CN110464334A - A kind of electrocardiographic abnormality detection method based on composite depth learning network - Google Patents
A kind of electrocardiographic abnormality detection method based on composite depth learning network Download PDFInfo
- Publication number
- CN110464334A CN110464334A CN201910667783.1A CN201910667783A CN110464334A CN 110464334 A CN110464334 A CN 110464334A CN 201910667783 A CN201910667783 A CN 201910667783A CN 110464334 A CN110464334 A CN 110464334A
- Authority
- CN
- China
- Prior art keywords
- window
- abnormality detection
- detection method
- electrocardiographic abnormality
- layer
- 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
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- Surgery (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Cardiology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a kind of electrocardiographic abnormality detection methods based on composite depth learning network, comprising: (a) does global abnormal detection to the aligned region in 12 leads by multiple slide window implementations；(b) combine long Memory Neural Networks in short-term to the modeling of each lead and fusion feature vector；(c) eventually by activation primitive by its map classification result；(d) it will test result to compare and analyze with artificial calibration result.Can prevent over-fitting, be added random drop layer in operation, random drop layer will random drop unit lattice parameter by a certain percentage in the training process, the over-fitting that can be effectively prevent in iterative process.
Description
Technical field
The present invention relates to detection field, in particular to a kind of electrocardiographic abnormality detection side based on composite deep learning network
Method.
Background technique
In recent years, with the trend of population aging, chronic cardiovascular disease is in high-incidence situation.According to incompletely statistics, entirely
It is heart disease that the world, which is died in population there are about the cause of the death of one third, and U.S. number Zhan dead due to coronary heart disease every year is always dead
The 24.7% of number, about 540,000 people die of sudden cardiac death every year in China.It can be seen that heart disease just seriously threatens
The health of the mankind [1].Most of heart diseases can by electrocardiogram (Electrocardiograph, ECG) test come into
Row detection, therefore accurately ECG results are carried out abnormality detection and have a very important significance [2].
Electrocardiogram is that electrocardiosignal is recorded by placing an electrode within human body the preceding paragraph time, and electrocardiosignal reflects heart
The potential change [3] that electron ion is generated in transfer process inside and outside the cardiac muscle cell when work.Electrocardiogram is detection Cardiovascular abnormality
One of main foundation, the analysis detection of ECG signal has biggish diagnostic value [4] to doctor.One is subnormal
Electrocardio-activity rise from sinoatrial node, the signal of generation passes to ventricle by atrium, and has caused current potential change in conductive process,
To various characteristic waves [5] in electrocardiogram occur.Electrocardiogram is exactly to be made of the waveform of this series of periodic, each
Period corresponds to a heart activity, is the electrical activity situation for reflecting entire heart comprehensively, and clinic commonly uses 12 lead set adults
Table electrocardiogram, lead used are Wilson system [6].12 leads of electrocardiogram include 6 limb leads (I, II, III,
AVR, aVL, aVF) and 6 chest leads (V1~V6), limb leads include standard bipolar leads (I, II and III) and augmented-lead
(aVR, aVL and aVF) [7].When using electrocardiosignal Diagnosing Cardiac exception, by the practical electrocardiogram of patient and the heart of standard
Electrograph compares, so that it may find out the anomalous ecg of patient, including atrium early contracting, ventricle morning contracting and chamber, bundle-branch block
Deng in particular for the diagnosis of myocardial infarction, electrocardiogram can provide important diagnostic message [8]；And in coronary heart disease Acute myocardial
When infarct, by continuous electrocardiogram monitoring, the arrhythmia cordis [9] of threat to life can be found and handled in time.
However, human body is consumed in interior about 100,000 cardiac cycle of normal electrocardiosignal generated for 24 hours so that manually checking very much
When, it is uninteresting, and mistaken diagnosis and rate of missed diagnosis are higher.With the rise of wearable device, the product of function is automatically analyzed with electrocardiogram
Demand is more and more, promotes area of computer aided ecg analysis and is able to extensive and in-depth research [10].To electrocardiosignal into
The conventional method that row automatically analyzes mainly includes 3 Signal Pretreatment, feature extraction and classification steps.Pass through filtering etc. first
Signal processing method eliminates various noises and false signal, signal subsection is then carried out feature extraction, finally by all kinds of algorithms
Complete classification.
