CN110327033B - Myocardial infarction electrocardiogram screening method based on deep neural network - Google Patents
Myocardial infarction electrocardiogram screening method based on deep neural network Download PDFInfo
- Publication number
- CN110327033B CN110327033B CN201910268945.4A CN201910268945A CN110327033B CN 110327033 B CN110327033 B CN 110327033B CN 201910268945 A CN201910268945 A CN 201910268945A CN 110327033 B CN110327033 B CN 110327033B
- Authority
- CN
- China
- Prior art keywords
- image
- neural network
- myocardial infarction
- electrocardiogram
- 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.)
- Active
Links
Images
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 method for screening a myocardial infarction electrocardiogram based on a deep neural network comprises the following steps: firstly, preprocessing an original electrocardiogram image, and removing a grid frame; step two, carrying out median filtering processing on the image in the step one, and eliminating the grid frame and then leaving pepper and salt noise points; enhancing contrast by a histogram equalization method, scaling the electrocardiogram to unify the size into 224 x 224, and normalizing the data into [0,1 ]; and step four, constructing a deep neural network, and training by utilizing an ECG training set. Obtaining a two-class neural network capable of diagnosing whether myocardial infarction occurs; and fifthly, carrying out auxiliary screening on the myocardial infarction of the electrocardiogram by using the deep neural network obtained by training. The invention can assist to judge whether the myocardial infarction is caused or not according to the electrocardiogram.
Description
Technical Field
The invention relates to the field of medical image analysis and machine learning, in particular to an electrocardiogram myocardial infarction screening method, and belongs to the field of deep learning-based medical image analysis.
Background
Myocardial Infarction (MI) is a common heart disease. Early detection is crucial for Percutaneous Coronary Intervention (PCI) or coronary bypass surgery to effectively treat acute myocardial infarction. Myocardial infarction is usually diagnosed by clinical manifestations, laboratory results and electrocardiograms. An electrocardiogram is produced by recording the time-varying electrical potentials at specific locations on the body surface, which represent the electrical activity of the heart. Deviations from the usual shape of the cardioelectric curve may indicate myocardial infarction as well as many other cardiac and non-cardiac diseases. Electrocardiograms (ECGs) are a popular diagnostic tool because they are non-invasive, inexpensive to produce, and offer high diagnostic value.
In the process of myocardial infarction judgment based on the electrocardiogram, doctors need to spend a great deal of time to carefully diagnose the electrocardiogram generated by each lead, which has high requirements on the energy of the doctors and the expert knowledge of the doctors. And the wave band characteristic difference in the electrocardiogram is small, which easily causes missed diagnosis and misdiagnosis with certain probability. And in some regions, doctors engaged in diagnosis of patient's conditions based on electrocardiography are limited due to conditions. In recent years, wavelet change and k-nearest neighbor methods are mostly adopted in traditional computer-aided diagnosis methods for diagnosis, but the methods have high requirements on the quality of an electrocardiogram, and the quality of a sample easily influences the result of a model. There are some mature studies of various arrhythmias, but there are few studies specifically on myocardial infarction diseases. In addition, myocardial infarction presents with different symptoms in the ECG, such as ST elevation and non-ST elevation. This places high demands on the algorithm for identifying electrocardiograms of multi-symptom myocardial infarction manifestations. These adverse factors make diagnosis difficult, make it difficult to obtain effective features using statistical methods, and make diagnosis difficult in conventional methods due to the fact that there are a lot of noise such as grid frames in the images.
Disclosure of Invention
In order to overcome the defects of high difficulty, low efficiency and low precision of the conventional ECG myocardial infarction diagnosis mode, the invention provides an auxiliary screening method of an ECG based on a deep neural network, which has high speed, high efficiency and high precision, realizes automatic analysis of an electrocardiogram, and can effectively perform auxiliary judgment on the myocardial infarction condition of the electrocardiogram.
In order to solve the technical problem, the invention adopts the technical scheme that:
a screening method of myocardial infarction electrocardiograms based on a deep neural network comprises the following steps:
firstly, preprocessing an original electrocardiogram image, and removing a grid frame;
step two, carrying out median filtering processing on the image in the step one, and eliminating the grid frame and then leaving pepper and salt noise points;
enhancing contrast by a histogram equalization method, scaling the electrocardiogram to unify the size into 224 x 224, and normalizing the data into [0,1 ];
and step four, constructing a deep neural network, and training by utilizing an ECG training set. Obtaining a two-classification neural network capable of judging whether the patient suffers from myocardial infarction;
and fifthly, carrying out auxiliary screening on the myocardial infarction of the electrocardiogram by using the deep neural network obtained by training.
