CN110327033A - A kind of screening method of the myocardial infarction electrocardiogram based on deep neural network - Google Patents
A kind of screening method of the myocardial infarction electrocardiogram based on deep neural network Download PDFInfo
- Publication number
- CN110327033A CN110327033A CN201910268945.4A CN201910268945A CN110327033A CN 110327033 A CN110327033 A CN 110327033A CN 201910268945 A CN201910268945 A CN 201910268945A CN 110327033 A CN110327033 A CN 110327033A
- Authority
- CN
- China
- Prior art keywords
- image
- electrocardiogram
- neural network
- myocardial infarction
- deep neural
- 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.)
- Granted
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
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—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)
- Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Public Health (AREA)
- Surgery (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Cardiology (AREA)
- Power Engineering (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
A kind of screening method of the myocardial infarction electrocardiogram based on deep neural network, comprising the following steps: step 1 pre-processes original electrocardiographicdigital image, removes grid frame；Step 2 carries out median filter process to the image in step 1, leaves spiced salt noise after eliminating removal grid frame；Step 3 enhances contrast by histogram equalization method, and size is unified for 224*224 by scaling electrocardiogram, and is [0,1] by data normalization；Step 4 is constructed deep neural network, and is trained using ECG training set.Obtain that two Classification Neurals for whether suffering from myocardial infarction can be diagnosed；Step 5 carries out auxiliary screening to electrocardiogram myocardial infarction using the deep neural network that training obtains.Whether the present invention can be myocardial infarction according to electrocardiogram auxiliary judgment.
Description
Technical field
The present invention relates to medical image analysis field and machine learning field, in particular to a kind of electrocardiogram myocardial infarction sieve
Checking method belongs to the medical image analysis field based on deep learning.
Background technique
Myocardial infarction (Myocardial Infarction, MI) is a kind of relatively conventional heart disease.Early detection pair
It is most important that acute myocardial infarction AMI is effectively treated in percutaneous coronary intervention (pci) (PCI) or coronary artery bypass surgery.The heart
Flesh infarct is usually to be diagnosed by clinical manifestation, laboratory result and electrocardiogram.Electrocardiogram is specific by record body surface
Potential that position changes over time and generate, potential represents the electrical activity of heart.Ecg curve deviates common shape can be with
Show myocardial infarction and many other hearts and non-cardiac disease.Electrocardiogram (Electrocardiograph, ECG) is a kind of
Popular diagnostic tool because they be it is non-invasive, production cost is low, but provides very high diagnostic value.
In the myocardial infarction deterministic process based on electrocardiogram, doctor requires a great deal of time carefully diagnosis each
The electrocardiogram that lead generates, this has higher requirement to the energy of doctor and the expertise of itself.And electrocardiogram medium wave band
Feature difference is smaller, this be easy to cause certain probability fail to pinpoint a disease in diagnosis and mistaken diagnosis.And it is engaged in some areas since condition limits
Doctor based on the electrocardiographic diagnosis sufferer state of an illness is limited.In recent years, traditional computer aided diagnosis method mostly uses Wavelet transformation
It is diagnosed with k near neighbor method, but these methods are higher to electrocardiogram quality requirement, the quality of sample is easy to influence the result of model.
There is the research to a variety of arrhythmia cordis of some maturations at present, but it is specially seldom to the research of myocardial infarction disease.In addition,
Myocardial infarction Symptoms in ECG are different, such as ST sections of elevations and Non-ST Elevation Acute type.For the more symptom cardiac muscles stalks of identification
The electrocardiogram extremely showed, this is to the more demanding of algorithm.These unfavorable factors to diagnose difficulty height, it is difficult to use statistical method
Validity feature is obtained, in addition there are the noises such as a large amount of grid frames in image, so that the diagnosis difficulty of conventional method is big.
Summary of the invention
Existing ECG myocardial infarction diagnosis mode difficulty is big, inefficiency, the lower deficiency of precision in order to overcome, the present invention
A kind of fast, high-efficient, the higher heart infarction electrocardiogram based on deep neural network of precision the auxiliary screening method of speed is proposed,
It realizes and electrocardiogram is automatically analyzed, auxiliary judgment effectively can be carried out to electrocardiogram myocardial infarction disease condition.
