CN112587151B - 12-lead T wave extraction device fusing U-net network and filtering method - Google Patents
12-lead T wave extraction device fusing U-net network and filtering method Download PDFInfo
- Publication number
- CN112587151B CN112587151B CN202011252306.8A CN202011252306A CN112587151B CN 112587151 B CN112587151 B CN 112587151B CN 202011252306 A CN202011252306 A CN 202011252306A CN 112587151 B CN112587151 B CN 112587151B
- Authority
- CN
- China
- Prior art keywords
- wave
- peak position
- wave peak
- lead
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- 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
- A61B5/72—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)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Biophysics (AREA)
- Animal Behavior & Ethology (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Power Engineering (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention relates to a 12-lead T wave extraction method fusing a U-net network and a filtering method, which divides the prediction process of the T wave peak position into two parts, and comprehensively judges the results of the two parts to obtain the final predicted value. The first part obtains the results of 12T wave peak position vectors by using a U-net network model trained by each lead; the second part obtains the result of 12T wave peak position vectors by carrying out the same filtering mode on the electrocardio data of each lead; and (4) calculating the variance after the difference is made between the finally obtained 24T wave peak positions and the R wave peak position vector of each electrocardiosignal, wherein the minimum variance is the finally predicted T wave peak position.
Description
Technical Field
The application belongs to the technical field of electrocardiogram waveform extraction, and particularly relates to a 12-lead T wave extraction method fusing a U-net network and a filtering method.
Background
Various signs caused by abnormalities in the cardiac electrical conduction system are called arrhythmias. The electrocardiosignal is an important tool for detecting arrhythmia diseases and is an electric signal generated when the heart of a human body moves. The T wave is used as an important component of an electrocardiographic waveform, reflects the repolarization process of ventricular muscle, physiological and pathological factors can change the form of the T wave, and each T wave change has different clinical significance and contains rich pathological information.
At present, identification aiming at T wave abnormality is widely applied to diagnosis and detection of diseases such as myocardial infarction, coronary heart disease, hypertension and the like. Therefore, the position of the peak of the T wave can be timely and accurately detected and positioned, and the front and back wave forms can be analyzed according to the position, so that the method has important significance for diagnosing corresponding heart diseases.
There have been some studies on extraction algorithms for individual waveforms of an electrocardiogram, for example, extracting ST segments by applying a fixed time window to the electrocardiogram, but this extraction method is not accurate enough when the heart rate varies.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: in order to solve the problem of inaccurate electrocardiogram T wave extraction in the prior art, a 12-lead T wave extraction method which integrates a U-net network and a filtering method is provided.
In order to solve the technical problems, the invention provides a 12-lead T wave extraction device fusing a U-net network and a filtering method, which predicts the position of a T wave by respectively utilizing a U-net network model and filtering processing, calculates variance with the position of a R wave peak, comprehensively analyzes and compares the variance values obtained by the two prediction methods, and finally takes the position of the T wave peak corresponding to the minimum variance value as a final judgment result.
The technical scheme adopted by the invention for solving the technical problems is as follows:
the invention provides a 12-lead T wave extraction device fusing U-net network and filtering method, comprising:
the data acquisition module is used for collecting a plurality of 12-lead electrocardiosignals which contain different T-wave waveforms and represent different arrhythmia types, wherein the 12-lead electrocardiosignals are marked with T-wave peaks and R-wave peaks;
the data preprocessing module is used for marking N sampling points of each 12-lead electrocardiosignal according to the T wave peak position information marked on each electrocardiosignal and marking a data label which represents the T wave band position for the corresponding electrocardiosignal;
the network training module is used for forming the electrocardiosignal data of each lead setting data label into a training set and training the U-net network corresponding to the lead;
the network prediction module is used for acquiring 12-lead electrocardiosignal data to be detected in real time, preprocessing the 12-lead electrocardiosignal data according to S2, inputting the preprocessed data into a U-net network model corresponding to leads, outputting N x 1 label vectors corresponding to the 12-lead electrocardiosignals, and identifying all T wave bands of the 12-lead electrocardiosignals and T wave peak positions corresponding to the T wave bands so as to obtain a first T wave peak position vector Tloc _1 corresponding to each lead;
the filtering prediction module is used for filtering the preprocessed 12-lead electrocardiosignal data to be detected in real time according to S2, identifying the T wave peak position of the filtered 12-lead electrocardiosignal by utilizing R wave peak position information, and outputting a second T wave peak position vector Tloc _2 corresponding to each lead;
the first variance module is used for solving a variance var _1 after the difference is made between the Tloc _1 and the R wave peak position vector;
the second variance module is used for calculating a variance var _2 after the difference is made between the Tloc _2 and the R wave peak position vector;
the judging module is used for judging that the final T wave peak position is Tloc _1 if var _1 is smaller than var _ 2; and if var _1> var _2, the determined final T wave peak position is Tloc _ 2.