Further, in the step one, the process of preprocessing the electrocardiogram comprises the following steps: and converting the image into a gray level image, carrying out bit-wise inversion operation on the gray level image, and then carrying out self-adaptive threshold, namely adopting different thresholds for different areas of the image. Obtaining a binary label image; performing morphological closed operation on the communication area to reduce independent noise and fill the holes in the communication area; and performing addition operation with the original image to obtain an image with the grid frame removed.
Preferably, in the first step, for the adaptive threshold, an adaptive gaussian method is adopted, and the adaptive threshold T (x, y) is calculated by:
step1.1, obtaining corresponding weight values by the distance from the surrounding pixels to the center point of the surrounding pixels through a Gaussian function;
step1.2 calculates the weighted mean of (15 × 15) size pixel blocks around each pixel and subtracts a constant 10 to obtain the threshold T (x, y).
Still preferably, in the first step, the structural element for the close operation is set to cv2.morph _ RECT, and the sizes of the structural elements in the horizontal line and vertical line identification are respectively set to:
in the formula, H and W represent the height and width of the grayscale image, respectively, and scale represents the scaling factor of the structural element.
More preferably, in the first step, the grid frame removing process:
setp1.1 in the image after adaptive thresholding, by means of the closing operation and the structural element cv2.MORPH _ RECT and the size
Identifying a transverse line;
setp1.2 in the image after adaptive thresholding, by means of the closing operation and the structural element cv2.MORPH _ RECT and the size
Identifying a vertical line;
the setp1.3 is fused through the images obtained by the setp1.1 and the setp1.2, namely, a grid part in the original image is extracted;
and the setp1.4 carries out add operation on the image obtained by the setp1.3 and the original image.
In the fourth step, the process of constructing the convolutional neural network is as follows:
step 4.1 inputting a set of electrocardiogram images with the size of 224 x 3;
step 4.2, performing convolution operation with the size of 7 × 7, and then performing batchnormalization and Relu activation function operation;
4.3, extracting main characteristics through Max scaling operation;
4.4, the extracted features are processed by a residual convolution module, wherein the residual convolution module comprises 2 groups of convolution operations of 3 × 3 and batch normalization;
step 4.5: repeating the step 3.3 three times, and performing average pooling with sizes of 7 × 7;
step 4.6: and finally, passing the characteristics through a full connection layer FullyConnectedLayer and a Softmax function to obtain an output binary result.
In the fourth step, the deep neural network architecture consists of 1 convolutional layer, 1 maximum pooling layer, 4 residual convolutional modules (Block modules), 1 average pooling layer and 1 full-link layer, the characteristics are standardized after each convolutional layer, the training speed is increased, and the nonlinear expression of the network is improved through a ReLU activation function; each residual convolution module comprises two convolution layers and a shortcut branch, the starting point of the shortcut branch is input, and the end point of the shortcut branch is the adding operation after the second convolution layer, so that the input characteristic can be directly added with the characteristic extracted by the second convolution layer; therefore, the network only needs to calculate the residual error compared with the original input, and the training difficulty is reduced; the output size of the last full-connection layer is 2, and the two classifications correspond to two classifications of screening results, wherein the two classifications are myocardial infarction electrocardiograms or non-myocardial infarction electrocardiograms.
The method is based on the auxiliary screening of the myocardial infarction electrocardiogram of the deep neural network, and utilizes the convolutional neural network to extract the ECG image characteristics so as to realize the judgment of the myocardial infarction electrocardiogram. Compared with the prior art, the method has the advantages that:
1. the electrocardiogram is analyzed through the convolutional neural network, and the myocardial infarction electrocardiogram is automatically screened in an auxiliary mode.
2. In the preprocessing of the electrocardiogram, the method introduces the elimination of grid frames and eliminates the noise in the image.
3. And extracting effective characteristics of the image by introducing a residual convolution module structure. Meanwhile, due to the existence of the residual shortcut branch, the training difficulty of the network is reduced.
Drawings
Fig. 1 is a flow chart of the identification of an electrocardiograph of myocardial infarction based on a deep neural network.
Fig. 2 is a schematic diagram of a neural network structure for diagnosing myocardial infarction.
