CN114176602B - Method for simultaneously positioning electrocardiograph P wave, QRS wave and T wave based on deep learning multi-target detection - Google Patents
Method for simultaneously positioning electrocardiograph P wave, QRS wave and T wave based on deep learning multi-target detection Download PDFInfo
- Publication number
- CN114176602B CN114176602B CN202111662087.5A CN202111662087A CN114176602B CN 114176602 B CN114176602 B CN 114176602B CN 202111662087 A CN202111662087 A CN 202111662087A CN 114176602 B CN114176602 B CN 114176602B
- Authority
- CN
- China
- Prior art keywords
- wave
- qrs
- data
- deep learning
- target detection
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 80
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000013135 deep learning Methods 0.000 title claims abstract description 41
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 18
- 238000004364 calculation method Methods 0.000 claims description 23
- 238000000605 extraction Methods 0.000 claims description 20
- 238000012545 processing Methods 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 7
- 230000001629 suppression Effects 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 238000002156 mixing Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 5
- 208000024172 Cardiovascular disease Diseases 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000004807 localization Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 206010003119 arrhythmia Diseases 0.000 description 1
- 230000006793 arrhythmia Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000002565 electrocardiography Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 208000031225 myocardial ischemia Diseases 0.000 description 1
- 230000003183 myoelectrical effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
-
- 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
- A61B5/353—Detecting P-waves
-
- 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
- A61B5/355—Detecting T-waves
-
- 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
-
- 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
- A61B5/7235—Details of waveform analysis
-
- 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
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- 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)
- Cardiology (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Molecular Biology (AREA)
- Veterinary Medicine (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a method for simultaneously positioning the P wave, the QRS wave and the T wave positions based on deep learning multi-target detection, solves the problem that the P wave and the T wave are required to be detected by relying on the QRS wave in the traditional algorithm, can directly obtain the starting and ending positions of the P wave, the QRS wave and the T wave, and compared with the scheme of the currently disclosed deep learning heart beat detection, the method completes the simultaneous detection of the P wave and the T wave, and aims at the scheme of partial deep learning heart beat detection.
Description
Technical Field
The invention relates to the technical field of artificial intelligence one-dimensional data and the technical field of medical electrocardiograph detection, in particular to a detection method for simultaneously positioning the positions of an electrocardiograph signal P wave, a QRS wave and a T wave based on deep learning target detection.
Background
Electrocardiogram is the simplest and effective tool for diagnosing cardiovascular diseases, and various cardiovascular diseases such as arrhythmia, myocardial ischemia and the like can be diagnosed by using the electrocardiosignal. The electrocardio heart beat mainly comprises a P wave, a QRS wave group and a T wave, different pathological states of the P wave, the QRS wave and the T wave have different characteristics, and the different characteristics reflect different cardiovascular diseases, so that the detection and the identification of the P wave, the QRS wave and the T wave are particularly important. The results of automatic electrocardiography are typically reviewed and corrected by clinicians with the aid of electrocardiography software. If the accuracy of electrocardio identification can be improved, the workload of doctors is greatly reduced, which is beneficial to the accurate screening of cardiovascular diseases.
At present, the traditional electrocardiosignal P wave, QRS wave and T wave identification and positioning method is to artificially design a filter to perform noise reduction and enhancement treatment on the electrocardiosignal according to the basic rules of the electrocardiosignal and the electrophysiological activity of the heart. The artificial features are designed through differential calculation, area calculation and the like, a proper threshold is set through an empirical value, and the P wave and the T wave are detected after the QRS wave is detected and then the position of the QRS wave is relied on. However, the physical condition states of each person are different, so that the electrocardiosignal heart beats have individual differences, and the electrocardiosignals possibly have noise interference in the acquisition process, so that the traditional signal analysis method is difficult to consider all conditions based on the characteristics of manual design, has low self-adaptability, and reduces the accuracy of heart beat detection when the electrocardiosignals outside the characteristics of manual design appear.
The deep learning has the characteristics of extremely strong autonomous learning capability and high nonlinear mapping, and can learn electrocardiosignal characteristics in a large amount of training data, and the self-adaptability is high, so that the method provides possibility for designing complex high-precision electrocardiosignal P wave, QRS wave and T wave detection and identification models. At present, some researches realize the vertex detection of the QRS wave by using an artificial intelligence technology, and realize the detection of the starting and ending positions of the QRS wave by using a disclosed deep learning target detection algorithm, and an electrocardiosignal characteristic extraction part also uses a disclosed characteristic extraction network such as AlexNet, VGG, resNet, squeezeNet, mobileNet and the like. Because the time window of the P wave is shorter than that of the QRS wave, the P wave characteristics can be weakened or vanished after the electrocardio data is compressed by using the characteristic extraction network, and the P wave detection is not facilitated.