In order to solve its technical problem the technical scheme adopted by the invention is that:
A kind of screening method of the heart infarction electrocardiogram based on deep neural network, comprising the following steps:
Step 1 pre-processes original electrocardiographicdigital image, removes grid frame；
Step 2 carries out median filter process to the image in step 1, leaves spiced salt noise after eliminating removal grid frame；
Step 3 enhances contrast by histogram equalization method, scales electrocardiogram for size and is unified for 224*224,
And by data normalization be [0,1]；
Step 4 is constructed deep neural network, and is trained using ECG training set.It obtains being capable of deciding whether to suffer from
Two Classification Neurals of myocardial infarction；
Step 5 carries out auxiliary screening to electrocardiogram myocardial infarction using the deep neural network that training obtains.
Further, in the step 1, to the pretreated process of electrocardiogram are as follows: grayscale image is converted the image into, to gray scale
Figure carries out step-by-step inversion operation, then carries out adaptive threshold, i.e., image different zones use different threshold values.Obtain binaryzation mark
Sign image；Closing operation of mathematical morphology operation is carried out to it, is reduced independent noise, is filled the hole of connected region；It is carried out with original image
Add operation obtains the image of removal grid frame.
Preferably, in the step 1, in adaptive threshold use adaptive Gauss method, adaptive threshold T (x,
Y) calculating process are as follows:
The distance of step 1.1 (x, y) surrounding pixel to its central point obtains corresponding weight by Gaussian function；
Step 1.2, which calculates around each pixel the weighted mean of (15 × 15) size block of pixels and subtracts constant 10, obtains threshold
Value T (x, y).
Further preferably, in the step 1, cv2.MORPH_RECT, structure are set as the structural element of closed operation
The size of element is respectively set in horizontal line and vertical line identification are as follows:
H in formula, W respectively indicate the height and width of gray level image, and scale indicates the zoom factor of structural element.
It is furthermore preferred that in the step 1, the process of grid frame removal:
Setp1.1 is in the image after adaptive threshold, by closed operation and structural element cv2.MORPH_RECT and greatly
It is small to beIdentify horizontal line；
Setp1.2 is in the image after adaptive threshold, by closed operation and structural element cv2.MORPH_RECT and greatly
It is small to beIdentify vertical line；
The image that setp1.3 is obtained by setp1.1 and setp1.2, is merged, i.e. lattice portion in extraction original image
Point；
The setp1.3 image obtained and original image are carried out add operation by setp1.4.
In the step 4, the process of convolutional neural networks is constructed are as follows:
Step 4.1 inputs the electrocardiogram image that one group of size is 224*224*3；
Step 4.2 first passes through the convolution operation of 7*7 size, then carries out batchnormalization, and Relu activates letter
Number operation；
Step 4.3 is operated by MaxPooling, extracts main feature；
The feature that step 4.4 is extracted passes through residual error convolution module, which includes the convolution operation and batch of 2 groups of 3*3
normalization；
Step 4.5: step 3.3 is repeated three times, using the average pond of 7*7 size；
Step 4.6: by features described above by full articulamentum FullyConnectedLayer and Softmax function, finally
To exporting two classification results.
In the step 4, deep neural network framework is by 1 convolutional layer, 1 maximum pond layer, 4 residual error convolution moulds
Block (Block module), 1 average pond layer, 1 full articulamentum form, and are standardized behaviour to feature after each convolutional layer
Make, improves training speed, and pass through ReLU activation primitive, improve the non-linear expression of network；Packet in each residual error convolution module
Containing two convolutional layers and a quick branch, quick branch starting point is input, and terminal is the add operation after second convolutional layer, is made
The feature progress numerical value that obtaining input feature vector can directly extract with second convolutional layer be added；Network is only needed to calculate and be compared in this way
The residual error of original input, reduces trained difficulty；The Output Size of the last one full articulamentum is 2, two points of corresponding screening results
Class, two classification is myocardial infarction electrocardiogram or non-salary motivation electrocardiogram.
The present invention is based on the auxiliary screenings of the heart infarction electrocardiogram of deep neural network, extract ECG using convolutional neural networks
Characteristics of image, to realize the judgement to myocardial infarction electrocardiogram.Compared with the conventional method, the beneficial effect is that:
1. analyzing by convolutional neural networks electrocardiogram, from dynamic auxiliary screening heart infarction electrocardiogram, tradition side is compared
Method is high-efficient, and speed is fast.