FIG. 3 is a schematic diagram of a residual module structure in a neural network.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
Referring to fig. 1 to 3, a method for screening an electrocardiograph of myocardial infarction based on a deep neural network includes the following steps:
firstly, preprocessing an original electrocardiogram image, and removing a grid frame;
step two, carrying out median filtering processing on the image in the step one, and eliminating the grid frame and then leaving pepper and salt noise points;
enhancing contrast by a histogram equalization method, scaling the electrocardiogram to unify the size into 224 x 224, and normalizing the data into [0,1 ];
and step four, constructing a deep neural network, and training by utilizing an ECG training set. Obtaining a two-class neural network capable of diagnosing whether myocardial infarction occurs;
and fifthly, performing auxiliary diagnosis on the myocardial infarction of the electrocardiogram by using the deep neural network obtained by training.
Further, in the step one, the process of preprocessing the electrocardiogram comprises the following steps: and converting the image into a gray level image, carrying out bit-wise inversion operation on the gray level image, and then carrying out self-adaptive threshold, namely adopting different thresholds for different areas of the image. And obtaining a binary label image. And performing morphological closing operation on the communication area to reduce independent noise and fill the holes in the communication area. And performing addition operation with the original image to obtain an image with the grid frame removed.
Preferably, in the first step, for the adaptive threshold, an adaptive gaussian method is adopted, and the adaptive threshold T (x, y) is calculated by:
step1.1, obtaining corresponding weight values by the distance from the surrounding pixels to the center point of the surrounding pixels through a Gaussian function;
step1.2 calculates the weighted mean of (15 × 15) size pixel blocks around each pixel and subtracts a constant 10 to obtain the threshold T (x, y).
Still preferably, in the first prime number step, the size of the structural element for the close operation is set to cv2.morph _ RECT, and the size of the structural element is set to:
in the formula, H and W represent the height and width of the grayscale image, respectively, and scale represents the scaling factor of the structural element.
More preferably, in the first step, the grid frame removing process:
step1.1 in the image after adaptive thresholding, by means of the closing operation and the structural element cv2.MORPH _ RECT and the size
Identifying a transverse line;
step1.2 in the image after adaptive thresholding, by means of the closing operation and the structural element cv2.MORPH _ RECT and the size
Identifying a vertical line;
step1.3, fusing the images obtained by step1.1 and step1.2, namely extracting a grid part in the original image;
step1.4 performs add operation on the image obtained from step1.3 and the original image.
In the fourth step, the process of constructing the convolutional neural network is as follows:
step 4.1 inputting a set of electrocardiogram images with the size of 224 x 3;
step 4.2, performing convolution operation with the size of 7 × 7, and then performing batchnormalization and Relu activation function operation;
4.3, extracting main characteristics through Max scaling operation;
4.4, the extracted features are processed by a residual convolution module, wherein the residual convolution module comprises 2 groups of convolution operations of 3 × 3 and batch normalization;
step 4.5 repeat step 3.3 three times, and then average pooling 7 x 7 times;
and 4.6, passing the characteristics through a FullyConnectedLayer function and a Softmax function of the full connection layer, and finally obtaining an output classification result.
In the fourth step, the deep neural network architecture mainly comprises a network architecture mainly composed of 1 convolutional layer, 1 maximum pooling layer, 4 residual convolutional modules (Block modules), 1 average pooling layer and 1 full-connection layer. After each convolution layer, the characteristics are subjected to standardized operation, the training speed is improved, and the nonlinear expression of the network is improved through a ReLU activation function. As shown in fig. 3, two convolutional layers are included in each residual convolutional block, one shortcut. The starting point of the shortcut branch is input, and the end point of the shortcut branch is addition operation after the second convolution layer, so that the input characteristic can be directly added with the characteristic extracted by the second convolution layer in a numerical value mode. Therefore, the network only needs to calculate the residual error compared with the original input, and the training difficulty is reduced. The output size of the last full-connection layer is 2, which corresponds to two categories of the screening result (myocardial infarction electrocardiogram or non-myocardial infarction electrocardiogram).
Example (c): the electrocardiograms used in this case are of the type 2, i.e., myocardial infarction electrocardiograms or non-myocardial infarction electrocardiograms. There were a total of 2282 samples with the same number of infarct and non-infarct electrocardiograms. 913 samples are randomly selected from the positive and negative samples respectively to serve as a training set, 114 samples serve as a verification set, and 114 samples serve as a test set. The following describes the electrocardiogram grid frame removal, model training and testing process.
Step one, removing a grid frame from the electrocardiogram.
Step1.1, converting the image into a gray-scale image, carrying out bit-wise negation operation on the gray-scale image, and then carrying out self-adaptive threshold;
step1.2, performing morphological closed operation on the image obtained in the step 1.1;
step1.3, performing add operation on the image obtained in the step1.2 and the original image to obtain an image with the grid frame removed;
step two, the construction and training of the neural network, the specific structure is shown in fig. 2.