Disclosure of Invention
In view of the foregoing, it is an object of a first aspect of the present invention to provide a method for simultaneous localization of the position of an electrocardiographic P-wave, QRS-wave, T-wave based on deep learning multi-target detection.
The object of the first aspect of the present invention is achieved by the following technical solutions:
the method for simultaneously positioning the positions of the P wave, the QRS wave and the T wave of the electrocardio based on deep learning multi-target detection comprises the following steps:
step S1: collecting electrocardiosignals, and carrying out noise reduction treatment on the electrocardiosignals;
Step S2: intercepting electrocardiograph data for a set time length, ensuring that an electrocardiograph sample data set contains different conditions of all electrocardiograph data P waves, QRS waves and T waves, and marking the starting and ending positions of the electrocardiograph data set to obtain a marked data set;
Step S3: performing prior frame processing in deep learning target detection on the intercepted and marked sample data, wherein the width of the prior frame is related to the widths of the P wave, the QRS wave and the T wave, so that a plurality of prior frames are generated on the sample data; according to the deep learning target detection, converting the prior frame and the positions of the actual marked P wave, QRS wave and T wave into the relation of offset, thereby obtaining a data set required by model training;
step S4: a feature extraction network for use in heart beat detection identification using deep learning object detection;
Step S5: inputting the training set data manufactured in the step S3 into a model according to a suggestion frame detection algorithm in a deep learning target detection algorithm, and distinguishing which parts in electrocardiograph data with set time length are suggestion frames containing P wave, QRS wave and T wave positions through regression and classification operation of the model;
step S6: performing suggested frame offset dataset processing;
Step S7: carrying out secondary processing on the suggestion frames, carrying out cross-correlation calculation on the suggestion frames obtained in the last step and the positions of the true marked P waves, QRS waves and T waves, taking the suggestion frames which are larger than a certain threshold value as positive samples, marking P waves, QRS waves and T wave categories of the positive samples, taking negative samples which are smaller than a certain threshold value as negative samples, and encoding positive and negative sample suggestion frames into coefficient relations of offset with the true frames to be used as classification layer sample data;
step S8: inputting the output characteristics obtained through the characteristic extraction network and the classified sample data in the step S7 into a classification layer, intercepting the extracted characteristics through a proposal frame of secondary processing, and inputting the classification layer after changing the length of the proposal frame into the same size because the lengths of the proposal frames are different, thereby obtaining the predicted frame offset and the corresponding predicted frame category through the classification layer;
step S9: finally, the electrocardiographic data are predicted through a model to obtain the starting and ending positions and categories of the P wave, the QRS wave and the T wave.
Further, in the step S4, the feature extraction network uses a ResNet structure;
furthermore, in the step S4, a pyramid-like network with feature layers superimposed up and down is used to realize an electrocardiographic data feature extraction network, two feature layers with different sizes are constructed, the set priori frame is ensured to contain the positions of the actually marked P wave, QRS wave and T wave, and the electrocardiographic data within the intercepted set time length is input into the feature extraction network to obtain electrocardiographic data features.
In step S6, according to the situation that there are multiple repetitions or overlaps between the positions of the proposed frames obtained in the previous layer, the proposed frames are cleared in a non-maximum suppression manner in the deep learning target detection algorithm, but because the proposed frame processing layer is not a final result, the threshold of the non-maximum suppression used is larger, and a part of proposed frames are reserved for further processing; .
Further, in step S9, the obtained prediction frames may overlap, and these overlapping prediction frames may be eliminated by using non-maximum suppression.
Further, in the step S1, the set time period is 8-20S.
In step S3, a plurality of pre-selected windows are obtained at the determined sampling frequency, the pre-selected windows and the marked data real windows are used for performing the calculation of the io u blending ratio in the target detection, the calculation mode of the blending ratio is the method type in the target detection of the general image, the two-dimensional calculation is reduced to the one-dimensional calculation in the electrocardiographic data, the blending ratio is obtained through the io u calculation, the prediction windows in different blending ratio ranges are divided into positive samples, negative samples and backgrounds, marking is performed respectively, and the offset of the pre-selected frames and the real frames in the positive samples to the real windows is obtained through calculation.
Further, the positive sample is marked as 1, the background is 0, and the negative sample is-1;
It is an object of a second aspect of the present invention to provide an apparatus for simultaneous localization of the position of an electrocardiographic P-wave, QRS-wave, T-wave based on deep learning multi-target detection, comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, which processor, when executing the computer program, implements the method as described above.
