CN113598784B - Arrhythmia detection method and system - Google Patents
Arrhythmia detection method and system Download PDFInfo
- Publication number
- CN113598784B CN113598784B CN202110983506.9A CN202110983506A CN113598784B CN 113598784 B CN113598784 B CN 113598784B CN 202110983506 A CN202110983506 A CN 202110983506A CN 113598784 B CN113598784 B CN 113598784B
- Authority
- CN
- China
- Prior art keywords
- electrocardiosignal
- hilbert
- time sequence
- original
- arrhythmia
- 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 74
- 206010003119 arrhythmia Diseases 0.000 title claims abstract description 49
- 230000006793 arrhythmia Effects 0.000 title claims abstract description 49
- 238000001228 spectrum Methods 0.000 claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 23
- 230000006870 function Effects 0.000 claims description 37
- 238000013527 convolutional neural network Methods 0.000 claims description 27
- 238000004590 computer program Methods 0.000 claims description 18
- 238000010606 normalization Methods 0.000 claims description 15
- 230000004927 fusion Effects 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 11
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 125000004122 cyclic group Chemical group 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 238000005520 cutting process Methods 0.000 claims description 4
- 230000003595 spectral effect Effects 0.000 claims 1
- 238000012549 training Methods 0.000 abstract description 16
- 239000000463 material Substances 0.000 abstract description 4
- 238000010183 spectrum analysis Methods 0.000 abstract description 2
- 238000009826 distribution Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 10
- 230000004913 activation Effects 0.000 description 8
- 238000006243 chemical reaction Methods 0.000 description 5
- 230000000052 comparative effect Effects 0.000 description 5
- 238000003860 storage Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000012216 screening Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 206010003658 Atrial Fibrillation Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 208000037891 myocardial injury Diseases 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000005180 public health Effects 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 208000003663 ventricular fibrillation Diseases 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
- A61B5/72—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
- A61B5/7253—Details of waveform analysis characterised by using transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Cardiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Signal Processing (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 provides a method and a system for detecting arrhythmia, which belong to the technical field of arrhythmia detection equipment and comprise the following steps: acquiring an original electrocardiosignal to be detected; preprocessing an original electrocardiosignal to respectively obtain a time sequence electrocardiosignal and a Hilbert spectrogram; based on the time sequence electrocardiosignal and the Hilbert spectrogram, respectively extracting the time sequence electrocardiosignal characteristic and the Hilbert spectrogram characteristic; processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not. The invention fully utilizes the existing big data to carry out model training on the electrocardiographic data of a plurality of databases, thereby solving the problem that a great deal of manpower and material resources are consumed to uniformly divide the standard; aiming at the randomness problem of electrocardiosignals, hilbert spectrum analysis is introduced, so that a model obtains richer information, and the accuracy of arrhythmia detection is further improved.
Description
Technical Field
The invention relates to the technical field of heart rhythm detection equipment, in particular to a method and a system for detecting arrhythmia.
Background
The death rate of patients with myocardial injury and arrhythmia is high, especially for patients with frequent arrhythmia (atrial fibrillation, ventricular fibrillation, etc.), arrhythmia detection (or early warning) is of great importance for rescuing patients with severe emergency, and has great significance for public health safety prevention and control or treatment of epidemic diseases.
The electrocardiosignal is a typical non-stationary medical signal especially for critical patients and can be used as the diagnosis basis of arrhythmia. In arrhythmia detection, conventional detection methods generally analyze an electrocardiographic signal as a stationary random signal directly from the time domain or the frequency domain.
The existing large amount of electrocardiograms are directly acquired data without expert marks, the problem that the label standards are inconsistent exists in the public data with the expert marks, and the direct model training of the electrocardiograms of a plurality of databases by using a supervised learning algorithm is difficult, so that the arrhythmia detection result is inaccurate and unreliable.