Currently, the method based on electrocardiosignal diagnosis of heart disease is roughly divided into two major classes: one kind is with clinical cardiac diagnosis
It learns to instruct, is judged by corresponding time limit threshold value or amplitude threshold, such as method of self-organizing feature map [11], automatic time
Return modelling [12], C means Method [13], multilayer perceptron method [14] etc..Another kind of is to be with the patient cases having confirmed that
Training set carries out the classification learning of all kinds of heart diseases by statistical learning and mode detection.As Fan Chengzhu [15] utilize convolution
Neural network (Convolutional Neural Network, CNN) carries out arrhythmia detection, Malhotra etc. [16] benefit
Arrhythmia detection classification is carried out with long Memory Neural Networks (Long Short-Term Memory, LSTM) in short-term,
Chauhan etc. [17] devises depth LSTM network and carries out detection classification, and the above research achieves certain effect.
As deep learning algorithm becomes the further investigation and research of research hotspot and people to artificial intelligence, depth is used
The detection classification that degree learning method carries out electrocardiosignal has the advantage that first is that only carry out less pretreatment to signal, and not
It needs pre- first pass through to calculate acquirement characteristic signal, can directly carry out calculating classification；If second is that there is suitable learning training collection,
Its type classified can constantly extend.But every kind of algorithm have the characteristics that it is respective, it is right if CNN algorithm is chiefly used in image detection
The waveform shape of electrocardiogram is more sensitive；LSTM algorithm is chiefly used in the sequential relationship detection to data, to each electrocardiographic wave
Relative positional relationship is more sensitive, but the details modeling ability of LSTM spatially is weaker than CNN, therefore only uses LSTM and be difficult to
Local fineization analyzes abnormal electrocardiogram signal, it is made to be easy to ignore local electrocardiosignal exception so as to cause erroneous judgement.
Summary of the invention
The electrocardiographic abnormality detection based on composite deep learning network that the technical problem to be solved in the present invention is to provide a kind of
Method can prevent over-fitting, and random drop layer is added in operation, and random drop layer will be in the training process by certain ratio
Example random drop unit lattice parameter, the over-fitting that can be effectively prevent in iterative process.
In order to solve the above-mentioned technical problem, the technical solution of the present invention is as follows:
A kind of electrocardiographic abnormality detection method based on composite deep learning network characterized by comprising
(a) global abnormal detection is done to the aligned region in 12 leads by multiple slide window implementations；
(b) combine long Memory Neural Networks in short-term to the modeling of each lead and fusion feature vector；
(c) eventually by activation primitive by its map classification result；
(d) it will test result to compare and analyze with artificial calibration result.
Preferably, the length in short-term Memory Neural Networks by it is multiple it is unit cascaded form, each unit include one forget door,
One input gate and an out gate.
Preferably, the sliding window shares three layers, and Memory Neural Networks share one layer to the length in short-term, wherein the cunning
Dynamic each layer of window extracts ecg characteristics using different windows.
Preferably, the sliding window triple-window mouth size is respectively: 50,30,10.
Preferably, an equal amount of window is taken to 12 lead signals in the sliding window simultaneously, carries out spatial alignment
Feature extraction.
Preferably, sliding is carried out to all sampled points on 12 lead electrocardiosignals with an equal amount of window and calculates spy
Sign, and it is directed to each window, the decision value after activation primitive is exported, for whether adjudicating the electrocardiosignal in the window
There are exceptions.
Preferably, among step (b) further include:
(b1) spy of the feature vector of length Memory Neural Networks layer output in short-term and three layers of sliding window output
Vector splicing fusion is levied, converts one-dimensional data for multidimensional data；
(b2) classification output is completed using activation primitive；
(b3) random drop layer is added in operation, the random drop layer will be random by a certain percentage in the training process
Abandon unit lattice parameter.