It is an object of a third aspect of the invention to provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method as described above.
The beneficial effects of the invention are as follows: the invention solves the problem that the detection of the P wave and the T wave is needed to be carried out by relying on the QRS wave in the traditional algorithm, and can directly obtain the starting and ending positions of the P wave, the QRS wave and the T wave. Compared with the scheme of deep learning heart beat detection disclosed at present, the invention completes the task of detecting P wave and T wave simultaneously. Aiming at the scheme of partial deep learning heart beat detection, the invention uses the two-stage scheme in deep learning target detection and uses the feature extraction layer similar to pyramid superposition, thereby achieving the technical effect of simultaneously detecting small target P waves and relatively wider QRS waves and T waves compared with the P waves.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof.
Drawings
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings, in which:
FIG. 1 is a schematic flow chart of the method of the present invention;
FIG. 2 is a diagram of a model structure employed in an embodiment of the present invention;
FIG. 3 is a graph of markers for single heart beat P, QRS, T waves;
fig. 4 is a sample plot after 10 seconds of electrocardiographic data labeling.
Detailed Description
Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be understood that the preferred embodiments are presented by way of illustration only and not by way of limitation.
As shown in fig. 1, the method for simultaneously locating the positions of the electrocardiograph P wave, the QRS wave and the T wave based on deep learning multi-target detection comprises the following steps:
step S1: electrocardiosignal collection and pretreatment
Carrying out noise reduction treatment on the electrocardiosignal through a band-pass filter or other modes (including but not limited to a conventional signal treatment mode of removing power frequency interference, baseline drift, myoelectric interference and the like), obtaining electrocardiosignal data after the noise reduction treatment, and carrying out normalization treatment on the electrocardiosignal data;
Step S2: the electrocardiographic signal is sliced into pieces of set time length (in this embodiment, the length is 10 seconds, other ranges can be selected in the actual use process, the proposed set time length is 8-20 s), the piece data form an electrocardiographic data set, and the electrocardiographic data set is ensured to contain different P waves, QRS waves, T waves with different forms and different intervals.
And marking the starting and ending positions of the P wave, the QRS wave and the T wave of the electrocardiograph data set to obtain a marked data set.
Step S3: constructing target data required by model training according to marked data samples
According to the electrocardiograph knowledge that the duration of the P wave is generally less than 0.11s, the duration of the QRS wave is 0.11-0.25s, the duration of the T wave is about 0.25s, the P wave and the QRS wave are obtained, the standard window of the T wave is the electrocardiograph data sampling rate (fs), the waveform duration(s) is rounded, taking 500hz sampling as an example, two windows are obtained and are respectively 55 and 125 (500 x 0.11=55 and 500 x 0.25=125), an abnormal waveform can appear in the real electrocardiograph, the actual waveform length is various, therefore, 6 preselected windows used for prolonging or shortening the two windows to obtain a model are respectively 27, 55, 68, 83, 125 and 187, a priori frame can be generated on electrocardiograph data by using the windows with the 6 length sizes, and the 6 size windows are the size of the priori frame. The window can also obtain a window range with more concentrated distribution by clustering the marked data.
And (3) performing the calculation of the io u blending proportion in target detection by using the preselected window and the marked data real window, wherein the calculation mode of the blending proportion is a method type in general image target detection, the two-dimensional calculation in the picture is reduced to one-dimensional calculation in the electrocardiographic data, the blending proportion is obtained by the io u calculation, the prediction windows in different blending proportion ranges are divided into positive samples, negative samples and backgrounds, the positive samples are marked as 1, the background is 0, and the negative samples are-1. And calculating the pre-selection frame and the real frame in the forward sample to obtain the offset of each pre-selection window to the real window.
The 6 pre-selected windows described above are divided into two classes, one class detecting P-waves 27, 55, 68 and the second class detecting QRS-waves and T-waves 83, 125, 187. The positive and negative samples and the background samples obtained through the iou are divided, offset is calculated with a real window of a real mark, and the two groups are stored into one array, so that model prediction target data is obtained. The purpose of this is to train on the different feature layers corresponding in the model, so that more accurate results are obtained.
Step S4: model construction
The feature extraction network used for detecting and identifying heart beat by using deep learning target detection is mainly a AlexNet, VGG, inception, resNet, squeezeNet, mobileNet public mature network. But the P wave is not detected after the electrocardio data is compressed, so that the method uses a pyramid-like network with feature layers overlapped up and down to realize an electrocardio data feature extraction network, constructs two feature layers with different sizes, ensures that a set priori frame can contain the positions of the actually marked P wave, QRS wave and T wave, and inputs the intercepted 10 seconds electrocardio data into the feature extraction network to obtain electrocardio data features.