2. method introduces the cancellation of grid frame in the pretreatment of electrocardiogram, the noise in image is eliminated.
3. extracting the validity feature of image by introducing residual error convolution module structure.The presence of the quick branch of residual error simultaneously
Reduce the training difficulty of network.
Detailed description of the invention
The identification process figure of heart infarction electrocardiogram of the Fig. 1 based on deep neural network.
Fig. 2 is used for the neural network structure schematic diagram of heart infarction electrocardiographic diagnosis.
Residual error modular structure schematic diagram in Fig. 3 neural network.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
Referring to Fig.1~Fig. 3, a kind of screening method of the heart infarction electrocardiogram based on deep neural network, comprising the following steps:
Step 1 pre-processes original electrocardiographicdigital image, removes grid frame；
Step 2 carries out median filter process to the image in step 1, leaves spiced salt noise after eliminating removal grid frame；
Step 3 enhances contrast by histogram equalization method, scales electrocardiogram for size and is unified for 224*224,
And by data normalization be [0,1]；
Step 4 is constructed deep neural network, and is trained using ECG training set.It obtains to diagnose and whether suffer from
Two Classification Neurals of myocardial infarction；
Step 5 carries out auxiliary diagnosis to electrocardiogram myocardial infarction using the deep neural network that training obtains.
Further, in the step 1, to the pretreated process of electrocardiogram are as follows: grayscale image is converted the image into, to gray scale
Figure carries out step-by-step inversion operation, then carries out adaptive threshold, i.e., image different zones use different threshold values.Obtain binaryzation mark
Sign image.Closing operation of mathematical morphology operation is carried out to it, is reduced independent noise, is filled the hole of connected region.It is carried out with original image
Add operation obtains the image of removal grid frame.
Preferably, in the step 1, in adaptive threshold use adaptive Gauss method, adaptive threshold T (x,
Y) calculating process are as follows:
The distance of step 1.1 (x, y) surrounding pixel to its central point obtains corresponding weight by Gaussian function；
Step 1.2, which calculates around each pixel the weighted mean of (15 × 15) size block of pixels and subtracts constant 10, obtains threshold
Value T (x, y).
Further preferably, in prime number step 1, cv2.MORPH_RECT, structure are set as the structural element of closed operation
The size of element is respectively set in horizontal line and vertical line identification are as follows:
H in formula, W respectively indicate the height and width of gray level image, and scale indicates the zoom factor of structural element.
It is furthermore preferred that in the step 1, the process of grid frame removal:
Step1.1 is in the image after adaptive threshold, by closed operation and structural element cv2.MORPH_RECT and greatly
It is small to beIdentify horizontal line；
Step1.2 is in the image after adaptive threshold, by closed operation and structural element cv2.MORPH_RECT and greatly
It is small to beIdentify vertical line；
The image that step1.3 is obtained by step1.1 and step1.2, is merged, i.e. lattice portion in extraction original image
Point；
The step1.3 image obtained and original image are carried out add operation by step1.4.
In the step 4, the process of convolutional neural networks is constructed are as follows:
Step 4.1 inputs the electrocardiogram image that one group of size is 224*224*3；
Step 4.2 first passes through the convolution operation of 7*7 size, then carries out batchnormalization, and Relu activates letter
Number operation；
Step 4.3 is operated by MaxPooling, extracts main feature；
The feature that step 4.4 is extracted passes through residual error convolution module, which includes the convolution operation and batch of 2 groups of 3*3
normalization；
Step 4.5 repeats step 3.3 three times, using the average pond of 7*7 size；
Step 4.6 by full articulamentum FullyConnectedLayer and Softmax function, finally obtains features described above
To exporting two classification results.
In the step 4, deep neural network framework is mainly by the network architecture mainly by 1 convolutional layer, 1 maximum pond
Change layer, 4 residual error convolution modules (Block module), 1 average pond layer, 1 full articulamentum composition.After each convolutional layer
Operation is standardized to feature, training speed is improved, and pass through ReLU activation primitive, improves the non-linear expression of network.Such as
It include two convolutional layers, a quick branch in each residual error convolution module shown in Fig. 3.Quick branch starting point is input, terminal
For the add operation after second convolutional layer, allows input feature vector directly and the feature of second convolutional layer extraction carries out numerical value
It is added.Network only needs to calculate the residual error compared to former input in this way, reduces trained difficulty.The output of the last one full articulamentum
Having a size of 2, the two of corresponding screening results classify (being myocardial infarction electrocardiogram or non-salary motivation electrocardiogram).