Step 2.1 the network architecture is mainly composed of 1 convolutional layer, 1 maximal pooling layer, 4 residual convolutional modules (Block modules), 1 average pooling layer and 1 full-link layer.
Step 2.2 the convolution kernel size of the first convolution layer is 7 x 7, the sliding step is 2, and padding is 3. The convolution kernels in the residual modules are all 3 x 3, except the sliding step of the first residual module is 1, the other convolution kernels are all 2, and the convolution kernels connected among the residual modules are 1 x 1. The number of convolution kernels increases as one goes into the residual block, 64,128,256,512, respectively. After each output convolution layer, the characteristics are subjected to batch standardization operation, the training speed is improved, and the nonlinear expression of the network is improved through a ReLU activation function. The largest pooling layer after the convolutional layer will reduce the feature size.
Step 2.3, all the parameter weights in the convolutional layer are initialized to be initialized random orthogonal matrixes in a weight regularization mode of L2 regularization, and the bias value is initialized to be 0. In the fully-connected layer, the weight is initialized to be random normal distribution, the weight regularization mode is L2 regularization, and the bias value is initialized to be 0.
And 2.4, constructing a network by using the Pythrch framework. The model adopts a batch training mode. The sample number of each batch of the training set generator and the verification set generator is 16, the training set generator returns data for 40 times as one round (epoch), after one round of training is completed, the generator returns 5 times and calculates the loss of the verification set, and the loss function is a cross entropy loss function. The model optimizer Adam, the parameters lr being 0.001, the beta being (0.9,0.999), eps being 1e-08, weight _ decade being 0, and amsgrad being False. The maximum training round of the model is 60, the training is stopped after the verification and the training loss are converged, and the model is stored as a pkl file to be used as a final training result.
Step three, testing the neural network model
And loading the model, inputting the preprocessed electrocardiogram test set sample into the model for analysis, and comparing the recognition result with the label to obtain the recognition accuracy of the model.
Through the operation of the steps, the construction, training and testing of the deep neural network for screening the myocardial infarction electrocardiogram can be realized.
The above detailed description is intended to illustrate the objects, aspects and advantages of the present invention, and it should be understood that the above detailed description is only exemplary of the present invention, and is not intended to limit the scope of the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (4)
1. A deep neural network construction method for electrocardiogram screening of myocardial infarction is characterized by comprising the following steps:
firstly, preprocessing an original electrocardiogram image, and removing a grid frame;
the electrocardiogram preprocessing process comprises the following steps: converting the image into a gray scale image, carrying out bitwise negation operation on the gray scale image, then carrying out self-adaptive threshold values, namely adopting different threshold values for different areas of the image to obtain a binary label image, carrying out morphological closed operation on the binary label image to reduce independent noise, filling holes in a communicated area, and carrying out add operation with an original image to obtain an image without a grid frame;
step two, carrying out median filtering processing on the image in the step one, and eliminating the grid frame and then leaving pepper and salt noise points;
enhancing contrast by a histogram equalization method, scaling the electrocardiogram to unify the size into 224 x 224, and normalizing the data into [0,1 ];
step four, constructing a deep neural network, and training by using an ECG training set to obtain a two-class neural network capable of diagnosing whether the patient has myocardial infarction;
the deep neural network architecture consists of 1 convolutional layer, 1 maximum pooling layer, 4 residual convolution modules, 1 average pooling layer and 1 full-connection layer, the characteristics are subjected to standardized operation after each convolutional layer, the training speed is increased, and the nonlinear expression of the network is improved through a ReLU activation function; each residual convolution module comprises two convolution layers and a shortcut branch, the starting point of the shortcut branch is input, and the end point of the shortcut branch is the adding operation after the second convolution layer, so that the input characteristic can be directly added with the characteristic extracted by the second convolution layer; the output size of the last full-connection layer is 2, and the two classifications correspond to two classifications of screening results, wherein the two classifications are myocardial infarction electrocardiograms or non-myocardial infarction electrocardiograms.
2. The method for constructing a deep neural network for electrocardiographic screening of myocardial infarction as claimed in claim 1, wherein the method comprises the following steps: in the first step, for the adaptive threshold, an adaptive gaussian method is adopted, and the adaptive threshold T (x, y) is calculated by:
step1.1, obtaining corresponding weight values by the distance from the surrounding pixels to the center point of the surrounding pixels through a Gaussian function;
step1.2 calculates the weighted mean of (15 × 15) size pixel blocks around each pixel and subtracts a constant 10 to obtain the threshold T (x, y).
3. The method for constructing a deep neural network for electrocardiographic screening of myocardial infarction as claimed in claim 1, wherein the method comprises the following steps: in the first step, H and W respectively represent the height and width of the grayscale image, scale represents the scaling factor of the structural element, and the grid frame removal process is as follows:
step1.1 in the image after adaptive thresholding, by means of the closing operation and the structural element cv2.MORPH _ RECT and the size
Identifying a transverse line;
step1.2 in the image after adaptive thresholding, by means of the closing operation and the structural element cv2.MORPH _ RECT and the size
Identifying a vertical line;
step1.3, fusing the images obtained by step1.1 and step1.2, namely extracting a grid part in the original image;
step1.4 performs an addition operation on the image obtained by step1.3 and the original image.
4. The method for constructing the deep neural network for the electrocardiographic screening of the myocardial infarction as claimed in one of claims 1 to 3, wherein the method comprises the following steps: in the fourth step, the process of constructing the convolutional neural network is as follows:
step 4.1 inputting a set of electrocardiogram images with the size of 224 x 3;
step 4.2, performing convolution operation with the size of 7 × 7, and then performing batch normalization and Relu activation function operation;
4.3, extracting main characteristics through Max scaling operation;
4.4, the extracted features are processed by a residual convolution module, wherein the residual convolution module comprises 2 groups of convolution operations of 3 × 3 and batch normalization;
step 4.5: repeating the step 3.3 three times, and performing average pooling with sizes of 7 × 7;
step 4.6: and finally, passing the characteristics through a full connection layer FullyConnectedLayer and a Softmax function to obtain an output binary result.
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 CN110327033A (en) | 2019-10-15 |
| CN110327033B true 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) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN111640219B (en) * | 2020-06-04 | 2022-11-18 | 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 |
|---|
| Designing a Novel ECG Simulator Multi-Modality Electrocardiography into a Three-Dimensional Wire Cube Network;Cho, Sohyung;《IEEE TECHNOLOGY AND SOCIETY MAGAZINE》;20160413;第75-84页 * |
| 人工神经网络在急性心肌梗死诊断中的应用;徐晶;《心血管康复医学杂志》;20040130;第493-495页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110327033A (en) | 2019-10-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110327033B (en) | Myocardial infarction electrocardiogram screening method based on deep neural network | |
| CN109480824B (en) | Method and device for processing electrocardio waveform data and server | |
| CN108511055B (en) | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules | |
| CN110613483B (en) | System for detecting fetal craniocerebral abnormality based on machine learning | |
| CN111329445A (en) | Atrial fibrillation identification method based on group convolution residual error network and long-term and short-term memory network | |
| CN111291727A (en) | Method and device for detecting signal quality by photoplethysmography | |
| CN115281688A (en) | Cardiac hypertrophy multi-label detection system based on multi-mode deep learning | |
| CN116758336A (en) | Medical image intelligent analysis system based on artificial intelligence | |
| Saad et al. | Addressing the intra-class mode collapse problem using adaptive input image normalization in GAN-based X-ray images | |
| CN111899272B (en) | Fundus image blood vessel segmentation method based on coupling neural network and line connector | |
| Hassan et al. | RRI-Net: classification of multi-class retinal diseases with deep recurrent residual inception network using OCT scans | |
| CN113408603A (en) | Coronary artery stenosis degree identification method based on multi-classifier fusion | |
| CN112932431B (en) | Heart rate identification method based on 1DCNN + Inception Net + GRU fusion network | |
| Pushpa et al. | Detection and stagewise classification of Alzheimer disease using deep learning methods | |
| CN113516641A (en) | End-to-end brain image data processing method and device based on deep learning | |
| Ion et al. | Breast Cancer Images Segmentation using Fuzzy Cellular Automaton | |
| CN114224354A (en) | Arrhythmia classification method, device and readable storage medium | |
| CN114098757B (en) | ECG signal monitoring method based on quantum particle swarm optimization | |
| Roy et al. | Identification and classification of human mental stress using physiological data: A low-power hybrid approach | |
| George et al. | Kidney Abnormalities Detection and Classification Using CNN-based Feature Extraction | |
| Gupta et al. | Brain Tumor Classification using MR Images and Transfer Learning | |
| CN113349790B (en) | Heart beat classification method and system of 2D convolutional neural network based on S-shaped reconstruction | |
| CN116740041B (en) | CTA scanning image analysis system and method based on machine vision | |
| CN116705252B (en) | Construction method, image classification method, device and medium for prostate cancer diagnosis model | |
| Prashantha et al. | Feature level fusion framework for brain MR image classification using supervised deep learning and hand crafted features |
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 |