CN110916645A - QRS wave identification method combining wavelet transformation and image segmentation network - Google Patents
QRS wave identification method combining wavelet transformation and image segmentation network Download PDFInfo
- Publication number
- CN110916645A CN110916645A CN201911255275.9A CN201911255275A CN110916645A CN 110916645 A CN110916645 A CN 110916645A CN 201911255275 A CN201911255275 A CN 201911255275A CN 110916645 A CN110916645 A CN 110916645A
- Authority
- CN
- China
- Prior art keywords
- image
- qrs wave
- data
- network
- training
- 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
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
- A61B5/24—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]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
- A61B5/366—Detecting abnormal QRS complex, e.g. widening
Abstract
The invention discloses a QRS wave identification method combining wavelet transformation and an image segmentation network, and mainly relates to the field of deep learning and medical image processing. The technological scheme of the invention mainly includes the following main points: 1) one-dimensional electrocardiosignals are stretched into a two-dimensional frequency spectrum image through a wavelet transform method to be used as shallow features of the electrocardiosignals; 2) then, reversing and cutting the two-dimensional frequency spectrum image to adapt to a later training network; 3) inputting the processed spectrogram into an image segmentation network (U-net) to perform pixel-by-pixel binary classification prediction; 4) and finally, carrying out positioning calculation on the QRS wave of the image result output by the segmentation network to obtain a specific positioning point of the QRS wave. Compared with the traditional methods P & T, XQRS and GQRS, the method comprehensively utilizes the characteristics of wavelet transformation anti-interference capability and fast deep learning network speed and high precision, thereby achieving the purpose of improving QRS wave identification precision.
Description
Technical Field
The invention belongs to the field of medical image processing and deep learning, and relates to a QRS wave identification method combining wavelet transformation and an image segmentation network.
Background
The electrocardiogram is one of the most widely used clinical examination methods at present, can effectively reflect the skin electrical changes caused by the electrical activity of the heart of a human body, and not only can directly help doctors diagnose cardiovascular diseases, but also can indirectly diagnose the problem of influencing the heart activity. The QRS wave is the most significant part in the EGG, reflects the micro-current signal of the heart when the ventricle contracts, has energy accounting for a large percentage of the electrocardiosignal, is an important detection object for judging the health degree of the heart, and has important research value. Computer-based electrocardiographic analysis is typically performed on the QRS complex, which is the waveform with the largest amount of information in the electrocardiographic data and is most visually significant.
For the QRS wave detection, the position of the QRS wave is located in the one-dimensional electrocardiosignal. Until now, many famous scholars at home and abroad are dedicated to the research of QRS wave recognition algorithm, and the traditional method for solving the problems mainly comprises the following steps: derivative, filter bank, wavelet transform, P & T, GQRS algorithm positioning, XQRS detection algorithm, etc. The filter bank is expensive in design cost and poor in flexibility. The wavelet transformation has relatively strong noise resistance and is relatively flexible. The WFDB is also widely applied to electrocardio detection tools (such as GQRS algorithm positioning, XQRS detection algorithm and the like) by virtue of the advantage of high speed. However, these algorithms are all focused on how to process one-dimensional electrocardiosignals, and the detection precision can not reach an ideal value.
At present, the methods found in the patent systems mainly include: a QRS wave identification method based on improved wavelet transformation and an integrated electrocardiosignal denoising and QRS wave identification rapid algorithm are provided. Recently, algorithms for detecting QRS waves based on wavelet transformation are diversified, so that the method for wavelet transformation is strong in practicability, shows strong flexibility in the field, has certain capability of resisting external signal interference, completely depends on QRS wave identification of wavelet transformation, and cannot achieve good effects in processing speed and identification precision.
Disclosure of Invention
The invention aims to provide a new idea for solving the problem that the current QRS wave positioning method is limited in identification precision, and provides a new QRS wave identification method combining wavelet transformation and an image segmentation network. According to the method, a one-dimensional electrocardiosignal is stretched into a two-dimensional frequency spectrum image by a wavelet transformation method, the two-dimensional frequency spectrum image is input into an image segmentation network, a positioning result is obtained through training, and the QRS wave recognition rate is finally improved.
In order to achieve the purpose, the invention adopts the technical scheme that:
a QRS wave identification method combining wavelet transformation and image segmentation network comprises the following steps:
(1) performing noise reduction processing on each one-dimensional electrocardiogram data in a training sample library to obtain good training data samples, wherein the size of each one-dimensional electrocardiogram data sample is 5000;
(2) in order to obtain frequency domain information of the electrocardiosignals, performing wavelet transformation on the training data sample in the step (1), and stretching the one-dimensional electrocardiosignals into two-dimensional frequency spectrum images, wherein the size of each image is 5000 x 512;
(3) in order to enlarge the sample capacity and reduce the training calculation amount of the later-stage network, the two-dimensional frequency spectrum image obtained in the step (2) is cut into 5 minutes, and the size of each cut frequency spectrum image is 1200 x 512;
(4) inputting the spectrogram obtained in the step (3) after cutting into an image segmentation network, and training the image segmentation network to obtain a gray scale image which can be subjected to pixel-by-pixel binary prediction;
(5) and (4) carrying out QRS wave positioning calculation on the gray scale image predicted in the step (4) to obtain a specific QRS wave positioning point.
As a further improvement of the method, the denoising processing in the step (1) specifically includes: the method comprises the steps of removing electromyographic interference signals by using a low-pass filter, restraining power frequency interference signals by using a stuffing filter, and correcting baseline drift by using a zero phase shift filter.
As a further improvement of the method, the wavelet transform in step (2) specifically includes: for the wavelet function the method selects the Mexican Hat function, setting the width of the wavelet transform to 256. The target detection network has a bias towards the center position of the picture, so that the spectrogram after wavelet transformation is subjected to overturning splicing, and the QRS wave position is always positioned at the center of the picture.
As a further improvement of the method, the image cutting processing in step (3) specifically includes: and (3) performing sliding window cutting on the two-dimensional spectrum image with the resolution of 5000 × 512 in the step (2), setting the window size to be 1200, filling 0 pixel values of 100 units into the initial position and the end position of the image respectively, and finally obtaining the central position image data of the 1200 × 512-resolution image, namely removing the data of 100 pixel values in front and back, leaving only the data of the central QRS wave, wherein each original image is divided into 5 two-dimensional electrocardiogram images with the resolution of 1000 × 512. When the picture is sliced by adopting a sliding window cutting method, the image edge data can be well moved to the center, the difficulty of edge detection is reduced, and the training precision is improved.
As a further improvement of the method, the data input image segmentation network in step (4) specifically includes:
step 1, making a data set from the frequency spectrum image data cut in the step 3, setting the width of 64 and the height of 512 as a label of an input image segmentation network (U-net) by taking the position of an original QRS wave as a center, randomly selecting 80% of original electrocardio two-dimensional image data as a training verification set of the U-net, and taking the rest as a test set;
step 2, inputting the training verification set in the step 1 into a U-net network for training;
step 3, inputting the test set in the step 1 into a U-net network for testing to obtain a gray scale image which can be subjected to pixel-by-pixel classification prediction;
as a further improvement of the method, the QRS wave location calculation in step (5) specifically includes: for each of the two classified predicted gray-scale maps obtained from the test image input U-net network in (5), the following operations are performed: judging from the first row of the image, if 70% or more of pixel points in the first row are all 1, the first row belongs to the range of QRS waves, and recording the position of the first row; otherwise, the information of the first column is not recorded, and the next column is switched to. If the previous column of information is recorded and the current column still belongs to the range of the QRS wave, the information of the current column is not recorded; and if the current column does not belong to the range of the QRS wave, recording the information of the current column. If the previous column of information is not recorded, if the current column belongs to the range of the QRS wave, recording the information of the current column; and if the current list does not belong to the range of the QRS wave, the information of the current list is not recorded. And so on until all columns of the picture are traversed. And obtaining the positions of the columns meeting the conditions, and taking the mean value of every two adjacent terms to obtain the specific position for positioning the QRS wave.
The innovation of the invention is that: the invention provides a method for carrying out pixel-by-pixel classification prediction on a two-dimensional frequency spectrum image by combining a traditional electrocardio recognition method with image segmentation in deep learning, positioning and expanding the traditional QRS wave on one dimension to recognize the two-dimensional frequency spectrum image and combining a deep learning image segmentation technology. In conclusion, the method expands the existing QRS wave identification method, and can effectively analyze the singularity of the signal, including the edge point of the image. The method can comprehensively utilize the anti-noise interference capability of wavelet transformation and improve the identification accuracy by utilizing deep learning.
Drawings
The attached drawing is a flow chart of the whole process of the invention
Detailed Description
The present invention will be described in detail below with reference to embodiments shown in the drawings. These embodiments are not intended to limit the present invention, and structural, methodological, or functional changes made by those skilled in the art according to these embodiments are included in the scope of the present invention.
The invention discloses a QRS wave identification method combining wavelet transformation and an image segmentation network. The specific implementation steps comprise:
(1) performing noise reduction processing on each one-dimensional electrocardiogram data in a training sample library to obtain good training data samples, wherein the size of each one-dimensional electrocardiogram data sample is 5000;
(2) in order to obtain frequency domain information of the electrocardiosignals, performing wavelet transformation on the training data sample in the step (1), and stretching the one-dimensional electrocardiosignals into two-dimensional frequency spectrum images, wherein the size of each image is 5000 x 512;
(3) in order to enlarge the sample capacity and reduce the training calculation amount of the later-stage network, the two-dimensional frequency spectrum image obtained in the step (2) is cut into 5 minutes, and the size of each cut frequency spectrum image is 1200 x 512;
(4) inputting the spectrogram obtained in the step (3) after cutting into an image segmentation network, and training the image segmentation network to obtain a gray scale image which can be subjected to pixel-by-pixel binary prediction;
(5) and (4) carrying out QRS wave positioning calculation on the gray scale image predicted in the step (4) to obtain a specific QRS wave positioning point.
As a further improvement of the method, the denoising processing in the step (1) specifically includes: the method comprises the steps of removing electromyographic interference signals by using a low-pass filter, restraining power frequency interference signals by using a stuffing filter, and correcting baseline drift by using a zero phase shift filter.
As a further improvement of the method, the wavelet transform in step (2) specifically includes: for the wavelet function the method selects the Mexican Hat function, setting the width of the wavelet transform to 256. The target detection network has a bias towards the center position of the picture, so that the spectrogram after wavelet transformation is subjected to overturning and splicing, and QRS waves are always displayed in the center of the picture.
As a further improvement of the method, the image cutting processing in step (3) specifically includes: and (3) performing sliding window cutting on the two-dimensional spectrum image with the resolution of 5000 × 512 in the step (2), setting the window size to be 1200, filling the initial position and the end position of the image with 0 pixel value of 100 units respectively, and finally obtaining the central position image data of the 1200 × 512-resolution picture, namely removing 100 pixel value data in front and back to leave only the data of the central QRS wave, wherein each original image is divided into 5 two-dimensional electrocardiogram images with the resolution of 1000 × 512. When the picture is sliced by adopting a sliding window cutting method, the image edge data can be well moved to the center, the difficulty of edge detection is reduced, and the training precision is improved.
As a further improvement of the method, the data input image segmentation network in step (4) specifically includes:
step 1, making a data set from the frequency spectrum image data cut in the step 3, setting the width of 64 and the height of 512 as a label of an input image segmentation network (U-net) by taking the position of an original QRS wave as a center, randomly selecting 80% of original electrocardio two-dimensional image data as a training verification set of the U-net, and taking the rest as a test set;
step 2, inputting the training verification set in the step 1 into a U-net network for training;
and 3, inputting the test set in the step 1 into a U-net network for testing to obtain a gray scale image which can be subjected to pixel-by-pixel two-classification prediction.
As a further improvement of the method, the QRS wave location calculation in step (5) specifically includes: for each of the two classified predicted gray-scale maps obtained from the test image input U-net network in (5), the following operations are performed: judging from the first row of the image, if 70% or more of pixel points in the first row are all 1, the first row belongs to the range of QRS waves, and recording the position of the first row; otherwise, the information of the first column is not recorded, and the next column is switched to. If the previous column of information is recorded and the current column still belongs to the range of the QRS wave, the information of the current column is not recorded; and if the current column does not belong to the range of the QRS wave, recording the information of the current column. If the previous column of information is not recorded, if the current column belongs to the range of the QRS wave, recording the information of the current column; and if the current list does not belong to the range of the QRS wave, the information of the current list is not recorded. And so on until all columns of the picture are traversed. And obtaining the positions of the columns meeting the conditions, and taking the mean value of every two adjacent terms to obtain the specific position for positioning the QRS wave.
The effects of the invention can be further illustrated by the following experiments:
1) conditions of the experiment
The experimental simulation environment is as follows: pycharm compiler and python language, single CPU kernel in 3.70Ghz i7-8700K CPU.
2) Content of the experiment
2.1) sources of Experimental data
The experiment was collected from 11 major hospitals using the electrocardiographic data set published by the second chinese physiological signal challenge (CPSC 2019), which included normal and ECG data representing eight categories of conditions. The data set consists of two parts, 2000 groups for each ECG data length 5000, and the position of the QRS wave.
2.2) Experimental procedure:
firstly, each one-dimensional electrocardio data in a training sample library is subjected to noise reduction processing to obtain a good training data sample, and wavelet transformation is carried out on the good training data sample to stretch one-dimensional electrocardio signals into two-dimensional frequency spectrum images. And secondly, performing sliding window overlapping cutting processing on the two-dimensional frequency spectrum image, dividing the two-dimensional frequency spectrum image into 5 minutes, wherein the size of the frequency spectrum image after each cutting is 1200 × 512. And finally, making a data set by the frequency spectrum image data after cutting, setting the width of 64 and the height of 512 as label of an input image segmentation network (U-net) by taking the position of the original QRS wave as a center, randomly selecting 80% of the original electrocardio two-dimensional image data as a training verification set of the U-net, and inputting the data into the U-net network to train the data to obtain a gray scale image capable of carrying out pixel-by-pixel two-classification prediction.
Analysis of Experimental results
The following are the results of the comparison of the method of the present invention with the P & T, XQRS, GQRS methods.
Method of producing a composite material | P&T | XQRS | GQRS | Ours |
Identification precision (%) | 32.66 | 41.18 | 34.06 | 69.00 |
TABLE 1 CPSC 2019 data test results
As can be seen from Table 1, the recognition rate of the method of the present invention is much higher than that of the conventional P & T, XQRS and GQRS methods. The method has better performance in the aspect of identification accuracy in the aspect of QRS wave positioning.
It should be understood that although the present description refers to embodiments, not every embodiment contains only a single technical solution, and such description is for clarity only, and those skilled in the art should make the description as a whole, and the technical solutions in the embodiments can also be combined appropriately to form other embodiments understood by those skilled in the art.
The above-listed detailed description is only a specific description of a possible embodiment of the present invention, and they are not intended to limit the scope of the present invention, and equivalent embodiments or modifications made without departing from the technical spirit of the present invention should be included in the scope of the present invention.
Claims (6)
1. A method of QRS wave identification combining wavelet transform and image segmentation networks, the method comprising:
the method comprises the following steps of (1) carrying out noise reduction processing on each one-dimensional electrocardiogram data in a training sample library to obtain good training data samples, wherein the size of each one-dimensional electrocardiogram data sample is 5000;
in the step (2), in order to obtain frequency domain information of the electrocardiosignals, performing wavelet transformation on the training data sample in the step (1), and stretching the one-dimensional electrocardiosignals into two-dimensional frequency spectrum images, wherein the size of each image is 5000 x 512;
step (3) is to enlarge the sample capacity and reduce the training calculation amount of the later network, and the two-dimensional frequency spectrum image obtained in the step (2) is cut into 5 minutes, and the size of each cut frequency spectrum image is 1200 x 512;
step (4) inputting the spectrogram cut in the step (3) into an image segmentation network, and training the image segmentation network to obtain a gray scale image which can be subjected to pixel-by-pixel binary prediction;
and (5) carrying out QRS wave positioning calculation on the gray scale image predicted in the step (4) to obtain a specific QRS wave positioning point.
2. The method according to claim 1, wherein the denoising processing in step (1) specifically comprises: the method comprises the steps of removing electromyographic interference signals by using a low-pass filter, restraining power frequency interference signals by using a stuffing filter, and correcting baseline drift by using a zero phase shift filter.
3. The method according to claim 1, wherein the wavelet transform in step (2) specifically comprises: selecting Mexican Hat function for the wavelet function method, setting the width of wavelet transformation to be 256, and performing upset splicing on a spectrogram after wavelet transformation because a target detection network has bias on the center position of a picture, so that the QRS wave position is always positioned at the center of the picture.
4. The method according to claim 1, wherein the process in step (3) specifically comprises: and (3) performing sliding window cutting on the two-dimensional frequency spectrum image with the resolution of 5000 × 512 in the step (2), setting the window size to be 1200, filling 0 pixel values of 100 units into the initial position and the final position of the image respectively, and finally obtaining the image data of the central position of the 1200 × 512 resolution image, namely removing 100 pixel values before and after data, only leaving data of a central QRS wave, dividing each original image into 5 two-dimensional electrocardio images with the resolution of 1000 × 512, and when the image is sliced by adopting a sliding window cutting method, moving the image edge data to the center well, reducing the difficulty of edge detection and improving the training precision.
5. The method according to claim 1, wherein the data input image segmentation network in the step (4) specifically comprises: making a data set acceptable by a network model from the data of the frequency spectrum image cut in the step (3), setting the width of 64 and the height of 512 as label of an input image segmentation network (U-net) by taking the position of an original QRS wave as a center, randomly selecting 80% of original electrocardio two-dimensional image data as a training verification set of the U-net, using the rest data as a network test set, inputting the prepared training verification set into the U-net network for data training, inputting the test set into the U-net network for testing the model after the network training is finished, and obtaining a gray scale image capable of carrying out pixel-by-pixel two-classification prediction.
6. The method according to claim 1, wherein the QRS wave location calculation in step (5) specifically comprises: and (5) performing the following operation on each of the two classified prediction gray level maps obtained by inputting the test image into the U-net network in the step (5): judging from the first row of the image, if 70% or more of pixel points in the first row are all 1, the first row belongs to the range of QRS waves, and recording the position of the first row; otherwise, the information of the first column is not recorded, and the next column is switched to. If the previous column of information is recorded and the current column still belongs to the range of the QRS wave, the information of the current column is not recorded; if the current list does not belong to the range of the QRS wave, recording the information of the current list, and if the information of the previous list is not recorded, recording the information of the current list when the current list belongs to the range of the QRS wave; and if the current column does not belong to the range of the QRS wave, the information of the current column is not recorded, and the process is analogized until all columns of the picture are traversed to obtain the positions of the columns meeting the conditions, and the specific position for positioning the QRS wave is obtained by averaging every two adjacent columns.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911255275.9A CN110916645A (en) | 2019-12-10 | 2019-12-10 | QRS wave identification method combining wavelet transformation and image segmentation network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911255275.9A CN110916645A (en) | 2019-12-10 | 2019-12-10 | QRS wave identification method combining wavelet transformation and image segmentation network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110916645A true CN110916645A (en) | 2020-03-27 |
Family
ID=69858034
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911255275.9A Pending CN110916645A (en) | 2019-12-10 | 2019-12-10 | QRS wave identification method combining wavelet transformation and image segmentation network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110916645A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112817049A (en) * | 2020-12-28 | 2021-05-18 | 中海油田服务股份有限公司 | Method for calculating sound wave time difference |
CN114190950A (en) * | 2021-11-18 | 2022-03-18 | 电子科技大学 | Intelligent electrocardiogram analysis method and electrocardiograph for containing noise label |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19517138A1 (en) * | 1995-05-10 | 1996-11-14 | Manfred Dipl Phys Dr Joswig | Detecting P, T and QRS complexes in Holter electrocardiogram |
US20060052717A1 (en) * | 2004-08-31 | 2006-03-09 | Mugler Dale H | Method and system for evaluating cardiac ischemia |
CN105748066A (en) * | 2016-03-03 | 2016-07-13 | 深圳竹信科技有限公司 | Method and device for extracting feature point of electrocardiosignal waveform |
CN105769177A (en) * | 2016-04-01 | 2016-07-20 | 佛山市质量计量监督检测中心 | Electrocardiograph signal examination record analyzing device |
CN106063704A (en) * | 2016-05-27 | 2016-11-02 | 浙江好络维医疗技术有限公司 | A kind of QRS wave starting and terminal point localization method based on regularization least square recurrence learning |
CN107103182A (en) * | 2017-03-28 | 2017-08-29 | 南京医科大学 | A kind of heart disease Warning System and method based on deep learning algorithm |
CN107688553A (en) * | 2017-08-16 | 2018-02-13 | 安徽心之声医疗科技有限公司 | Method based on wavelet transformation and logistic regression algorithm detection ecg wave form feature |
CN107951485A (en) * | 2017-11-27 | 2018-04-24 | 乐普(北京)医疗器械股份有限公司 | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study |
CN108272451A (en) * | 2018-02-11 | 2018-07-13 | 上海交通大学 | A kind of QRS wave recognition methods based on improvement wavelet transformation |
CN109745033A (en) * | 2018-12-25 | 2019-05-14 | 东南大学 | Dynamic electrocardiogram method for evaluating quality based on time-frequency two-dimensional image and machine learning |
CN109864736A (en) * | 2019-03-22 | 2019-06-11 | 深圳市理邦精密仪器股份有限公司 | Processing method, device, terminal device and the medium of electrocardiosignal |
CN109886935A (en) * | 2019-01-28 | 2019-06-14 | 南京威翔科技有限公司 | A kind of road face foreign matter detecting method based on deep learning |
CN109893118A (en) * | 2019-03-05 | 2019-06-18 | 武汉大学 | A kind of electrocardiosignal classification diagnosis method based on deep learning |
US10426364B2 (en) * | 2015-10-27 | 2019-10-01 | Cardiologs Technologies Sas | Automatic method to delineate or categorize an electrocardiogram |
CN110353665A (en) * | 2019-08-14 | 2019-10-22 | 苏州大学 | ECG signal signature waveform detection method is singly led based on improved one-dimensional U-net |
-
2019
- 2019-12-10 CN CN201911255275.9A patent/CN110916645A/en active Pending
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19517138A1 (en) * | 1995-05-10 | 1996-11-14 | Manfred Dipl Phys Dr Joswig | Detecting P, T and QRS complexes in Holter electrocardiogram |
US20060052717A1 (en) * | 2004-08-31 | 2006-03-09 | Mugler Dale H | Method and system for evaluating cardiac ischemia |
US10426364B2 (en) * | 2015-10-27 | 2019-10-01 | Cardiologs Technologies Sas | Automatic method to delineate or categorize an electrocardiogram |
CN105748066A (en) * | 2016-03-03 | 2016-07-13 | 深圳竹信科技有限公司 | Method and device for extracting feature point of electrocardiosignal waveform |
CN105769177A (en) * | 2016-04-01 | 2016-07-20 | 佛山市质量计量监督检测中心 | Electrocardiograph signal examination record analyzing device |
CN106063704A (en) * | 2016-05-27 | 2016-11-02 | 浙江好络维医疗技术有限公司 | A kind of QRS wave starting and terminal point localization method based on regularization least square recurrence learning |
CN107103182A (en) * | 2017-03-28 | 2017-08-29 | 南京医科大学 | A kind of heart disease Warning System and method based on deep learning algorithm |
CN107688553A (en) * | 2017-08-16 | 2018-02-13 | 安徽心之声医疗科技有限公司 | Method based on wavelet transformation and logistic regression algorithm detection ecg wave form feature |
CN107951485A (en) * | 2017-11-27 | 2018-04-24 | 乐普(北京)医疗器械股份有限公司 | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study |
CN108272451A (en) * | 2018-02-11 | 2018-07-13 | 上海交通大学 | A kind of QRS wave recognition methods based on improvement wavelet transformation |
CN109745033A (en) * | 2018-12-25 | 2019-05-14 | 东南大学 | Dynamic electrocardiogram method for evaluating quality based on time-frequency two-dimensional image and machine learning |
CN109886935A (en) * | 2019-01-28 | 2019-06-14 | 南京威翔科技有限公司 | A kind of road face foreign matter detecting method based on deep learning |
CN109893118A (en) * | 2019-03-05 | 2019-06-18 | 武汉大学 | A kind of electrocardiosignal classification diagnosis method based on deep learning |
CN109864736A (en) * | 2019-03-22 | 2019-06-11 | 深圳市理邦精密仪器股份有限公司 | Processing method, device, terminal device and the medium of electrocardiosignal |
CN110353665A (en) * | 2019-08-14 | 2019-10-22 | 苏州大学 | ECG signal signature waveform detection method is singly led based on improved one-dimensional U-net |
Non-Patent Citations (1)
Title |
---|
RUNZE YU,YINGGUO GAO,XIAOHUI DUAN等: "QRS Detection and Measurement Method of ECG Paper Based on Convolutional Neural Networks", 《ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. IEEE ENGINEERING IN MEDICINE AND BIOLOGY SOCIETY. ANNUAL INTERNATIONAL CONFERENCE》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112817049A (en) * | 2020-12-28 | 2021-05-18 | 中海油田服务股份有限公司 | Method for calculating sound wave time difference |
CN112817049B (en) * | 2020-12-28 | 2022-08-02 | 中海油田服务股份有限公司 | Method for calculating sound wave time difference |
CN114190950A (en) * | 2021-11-18 | 2022-03-18 | 电子科技大学 | Intelligent electrocardiogram analysis method and electrocardiograph for containing noise label |
CN114190950B (en) * | 2021-11-18 | 2023-07-28 | 电子科技大学 | Electrocardiogram intelligent analysis method for noise-containing label and electrocardiograph |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108714026B (en) | Fine-grained electrocardiosignal classification method based on deep convolutional neural network and online decision fusion | |
CN110840402B (en) | Atrial fibrillation signal identification method and system based on machine learning | |
CN109171712B (en) | Atrial fibrillation identification method, atrial fibrillation identification device, atrial fibrillation identification equipment and computer readable storage medium | |
CA2976568C (en) | Methods and systems for disease analysis based on tranformations of diagnostic signals | |
CN109117730B (en) | Real-time electrocardiogram atrial fibrillation judgment method, device and system and storage medium | |
CN108647565B (en) | Data preprocessing method for classifying electrocardiosignals based on deep learning model | |
Wu et al. | On the adaptive detection of blood vessels in retinal images | |
CN104706349B (en) | Electrocardiosignal construction method based on pulse wave signals | |
CN111481192B (en) | Electrocardiosignal R wave detection method based on improved U-Net | |
CN111110228B (en) | Electrocardiosignal R wave detection method and device | |
CN109948396B (en) | Heart beat classification method, heart beat classification device and electronic equipment | |
CN111449644A (en) | Bioelectricity signal classification method based on time-frequency transformation and data enhancement technology | |
CN109124620B (en) | Atrial fibrillation detection method, device and equipment | |
CN108567418A (en) | A kind of pulse signal inferior health detection method and detecting system based on PCANet | |
CN112932498B (en) | T waveform state classification system with generalization capability based on deep learning | |
CN104905786A (en) | Electrocardiographic artifact on-line removal algorithm | |
CN111419213A (en) | ECG electrocardiosignal generation method based on deep learning | |
CN110916645A (en) | QRS wave identification method combining wavelet transformation and image segmentation network | |
CN114469124A (en) | Method for identifying abnormal electrocardiosignals in motion process | |
Guan et al. | Low-dimensional denoising embedding transformer for ECG classification | |
Peláez et al. | Ischemia classification via ECG using MLP neural networks | |
Tseng et al. | Sliding large kernel of deep learning algorithm for mobile electrocardiogram diagnosis | |
CN114652322A (en) | Electrocardiosignal classification method and system based on multi-domain feature learning | |
CN113486752B (en) | Emotion recognition method and system based on electrocardiosignal | |
Ziani | Contribution to Single-Channel Fetal Electrocardiogram Identification. |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200327 |