Example: electrocardiogram used in present case totally 2 class, i.e. heart infarction electrocardiogram or non-heart infarction electrocardiogram.Electrocardiogram is shared
2282 samples, heart infarction electrocardiogram and non-heart infarction electrocardiogram same number.Randomly select 913 samples respectively from positive negative sample
As training set, 114 samples are as verifying collection, and 114 samples are as test set.Lower mask body introduces electrocardiogram removal grid
Frame, the training of model and test process.
Step 1, electrocardiogram remove grid frame.
Step 1.1 converts the image into grayscale image, carries out step-by-step inversion operation to grayscale image, then carry out adaptive threshold；
Step 1.2 carries out closing operation of mathematical morphology operation to the image that step 1.1 obtains；
Step 1.3 carries out add operation with original image to the image that step 1.2 obtains, and obtains the image of removal grid frame；
Step 2, the building and training of neural network, specific structure are as shown in Figure 2.
Step 2.1 network architecture is mainly by 1 convolutional layer, 1 maximum pond layer, 4 residual error convolution module (Block moulds
Block), 1 average pond layer, 1 full articulamentum composition.
The convolution kernel size of step 2.2 first layer convolutional layer is 7*7, sliding step 2, padding 3.Residual error mould
Convolution kernel size is all 3*3 in block, and in addition to the sliding step of first residual error module is 1, other are all 2, and residual error module
Between be connected convolutional layer convolution kernel size be 1*1.Convolution nucleus number becomes more with residual error module is entered, and respectively 64,128,
256,512.Batch normalizing operation is carried out to feature after each convolutional layer of output, improves training speed, and swash by ReLU
Function living, promotes the non-linear expression of network.Maximum pond layer after convolutional layer will reduce characteristic pattern size.
All parameters weightings are initialized as random orthogonal matrix initialisation, weight regularization mode in step 2.3 convolutional layer
For L2 canonical, bias is initialized as 0.In full articulamentum, weights initialisation is random normal distribution, and weight regularization mode is
L2 canonical, bias are initialized as 0.
Step 2.4 builds network using Pytorch frame.Model is by the way of batch training.Training set generator and
The sample number batch size of the verifying collection each batch of generator is 16, the 40 data conducts of every return of training set generator
One round (epoch), after the completion of a round training, generator can return 5 times and calculate verifying collection loss, and loss function is
Cross entropy loss function.Model optimizer is Adam, parameter lr=0.001, betas=(0.9,0.999), eps=1e-
08, weight_decay=0, amsgrad=False.Model maximum training round is 60, is stopped after verifying and training loss convergence
It only trains, and preservation model is .pkl file, as final training result.
Step 3, neural network model test
It is loaded into model, the electrocardiogram test set sample input model that pretreatment is finished is analyzed, and recognition result is marked with it
Label comparison obtains the recognition accuracy of model.
By the operation of above-mentioned steps, building, the instruction of the deep neural network for screening heart infarction electrocardiogram can be realized
Practice and tests.
Above-described specific descriptions have carried out further specifically the purpose of invention, technical scheme and beneficial effects
It is bright, it should be understood that above is only a specific embodiment of the present invention, being used to explain the present invention, it is not used to limit this
The protection scope of invention, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should all
It is included within protection scope of the present invention.
Claims (6)
1. a kind of screening method of the myocardial infarction electrocardiogram based on deep neural network, which is characterized in that the method includes
Following steps:
Step 1 pre-processes original electrocardiographicdigital image, removes grid frame；
Step 2 carries out median filter process to the image in step 1, leaves spiced salt noise after eliminating removal grid frame；
Step 3 enhances contrast by histogram equalization method, and size is unified for 224*224 by scaling electrocardiogram, and will
Data normalization is [0,1]；
Step 4 is constructed deep neural network, and is trained using ECG training set.It obtains whether to diagnose with cardiac muscle
Two Classification Neurals of infarct；
Step 5 carries out auxiliary screening to myocardial infarction electrocardiogram using the deep neural network that training obtains.