As shown in fig. 2, in this embodiment, a ResNet structure is used in the feature extraction network, unlike in the disclosed deep learning electrocardiographic target detection algorithm, the present invention uses network layers of 1/4 (denoted as Stage 1) and 1/2 (Stage 2) of the original data length and the last layer (Stage 3) of ResNet in the model structure as the output of the feature extraction layer, and different Stage layers are used to facilitate the superposition of the following feature layers, stage2 adjusts the number of output channels to be equal to Stage3 through convolution of 1*1, stage3 aligns the length with Stage2 through upsampling, and then superposes Stage2 with Stage3 to obtain the first target detection layer P2. The Stage1 is convolved to adjust the number of output channels to be equal to P2 by 1*1, the length of the P2 is aligned with the Stage1 by up-sampling, and then the P2 and the Stage1 are overlapped to obtain a second target detection layer P1. The length of the P2 target detection layer obtained through the operation is equal to 1/4 of the length of the original data, and the length is used as a target detection layer of the QRS wave and the T wave. The length of the P1 target detection layer is equal to 1/2 of the length of the original data, and the P1 target detection layer is used as a target detection layer of the P wave. Different characteristic layers are used for detecting different waveforms, the QRS wave and the T wave are longer in duration, the characteristic layer P2 with small size is used, and the characteristic layer P1 with larger size is used for shorter duration of the P wave. The invention is characterized in that for different characteristics of P wave, QRS wave and T wave, different characteristic layers are used for detection, if P2 characteristic layers are used for detecting the P wave, the P wave characteristics disappear due to the fact that the size of the P2 characteristic layers is reduced after compression, and the P wave detection is not facilitated.
The object detection layer is followed by a rpn layer, and the rpn layer has two groups of outputs, one group is used for distinguishing whether one waveform data cls_score of P wave, QRS wave and T wave exists in the electrocardio number of the preselected window, and the other group is a suggestion frame cls_ bboxes of the P wave, the QRS wave and the T wave. In the layer model calculation, the judgment of the marked array in the step (III (3)) is carried out, and if the first group of data in the array detects the P wave, the calculation is carried out in the corresponding P1 target detection layer. Similarly, detecting QRS waves and T waves can be calculated in the corresponding P2 target detection layer.
And S5, inputting the training set data manufactured in the step S3 into a model according to a suggestion frame detection algorithm in the deep learning target detection algorithm, and distinguishing which parts in the electrocardiographic data with the set time length are suggestion frames containing P wave, QRS wave and T wave positions through regression and classification operation of the model.
Regression and classification operation are carried out through a model to distinguish which parts in the 10s electrocardiograph data are suggestion frames containing P wave, QRS wave and T wave positions, and the main function of the layer is to distinguish which pre-selection frames in the electrocardiograph data contain the P wave, the QRS wave and the T wave positions and do not have the P wave, the QRS wave, the T wave or the incomplete positions of the waveforms of the P wave, the QRS wave and the T wave, and the layer is notable to distinguish the data which do not contain the P wave, the QRS wave and the T wave in a piece of electrocardiograph data.
The invention uses the two-stage scheme (compared with the one-stage scheme, the precision is higher than the one-stage, but the speed is slower), and the pre-selection window obtained in the step S2 contains one of the waveforms of P wave, QRS wave and T wave, and is not distinguished from the detected waveforms. Thus, at rpn layers, a classifier_position layer follows. The corresponding target waveforms are found on P1 and P2 through the suggestion frame, the length of the intercepted target waveforms is unequal, the intercepted target waveform data is required to be scaled to the same length, and then the detected target waveforms are classified and determined to be one of P waves, QRS waves and T waves. And calculating the target window through the layer to obtain more accurate position information to obtain a prediction window.
Step S6: performing suggested frame offset dataset processing
There are multiple cases of repetition or superposition according to the suggested frame position obtained in the previous layer. In the deep learning target detection algorithm, the non-maximum suppression mode is used for clearing the suggestion frame, but because the suggestion frame processing layer is not a final result, the threshold value of the non-maximum suppression is larger, part of suggestion frame is reserved for further processing, and the suggestion frame obtained by the layer has small overlapping degree with the positions of the real P wave, the real QRS wave and the real T wave, so that the suggestion frame can be further processed.
The invention uses the detection scheme of two-stage, so that the output result of rpn layers can be used as an optimization target to calculate the loss value together with the positive and negative samples and the background calculated in the step S2 and the offset of a preselected window. The background part is discarded and only positive and negative samples are focused on by calculating the offset of the pre-selected window and whether the pre-selected window contains one waveform of the P wave, the QRS wave and the T wave. The pre-selection window offset is obtained through rpn layers, and whether the pre-selection window contains P waves, QRS waves and T waves or not is obtained.