By adopting the above technical scheme, the beneficial effects of the present invention are:
The first, finally the feature vector that LSTM layers export is merged with the feature vector splicing that three-layer coil lamination exports, is made
With Flatten layers, one-dimensional data is converted by multidimensional data；Use Sigmoid function as sharp in full articulamentum Dense
Function living completes classification output.To prevent over-fitting, random drop layer is added in operation, random drop layer will train
Random drop unit lattice parameter by a certain percentage in journey, the over-fitting that can be effectively prevent in iterative process；
The second, it especially suitable for being spaced in processing and forecasting sequence and postponing relatively long critical event, is turning at present
It translates, image analysis, speech detection, the fields such as documentation summary all achieve better effects.Since electrocardiosignal is continuous signal,
And there are very strong timing dependences, therefore electrocardiosignal feature can sufficiently be captured from timing by LSTM modeling.
Detailed description of the invention
Fig. 1 is LSTM cellular construction schematic diagram of the invention；
Fig. 2 is depth Hybrid Learning network structure of the invention；
Fig. 3 is 12 lead electrocardiogram schematic diagram datas of the invention.
Specific embodiment
Specific embodiments of the present invention will be further explained with reference to the accompanying drawing.It should be noted that for
The explanation of these embodiments is used to help understand the present invention, but and does not constitute a limitation of the invention.In addition, being retouched below
Technical characteristic involved in each embodiment of the present invention stated as long as they do not conflict with each other can be mutual group
It closes.
The present invention provides a kind of electrocardiographic abnormality detection methods based on composite deep learning network, wherein described compound
Deep learning network includes a kind of sliding window CNN algorithm and long Memory Neural Networks LSTM algorithm in short-term, and wherein CNN is logical
After crossing convolution kernel and detected data progress convolutional calculation, is iterated and extract correlated characteristic point, then pass through neural network meter
Classification is realized after calculation, is widely used in image recognition at present and brain wave identifies field.For 2-D data, convolution expression
Formula such as following formula:
By mathematic(al) representation it can be seen that convolution operation is substantially exactly the 2-D data D inputted, with each pixel of convolution kernel K
Summation obtains convolution results R after multiplication.Convolution algorithm is by being weighted expression to local feature, thus to different characteristic
Different response ratios is played, to achieve the purpose that feature extraction, the different characteristic of input layer is extracted with multiple convolution kernels.
Herein for electrocardiogram, using two-dimensional convolution, it can not only play the role of feature extraction in sample direction, it can right
Electrocardiosignal in some lead carries out feature extraction, but also can play space on the space constructed by multi-lead
The effect that fusion is extracted with alignment feature, to play the value of the comprehensive judgement of 12 leads.
LSTM [19] is a kind of Recognition with Recurrent Neural Network (Recurrent Neural Networks, RNN) of specific type,
It can learn the information relied on for a long time, unit cascaded be formed by multiple altogether.As shown in Figure 1, each unit mainly passes through 3 doors
Structure realizes protection, storage and control unit state.
LSTM network base units structure is as shown in Figure 2.The first step is to forget door, by a function, i.e. activation primitive
Sigmoid is constituted, and exports the numerical value between 0 to 1, shows the program of forgetting to upper level network cell, and 1 indicates " complete
All risk insurance stays ", 0 indicates " giving up completely ", it can determine timing information length to be treated；Second step is input gate, activation
One real number input is mapped in [- 1,1] range by function Tanh, and input gate is learned by Sigmoid function and Tanh function
It is to be updated to practise which information；It is finally out gate, learns which stored information is able to use, and exports to next stage
Unit.LSTM is a kind of time recurrent neural network, especially suitable for being spaced and postponing relatively long in processing and forecasting sequence
Critical event, all achieve better effects in fields such as translation, image analysis, speech detection, documentation summaries at present.Due to
Electrocardiosignal is continuous signal, and there are very strong timing dependences, therefore can sufficiently be captured from timing by LSTM modeling
Electrocardiosignal feature.