Disclosure of Invention
The invention aims to provide an arrhythmia detection method and system for further improving arrhythmia detection precision by introducing Hilbert spectrum and ensuring to acquire more abundant information so as to solve at least one technical problem in the background technology.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
in one aspect, the present invention provides a method of arrhythmia detection comprising:
acquiring an original electrocardiosignal to be detected;
preprocessing an original electrocardiosignal to respectively obtain a time sequence electrocardiosignal and a Hilbert spectrogram;
based on the time sequence electrocardiosignal and the Hilbert spectrogram, respectively extracting the time sequence electrocardiosignal characteristic and the Hilbert spectrogram characteristic;
processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
Preferably, acquiring the original electrocardiographic signal to be detected includes: the method comprises the steps of collecting an original electrocardiosignal of a tested person, and removing baseline drift, high-frequency noise and power frequency interference by using Hilbert-Huang transform (HHT transform for short) to obtain the original electrocardiosignal to be detected.
Preferably, the original electrocardiosignals are normalized by using a maximum and minimum normalization method to obtain time sequence electrocardiosignals; and obtaining an electrocardiosignal by HHT conversion and converting the electrocardiosignal into a Hilbert spectrogram.
Preferably, based on the positive sample and the negative sample, the contrast prediction coding CPC model is used for training, the trained contrast prediction coding CPC model is used for processing the time sequence electrocardiosignal, and the time sequence electrocardiosignal characteristics are extracted.
Preferably, a convolutional neural network is utilized to extract the characteristic of a certain moment in the time sequence electrocardiosignal; generating fusion history information by using a cyclic neural network to generate electrocardiosignal characteristics before the certain moment; the fusion history information and the predicted electrocardiosignal characteristic after a certain moment form a positive sample, and the fusion history information and the characteristic representation of any sampled sequence point form a negative sample.
Preferably, the mutual information formula is introduced according to the electrocardiosignal characteristic representation before a certain moment and based on the electrocardiosignal context relation, so as to predict the electrocardiosignal characteristic after the certain moment.
Preferably, a trained contrast learning SimCLR model is used for processing the Hilbert spectrum, and Hilbert spectrum characteristics are extracted.
Preferably, the basic framework of the contrast learning SimCLR model comprises a data enhancement module, a convolutional neural network, a fully connected network and a contrast loss function; the data enhancement module is used for converting the picture, including turning, rotating, zooming and random cutting; the convolutional neural network is used for extracting respective characteristic representations of the converted pictures; the fully connected network is used to map the feature representation learned by the convolutional neural network to a one-dimensional feature representation.
In a second aspect, the present invention provides an arrhythmia detection system comprising:
the acquisition module is used for acquiring an original electrocardiosignal to be detected;
the preprocessing module is used for preprocessing the original electrocardiosignals to respectively obtain time sequence electrocardiosignals and Hilbert spectrograms;
the extraction module is used for respectively extracting the time sequence electrocardiosignal characteristics and the Hilbert spectrum characteristics based on the time sequence electrocardiosignal and the Hilbert spectrum;
the detection module is used for processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
In a third aspect, the present invention provides a non-transitory computer readable storage medium for storing computer instructions which, when executed by a processor, implement the arrhythmia detection method as described above.
In a fourth aspect, the present invention provides an electronic device comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and wherein the computer program is stored in the memory, said processor executing the computer program stored in said memory when the electronic device is running, to cause the electronic device to execute instructions implementing the arrhythmia detection method as described above.
The invention has the beneficial effects that: the existing big data are fully utilized to carry out model training on the electrocardiographic data of a plurality of databases, so that the problem that a great deal of manpower and material resources are consumed to uniformly divide the standard is solved; aiming at the problem of nonstationary randomness of electrocardiosignals, hilbert spectrum analysis is introduced, so that a model obtains richer information, and the arrhythmia detection precision is further improved.
Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a basic flow diagram of a contrast learning-based arrhythmia detection method according to an embodiment of the present invention;
FIG. 2 is a frame structure diagram of a CPC model based on comparative predictive coding according to an embodiment of the present invention;
FIG. 3 is a schematic view of sample 1 according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of sample 2 according to an embodiment of the present invention;
fig. 5 is a network frame structure diagram of a comparative learning SimCLR model according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements throughout or elements having like or similar functionality. The embodiments described below by way of the drawings are exemplary only and should not be construed as limiting the invention.
It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless expressly stated otherwise, as understood by those skilled in the art. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, and/or groups thereof.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, the different embodiments or examples described in this specification and the features of the different embodiments or examples may be combined and combined by those skilled in the art without contradiction.
In order that the invention may be readily understood, a further description of the invention will be rendered by reference to specific embodiments that are illustrated in the appended drawings and are not to be construed as limiting embodiments of the invention.
It will be appreciated by those skilled in the art that the drawings are merely schematic representations of examples and that the elements of the drawings are not necessarily required to practice the invention.
Example 1
Embodiment 1 provides an arrhythmia detection system, comprising:
the acquisition module is used for acquiring an original electrocardiosignal to be detected;
the preprocessing module is used for preprocessing the original electrocardiosignals to respectively obtain time sequence electrocardiosignals and Hilbert spectrograms;
the extraction module is used for respectively extracting the time sequence electrocardiosignal characteristics and the Hilbert spectrum characteristics based on the time sequence electrocardiosignal and the Hilbert spectrum;
the detection module is used for processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
In this embodiment 1, with the arrhythmia detection system as described above, an arrhythmia detection method is realized, the method comprising:
acquiring an original electrocardiosignal to be detected by using an acquisition module;
preprocessing the original electrocardiosignals by utilizing a preprocessing module to respectively obtain time sequence electrocardiosignals and Hilbert spectrograms;
the method comprises the steps that an extraction module is utilized to respectively extract time sequence electrocardiosignal characteristics and Hilbert spectrum characteristics based on the time sequence electrocardiosignal and the Hilbert spectrum;
processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics based on a trained detection model by using a detection module to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
In this embodiment 1, acquiring the original electrocardiographic signal to be detected includes: the method comprises the steps of collecting original electrocardiosignals of a tested person, removing baseline drift and high-frequency noise interference by using HHT conversion, and removing power frequency interference by using a trap to obtain the original electrocardiosignals to be detected.
In the embodiment 1, the original electrocardiograph signal is normalized by using a maximum and minimum normalization method to obtain a time sequence electrocardiograph signal; hilbert spectrograms of electrocardiosignals are obtained by HHT conversion.
In this embodiment 1, the training of the comparative predictive coding CPC model is performed based on the positive and negative samples, and the time-series electrocardiograph signals are processed by using the trained comparative predictive coding CPC model to extract the time-series electrocardiograph signal characteristics.
In the embodiment 1, a convolutional neural network is used to extract the characteristic of a certain moment in time-series electrocardiosignals; generating fusion history information by using a cyclic neural network to generate electrocardiosignal characteristics before the certain moment; the fusion history information and the predicted electrocardiosignal characteristic after a certain moment form a positive sample, and the fusion history information and the characteristic representation of any sampled sequence point form a negative sample.
And introducing a mutual information formula according to the electrocardiosignal characteristic representation before a certain moment based on the electrocardiosignal context relation, and predicting the electrocardiosignal characteristic after the certain moment.
In the embodiment 1, a trained contrast learning SimCLR model is used to process Hilbert spectra and extract Hilbert spectrum characteristics.
In this embodiment 1, the basic framework of the contrast learning SimCLR model includes a data enhancement module, a convolutional neural network, a fully connected network, and a contrast loss function; the data enhancement module is used for converting the picture, including turning, rotating, zooming and random cutting; the convolutional neural network is used for extracting respective characteristic representations of the converted pictures; the fully connected network is used to map the feature representation learned by the convolutional neural network to a one-dimensional feature representation.