2. a kind of screening method of the myocardial infarction electrocardiogram based on deep neural network as described in claim 1, feature
It is: in the step 1, the pretreated process of electrocardiogram are as follows: convert the image into grayscale image, step-by-step is carried out to grayscale image and is taken
Inverse operations, then carry out adaptive threshold, i.e., image different zones use different threshold values.Obtain binaryzation label image.To it
Closing operation of mathematical morphology operation is carried out, independent noise is reduced, fills the hole of connected region.Add operation is carried out with original image, is obtained
Remove the image of grid frame.
3. a kind of screening method of the myocardial infarction electrocardiogram based on deep neural network as claimed in claim 2, feature
It is: in the step 1, for using adaptive Gauss method, the calculating of adaptive threshold T (x, y) in adaptive threshold
Journey are as follows::
The distance of step 1.1 (x, y) surrounding pixel to its central point obtains corresponding weight by Gaussian function；
Step 1.2, which calculates around each pixel the weighted mean of (15 × 15) size block of pixels and subtracts constant -10, obtains threshold value T
(x,y)。
4. a kind of screening method of the myocardial infarction electrocardiogram based on deep neural network as claimed in claim 2, feature
It is: in the step 1, the process of grid frame removal are as follows:
Step1.1 is in the image after adaptive threshold, by closed operation and structural element cv2.MORPH_RECT and sizeIdentify horizontal line；
Step1.2 is in the image after adaptive threshold, by closed operation and structural element cv2.MORPH_RECT and sizeIdentify vertical line；
The image that step1.3 is obtained by step1.1 and step1.2, is merged, i.e. meshing in extraction original image；
The step1.3 image obtained and original image are carried out add operation by step1.4.
5. a kind of screening method of myocardial infarction electrocardiogram based on deep neural network as described in one of Claims 1 to 4,
It is characterized by: constructing the process of convolutional neural networks in the step 4 are as follows:
Step 4.1 inputs the electrocardiogram image that one group of size is 224*224*3；
Step 4.2 first passes through the convolution operation of 7*7 size, then carries out batch normalization, Relu activation primitive behaviour
Make；
Step 4.3 is operated by MaxPooling, extracts main feature；
The feature that step 4.4 is extracted passes through residual error convolution module, which includes the convolution operation and batch of 2 groups of 3*3
normalization；
Step 4.5: step 3.3 is repeated three times, using the average pond of 7*7 size；
Step 4.6: by features described above by full articulamentum FullyConnectedLayer and Softmax function, finally obtaining defeated
Two classification results out.
6. a kind of screening method of myocardial infarction electrocardiogram based on deep neural network as described in one of Claims 1 to 4,
It is characterized by: deep neural network framework is by 1 convolutional layer, 1 maximum pond layer, 4 residual error convolution in the step 4
Module, 1 average pond layer, 1 full articulamentum form, are standardized operation to feature after each convolutional layer, improve instruction
Practice speed, and pass through ReLU activation primitive, improves the non-linear expression of network；It include two convolution in each residual error convolution module
Layer and a quick branch, quick branch starting point are input, and terminal is the add operation after second convolutional layer, so that input feature vector
Numerical value directly can be carried out with the feature that second convolutional layer extracts to be added；The Output Size of the last one full articulamentum is 2, right
Two classification of screening results are answered, two classification is myocardial infarction electrocardiogram or non-salary motivation electrocardiogram.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910268945.4A CN110327033B (en) | 2019-04-04 | 2019-04-04 | Myocardial infarction electrocardiogram screening method based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910268945.4A CN110327033B (en) | 2019-04-04 | 2019-04-04 | Myocardial infarction electrocardiogram screening method based on deep neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110327033A true CN110327033A (en) | 2019-10-15 |