Step S7: proposal frame secondary treatment
Before entering a classification layer (classification_position layer), the data is subjected to TARGETSTAGE to reversely encode the result after rpn layers are optimized, the offset is converted into the window width, nms non-maximum suppression is further carried out, and suggestion boxes with higher overlapping rate and lower score are removed so as to prevent imbalance of positive and negative samples. For electrocardiographic signal data within 10 seconds, the maximum normal heart beat number does not exceed 18 times, and the P-wave and T-wave waveform data are added about 54 times, so that a set of data is set as 100 targets. Specifically, performing cross-over ratio calculation on a rpn-layer output offset reverse coding obtained suggestion frame and an actual marked real frame, reserving a window with the cross-over ratio being more than 75% as positive samples, screening out first 54 positive samples, and filling negative samples with the cross-over ratio being less than 50% and more than 0 of less than 54 into 100 targets. The positive samples in these 100 targets are distinguished by comparing with the true labeled data, where the positive samples belong to the P wave, QRS wave, T wave and are labeled 1,2,3 by classification, and the negative samples are labeled 0. And meanwhile, the offset is calculated again as optimization target data of a classifier_position layer by obtaining a suggested frame and an actual frame of a real mark through reverse coding.
Step S8: the output features obtained through the feature extraction network and the classified sample data in the step S7 are input into a classification layer, extracted features are intercepted through a twice processed suggestion frame, and the length of the suggestion frame is changed into the same size because of different lengths of the suggestion frame, and then the same size is input into the classification layer, so that the prediction frame offset and the corresponding prediction frame category are obtained through the classification layer.
Specifically, the feature layers P1 and P2 in step S4 and the feature layer P2 in step S7 are reversely encoded by rpn layer offsets to 100 target data input classifier_position layers with window widths. And intercepting the length data on different characteristic layers according to different window sizes in 100 targets, and calculating offset again at the layer by using the classification marks of the 100 targets and the actual frames of the real marks in the step S4 as an optimization target to calculate a loss value. By this step, a prediction window closer to the true mark is obtained, and the targets in the prediction window belong to the P wave, the QRS wave and the T wave.
Finally, the electrocardiographic data are predicted through a model to obtain the starting and ending positions and categories of the P wave, the QRS wave and the T wave.
Compared with the traditional method for detecting P wave and T wave, the invention has the following differences: the traditional detection of the P wave and the T wave depends on the prior determination of the QRS wave by using differential methods and the like, and the detection of the P wave and the T wave can be influenced if the detection of the QRS wave is inaccurate. After the QRS wave is detected, the P wave and the T wave are determined by calculating a differential threshold line and a vertex area threshold, but the P wave is easy to interfere due to smaller amplitude in the electrocardiosignal, so that the P wave is difficult to detect. In addition, the traditional method for detecting the P wave, the T wave and the QRS wave is in a mode of using an empirical threshold, and the threshold setting adaptability is low. The accuracy of the data decreases when an artificial design experience threshold is exceeded. However, deep learning techniques may improve the adaptability of the algorithm through big data learning.
Compared with the currently disclosed deep learning electrocardiosignal detection algorithm, the method disclosed by the invention is characterized in that: most of the deep learning detection algorithms disclosed at present are used for detecting single QRS waves or P waves, and the invention realizes that a single model can detect the P waves, the QRS waves and the T waves simultaneously. Compared with a single deep learning P slice wave detection algorithm, the standard P wave duration is 0.11s, but interference or special P waves can occur in an actual electrocardiosignal, so that the P wave duration is greater than or less than 0.11s, different preselected windows for P wave detection are designed, and P wave detection with different durations is met. In order to achieve simultaneous detection of P-waves, QRS-waves and T-waves, the present invention differs from the presently disclosed single wave detection deep learning algorithm. According to the P-wave duration generally being less than 0.11s, the QRS-wave duration being 0.11-0.25s, the T-wave duration being about 0.25s, the P-wave and QRS-wave, T-wave are detected using different size feature layers, and different pre-selected windows for QRS-wave, T-wave detection are designed.
It should be appreciated that embodiments of the invention may be implemented or realized by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer readable memory. The methods may be implemented in a computer program using standard programming techniques, including a non-transitory computer readable storage medium configured with a computer program, where the storage medium so configured causes a computer to operate in a specific and predefined manner, in accordance with the methods and drawings described in the specific embodiments. Each program may be implemented in a high level procedural or object oriented programming language to communicate with a computer system. However, the program(s) can be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language. Furthermore, the program can be run on a programmed application specific integrated circuit for this purpose.
Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes (or variations and/or combinations thereof) described herein may be performed under control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications), by hardware, or combinations thereof, collectively executing on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.
Further, the method may be implemented in any type of computing platform operatively connected to a suitable computing platform, including, but not limited to, a personal computer, mini-computer, mainframe, workstation, network or distributed computing environment, separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, and so forth. Aspects of the invention may be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and/or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer, which when read by a computer, is operable to configure and operate the computer to perform the processes described herein. Further, the machine readable code, or portions thereof, may be transmitted over a wired or wireless network. When such media includes instructions or programs that, in conjunction with a microprocessor or other data processor, implement the steps described above, the invention described herein includes these and other different types of non-transitory computer-readable storage media. The invention also includes the computer itself when programmed according to the methods and techniques of the present invention.
The computer program can be applied to the input data to perform the functions described herein, thereby converting the input data to generate output data that is stored to the non-volatile memory. The output information may also be applied to one or more output devices such as a display. In a preferred embodiment of the invention, the transformed data represents physical and tangible objects, including specific visual depictions of physical and tangible objects produced on a display.
Finally, it is noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the present invention, which is intended to be covered by the claims of the present invention.
Claims (8)
1. A method for simultaneously positioning the positions of an electrocardiograph P wave, a QRS wave and a T wave based on deep learning multi-target detection is characterized by comprising the following steps: the method comprises the following steps:
Step S1: acquiring electrocardiosignals, and carrying out baseline drift and noise reduction treatment on the electrocardiosignals; step S2: intercepting marked electrocardiograph data for a set time length, ensuring that the electrocardiograph data set contains different conditions of all electrocardiograph data P waves, QRS waves and T waves, and marking the starting and ending positions of the electrocardiograph data set P waves, the QRS waves and the T waves to obtain a marked data set;
Step S3: performing prior frame processing in deep learning target detection on the intercepted and marked sample data, wherein the width of the prior frame is related to the widths of the P wave, the QRS wave and the T wave, so that a plurality of prior frames are generated on the sample data; according to the deep learning target detection, converting the prior frame and the positions of the actual marked P wave, QRS wave and T wave into the relation of offset, thereby obtaining a data set required by model training;
step S4: a feature extraction network for use in heart beat detection identification using deep learning object detection;
in the step S4, the feature extraction network uses a ResNet structure;
In the step S4, an electrocardiograph data feature extraction network is realized by using a pyramid-like network with feature layers overlapped up and down, two feature layers with different sizes are constructed, the positions of the P wave, the QRS wave and the T wave which are actually marked can be contained in the set prior frame, and the electrocardiograph data within the intercepted set time length is input into the feature extraction network to obtain electrocardiograph data features;
Step S5: inputting the training set data manufactured in the step S3 into a model according to a suggestion frame detection algorithm in a deep learning target detection algorithm, and distinguishing which parts in electrocardiograph data with set time length are suggestion frames containing P wave, QRS wave and T wave positions through regression and classification operation of the model;
step S6: performing suggested frame offset dataset processing;
Step S7: carrying out secondary processing on the suggestion frames, carrying out cross-correlation calculation on the suggestion frames obtained in the last step and the positions of the true marked P waves, QRS waves and T waves, taking the suggestion frames which are larger than a certain threshold value as positive samples, marking P waves, QRS waves and T wave categories of the positive samples, taking negative samples which are smaller than a certain threshold value as negative samples, and encoding positive and negative sample suggestion frames into coefficient relations of offset with the true frames to be used as classification layer sample data;
step S8: inputting the output characteristics obtained through the characteristic extraction network and the classified sample data in the step S7 into a classification layer, intercepting the extracted characteristics through a proposal frame of secondary processing, and inputting the classification layer after changing the length of the proposal frame into the same size because the lengths of the proposal frames are different, thereby obtaining the predicted frame offset and the corresponding predicted frame category through the classification layer;
step S9: finally, the electrocardiographic data are predicted through a model to obtain the starting and ending positions and categories of the P wave, the QRS wave and the T wave.
2. The method for simultaneously locating the positions of the electrocardiographic P wave, the QRS wave and the T wave based on deep learning multi-target detection according to claim 1, wherein the method comprises the following steps of: in step S6, according to the situation that there are multiple repetitions or overlaps of the positions of the suggestion frames obtained in the previous layer, the multiple repetitions or overlaps of the suggestion frames are cleared by using the manner of suppressing the non-maxima in the deep learning target detection algorithm, but since the suggestion frame processing layer is not the final result, the threshold of suppressing the non-maxima used is larger, and part of the suggestion frames are reserved for further processing.