It is diagnosed and is required according to Electrocardiography, the present invention is devised to be answered by what two-dimentional sliding window CNN and LSTM was combined
Close network.CNN can be by two-dimensional convolution core iterative learning, effectively extraction signal characteristic, but models to global sequential relationship
Ability is relatively weak.In electrocardio map interpretation, main foundation: first is that effectively to detect form, the direction etc. of each lead waveform
Feature, second is that judge the relative position and distance between each lead waveform.LSTM due to long short-term memory characteristic, to spy
Position and distance relation between sign is especially sensitive, and two-dimentional CNN detects opposite with the upper ability of part fining detection in spatial alignment
It is relatively strong, therefore the calculated feature vector of the two institute is after splicing fusion, can effectively meet two judgements of electrocardiogram according to
According to.
Network is divided into four layers, and first layer is LSTM layers, and excess-three layer is CNN layers: using different windows for each layer in CNN
W extracts ecg characteristics, and triple-window mouth size is respectively: 50,30,10, the design considerations of window size is adopted in 500Hz
Under sample frequency, window size required for an abnormal electrocardiogram signal is differentiated.Then due to the pond of network, window
It also needs to scale, to reduce error.
Wherein two dimensional character extraction CNN selects Tanh as activation primitive, passes through the spy of two-dimentional CNN iterative extraction electrocardiogram
Sign, and by the step-length of control convolution kernel to play the effect in dynamic pond, reduction while remaining validity feature point
Calculation amount, improves the abstracting power of network.
Two-dimentional sliding window CNN selects Sigmoid as activation primitive, and convolution nuclear shape is that [12, W] lead 12
Connection signal takes the window of onesize W simultaneously, carries out spatial alignment feature extraction.It is [12, W] that specific implementation, which is with size,
Sliding window carries out sliding to all sampled points on 12 lead electrocardiosignals and calculates feature, and is directed to each window, defeated
The value that decision value after Sigmoid out, i.e. each window obtain one [0,1], for adjudicating the letter of the electrocardio in the window
Number with the presence or absence of abnormal.
The feature vector that LSTM layers export is merged with the feature vector splicing that three-layer coil lamination exports finally, is used
Flatten layers, one-dimensional data is converted by multidimensional data；Use Sigmoid function as activation in full articulamentum Dense
Function completes classification output.To prevent over-fitting, random drop layer is added in operation, random drop layer will be in training process
In random drop unit lattice parameter by a certain percentage, the over-fitting that can be effectively prevent in iterative process.
Specifically, according to above content, the present invention provides a kind of electrocardiographic abnormality detection methods, comprising:
(a) global abnormal detection is done to the aligned region in 12 leads by multiple slide window implementations；
(b) combine long Memory Neural Networks in short-term to the modeling of each lead and fusion feature vector；
(c) eventually by activation primitive by its map classification result；
(d) it will test result to compare and analyze with artificial calibration result.
In the above-mentioned methods, the length in short-term Memory Neural Networks by it is multiple it is unit cascaded form, each unit include one
Forget door, an input gate and an out gate.
In the above-mentioned methods, the sliding window shares three layers, and Memory Neural Networks share one layer to the length in short-term, wherein
Each layer of the sliding window extracts ecg characteristics using different windows.
In the above-mentioned methods, the sliding window triple-window mouth size is respectively: 50,30,10.
In the above-mentioned methods, an equal amount of window is taken to 12 lead signals in the sliding window simultaneously, is carried out empty
Between alignment feature extract.Sliding calculating is carried out to all sampled points on 12 lead electrocardiosignals with an equal amount of window
Feature, and it is directed to each window, the decision value after activation primitive is exported, is for adjudicating the electrocardiosignal in the window
It is no to there is exception.
Among step (b) further include:
(b1) spy of the feature vector of length Memory Neural Networks layer output in short-term and three layers of sliding window output
Vector splicing fusion is levied, converts one-dimensional data for multidimensional data；
(b2) classification output is completed using activation primitive；
(b3) random drop layer is added in operation, the random drop layer will be random by a certain percentage in the training process
Abandon unit lattice parameter.
According to the above method, specific experimental method is as follows:
12 lead electrocardiogram data used in testing are by the diagnostic result of cardiovascular doctor as goldstandard, entire data
For collection altogether comprising 600 records, every record all includes 12 leads, sample frequency 500Hz.One of subject's
The effect of visualization of 12 lead electrocardiogram data is as shown in Figure 3.
The method for taking interval to extract data chooses 80% data for training, and 20% data are for testing.In net
In network design, we develop two kinds of network plans for sliding conventional part: fricton-tight window and having sliding window, be used for
Explore effect of the two-dimentional sliding window in the detection of multi-lead spatial alignment.In deep learning usually using accuracy rate (P),
Recall rate (R) and F1 value evaluate the test result of neural network as evaluation index.P, the calculation formula of R, F1 is such as
Under:
Wherein, TpFor true positives, FpFor false positive, FnFor false negative.50 electrocardiosignals of selection are a collection of when test, 300
After secondary iteration, test result is shown in Table 1.
1 fatigue monitoring glasses monitoring result of table
| P | R | F1 | |
| Fricton-tight window | 77.26% | 74.12% | 75.65% |
| There is sliding window | 92.91% | 91.33% | 92.11% |
This algorithm has preferable effect in the abnormality detection of electrocardiogram, and F1 value has reached 92.11%, demonstrates this
The validity of algorithm.Current electrocardiogram (ECG) data is analyzed to be difficult to obtain higher F1 value to be the electrocardiosignal wave because of (1) 12 lead
Shape is close, is not easily distinguishable, and exists simultaneously larger interference, brings certain difficulty to classification.(2) data volume in training set is also
It needs to continue to expand, one of the characteristics of this is also deep learning algorithm, the raising of accuracy rate and the quantity of learning sample and matter
It measures related.
In conjunction with attached drawing, the embodiments of the present invention are described in detail above, but the present invention is not limited to described implementations
Mode.For a person skilled in the art, in the case where not departing from the principle of the invention and spirit, to these embodiment party
Formula carries out a variety of change, modification, replacement and modification, still falls in protection scope of the present invention.
Claims (7)
1. a kind of electrocardiographic abnormality detection method based on composite deep learning network characterized by comprising
(a) global abnormal detection is done to the aligned region in 12 leads by multiple slide window implementations；
(b) combine long Memory Neural Networks in short-term to the modeling of each lead and fusion feature vector；
(c) eventually by activation primitive by its map classification result；
(d) it will test result to compare and analyze with artificial calibration result.
2. electrocardiographic abnormality detection method according to claim 1, which is characterized in that length Memory Neural Networks in short-term
By it is multiple it is unit cascaded form, each unit includes one forgetting door, an input gate and an out gate.
3. electrocardiographic abnormality detection method according to claim 2, which is characterized in that the sliding window shares three layers,
Memory Neural Networks share one layer to the length in short-term, wherein each layer of the sliding window is extracted electrocardio using different windows
Feature.
4. electrocardiographic abnormality detection method according to claim 3, which is characterized in that the sliding window triple-window mouth
Size is respectively: 50,30,10.
5. electrocardiographic abnormality detection method according to claim 1, which is characterized in that led in the sliding window to 12
Connection signal takes an equal amount of window simultaneously, carries out spatial alignment feature extraction.
6. electrocardiographic abnormality detection method according to claim 5, which is characterized in that with an equal amount of window at 12
Sliding is carried out to all sampled points on lead electrocardiosignal and calculates feature, and is directed to each window, after output activation primitive
Decision value, for adjudicating the electrocardiosignal in the window with the presence or absence of abnormal.