Example 2
Embodiment 2 provides an arrhythmia detection system, comprising:
the acquisition module is used for acquiring an original electrocardiosignal to be detected;
the preprocessing module is used for preprocessing the original electrocardiosignals to respectively obtain time sequence electrocardiosignals and Hilbert spectrograms;
the extraction module is used for respectively extracting the time sequence electrocardiosignal characteristics and the Hilbert spectrum characteristics based on the time sequence electrocardiosignal and the Hilbert spectrum;
the detection module is used for processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
In this embodiment 2, with the arrhythmia detection system as described above, an arrhythmia detection method is realized, the method comprising:
acquiring an original electrocardiosignal to be detected by using an acquisition module;
preprocessing the original electrocardiosignals by utilizing a preprocessing module to respectively obtain time sequence electrocardiosignals and Hilbert spectrograms;
the method comprises the steps that an extraction module is utilized to respectively extract time sequence electrocardiosignal characteristics and Hilbert spectrum characteristics based on the time sequence electrocardiosignal and the Hilbert spectrum;
processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics based on a trained detection model by using a detection module to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
In this embodiment 2, acquiring the original electrocardiographic signal to be detected includes: and acquiring an original electrocardiosignal of the tested person, and removing baseline drift, high-frequency noise interference and power frequency interference by using HHT conversion to obtain the original electrocardiosignal to be detected.
The HHT transformation comprises the following steps:
(1) Finding out local maximum value points and local minimum value points of the signal x (t), and then respectively interpolating the local maximum value points and the local minimum value points by using a cubic spline function to obtain an upper envelope u (t) and a lower envelope l (t) of the signal x (t); let m 1 (t)＝[u(t)+l(t)]2, then m 1 And (t) is the average of the sum of the upper and lower envelopes.
(2) Let h 1 (t)＝x(t)-m 1 (t) likewise find h 1 (t) local maximum value point and local minimum value point, and respectively interpolating them by using cubic spline function to obtain upper and lower envelopes u 11 (t) and l 11 (t) determining the mean curve m of them 11 (t) to obtain h 11 (t)＝h 1 (t)-m 11 (t)。
Inspection h 11 (t) if the following condition of the Intrinsic Mode Function (IMF) is met, if not, continuing the iterative process until h 1k (t)＝h 1(k-1) (t)-m 1k (t) meets the conditions of IMF. And let c 1 (t)＝h 1k (t), then c 1 (t) is the first IMF component to be screened, and the above procedure completes the first screening. Wherein, the stopping criterion is:
where T is the data length. The value is the normalized standard deviation of two continuous iterative processes, and the reference value is 0.2-0.3. Once SD is k Less than this value, iteration may be stopped.
(3) Let r 1 (t)＝x(t)-c 1 (t). Obviously r 1 (t) is the original signal x (t) and the first IMF component c 1 (t) difference between the two. Will c 1 (t) is regarded as the original signal x (t), for c 1 (t) repeating the steps (1) and (2) to obtain
r 2 (t)＝r 1 (t)-c 2 (t)
...
r m (t)＝r m-1 (t)-c m (t)
In c 2 (t)，…，c m (t) is the newly screened IMF component. The decomposition process is up to r m (t) becomes a monotonic function or stops when it contains only one extreme point. Thus, the original signal x (t) is decomposed into m IMF components and the final residual r m The sum of (t), i.eIn practice, r m (t) is a simple trend function, or a constant.
(4) For each IMF component c k (t) solving for Hilbert transform thereof
Thereby construct c k Analytical signal of (t)
In the method, in the process of the invention,for time-varying amplitude>Is the instantaneous frequency. Then c k The time-varying energy of (t) is E k (t)＝|a k (t)| 2
(5) After each IMF component of the original signal x (t) and the corresponding time-varying energy are obtained, the time-varying energy is summed to obtain the Hilbert spectrum.
In this embodiment 2, the time-series electrocardiograph signal is obtained by normalizing the original electrocardiograph signal by using the maximum and minimum normalization method.