3. The method for simultaneously locating the positions of the electrocardiographic P wave, the QRS wave and the T wave based on deep learning multi-target detection according to claim 1, wherein the method comprises the following steps of: in step S9, the obtained prediction frames may overlap, and these overlapping prediction frames are eliminated by using non-maximum suppression.
4. The method for simultaneously locating the positions of the electrocardiographic P wave, the QRS wave and the T wave based on deep learning multi-target detection according to claim 1, wherein the method comprises the following steps of: in the step S1, the set time length is 8-20S.
5. The method for simultaneously locating the positions of the electrocardiographic P wave, the QRS wave and the T wave based on deep learning multi-target detection according to claim 1, wherein the method comprises the following steps of: in the step S3, under the determined sampling frequency, a plurality of pre-selected windows are obtained, the pre-selected windows and the marked data real windows are used for performing the io u merging proportion calculation in the target detection, the merging proportion calculation mode is a method type in the general image target detection, the two-dimensional calculation is reduced to one-dimensional calculation in the electrocardiographic data, the merging proportion is obtained through the io u calculation, the prediction windows in different merging proportion ranges are divided into positive samples, negative samples and backgrounds and marked respectively, and the offset of each pre-selected window to the real window is obtained through calculation of a pre-selected frame and a real frame in the positive samples.
6. The method for simultaneously locating the positions of the electrocardiographic P wave, the QRS wave and the T wave based on deep learning multi-target detection according to claim 5, wherein the method comprises the following steps: positive samples were marked as 1, background as 0, negative samples as-1.
7. An apparatus for simultaneously locating the position of an electrocardiograph P wave, QRS wave and T wave based on deep learning multi-target detection, comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein: the processor, when executing the computer program, implements the method of any of claims 1-3.
8. A computer-readable storage medium having stored thereon a computer program, characterized by: the computer program implementing the method according to any of claims 1-3 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111662087.5A CN114176602B (en) | 2021-12-30 | 2021-12-30 | Method for simultaneously positioning electrocardiograph P wave, QRS wave and T wave based on deep learning multi-target detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111662087.5A CN114176602B (en) | 2021-12-30 | 2021-12-30 | Method for simultaneously positioning electrocardiograph P wave, QRS wave and T wave based on deep learning multi-target detection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114176602A CN114176602A (en) | 2022-03-15 |
| CN114176602B true CN114176602B (en) | 2024-04-26 |
Family
ID=80545439
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111662087.5A Active CN114176602B (en) | 2021-12-30 | 2021-12-30 | Method for simultaneously positioning electrocardiograph P wave, QRS wave and T wave based on deep learning multi-target detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114176602B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115956925B (en) * | 2023-02-10 | 2023-06-06 | 合肥心之声健康科技有限公司 | QRS wave detection method and system based on multistage smooth envelope |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101919695A (en) * | 2010-08-06 | 2010-12-22 | 李楚雅 | Electrocardiosignal QRS complex detection method based on wavelet transform |
| JP2012210235A (en) * | 2011-03-30 | 2012-11-01 | Sony Corp | Signal processing device, signal processing method and program, and information processing device |
| CN103110417A (en) * | 2013-02-28 | 2013-05-22 | 华东师范大学 | Automatic electrocardiogram recognition system |
| CN110013247A (en) * | 2019-05-24 | 2019-07-16 | 东北大学 | A kind of detection, differentiation and the localization method of P wave of electrocardiogram and T wave |
| CN111184508A (en) * | 2020-01-19 | 2020-05-22 | 武汉大学 | Electrocardiosignal detection device and analysis method based on joint neural network |
| CN111345808A (en) * | 2018-12-24 | 2020-06-30 | Zoll医疗公司 | Method for processing electrocardiosignal, electrocardiosignal monitoring device and storage medium |
| CN112826513A (en) * | 2021-01-05 | 2021-05-25 | 华中科技大学 | Fetal heart rate detection system based on deep learning and specificity correction on FECG |
| CN112842355A (en) * | 2021-02-24 | 2021-05-28 | 推演医疗科技(北京)有限责任公司 | Electrocardiosignal heart beat detection and identification method based on deep learning target detection |
| CN113128585A (en) * | 2021-04-16 | 2021-07-16 | 重庆康如来科技有限公司 | Deep neural network based multi-size convolution kernel method for realizing electrocardiographic abnormality detection and classification |