7. electrocardiographic abnormality detection method according to claim 5, which is characterized in that among step (b) further include:
(b1) feature of the feature vector of length Memory Neural Networks layer output in short-term and three layers of sliding windows output to
Amount splicing fusion, converts one-dimensional data for multidimensional data；
(b2) classification output is completed using activation primitive；
(b3) it is added random drop layer in operation, the random drop layer will random drop by a certain percentage in the training process
Unit lattice parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910667783.1A CN110464334A (en) | 2019-07-23 | 2019-07-23 | A kind of electrocardiographic abnormality detection method based on composite depth learning network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910667783.1A CN110464334A (en) | 2019-07-23 | 2019-07-23 | A kind of electrocardiographic abnormality detection method based on composite depth learning network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110464334A true CN110464334A (en) | 2019-11-19 |
Family
ID=68509773
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910667783.1A Pending CN110464334A (en) | 2019-07-23 | 2019-07-23 | A kind of electrocardiographic abnormality detection method based on composite depth learning network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110464334A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111150393A (en) * | 2020-02-19 | 2020-05-15 | 杭州电子科技大学 | Electroencephalogram epilepsy spike discharge joint detection method based on LSTM multichannel |
| CN111772628A (en) * | 2020-07-16 | 2020-10-16 | 华中科技大学 | Electrocardiosignal atrial fibrillation automatic detection system based on deep learning |
| CN111857352A (en) * | 2020-07-30 | 2020-10-30 | 太原科技大学 | Gesture recognition method based on imagination type brain-computer interface |
| 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 |
| CN113288163A (en) * | 2021-06-04 | 2021-08-24 | 浙江理工大学 | Multi-feature fusion electrocardiosignal classification model modeling method based on attention mechanism |
| CN113419902A (en) * | 2021-06-29 | 2021-09-21 | 上海大学 | Multichannel electroencephalogram signal correlation analysis and data recovery method based on long-time and short-time memory network |
| TWI752798B (en) * | 2020-01-17 | 2022-01-11 | 長佳智能股份有限公司 | A method for establishing a heart rhythm classification model |
| CN114119968A (en) * | 2021-12-16 | 2022-03-01 | 安徽心之声医疗科技有限公司 | Paper electrocardiogram signal region extraction method and system based on deep neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170032221A1 (en) * | 2015-07-29 | 2017-02-02 | Htc Corporation | Method, electronic apparatus, and computer readable medium of constructing classifier for disease detection |
| CN106901723A (en) * | 2017-04-20 | 2017-06-30 | 济南浪潮高新科技投资发展有限公司 | A kind of electrocardiographic abnormality automatic diagnosis method |
| CN108030488A (en) * | 2017-11-30 | 2018-05-15 | 北京医拍智能科技有限公司 | The detecting system of arrhythmia cordis based on convolutional neural networks |
| CN108836307A (en) * | 2018-05-14 | 2018-11-20 | 广东工业大学 | A kind of intelligent ECG detection device, equipment and mobile terminal |
-
2019
- 2019-07-23 CN CN201910667783.1A patent/CN110464334A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170032221A1 (en) * | 2015-07-29 | 2017-02-02 | Htc Corporation | Method, electronic apparatus, and computer readable medium of constructing classifier for disease detection |
| CN106901723A (en) * | 2017-04-20 | 2017-06-30 | 济南浪潮高新科技投资发展有限公司 | A kind of electrocardiographic abnormality automatic diagnosis method |
| CN108030488A (en) * | 2017-11-30 | 2018-05-15 | 北京医拍智能科技有限公司 | The detecting system of arrhythmia cordis based on convolutional neural networks |
| CN108836307A (en) * | 2018-05-14 | 2018-11-20 | 广东工业大学 | A kind of intelligent ECG detection device, equipment and mobile terminal |
Non-Patent Citations (1)
| Title |
|---|
| 张坤 等: "基于深度学习的心律失常检测算法研究", 《医疗卫生装备》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI752798B (en) * | 2020-01-17 | 2022-01-11 | 長佳智能股份有限公司 | A method for establishing a heart rhythm classification model |
| CN111150393A (en) * | 2020-02-19 | 2020-05-15 | 杭州电子科技大学 | Electroencephalogram epilepsy spike discharge joint detection method based on LSTM multichannel |