| CN110327033B CN110327033B (en) | 2022-05-03 |
Family
ID=68139300
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910268945.4A Active CN110327033B (en) | 2019-04-04 | 2019-04-04 | Myocardial infarction electrocardiogram screening method based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110327033B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110731773A (en) * | 2019-10-28 | 2020-01-31 | 浙江工业大学 | abnormal electrocardiogram screening method based on fusion of global and local depth features of electrocardiogram |
| CN110974214A (en) * | 2019-12-20 | 2020-04-10 | 华中科技大学 | Automatic electrocardiogram classification method, system and equipment based on deep learning |
| CN111640219A (en) * | 2020-06-04 | 2020-09-08 | 许昌开普电气研究院有限公司 | Inspection robot control system and method based on overhead line |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3884221A (en) * | 1969-06-10 | 1975-05-20 | George Eastman | Method for the construction and diagnosis of three dimensional ectocariagrams |
| WO2001087143A2 (en) * | 2000-05-18 | 2001-11-22 | Commwell, Inc. | Chair and ancillary apparatus with medical diagnostic features in a remote health monitoring system |
| JP2003250786A (en) * | 2002-02-28 | 2003-09-09 | Konica Corp | Image-processing apparatus, image-processing method, program, and storage medium |
| CN101167105A (en) * | 2005-04-29 | 2008-04-23 | 皇家飞利浦电子股份有限公司 | Multi-surface modelling |
| CN104182625A (en) * | 2014-08-15 | 2014-12-03 | 重庆邮电大学 | Electrocardiosignal denoising method based on morphology and EMD (empirical mode decomposition) wavelet threshold value |
| CN104280035A (en) * | 2013-07-08 | 2015-01-14 | 厦门雅迅网络股份有限公司 | Method for elimination of electronic map interest point label covering of roads |
| CN104809702A (en) * | 2015-04-22 | 2015-07-29 | 上海理工大学 | Pulse diagnosis curve grid eliminating method based on frequency domain processing |
| CN105125206A (en) * | 2015-09-15 | 2015-12-09 | 中山大学 | Intelligent electrocardio monitoring method and device |
| CN107174232A (en) * | 2017-04-26 | 2017-09-19 | 天津大学 | A kind of electrocardiographic wave extracting method |
| WO2017215284A1 (en) * | 2016-06-14 | 2017-12-21 | 山东大学 | Gastrointestinal tumor microscopic hyper-spectral image processing method based on convolutional neural network |
| CN107590797A (en) * | 2017-07-26 | 2018-01-16 | 浙江工业大学 | A kind of CT images pulmonary nodule detection method based on three-dimensional residual error neutral net |
| CN107909555A (en) * | 2017-11-27 | 2018-04-13 | 北京大恒图像视觉有限公司 | A kind of gridding noise elimination method for keeping acutance |
| CN108647565A (en) * | 2018-03-28 | 2018-10-12 | 浙江工业大学 | A kind of data preprocessing method classified to electrocardiosignal based on deep learning model |
-
2019
- 2019-04-04 CN CN201910268945.4A patent/CN110327033B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3884221A (en) * | 1969-06-10 | 1975-05-20 | George Eastman | Method for the construction and diagnosis of three dimensional ectocariagrams |
| WO2001087143A2 (en) * | 2000-05-18 | 2001-11-22 | Commwell, Inc. | Chair and ancillary apparatus with medical diagnostic features in a remote health monitoring system |
| JP2003250786A (en) * | 2002-02-28 | 2003-09-09 | Konica Corp | Image-processing apparatus, image-processing method, program, and storage medium |
| CN101167105A (en) * | 2005-04-29 | 2008-04-23 | 皇家飞利浦电子股份有限公司 | Multi-surface modelling |
| CN104280035A (en) * | 2013-07-08 | 2015-01-14 | 厦门雅迅网络股份有限公司 | Method for elimination of electronic map interest point label covering of roads |
| CN104182625A (en) * | 2014-08-15 | 2014-12-03 | 重庆邮电大学 | Electrocardiosignal denoising method based on morphology and EMD (empirical mode decomposition) wavelet threshold value |
| CN104809702A (en) * | 2015-04-22 | 2015-07-29 | 上海理工大学 | Pulse diagnosis curve grid eliminating method based on frequency domain processing |
| CN105125206A (en) * | 2015-09-15 | 2015-12-09 | 中山大学 | Intelligent electrocardio monitoring method and device |
| WO2017215284A1 (en) * | 2016-06-14 | 2017-12-21 | 山东大学 | Gastrointestinal tumor microscopic hyper-spectral image processing method based on convolutional neural network |
| CN107174232A (en) * | 2017-04-26 | 2017-09-19 | 天津大学 | A kind of electrocardiographic wave extracting method |