A second aspect of the present invention provides a computer storage medium having a computer program stored thereon, the computer program, when executed by a processor, is for implementing the processes performed by the extraction apparatus of the first aspect of the present invention.
The invention has the beneficial effects that: the T wave extraction method of the invention integrates U-net network and filtering, and the prediction process of the wave peak position of the 12-lead T wave is divided into two parts, wherein the first part obtains the results of 12T wave peak position vectors by using the U-net network model trained by each lead; the second part obtains the result of 12T wave peak position vectors by carrying out the same filtering mode on the electrocardio data of each lead; and (4) calculating the variance after the difference is made between the finally obtained 24T wave peak positions and the R wave peak position vector of each electrocardiosignal, wherein the minimum variance is the finally predicted T wave peak position. The extraction method can accurately identify the wave peak of the T wave and extract the T wave in the electrocardiogram.
Drawings
The technical solution of the present application is further explained below with reference to the drawings and the embodiments.
FIG. 1 is a flow chart of the operation of an extraction device according to an embodiment of the present application;
FIG. 2 is a detailed structural diagram of an extraction apparatus according to an embodiment of the present application;
fig. 3 is a diagram of a U-net network structure according to an embodiment of the present application.
Detailed Description
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
The technical solutions of the present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
A first embodiment of the present invention provides a 12-lead T-wave extraction device that merges a U-net network and a filtering method, including:
the data acquisition module is used for collecting a plurality of 12-lead electrocardiosignals which contain different T-wave waveforms and represent different arrhythmia types, wherein the 12-lead electrocardiosignals are marked with T-wave peaks and R-wave peaks;
the data preprocessing module is used for marking N sampling points of each 12-lead electrocardiosignal according to the T wave peak position information marked on each electrocardiosignal and marking a data label which represents the T wave band position for the corresponding electrocardiosignal;
the network training module is used for forming the electrocardiosignal data of each lead setting data label into a training set and training the U-net network corresponding to the lead;
the network prediction module is used for acquiring 12-lead electrocardiosignal data to be detected in real time, preprocessing the 12-lead electrocardiosignal data according to S2, inputting the preprocessed data into a U-net network model corresponding to leads, outputting N x 1 label vectors corresponding to the 12-lead electrocardiosignals, and identifying all T wave bands of the 12-lead electrocardiosignals and T wave peak positions corresponding to the T wave bands so as to obtain a first T wave peak position vector Tloc _1 corresponding to each lead;
the filtering prediction module is used for filtering the preprocessed 12-lead electrocardiosignal data to be detected in real time according to S2, identifying the T wave peak position of the filtered 12-lead electrocardiosignal by utilizing R wave peak position information, and outputting a second T wave peak position vector Tloc _2 corresponding to each lead;
the first variance module is used for solving a variance var _1 after the difference is made between the Tloc _1 and the R wave peak position vector;
the second variance module is used for calculating a variance var _2 after the difference is made between the Tloc _2 and the R wave peak position vector;
the judging module is used for judging that the final T wave peak position is Tloc _1 if var _1 is smaller than var _ 2; and if var _1> var _2, the determined final T wave peak position is Tloc _ 2.
In the embodiment, 10s electrocardiographic signal data of clinical rest 12 leads, of which nearly 5000 known R-wave peak positions and T-wave peak positions are collected, and the sampling frequency of the electrocardiographic signal is 500 Hz. The length of the acquired electrocardiogram data is 10s, the clinical general measurement time is 1 minute, and a doctor selects a signal with less interference and good quality for diagnosis within 10 seconds. The sampling frequency is 500Hz, if not 500Hz, resampling to 500Hz is needed. The 12 leads are limb leads I, II and III, and the compression limb leads aVR, aVL and aVF and the chest leads V1-V6.
