythmia detection | |
| Iadarola et al. | A new method for sEMG envelope detection from reduced measurements | |
| Hegde et al. | A review on ECG signal processing and HRV analysis | |
| CN110960207A (en) | Tree model-based atrial fibrillation detection method, device, equipment and storage medium | |
| Liu et al. | Automatic Arrhythmia Detection Based on Convolutional Neural Networks. | |
| Fathail et al. | Ecg paper digitization and r peaks detection using fft | |
| Jiang et al. | Heartbeat classification system based on modified stacked denoising autoencoders and neural networks | |
| Lin et al. | Fractal QRS-complexes pattern recognition for imperative cardiac arrhythmias | |
| Khandait et al. | ECG signal processing using classifier to analyses cardiovascular disease | |
| Dembrani et al. | Accurate detection of ECG signals in ECG monitoring systems by eliminating the motion artifacts and improving the signal quality using SSG filter with DBE |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190618 |
|
| RJ01 | Rejection of invention patent application after publication |