| CN107590797A (en) * | 2017-07-26 | 2018-01-16 | 浙江工业大学 | A kind of CT images pulmonary nodule detection method based on three-dimensional residual error neutral net |
| CN107909555A (en) * | 2017-11-27 | 2018-04-13 | 北京大恒图像视觉有限公司 | A kind of gridding noise elimination method for keeping acutance |
| CN108647565A (en) * | 2018-03-28 | 2018-10-12 | 浙江工业大学 | A kind of data preprocessing method classified to electrocardiosignal based on deep learning model |
Non-Patent Citations (2)
| Title |
|---|
| CHO, SOHYUNG: "Designing a Novel ECG Simulator Multi-Modality Electrocardiography into a Three-Dimensional Wire Cube Network", 《IEEE TECHNOLOGY AND SOCIETY MAGAZINE》 * |
| 徐晶: "人工神经网络在急性心肌梗死诊断中的应用", 《心血管康复医学杂志》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110731773A (en) * | 2019-10-28 | 2020-01-31 | 浙江工业大学 | abnormal electrocardiogram screening method based on fusion of global and local depth features of electrocardiogram |
| CN110731773B (en) * | 2019-10-28 | 2022-10-28 | 浙江工业大学 | Abnormal electrocardiogram screening method based on fusion of global and local depth features of electrocardiogram |
| CN110974214A (en) * | 2019-12-20 | 2020-04-10 | 华中科技大学 | Automatic electrocardiogram classification method, system and equipment based on deep learning |
| CN111640219A (en) * | 2020-06-04 | 2020-09-08 | 许昌开普电气研究院有限公司 | Inspection robot control system and method based on overhead line |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110327033B (en) | 2022-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109998525B (en) | Arrhythmia automatic classification method based on discriminant deep belief network | |
| CN108714026B (en) | Fine-grained electrocardiosignal classification method based on deep convolutional neural network and online decision fusion | |
| CN111772619B (en) | Heart beat identification method based on deep learning, terminal equipment and storage medium | |
| CN111990989A (en) | Electrocardiosignal identification method based on generation countermeasure and convolution cyclic network | |
| CN110179453A (en) | Electrocardiogram classification method based on convolutional neural networks and shot and long term memory network | |
| CN107811626A (en) | A kind of arrhythmia classification method based on one-dimensional convolutional neural networks and S-transformation | |
| CN110403600B (en) | Intelligent analysis method and system for paroxysmal atrial fibrillation based on difference time scatter diagram | |
| CN109191457A (en) | A kind of pathological image quality validation recognition methods | |
| CN108511055B (en) | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules | |
| CN110327033A (en) | A kind of screening method of the myocardial infarction electrocardiogram based on deep neural network | |
| CN101667289B (en) | Retinal image segmentation method based on NSCT feature extraction and supervised classification | |
| CN110638430B (en) | Method for building cascade neural network ECG signal arrhythmia classification model | |
| CN111759345B (en) | Heart valve abnormality analysis method, system and device based on convolutional neural network | |
| CN110477909B (en) | Gender classification method based on resting electroencephalogram data | |
| CN112785551A (en) | Coronary artery segmentation method based on deep learning | |
| CN112906748A (en) | 12-lead ECG arrhythmia detection classification model construction method based on residual error network | |
| CN109902682A (en) | A kind of mammary gland x line image detection method based on residual error convolutional neural networks | |
| CN113509186B (en) | ECG classification system and method based on deep convolutional neural network | |
| CN113349793A (en) | Arrhythmia classification method based on layered attention model | |
| CN113171106A (en) | Electrocardio abnormality detection method based on VQ-VAE2 and deep neural network method | |
| Xu et al. | An ECG denoising method based on the generative adversarial residual network | |
| CN116361688A (en) | Multi-mode feature fusion model construction method for automatic classification of electrocardiographic rhythms | |
| CN115281688A (en) | Cardiac hypertrophy multi-label detection system based on multi-mode deep learning | |
| CN115035127A (en) | Retinal vessel segmentation method based on generative confrontation network | |
| CN111261283B (en) | Electrocardiosignal deep neural network modeling method based on pyramid convolution layer |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |