CN109846471A - A kind of myocardial infarction detection method based on BiGRU deep neural network - Google Patents
A kind of myocardial infarction detection method based on BiGRU deep neural network Download PDFInfo
- Publication number
- CN109846471A CN109846471A CN201910095803.2A CN201910095803A CN109846471A CN 109846471 A CN109846471 A CN 109846471A CN 201910095803 A CN201910095803 A CN 201910095803A CN 109846471 A CN109846471 A CN 109846471A
- Authority
- CN
- China
- Prior art keywords
- bigru
- neural network
- deep neural
- formula
- output
- 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
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention relates to a kind of myocardial infarction detection methods based on BiGRU deep neural network, the following steps are included: 1), data prediction, the baseline drift in original electro-cardiologic signals is filtered out using median filtering, Hz noise in original electro-cardiologic signals is filtered out using Butterworth digital band-reject filter, myoelectricity interference is filtered out using Chebyshev's wave digital lowpass filter；2), the heart claps segmentation, detects R crest value into Spline Wavelet Transform by two, and then calculate RR interphase and extract to QRS complex data；3), model training carries out deep learning classification to the waveform detected in the step 2) of detection by BiGRU deep neural network；The present invention has the advantages that myocardial infarction accurately detects classification, effectively carries out deep learning classification to electrocardiosignal.
Description
Technical field
The invention belongs to the hearts to clap detection sorting technique field, and in particular to a kind of heart based on BiGRU deep neural network
Flesh infarct detection method.
Background technique
Cardiovascular disease is to seriously threaten one of disease of human health, and in China, cardiovascular disease is increasingly becoming high-incidence
Disease.Myocardial infarction refers to that athero- variation has occurred in the coronary artery of cardiac nutrition, and the cholesterol plaques deposited on lumen wall fall off
Thrombus is formed, certain branch coronary artery is plugged, certain part cardiac muscle is made to cannot get blood supply for a long time, myocardial ischemia, damage just has occurred
Wound even necrosis.Heart infarction has the high death rate and disability rate.Myocardial infarction frequently-occurring disease was in six or seven ten years old the elderly in the past,
Due to too fast rhythm of life and undesirable living habit, so that the young man for suffering from the disease is more and more.For Acute myocardial
Infarction Patients restore the blood supply of cardiac muscle if the blood vessel of blocking can be got through in 2 hours, and most cardiac muscles can exempt from
In necrosis.If the blood vessel of blocking can be got through in 1 hour after the onset, mortality is only 1%.
The characteristic feature of myocardial infarction ECG waveform has: ST sections of oblique types are raised, and T wave height is alarmmed；The ST sections of back of a bow or horizontal type lift
It is high；T wave symmetry is inverted；Symmetry negative T wave is from depth to shallow；T wave restores normal or long-term unchanged；Deepen and broadening disease
Rationality Q wave.Pathologic Q wave or QS wave occur, and are caused by necrotic myocardium；ST sections in the back of a bow, shape is raised upwards, is drawn by myocardial damage
It rises；T wave is inverted, and is caused by myocardial ischemia.The method of traditional manual extraction feature determines R wave crest first, then with similar
Method positions Q wave starting point, and S wave terminal, peak point, the beginning and end of P wave and T wave, finally obtain several amplitudes and interphase is special
Sign.These are characterized in rule selection according to the doctor's diagnosis, it has one disadvantage in that, although QRS complex detection algorithm is accurate
Degree is very high, but the detection of R wave still remains error, is extracted due to the position that other characteristic values are all based on R wave, institute
Accumulated error can be generated with other characteristic values extracted based on this.Although the detection technique of QRS complex is more mature, can not also
Accomplish effective detection to waveforms such as P wave, T waves.If corresponding waveform cannot be effectively detected out, just can not accurately diagnose
The state of an illness out.Currently, ECG automatic identification algorithm can only identify several typical abnormal heart rhythms, and accuracy rate is also not up to faced
The requirement of bed diagnosis.
Hand-designed feature relies primarily on the priori knowledge of designer, is difficult with the advantage of big data, due to relying on hand
Work adjusting parameter, the parameter for only allowing to occur a small amount of in characteristic Design.
Summary of the invention
One kind is provided the purpose of the present invention is overcome the deficiencies in the prior art, and there is myocardial infarction accurately to detect classification, have
Imitate the myocardial infarction detection method based on BiGRU deep neural network that deep learning classification is carried out to electrocardiosignal.
Technical scheme is as follows:
A kind of myocardial infarction detection method based on BiGRU deep neural network, comprising the following steps:
1), data prediction filters out the baseline drift in original electro-cardiologic signals using median filtering, using Butterworth number
Word bandstop filter filters out the Hz noise in original electro-cardiologic signals, and it is dry to filter out myoelectricity using Chebyshev's wave digital lowpass filter
It disturbs；
2), the heart claps segmentation, detects R crest value into Spline Wavelet Transform by two, and then calculate RR interphase and to QRS complex
Data extract；
3), model training carries out depth to the waveform detected in the step 2) of detection by BiGRU deep neural network
Practise classification.
Preferably, the specific classification method of the step 3) are as follows:
Firstly, building BiGRU deep neural network, then using the BiGRU deep neural network of building between RR
Phase and QRS complex data carry out processing classification.
Further, using the output of this neuron inputted with a upper neuron as the defeated of GRU deep neural network
Enter, the output of this neuron is calculated by GRU deep neural network；
Specifically, the formula that the GRU deep neural network uses is as follows:
zt=σ (Wz·[ht-1, xt]) formula 1
rt=σ (Wr·[ht-1, xt]) formula 2
In the formula 1- formula 4:
ht-1Indicate the output of a neuron；
xtIndicate the input of this neuron；
WzIndicate the weight of update door；
σ indicates sigmoid function；
Update door ztThe information at moment is brought into the degree of current hidden state, z before controltBigger, the moment is hidden before
It is more to hide the information that node provides；
rtResetting door is indicated, as resetting door rtWhen close to 0, the information of concealed nodes before indicating to ignore, only by current time
Input as input, this mechanism can make model abandon some garbages an of neuron；
WrIndicate the weight of resetting door；
Indicate the candidate output valve of Current neural member；
The weight of W expression output state；
Tanh indicates hyperbolic tangent function；
htIndicate the output valve of this neuron；
Forward direction hidden layer state h corresponding to t moment BiGRU is calculated separately using aforementioned formula 1-4tWith reversed hidden layer state
ht', then to htAnd ht' weighted sum obtains the hidden layer state h of t momentt", formula are as follows:
ht"=wtht+vth′t+bjFormula 5
Wherein, wt、vtRespectively indicate t moment htAnd ht' corresponding weight, bjIt indicates to the corresponding biasing of training；
The heart for finally calculating output claps type of prediction, as follows using formula:
yjType of prediction is clapped for the heart of output；
wjIndicate weight coefficient matrix to be trained.
Further, t moment is calculated using the formula 1-4 calculate separately forward direction hidden layer state corresponding to t moment BiGRU
htWith reversed hidden layer state ht' when, the output h of (t-1) moment forward hidden layer state is used respectivelyt-1With reversed hidden layer state
Export ht-1Substitution formula 1-4 calculates corresponding htAnd ht′。
Further, it is measured between BiGRU deep neural network reality output and desired output using cross entropy loss function
Degree of closeness, is defined as:
Wherein, y is desired output,For the reality output of neuron, n is the sample size of training.
Compared with prior art, the beneficial effects of the present invention are:
1, the present invention using median filtering filters out baseline drift, Butterworth digital band-reject filter filters out Hz noise,
Myoelectricity interference noise is filtered out using the low cylinder filter of Chebyshev's number, effectively to the pole baseline drift of original electro-cardiologic signals and is made an uproar
Sound carries out filtering out processing, convenient for subsequent analysis and classification to electrocardiosignal；
2, the present invention is using BiGRU deep neural network and softmax function to the electrocardiosignal by handling and dividing
Deep learning classification is carried out, to improve myocardial infarction detection efficiency, and effectively improves the love classification of myocardial infarction deep learning
Accuracy；
In short, there is myocardial infarction accurately to detect classification, effectively carry out deep learning classification to electrocardiosignal by the present invention
Advantage.
Detailed description of the invention
Fig. 1 is beat classification flow chart of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
A kind of myocardial infarction detection method based on BiGRU deep neural network, comprising the following steps:
1), data prediction filters out the baseline drift in original electro-cardiologic signals using median filtering, using Butterworth number
Word bandstop filter filters out the Hz noise in original electro-cardiologic signals, and it is dry to filter out myoelectricity using Chebyshev's wave digital lowpass filter
It disturbs；
2), the heart claps segmentation, detects R crest value into Spline Wavelet Transform by two, and then calculate RR interphase and to QRS complex
Data extract；
3), model training carries out depth to the waveform detected in the step 2) of detection by BiGRU deep neural network
Practise classification.
Preferably, the specific classification method of the step 3) are as follows:
Firstly, building BiGRU deep neural network, then using the BiGRU deep neural network of building between RR
Phase and QRS complex data carry out processing classification.
Further, using the output of this neuron inputted with a upper neuron as the defeated of GRU deep neural network
Enter, the output of this neuron is calculated by GRU deep neural network；
Specifically, the formula that the GRU deep neural network uses is as follows:
zt=σ (Wz·[ht-1, xt]) formula 1
rt=σ (Wr·[ht-1, xt]) formula 2
In the formula 1- formula 4:
ht-1Indicate the output of a neuron；
xtIndicate the input of this neuron；
WzIndicate the weight of update door；
σ indicates sigmoid function；
Update door ztThe information at moment is brought into the degree of current hidden state, z before controltBigger, the moment is hidden before
It is more to hide the information that node provides；
rtResetting door is indicated, as resetting door rtWhen close to 0, the information of concealed nodes before indicating to ignore, only by current time
Input as input, this mechanism can make model abandon some garbages an of neuron；
WrIndicate the weight of resetting door；
Indicate the candidate output valve of Current neural member；
The weight of w expression output state；
Tanh indicates hyperbolic tangent function；
htIndicate the output valve of this neuron；
Forward direction hidden layer state h corresponding to t moment BiGRU is calculated separately using aforementioned formula 1-4tWith reversed hidden layer state
ht', then to htAnd ht' weighted sum obtains the hidden layer state h of t momentt", formula are as follows:
ht"=wtht+vth′t+bjFormula 5
Wherein, wt、vtRespectively indicate t moment htAnd ht' corresponding weight, bjIt indicates to the corresponding biasing of training；
The heart for finally calculating output claps type of prediction, as follows using formula:
yjType of prediction is clapped for the heart of output；
wjIndicate weight coefficient matrix to be trained.
Further, t moment is calculated using the formula 1-4 calculate separately forward direction hidden layer state corresponding to t moment BiGRU
htWith reversed hidden layer state ht' when, the output h of (t-1) moment forward hidden layer state is used respectivelyt-1With reversed hidden layer state
Export ht-1Substitution formula 1-4 calculates corresponding htAnd ht′。
Further, it is measured between BiGRU deep neural network reality output and desired output using cross entropy loss function
Degree of closeness, is defined as:
Wherein, y is desired output,For the reality output of neuron, n is the sample size of training.
Experimental verification:
1, experiment parameter
Noise suppression preprocessing carried out to the ECG signal of input first, then carry out the detection of R wave, by every about two minutes
Electrocardiosignal is divided into heart bat one by one, and front and back takes the sampled data of 250ms and 400ms respectively on the basis of R wave crest, constitutes
The electrocardial vector of one lead is respectively adopted identical mode to 8 lead electrocardiosignals and handles, and generates 8 electrocardial vectors.
Each electrocardial vector inputs a BiGRU network and is learnt, and the result of 8 BiGRU e-learnings inputs a full connection
SoftMax layer output category result.Network parameter is constrained using L2 regularization method, training process introduces
Dropout strategy prevents over-fitting, is used for model training using the SGD optimization method of batch.
Design parameter setting is as shown in the table:
2, evaluation index
In order to evaluate the performance of detection algorithm proposed by the present invention, we used four statistical indicators, they are respectively
Classification sensitivity (Sen), specific (Spe), precision (Ppr) and accuracy (Acc).The method that classification Accuracy evaluation is proposed
In the overall performance that all effective hearts are clapped.Since the quantity that the different type heart is clapped is different, Sen, Spe and Ppr classify in assessment
Device aspect of performance will appear lesser deviation.Four statistical indicators can calculate as follows:
Wherein TP is the quantity for being correctly detected as the MI heart of MI and clapping, and TN is the quantity for being correctly identified as the HC heart of HC and clapping, FN
It is the quantity for the MI heart bat that error detection is HC, FP is the HC heart umber of beats amount that error diagnosis is MI.
3, result and analysis
Classification experiments are carried out to PTB electrocardiosignal on TensorFlow platform herein, platform intergration CNN, RNN,
LSTM and GRU even deep learning model, wherein CPU is i7-7700, inside saves as 32GB, and GPU is NVIDIAGeForce GTX
1080, video memory 8GB, operating system are 64 Windows10.
It is in terms of accuracy, sensibility and specificity as a result, Sen=99.93%, Spe=in the present embodiment
99.72%, Acc=99.89%.
This paper presents the multi-lead myocardial infarction detection algorithms based on BiGRU, are believed first using filter group electrocardio
It number is pre-processed, R wave crest is then positioned using Quadric Spline small wave converting method, next every electrocardiosignal is divided
It is clapped at the independent heart, two classification is finally carried out using BiGRU deep learning method.With disclosed PTB ecg database to algorithm
Verified, and with other documents propose algorithm experimental result compare, the results showed that algorithm proposed in this paper have compared with
High sensitivity, accurate rate, accuracy, and there is universality.It can be from using deep learning frame proposed by the invention
Potential useful feature is extracted in the ECG signal of multi-lead.Method proposed in this paper can further be extended,
It realizes and similar classification task herein.Further work can explore training classifier on other ECG data collection, with
Detect various other heart diseases.In view of its excellent performance, the multi-lead heart infarction detection algorithm based on BiGRU can be applied
In computer-aided diagnosis platform, to assist true MI to detect.
Although the present invention is described in detail referring to the foregoing embodiments, for those skilled in the art,
It is still possible to modify the technical solutions described in the foregoing embodiments, or part of technical characteristic is carried out etc.
With replacement, all within the spirits and principles of the present invention, any modification, equivalent replacement, improvement and so on should be included in this
Within the protection scope of invention.
Claims (5)
1. a kind of myocardial infarction detection method based on BiGRU deep neural network, which comprises the following steps:
1), data prediction filters out the baseline drift in original electro-cardiologic signals using median filtering, using Butterworth number tape
Resistance filter filters out the Hz noise in original electro-cardiologic signals, filters out myoelectricity interference using Chebyshev's wave digital lowpass filter；
2), the heart claps segmentation, detects R crest value into Spline Wavelet Transform by two, and then calculate RR interphase and to QRS complex data
It extracts；
3), model training carries out deep learning point to the waveform detected in the step 2) of detection by BiGRU deep neural network
Class.
2. the myocardial infarction detection method according to claim 1 based on BiGRU deep neural network, which is characterized in that
The specific classification method of the step 3) are as follows:
Firstly, building BiGRU deep neural network, then using the BiGRU deep neural network of building to RR interphase and
QRS complex data carry out processing classification.
3. the myocardial infarction detection method according to claim 2 based on BiGRU deep neural network, it is characterised in that:
Using the input of the input of this neuron and a upper neuron exported as GRU deep neural network, pass through GRU depth
Degree neural computing goes out the output of this neuron；
Specifically, the formula that the GRU deep neural network uses is as follows:
zt=σ (Wz·[ht-1, xt]) formula 1
rt=σ (Wr·[ht-1, xt]) formula 2
In the formula 1-4:
ht-1Indicate the output of a neuron；
xtIndicate the input of this neuron；
WzIndicate the weight of update door；
σ indicates sigmoid function；
Update door ztThe information at moment is brought into the degree of current hidden state, z before controltBigger, the moment hides section before
The information that point provides is more；
rtResetting door is indicated, as resetting door rtWhen close to 0, the information of concealed nodes before indicating to ignore is only defeated by current time
Enter as input, this mechanism can make model abandon some garbages an of neuron；
wrIndicate the weight of resetting door；
Indicate the candidate output valve of Current neural member；
The weight of w expression output state；
Tanh indicates hyperbolic tangent function；
htIndicate the output valve of this neuron；
Forward direction hidden layer state h corresponding to t moment BiGRU is calculated separately using aforementioned formula 1-4tWith reversed hidden layer state ht',
Then to htAnd ht' weighted sum obtains the hidden layer state h of t momentt", formula are as follows:
ht"=wtht+vth′t+bjFormula 5
Wherein, wt、vtRespectively indicate t moment htAnd ht' corresponding weight, bjIt indicates to the corresponding biasing of training；
The heart for finally calculating output claps type of prediction, as follows using formula:
yj=softmax (wjh″t+bj) formula 6
yjType of prediction is clapped for the heart of output；
wjIndicate weight coefficient matrix to be trained.
4. the myocardial infarction detection method according to claim 3 based on BiGRU deep neural network, it is characterised in that:
T moment, which is calculated, using the formula 1-4 calculates separately forward direction hidden layer state h corresponding to t moment BiGRUtWith reversed hidden layer shape
State ht' when, the output h of (t-1) moment forward hidden layer state is used respectivelyt-1With the output h of reversed hidden layer statet-1Substitution formula
1-4 calculates corresponding htAnd ht′。
5. the myocardial infarction detection method according to claim 3 based on BiGRU deep neural network, which is characterized in that
Degree of closeness between BiGRU deep neural network reality output and desired output, definition are measured using cross entropy loss function
Are as follows:
Wherein, y is desired output,For the reality output of neuron, n is the sample size of training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910095803.2A CN109846471A (en) | 2019-01-30 | 2019-01-30 | A kind of myocardial infarction detection method based on BiGRU deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910095803.2A CN109846471A (en) | 2019-01-30 | 2019-01-30 | A kind of myocardial infarction detection method based on BiGRU deep neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109846471A true CN109846471A (en) | 2019-06-07 |
Family
ID=66897187
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910095803.2A Pending CN109846471A (en) | 2019-01-30 | 2019-01-30 | A kind of myocardial infarction detection method based on BiGRU deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109846471A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110141220A (en) * | 2019-06-20 | 2019-08-20 | 鲁东大学 | Myocardial infarction automatic testing method based on multi-modal fusion neural network |
| CN110141219A (en) * | 2019-06-20 | 2019-08-20 | 鲁东大学 | Myocardial infarction automatic testing method based on lead fusion deep neural network |
| CN110381524A (en) * | 2019-07-15 | 2019-10-25 | 安徽理工大学 | The mobile flow on-line prediction method of large scene based on Bi-LSTM, system and storage medium |
| CN111110228A (en) * | 2020-01-17 | 2020-05-08 | 武汉中旗生物医疗电子有限公司 | Electrocardiosignal R wave detection method and device |
| CN112842342A (en) * | 2021-01-25 | 2021-05-28 | 北京航空航天大学 | Electrocardiogram and magnetic signal classification method combining Hilbert curve and integrated learning |
| CN112989508A (en) * | 2021-02-01 | 2021-06-18 | 复旦大学 | Filter optimization design method based on deep learning algorithm |
| CN113080984A (en) * | 2021-03-25 | 2021-07-09 | 南京蝶谷健康科技有限公司 | Myocardial infarction identification and positioning method based on CNN and LSTM |
| CN113171104A (en) * | 2021-04-25 | 2021-07-27 | 安徽十锎信息科技有限公司 | Congestive heart failure automatic diagnosis method based on deep learning |
| CN113749666A (en) * | 2021-09-10 | 2021-12-07 | 郑州大学 | Myocardial infarction classification method based on fusion of ventricular regular features and XGboost |
| CN117281531A (en) * | 2023-11-27 | 2023-12-26 | 北京科技大学 | Psychological fatigue state identification method and system based on convolution long and short-time memory network |
| CN118177827A (en) * | 2024-04-22 | 2024-06-14 | 中国人民解放军南部战区总医院 | Myocardial infarction positioning method based on electrocardiograph vector diagram |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1539372A (en) * | 2003-10-24 | 2004-10-27 | �Ϻ���ͨ��ѧ | Method and device for early diagnosis of heart disease basaed on high-frequency waveform of cardiograph |
| WO2010077997A2 (en) * | 2008-12-16 | 2010-07-08 | Bodymedia, Inc. | Method and apparatus for determining heart rate variability using wavelet transformation |
| CN105411565A (en) * | 2015-11-20 | 2016-03-23 | 北京理工大学 | Heart rate variability feature classification method based on generalized scale wavelet entropy |
| CN108062388A (en) * | 2017-12-15 | 2018-05-22 | 北京百度网讯科技有限公司 | Interactive reply generation method and device |
| CN108830334A (en) * | 2018-06-25 | 2018-11-16 | 江西师范大学 | A kind of fine granularity target-recognition method based on confrontation type transfer learning |
| CN109214003A (en) * | 2018-08-29 | 2019-01-15 | 陕西师范大学 | The method that Recognition with Recurrent Neural Network based on multilayer attention mechanism generates title |
-
2019
- 2019-01-30 CN CN201910095803.2A patent/CN109846471A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1539372A (en) * | 2003-10-24 | 2004-10-27 | �Ϻ���ͨ��ѧ | Method and device for early diagnosis of heart disease basaed on high-frequency waveform of cardiograph |
| WO2010077997A2 (en) * | 2008-12-16 | 2010-07-08 | Bodymedia, Inc. | Method and apparatus for determining heart rate variability using wavelet transformation |
| CN105411565A (en) * | 2015-11-20 | 2016-03-23 | 北京理工大学 | Heart rate variability feature classification method based on generalized scale wavelet entropy |
| CN108062388A (en) * | 2017-12-15 | 2018-05-22 | 北京百度网讯科技有限公司 | Interactive reply generation method and device |
| CN108830334A (en) * | 2018-06-25 | 2018-11-16 | 江西师范大学 | A kind of fine granularity target-recognition method based on confrontation type transfer learning |
| CN109214003A (en) * | 2018-08-29 | 2019-01-15 | 陕西师范大学 | The method that Recognition with Recurrent Neural Network based on multilayer attention mechanism generates title |
Non-Patent Citations (1)
| Title |
|---|
| 博士学位论文编辑部: "《2007年上海大学博士学位论文 56 基于软计算机的故障诊断机理及其应用研究》", 31 December 2007 * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110141219A (en) * | 2019-06-20 | 2019-08-20 | 鲁东大学 | Myocardial infarction automatic testing method based on lead fusion deep neural network |
| CN110141220A (en) * | 2019-06-20 | 2019-08-20 | 鲁东大学 | Myocardial infarction automatic testing method based on multi-modal fusion neural network |
| CN110381524B (en) * | 2019-07-15 | 2022-12-20 | 安徽理工大学 | Bi-LSTM-based large scene mobile flow online prediction method, system and storage medium |
| CN110381524A (en) * | 2019-07-15 | 2019-10-25 | 安徽理工大学 | The mobile flow on-line prediction method of large scene based on Bi-LSTM, system and storage medium |
| CN111110228A (en) * | 2020-01-17 | 2020-05-08 | 武汉中旗生物医疗电子有限公司 | Electrocardiosignal R wave detection method and device |
| CN112842342A (en) * | 2021-01-25 | 2021-05-28 | 北京航空航天大学 | Electrocardiogram and magnetic signal classification method combining Hilbert curve and integrated learning |
| CN112842342B (en) * | 2021-01-25 | 2022-03-29 | 北京航空航天大学 | Electrocardiogram and magnetic signal classification method combining Hilbert curve and integrated learning |
| CN112989508A (en) * | 2021-02-01 | 2021-06-18 | 复旦大学 | Filter optimization design method based on deep learning algorithm |
| CN112989508B (en) * | 2021-02-01 | 2022-05-20 | 复旦大学 | Filter optimization design method based on deep learning algorithm |
| CN113080984A (en) * | 2021-03-25 | 2021-07-09 | 南京蝶谷健康科技有限公司 | Myocardial infarction identification and positioning method based on CNN and LSTM |
| CN113171104A (en) * | 2021-04-25 | 2021-07-27 | 安徽十锎信息科技有限公司 | Congestive heart failure automatic diagnosis method based on deep learning |
| CN113749666A (en) * | 2021-09-10 | 2021-12-07 | 郑州大学 | Myocardial infarction classification method based on fusion of ventricular regular features and XGboost |
| CN113749666B (en) * | 2021-09-10 | 2023-10-27 | 郑州大学 | Myocardial infarction classification method based on fusion of ventricular rule features and XGBoost |
| CN117281531A (en) * | 2023-11-27 | 2023-12-26 | 北京科技大学 | Psychological fatigue state identification method and system based on convolution long and short-time memory network |
| CN117281531B (en) * | 2023-11-27 | 2024-01-30 | 北京科技大学 | Psychological fatigue state identification method and system based on convolution long and short-time memory network |
| CN118177827A (en) * | 2024-04-22 | 2024-06-14 | 中国人民解放军南部战区总医院 | Myocardial infarction positioning method based on electrocardiograph vector diagram |
| CN118177827B (en) * | 2024-04-22 | 2024-10-18 | 中国人民解放军南部战区总医院 | Myocardial infarction positioning method based on electrocardiograph vector diagram |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109846471A (en) | A kind of myocardial infarction detection method based on BiGRU deep neural network | |
| Limam et al. | Atrial fibrillation detection and ECG classification based on convolutional recurrent neural network | |
| CN110619322A (en) | Multi-lead electrocardio abnormal signal identification method and system based on multi-flow convolution cyclic neural network | |
| CN111990989A (en) | Electrocardiosignal identification method based on generation countermeasure and convolution cyclic network | |
| CN106725428A (en) | A kind of electrocardiosignal sorting technique and device | |
| CN106108889A (en) | Electrocardiogram classification method based on degree of depth learning algorithm | |
| CN111297349A (en) | Machine learning-based heart rhythm classification system | |
| WO2021017313A1 (en) | Atrial fibrillation detection method and apparatus, computer device, and storage medium | |
| CN106214145A (en) | A kind of electrocardiogram classification method based on degree of depth learning algorithm | |
| CN111329445B (en) | Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network | |
| CN111887858B (en) | Ballistocardiogram signal heart rate estimation method based on cross-modal mapping | |
| CN107811649A (en) | A kind of more sorting techniques of heart sound based on depth convolutional neural networks | |
| ŞEN et al. | ECG arrhythmia classification by using convolutional neural network and spectrogram | |
| CN107239684A (en) | A kind of feature learning method and system for ECG identifications | |
| CN113057648A (en) | ECG signal classification method based on composite LSTM structure | |
| Chen et al. | Region aggregation network: improving convolutional neural network for ECG characteristic detection | |
| CN111368627A (en) | Heart sound classification method and system based on CNN combined with improved frequency wavelet slice transformation | |
| CN108309284A (en) | Electrocardiogram T wave end point detection method and device | |
| CN110313894A (en) | Arrhythmia cordis sorting algorithm based on convolutional neural networks | |
| CN110327055A (en) | A kind of classification method of the heart impact signal based on higher-order spectrum and convolutional neural networks | |
| CN108647584B (en) | Arrhythmia identification and classification method based on sparse representation and neural network | |
| CN109124620A (en) | A kind of atrial fibrillation detection method, device and equipment | |
| CN113499079A (en) | Atrial fibrillation detection method in electrocardiogram | |
| CN115363586A (en) | Psychological stress grade assessment system and method based on pulse wave signals | |
| CN113768514A (en) | Arrhythmia classification method based on convolutional neural network and gated cyclic unit |
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: 20190607 |
|
| RJ01 | Rejection of invention patent application after publication |