| CN111150393B (en) * | 2020-02-19 | 2023-03-28 | 杭州电子科技大学 | Electroencephalogram epilepsy spike discharge joint detection method based on LSTM multichannel |
| CN111772628A (en) * | 2020-07-16 | 2020-10-16 | 华中科技大学 | Electrocardiosignal atrial fibrillation automatic detection system based on deep learning |
| CN111857352A (en) * | 2020-07-30 | 2020-10-30 | 太原科技大学 | Gesture recognition method based on imagination type brain-computer interface |
| CN111857352B (en) * | 2020-07-30 | 2022-04-08 | 太原科技大学 | Gesture recognition method based on imagination type brain-computer interface |
| 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 |
| CN113288163A (en) * | 2021-06-04 | 2021-08-24 | 浙江理工大学 | Multi-feature fusion electrocardiosignal classification model modeling method based on attention mechanism |
| CN113419902A (en) * | 2021-06-29 | 2021-09-21 | 上海大学 | Multichannel electroencephalogram signal correlation analysis and data recovery method based on long-time and short-time memory network |
| CN114119968A (en) * | 2021-12-16 | 2022-03-01 | 安徽心之声医疗科技有限公司 | Paper electrocardiogram signal region extraction method and system based on deep neural network |
| CN114119968B (en) * | 2021-12-16 | 2022-05-24 | 安徽心之声医疗科技有限公司 | Paper electrocardiogram signal region extraction method and system based on deep neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110464334A (en) | A kind of electrocardiographic abnormality detection method based on composite depth learning network | |
| US11344243B2 (en) | Artificial intelligence self-learning-based static electrocardiography analysis method and apparatus | |
| CN111449645B (en) | Intelligent classification and identification method for electrocardiogram and heartbeat | |
| US11617528B2 (en) | Systems and methods for reduced lead electrocardiogram diagnosis using deep neural networks and rule-based systems | |
| CN107837082A (en) | Electrocardiogram automatic analysis method and device based on artificial intelligence self study | |
| CN110974214A (en) | Automatic electrocardiogram classification method, system and equipment based on deep learning | |
| CN111297349A (en) | Machine learning-based heart rhythm classification system | |
| CN107137072A (en) | A kind of ventricular ectopic beating detection method based on 1D convolutional neural networks | |
| Jin et al. | A novel interpretable method based on dual-level attentional deep neural network for actual multilabel arrhythmia detection | |
| Xiang et al. | ECG-based heartbeat classification using two-level convolutional neural network and RR interval difference | |
| CN109222963A (en) | A kind of anomalous ecg method for identifying and classifying based on convolutional neural networks | |
| CN110148466A (en) | A kind of heart impact signal atrial fibrillation computer aided diagnosing method based on transfer learning | |
| CN115281688B (en) | Multi-mode deep learning-based cardiac hypertrophy multi-label detection system | |
| US11571161B2 (en) | Systems and methods for electrocardiogram diagnosis using deep neural networks and rule-based systems | |
| CN104462744A (en) | Data quality control method suitable for cardiovascular remote monitoring system | |
| Lu et al. | An arrhythmia classification algorithm using C-LSTM in physiological parameters monitoring system under internet of health things environment | |
| CN111000551A (en) | Heart disease risk diagnosis method based on deep convolutional neural network model | |
| CN113349793A (en) | Arrhythmia classification method based on layered attention model | |
| CN115177262A (en) | Heart sound and electrocardiogram combined diagnosis device and system based on deep learning | |
| Hassan et al. | Performance comparison of CNN and LSTM algorithms for arrhythmia classification | |
| Wang et al. | Multiscale residual network based on channel spatial attention mechanism for multilabel ECG classification | |
| Zhao et al. | PVC recognition for wearable ECGs using modified frequency slice wavelet transform and convolutional neural network | |
| Irfan et al. | Deep learning based classification for healthcare data analysis system | |
| Hui et al. | Wearable devices acquired ECG signals detection method using 1D convolutional neural network | |
| CN112022149B (en) | Atrial fibrillation detection method based on electrocardiosignals |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191119 |