The acquired electrocardiosignal data comprise 12-lead electrocardiosignal data of different T-wave waveforms and different diseases, and the T-wave waveforms comprise waveform types of normal T-wave, low and flat T-wave, high and sharp T-wave, inverted T-wave, cut track T-wave and bidirectional T-wave.
After data acquisition is completed, data preprocessing is required, 5000 sampling points of each electrocardiosignal of each lead are marked through a data preprocessing process, each electrocardiosignal obtains a data label consisting of 5000 numbers, and the position of a T wave band can be identified according to the marking data of each sampling point in the data label.
And forming a training set for the preprocessed 12-lead electrocardiogram data, and training 12U-net networks corresponding to the 12 leads.
For 12-lead electrocardiogram data acquired in real time, after preprocessing, 5000 × 1 12-lead electrocardiogram data are obtained, and are respectively input into a U-net network corresponding to leads, 5000 × 1 label vectors corresponding to the 12 leads are output, and according to the data of the label vectors, all T wave bands of the 12-lead electrocardiogram data and T wave peak positions corresponding to the T wave bands are identified, so that 12 first T wave peak position vectors Tloc _1 corresponding to the 12 leads are obtained.
And performing filtering processing on the 12-lead electrocardiogram data with 5000 × 1 obtained after preprocessing, and identifying the T-wave peak positions by using the R-wave peak position information to obtain 12 second T-wave peak position vectors Tloc _2 corresponding to the 12 leads.
Calculating the variance after respectively making differences between the 12 first T wave peak position vectors Tloc _1 and the R wave peak position vector; calculating the variance after respectively subtracting the 12 second T wave peak position vectors Tloc _2 and the R wave peak position vector; and comprehensively analyzing and comparing the 24 variance values, and finally taking the T wave peak position corresponding to the minimum variance value as a final judgment result.
Optionally, the preprocessing module in this embodiment includes:
the filtering unit is used for carrying out wavelet filtering on the collected electrocardiosignals of each lead;
and the data marking unit is used for marking sampling points which correspond to the T wave peak positions towards the left 50 and the right 50 as 1 according to the T wave peak position information marked by the electrocardiosignals of each lead, and setting a data label which represents the T wave band position for each electrocardiosignal of each lead, wherein the rest sampling points are defaulted to be 0.
In this embodiment, each 10s electrocardiographic signal needs to be preprocessed, and a filtering unit is used for performing wavelet filtering; and marking each of 5000 points of each electrocardiosignal by using a data marking unit. The implementation manner of the data marking unit in this embodiment is as follows: according to the T wave peak position information marked by the electrocardiosignals, sampling points of the T wave peak position towards the left 50 and the right 50 are marked as 1, and the other points are defaulted to be 0.
For example, one cardiac signal is labeled with 5000 digital components [ 000000000000000111111111.. 1111000000.. 1111.. 11110000 ], where 111.. indicates that the segment is a band in which T waves exist; denote points other than the T-wave band (non-T-band). In order to facilitate the model training, the labels are subjected to the one-hot operation, the label with the length of 5000 corresponding to each electrocardiosignal is changed into a (5000,2) form, and 5000 rows and 2 columns are provided, wherein each row represents the type of one point.
Optionally, the network prediction module includes:
the T-wave band identification unit is used for judging that a wave band formed by the continuous set number of sampling points is a T-wave band if the label values of m continuous sampling points are greater than a set threshold value in the output N x 1 label vector, and m is greater than or equal to 50;
and the first T wave peak position identification unit is used for taking a midpoint of each T wave band, and the midpoint position is the peak position of the corresponding T wave band.
Taking the I lead as an example, the network prediction module is implemented as follows:
inputting a clinical I-lead electrocardiosignal to be detected for 10s into a trained I-lead U-net model, outputting a label vector of 5000 x 1 by the I-lead U-net model, and if the label data values (continuous 0.1s) of continuous 50 sampling points in the output label vector are greater than 0.5, determining that the wave band where the 50 sampling points are located is a T-wave band, taking a midpoint of each T-wave band in the T-wave bands, wherein the position corresponding to each midpoint is the position of each T-wave peak.
For example: the tag vector predicted by the U-net network model is [ 00000000.560.620.750.880.980.980.981111111.. 0.620.420000.. 00000.630.790.950.980.990.990.99111111.. 0.780.320.100000000.. wherein 50 consecutive points from 0.56-0.62 are greater than 0.5, the segment is considered to be a T-wave band, and assuming that there are 68 points from 0.56-0.42, the 34 th point is found where the point is located in the band, and this 34 th point is indexed 368 in a vector of 5000 lengths, i.e., 368 is the location of the first T-wave peak of the cardiac electrical signal.
By analogy, all the T wave peak positions of the 10s electrocardiosignals can be found to form a T wave peak position vector Tloc _ 1.
Optionally, in S5, the filtering prediction module includes:
the filtering unit is used for firstly carrying out 0.06s window median filtering; secondly, performing median filtering in a 0.06s window; then carrying out 0.1s mean value filtering;
and the second T wave peak position identification unit is used for marking the position of the maximum value which appears for the first time within 0.4s after the position of each R wave peak as the T wave peak position.
The filtering prediction module of this embodiment performs, by the filtering unit, first 0.06s window median filtering, then performs 0.06s window median filtering, and then performs 0.1s mean filtering.
And marking the position of the maximum value which appears for the first time within 0.4s (namely 200 sampling points after the R wave crest) after the position of each R wave crest as the position of each T wave crest by a second T wave crest position identification unit.
For example: the position of a certain R wave crest given by a doctor is 162, the size of 200 sampling points after the 162 th sampling point is judged, and the position of the point where the maximum value appears for the first time is found and recorded as the position of the T wave crest.
Optionally, the first variance module includes:
the first calculation unit is used for correspondingly differentiating the value in the R wave crest position vector and the value in the first T wave crest position vector Tloc _1 one by one to obtain a new vector if the number of elements in the first T wave crest position vector Tloc _1 is equal to the number of elements in the R wave crest position vector, and then solving the variance of the new vector to be marked as var _ 1;
the second calculation unit is used for respectively searching and storing element values which are larger than Ri and closest to Ri in the Tloc _1 for each element value Ri in the R wave peak position vector if the number of elements in the first T wave peak position vector Tloc _1 is larger than that of elements in the R wave peak position vector; the saved element values in Tloc _1 form a corrected T wave peak position vector, and then the variance var _1 is solved by adopting the method of S61;
a third calculating unit, configured to, if the number of elements in the first T-wave peak position vector Tloc _1 is smaller than the number of elements in the R-wave peak position vector, respectively search and store an element value that is smaller than Ti and closest to Ti in the R-wave peak position vector for each element value Ti in the T-wave peak position vector; the element values of the stored R-wave peak position vector constitute a corrected R-wave peak position vector, and then the variance var _1 is obtained by using the method of S61.
Tloc _1 in this embodiment is a first T-wave peak position vector finally determined by the U-net network model.
Through a first calculating unit, if the number of element values in the Tloc _1 is judged to be equal to the number of element values in the R wave crest position vector given by a doctor, the element value in each R wave crest position vector is subtracted by the element value in the corresponding first T wave crest position vector Tloc _1 to obtain a new vector, then the vector is subjected to variance calculation and is marked as var _1, and finally 12 variance values var _1 are obtained;
through the second calculating unit, if the number of the element values in the Tloc _1 is larger than the number of the element values in the R-wave peak position vector given by the doctor, deleting the element in the Tloc _1, wherein the deleting method comprises the following steps:
position R of each R wave peak in cyclic R wave peak position vectoriLooking for greater than R in Tloc _1iAnd is closest to RiThe value of (2) is saved. For example:
the R-wave peak position vector is [ 536136919372338 ], Tloc _1 is [ 6541523213425362912 ], and for the first element 536 in the R-wave peak position vector, greater than 536 in Tloc _1 and 654, the value closest to 536, is saved;
for the second element 1369 in the R-wave peak position vector, the value greater than 1369 in Tloc _1 and closest to 1369 is 1523, which is saved;
and sequentially looping, finishing searching each element in the R wave peak position vector, and finally obtaining the corrected Tloc _1 as [ 654152321342536 ].
And (4) solving the variance between the new Tloc _1 obtained after correction and the R wave peak position vector given by the doctor by adopting the method in (1) and recording the variance as var _ 1.
Through a third calculation unit, if the number of the element values in the Tloc _1 is less than the number of the element values in the R wave peak position vector given by the doctor, deleting the R wave peak position vector, wherein the deleting method comprises the following steps:
and circulating the wave peak position Ti of each T wave, finding out the value which is smaller than the Ti in the wave peak position vector of the R wave and is closest to the Ti, storing and deleting the rest. For example: the R wave peak position is [ 5361369193723382742 ], Tloc _1 is [ 654152321342536 ], if Ti is 654, the value which is smaller than 654 and closest to 654 is 536 is found in the R wave peak position vector, and the process is circulated in sequence until Ti is 2536, the value which is smaller than 2536 in the R wave peak position vector is 536,1369,1937,2338, but the closest is 2338, so 2338 is saved, and finally the extra points 2742 in the R wave peak position vector are deleted. And finally, performing difference on the Tloc _1 vector and the corrected R wave peak position vector given by the doctor in the same step, and solving the variance to be marked as var _ 1.
Optionally, a specific implementation manner of the second variance module is as follows:
and performing one-to-one correspondence difference between the value in the R wave peak position vector and the value in Tloc _2 to obtain a new vector, and calculating the variance of the new vector and recording the variance as var _ 2.
Because the Tloc _2 vector is determined by the maximum value appearing at 200 points after each R wave position, the situation that the number of the Tloc _2 vector is different from that of the R wave peak position vector does not exist, therefore, the second variance module directly uses the element value in each Tloc _2 corresponding to the element value in each R wave peak vector to obtain a new vector, and then calculates the variance of the new vector to be marked as var _ 2.
Optionally, as shown in fig. 3, the input layer of the U-net network is sequentially connected to the first convolution layer and the second convolution layer;
the output of the second convolution layer is the input of the first lower sampling layer, and the first lower sampling layer is sequentially connected with the third convolution layer and the fourth convolution layer;
the output of the fourth convolution layer is the input of a second down-sampling layer, and the second down-sampling layer is sequentially connected with the fifth convolution layer and the sixth convolution layer;
the output of the sixth convolution layer is the input of a third down-sampling layer, and the third down-sampling layer is sequentially connected with a seventh convolution layer and an eighth convolution layer;
the eighth convolution layer is connected with the first up-sampling layer, and the first up-sampling layer is sequentially connected with the ninth convolution layer and the tenth convolution layer;
the output of the tenth convolutional layer is the input of a second upsampling layer, and the second upsampling layer is sequentially connected with the eleventh convolutional layer and the twelfth convolutional layer;
the output of the twelfth convolutional layer is the input of the third upsampling layer, and the output of the third upsampling layer is sequentially connected with the thirteenth convolutional layer and the fourteenth convolutional layer, wherein the fourteenth convolutional layer is the output layer.
In the embodiment, the U-net network inputs 5000 × 1I-lead electrocardiographic signal data and outputs 5000 × 1 vectors, and the U-net network adopts a symmetrical U-shaped structure.
The down-sampling layer adopts a maximum pooling method, and the up-sampling layer adopts bilinear interpolation (upsamplle) and convolution to learn the characteristics of higher layers.
The number of filters in the first convolutional layer, the second convolutional layer, the third convolutional layer, the fourth convolutional layer, the fifth convolutional layer, the sixth convolutional layer, the seventh convolutional layer, the eighth convolutional layer, the ninth convolutional layer, the tenth convolutional layer, the eleventh convolutional layer, the twelfth convolutional layer, the thirteenth convolutional layer and the fourteenth convolutional layer is six, twelve, twenty-four, forty-eight, twenty-four, twelve, six and one in sequence. The convolution kernel sizes of the first convolution layer, the second convolution layer, the third convolution layer, the fourth convolution layer, the fifth convolution layer, the sixth convolution layer, the seventh convolution layer, the eighth convolution layer, the ninth convolution layer, the tenth convolution layer, the eleventh convolution layer, the twelfth convolution layer, the thirteenth convolution layer and the fourteenth convolution layer are all (3 x 1). The step sizes of the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth, eleventh, twelfth, thirteenth, and fourteenth convolutional layers are all 1, and the filling modes are all "same". Excitation functions of the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth, eleventh, twelfth, and thirteenth convolutional layers are all ReLU functions, and an excitation function of the fourteenth convolutional layer is a Softmax function. The first downsampling layer, the second downsampling layer and the third downsampling layer all adopt a maximum pooling method, and the pooling windows of the first downsampling layer, the second downsampling layer and the third downsampling layer are all set to be 2. The number of filters in the first upper sampling layer, the second upper sampling layer and the third upper sampling layer during convolution operation is twenty-four, twelve and six in sequence, the sizes of convolution kernels in the first upper sampling layer, the second upper sampling layer and the third upper sampling layer during convolution operation are all (2 x 1), the step lengths of convolution operations in the first upper sampling layer, the second upper sampling layer and the third upper sampling layer are all 1, the filling mode is 'same', and excitation functions in the first upper sampling layer, the second upper sampling layer and the third upper sampling layer during convolution operation are ReLU functions.
The loss functions all use the categorical _ crosssentryp. The training algorithm may be: a random gradient descent algorithm, an Adam algorithm, a RMSProp algorithm, an adagard algorithm, an adapelta algorithm, an Adamax algorithm, and the like.
The U-net network model of the embodiment can fuse low-layer features and high-layer features, and is beneficial to better learning the features. And the U-net network model is an automatic coding decoder, which can ensure the same input and output lengths, and is convenient for visually judging the position of the T wave in the whole section of electrocardiosignals.
In the embodiment, the prediction process of the T wave peak position is divided into two parts, and the results of the two parts are comprehensively judged to obtain the final predicted value. The first part obtains the results of 12T wave peak position vectors by using a U-net network model trained by each lead; the second part obtains the result of 12T wave peak position vectors by carrying out the same filtering mode on the electrocardio data of each lead; and (4) calculating the variance after the difference is made between the finally obtained 24T wave peak positions and the R wave peak position vector of each electrocardiosignal, wherein the minimum variance is the finally predicted T wave peak position.
The 12-lead T wave extraction algorithm fusing the Unet network and the filtering method provided by the embodiment can identify the T wave peak more accurately.
The second embodiment of the present invention also provides a computer storage medium, on which a computer program is stored, which, when being executed by a processor, is used for implementing the execution procedure of the apparatus according to the first embodiment of the present invention.
In light of the foregoing description of the preferred embodiments according to the present application, it is to be understood that various changes and modifications may be made without departing from the spirit and scope of the invention. The technical scope of the present application is not limited to the contents of the specification, and must be determined according to the scope of the claims.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Claims (6)
1. A12-lead T wave extraction device fusing a U-net network and a filtering method is characterized by comprising the following steps:
the data acquisition module is used for collecting a plurality of 12-lead electrocardiosignals which contain different T-wave waveforms and represent different arrhythmia types, wherein the 12-lead electrocardiosignals are marked with T-wave peaks and R-wave peaks;
a data preprocessing module for executing step S2, wherein step S2 is: marking N sampling points of each electrocardiosignal of 12 leads according to the T wave peak position information marked on each electrocardiosignal, and marking a data label representing the T wave band position for the corresponding electrocardiosignal;
the network training module is used for forming the electrocardiosignal data of each lead setting data label into a training set and training the U-net network corresponding to the lead;
the network prediction module is used for acquiring 12-lead electrocardiosignal data to be detected in real time, preprocessing the 12-lead electrocardiosignal data according to S2, inputting the preprocessed data into a U-net network model corresponding to leads, outputting N x 1 label vectors corresponding to the 12-lead electrocardiosignals, and identifying all T wave bands of the 12-lead electrocardiosignals and T wave peak positions corresponding to the T wave bands so as to obtain a first T wave peak position vector Tloc _1 corresponding to each lead;
the filtering prediction module is used for filtering the preprocessed 12-lead electrocardiosignal data to be detected in real time according to S2, identifying the T wave peak position of the filtered 12-lead electrocardiosignal by utilizing R wave peak position information, and outputting a second T wave peak position vector Tloc _2 corresponding to each lead;
the first variance module is used for solving a variance var _1 after the difference is made between the Tloc _1 and the R wave peak position vector; the first variance module comprises:
the first calculation unit is used for correspondingly differentiating the value in the R wave crest position vector and the value in the first T wave crest position vector Tloc _1 one by one to obtain a new vector if the number of elements in the first T wave crest position vector Tloc _1 is equal to the number of elements in the R wave crest position vector, and then solving the variance of the new vector to be marked as var _ 1;
a second calculation unit for calculating R for each element value in the R wave peak position vector if the number of elements in the first T wave peak position vector Tloc _1 is greater than the number of elements in the R wave peak position vectoriRespectively searching for greater than R in Tloc _1iAnd is closest to RiAnd saving the element values; the saved element values in Tloc _1 form a corrected T wave peak position vector, and then the corrected T wave peak position vector is differed with the R wave peak position vector to obtain a variance var _ 1;
a third calculating unit for calculating a value T of each element in the T-wave peak position vector if the number of elements in the first T-wave peak position vector Tloc _1 is less than the number of elements in the R-wave peak position vectoriRespectively searching element values which are smaller than Ti and closest to Ti in the R wave peak position vector, and storing the element values; the element values of the stored R wave peak position vector form a corrected R wave peak position vector, and then the corrected R wave peak position vector is differed with the R wave peak position vector to obtain a variance var _ 1;
the second variance module is used for calculating a variance var _2 after the difference is made between the Tloc _2 and the R wave peak position vector; the specific implementation manner of the second variance module is as follows:
making a difference between the value in the R wave peak position vector and the value in Tloc _2 in a one-to-one correspondence manner to obtain a new vector, and solving the variance of the new vector to be marked as var _ 2;
the judging module is used for judging that the final T wave peak position is Tloc _1 if var _1 is smaller than var _ 2; and if var _1> var _2, the determined final T wave peak position is Tloc _ 2.
2. The U-net network and filtering method converged 12-lead T-wave extraction device according to claim 1, wherein the network prediction module comprises:
the T-wave band identification unit is used for judging that a wave band formed by a set number of sampling points is a T-wave band if the label values of m continuous sampling points in the output N x 1 label vector are greater than a set threshold, and m is greater than or equal to 50;
and the first T wave peak position identification unit is used for taking a midpoint of each T wave band, and the midpoint position is the peak position of the corresponding T wave band.
3. The U-net network and filtering method fused 12-lead T-wave extraction device according to claim 1, wherein the filtering prediction module comprises:
the filtering unit is used for firstly carrying out 0.06s window median filtering; secondly, performing median filtering in a 0.06s window; then carrying out 0.1s mean value filtering;
and the second T wave peak position identification unit is used for marking the position of the maximum value which appears for the first time within 0.4s after the position of each R wave peak as the T wave peak position.
4. The U-net network and filtering method converged 12-lead T-wave extraction device according to claim 1, wherein the pre-processing module comprises:
the filtering unit is used for carrying out wavelet filtering on the collected electrocardiosignals of each lead;
and the data marking unit is used for marking sampling points which correspond to the T wave peak positions towards the left 50 and the right 50 as 1 according to the T wave peak position information marked by the electrocardiosignals of each lead, and setting a data label which represents the T wave band position for each electrocardiosignal of each lead, wherein the rest sampling points are defaulted to be 0.
5. The U-net network and filtering method integrated 12-lead T-wave extracting apparatus according to claim 1, wherein an input layer of the U-net network is sequentially connected to a first convolutional layer and a second convolutional layer;
the output of the second convolution layer is the input of the first lower sampling layer, and the first lower sampling layer is sequentially connected with the third convolution layer and the fourth convolution layer;
the output of the fourth convolution layer is the input of a second down-sampling layer, and the second down-sampling layer is sequentially connected with the fifth convolution layer and the sixth convolution layer;
the output of the sixth convolution layer is the input of a third down-sampling layer, and the third down-sampling layer is sequentially connected with a seventh convolution layer and an eighth convolution layer;
the eighth convolution layer is connected with the first up-sampling layer, and the first up-sampling layer is sequentially connected with the ninth convolution layer and the tenth convolution layer;
the output of the tenth convolutional layer is the input of a second upsampling layer, and the second upsampling layer is sequentially connected with the eleventh convolutional layer and the twelfth convolutional layer;
the output of the twelfth convolutional layer is the input of the third upsampling layer, and the output of the third upsampling layer is sequentially connected with the thirteenth convolutional layer and the fourteenth convolutional layer, wherein the fourteenth convolutional layer is the output layer.
6. A computer storage medium having a computer program stored thereon, wherein the computer program, when being executed by a processor, is adapted to carry out the T-wave extraction step using the apparatus according to any of the claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011252306.8A CN112587151B (en) | 2020-11-11 | 2020-11-11 | 12-lead T wave extraction device fusing U-net network and filtering method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011252306.8A CN112587151B (en) | 2020-11-11 | 2020-11-11 | 12-lead T wave extraction device fusing U-net network and filtering method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112587151A CN112587151A (en) | 2021-04-02 |
| CN112587151B true CN112587151B (en) | 2022-04-12 |
Family
ID=75183364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011252306.8A Active CN112587151B (en) | 2020-11-11 | 2020-11-11 | 12-lead T wave extraction device fusing U-net network and filtering method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112587151B (en) |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104921722A (en) * | 2015-07-15 | 2015-09-23 | 西南民族大学 | Double-lead combined with electrocardiogram QRS wave detection method |
| US10321837B2 (en) * | 2017-04-28 | 2019-06-18 | Biosense Webster (Israel) Ltd. | ECG machine including filter for feature detection |
| CN108309284A (en) * | 2018-03-15 | 2018-07-24 | 广州视源电子科技股份有限公司 | ECG T wave end-point detection method and device |
| CN109620206B (en) * | 2018-12-26 | 2020-07-07 | 上海数创医疗科技有限公司 | Atrial fibrillation artificial intelligence identification method and device with ectopic heartbeat judgment function |
| CN110353665B (en) * | 2019-08-14 | 2022-03-11 | 苏州大学 | Single lead electrocardiogram signal characteristic waveform detection method based on improved one-dimensional U-net |
| CN111481192B (en) * | 2020-04-22 | 2022-06-03 | 重庆邮电大学 | Electrocardiosignal R wave detection method based on improved U-Net |
| CN111657915A (en) * | 2020-04-30 | 2020-09-15 | 上海数创医疗科技有限公司 | Electrocardiogram form recognition model based on deep learning and use method thereof |
-
2020
- 2020-11-11 CN CN202011252306.8A patent/CN112587151B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112587151A (en) | 2021-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11134880B2 (en) | Automatic method to delineate or categorize an electrocardiogram | |
| JP7091451B2 (en) | Automatic ECG analysis method based on artificial intelligence self-learning, equipment used to execute the analysis method, computer program products and computer-readable storage media | |
| EP3367897B1 (en) | An automatic method to delineate or categorize an electrocardiogram | |
| CN110236520B (en) | Electrocardiogram type recognition device based on double convolution neural network | |
| CN108294745B (en) | Method and system for detecting P wave and T wave start and stop points in multi-lead electrocardiogram signal | |
| CN106815570B (en) | Electrocardiosignal ST-T segment identification method based on dynamic pattern identification | |
| US11331034B2 (en) | Automatic method to delineate or categorize an electrocardiogram | |
| CN115470828A (en) | Multi-lead electrocardiogram classification and identification method based on convolution and self-attention mechanism | |
| CN110717416B (en) | Neural network training method for ST segment classification recognition based on feature selection | |
| CN115337018B (en) | Electrocardiogram signal classification method and system based on overall dynamic characteristics | |
| CN112932498A (en) | T wave morphology classification system with strong generalization capability based on deep learning | |
| CN110693488A (en) | Electrocardiosignal processing system, electrocardiosignal processing method, electronic equipment and readable storage medium | |
| CN109770891B (en) | Electrocardiosignal preprocessing method and preprocessing device | |
| Srivastava et al. | Channel self-attention deep learning framework for multi-cardiac abnormality diagnosis from varied-lead ECG signals | |
| CN112587151B (en) | 12-lead T wave extraction device fusing U-net network and filtering method | |
| KR102216047B1 (en) | Method and apparatus for classifying cardiac arrhythmia using an auto associative neural network | |
| CN112587152B (en) | 12-lead T wave extraction method fusing U-net network and filtering method | |
| CN116226724A (en) | Non-invasive coronary heart disease detection system | |
| CN112587149A (en) | Atrial premature beat target detection device based on convolutional neural network | |
| Escrivães et al. | ECG classification using Artificial Intelligence: Model Optimization and Robustness Assessment | |
| CN117421678B (en) | Single-lead atrial fibrillation recognition system based on knowledge distillation | |
| CN111685754B (en) | Heart rate calculation method and system for wearable ECG acquisition equipment | |
| CN112587150B (en) | Electrocardiosignal type identification method and device by using fuzzification similarity template | |
| CN112587146B (en) | Heart rhythm type identification method and device based on neural network with improved loss function | |
| CN112587147A (en) | Atrial premature beat target detection method based on convolutional neural network |
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 |