| CN113491523A (en) * | 2021-07-30 | 2021-10-12 | 济南汇医融工科技有限公司 | Electrocardiosignal characteristic point detection method and system |
| CN113499079A (en) * | 2021-06-18 | 2021-10-15 | 南京信息工程大学 | Atrial fibrillation detection method in electrocardiogram |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150335288A1 (en) * | 2013-06-06 | 2015-11-26 | Tricord Holdings, Llc | Modular physiologic monitoring systems, kits, and methods |
-
2021
- 2021-12-30 CN CN202111662087.5A patent/CN114176602B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101919695A (en) * | 2010-08-06 | 2010-12-22 | 李楚雅 | Electrocardiosignal QRS complex detection method based on wavelet transform |
| JP2012210235A (en) * | 2011-03-30 | 2012-11-01 | Sony Corp | Signal processing device, signal processing method and program, and information processing device |
| CN103110417A (en) * | 2013-02-28 | 2013-05-22 | 华东师范大学 | Automatic electrocardiogram recognition system |
| CN111345808A (en) * | 2018-12-24 | 2020-06-30 | Zoll医疗公司 | Method for processing electrocardiosignal, electrocardiosignal monitoring device and storage medium |
| CN110013247A (en) * | 2019-05-24 | 2019-07-16 | 东北大学 | A kind of detection, differentiation and the localization method of P wave of electrocardiogram and T wave |
| CN111184508A (en) * | 2020-01-19 | 2020-05-22 | 武汉大学 | Electrocardiosignal detection device and analysis method based on joint neural network |
| CN112826513A (en) * | 2021-01-05 | 2021-05-25 | 华中科技大学 | Fetal heart rate detection system based on deep learning and specificity correction on FECG |
| CN112842355A (en) * | 2021-02-24 | 2021-05-28 | 推演医疗科技(北京)有限责任公司 | Electrocardiosignal heart beat detection and identification method based on deep learning target detection |
| CN113128585A (en) * | 2021-04-16 | 2021-07-16 | 重庆康如来科技有限公司 | Deep neural network based multi-size convolution kernel method for realizing electrocardiographic abnormality detection and classification |
| CN113499079A (en) * | 2021-06-18 | 2021-10-15 | 南京信息工程大学 | Atrial fibrillation detection method in electrocardiogram |
| CN113491523A (en) * | 2021-07-30 | 2021-10-12 | 济南汇医融工科技有限公司 | Electrocardiosignal characteristic point detection method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114176602A (en) | 2022-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110232383B (en) | Focus image recognition method and focus image recognition system based on deep learning model | |
| CN109117730B (en) | Real-time electrocardiogram atrial fibrillation judgment method, device and system and storage medium | |
| CN111340142B (en) | Epilepsia magnetoencephalogram spike automatic detection method and tracing positioning system | |
| CN104978725B (en) | Coronary artery segmentation method and device | |
| CN110796613B (en) | Automatic identification method and device for image artifacts | |
| EP2916737B1 (en) | System and method for automated detection of lung nodules in medical images | |
| CN110755076B (en) | Image acquisition method, device, equipment and medium of magnetic resonance scanner | |
| CN105913432B (en) | Aorta extracting method and device based on CT sequence images | |
| CN109493343A (en) | Medical image abnormal area dividing method and equipment | |
| CN104217418A (en) | Segmentation of a calcified blood vessel | |
| CN111951221A (en) | Glomerular cell image identification method based on deep neural network | |
| Maiora et al. | Abdominal CTA image analisys through active learning and decision random forests: Aplication to AAA segmentation | |
| CN114532993B (en) | Automatic detection method for electroencephalogram high-frequency oscillation signals of epileptic patients | |
| CN110432895B (en) | Training data processing method, electrocardiographic waveform detection method and electronic equipment | |
| Bourbakis | Detecting abnormal patterns in WCE images | |
| CN114176602B (en) | Method for simultaneously positioning electrocardiograph P wave, QRS wave and T wave based on deep learning multi-target detection | |
| CN105877739A (en) | Clinical examination method of electrocardio intelligent analyzing system | |
| Guo et al. | Emfn: Enhanced multi-feature fusion network for hard exudate detection in fundus images | |
| CN113128585B (en) | Deep neural network based multi-size convolution kernel method for realizing electrocardiographic abnormality detection and classification | |
| CN113160120A (en) | Liver blood vessel segmentation method and system based on multi-mode fusion and deep learning | |
| CN103839048B (en) | Stomach CT image lymph gland recognition system and method based on low-rank decomposition | |
| CN104463862A (en) | Method for fast segmenting kidney CT sequential image | |
| CN104915989A (en) | CT image-based blood vessel three-dimensional segmentation method | |
| CN116864140A (en) | Intracardiac branch of academic or vocational study postoperative care monitoring data processing method and system thereof | |
| KR20140097039A (en) | Method and apparatus for classifying cardiac arrhythmia using an auto associative 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 |