The maximum and minimum normalization method is a data normalization method. Typically, before modeling, data needs to be normalized to eliminate the influence of dimension. If the non-standardized data is directly modeled, the model may learn too much for variables with large logarithmic values, while the model may not work well because variables with small logarithmic values are not trained sufficiently. Common data normalization methods include maximum and minimum normalization, mean variance normalization, decimal scaling, quantitative feature binarization and the like.
The maximum and minimum normalization is to perform normalization processing by using the maximum value and the minimum value in the data column, wherein the normalized value is between 0 and 1, and the calculation mode is that the data and the minimum value in the column are subjected to difference and then divided by the extremely difference. The specific formula is as follows:
in the formula, x' represents the value of single data, min is the minimum value of the column of the data, and max is the maximum value of the column of the data.
The maximum and minimum normalization is easily affected by the extreme value, when the extreme value exists in a certain column of data, the extreme value or the abnormal value can be eliminated in advance according to the actual service scene, or the normalized data is transformed, such as taking the logarithm, so that the transformed data is close to normal distribution.
In this example 2, a Hilbert spectrum of electrocardiographic signals was obtained by HHT conversion.
The Hilbert spectrum of the hht function estimation signal can be directly called in MATLAB, and the length of the window represents the segment data length of each process.
In this embodiment 2, the training of the contrast predictive coding CPC model is performed based on the positive and negative samples, and the time-series electrocardiographic signals are processed by using the trained contrast predictive coding CPC model to extract the time-series electrocardiographic signal characteristics.
In the embodiment 2, a convolutional neural network is used to extract the characteristic of a certain moment in the time-series electrocardiosignal; generating fusion history information by using a cyclic neural network to generate electrocardiosignal characteristics before the certain moment; the fusion history information and the predicted electrocardiosignal characteristic after a certain moment form a positive sample, and the fusion history information and the characteristic representation of any sampled sequence point form a negative sample.
And introducing a mutual information formula according to the electrocardiosignal characteristic representation before a certain moment based on the electrocardiosignal context relation, and predicting the electrocardiosignal characteristic after the certain moment.
Specifically, in embodiment 2, the comparative predictive coding CPC model training specifically includes the following steps:
given an electrocardio time sequence of 10 seconds, assuming that the current time is t, c can be represented according to the electrocardio signal characteristics before the time t t Predicting an electrocardiograph signal x at a time k after a time t t+k If the condition generation model p (x) t+k c t ) The calculation degree is complex, the context relation of electrocardiosignals is easy to ignore, and the effect is poor. Therefore, in this embodiment 2, the mutual information formula is introduced as shown in (1):
wherein x is t+k The size is (B, C, L), B represents the batch processing number, and is generally set to be 32, C represents the electrocardio lead number, the default is 12, L is the sample point at the t+k time of the electrocardio time sequence, and the total length of the electrocardio sequence is 1000 (10 seconds electrocardio signal sampling frequency is 100 hz). c t The size is (B, L 1 ,H)，L 1 And the characteristic points at the time t after the processing of the cyclic neural network are represented, and H represents the number of hidden layers output by the cyclic neural network.
In this example 2, equation (1) can be regarded as the mathematical expectation of the discrete random variable function, and therefore, only the secret is calculatedRatio of degreeTo calculate the density ratio, a noise contrast estimation algorithm is introduced from p (x t+k |c t ) The data taken from the distribution is recorded as positive samples, the positive samples are the characteristic representation c of the electrocardiosignal before the known t moment t Under the condition, the electrocardiosignals at the moment k after t are sample points of real signals, and the label is 1.
From p (x) t+k ) Random taken data in the distribution is noted as negative sample, labeled 0, where p (x t+k ) Is with c t There is no associated noise profile. After obtaining two types of tagged data, a classifier needs to be trained to be able to distinguish between the two types of data.
Classifier loss function construction, assuming the set x= { X for a given positive and negative sample 1 ,x 2 ,...,x N A positive sample x t+k From distribution p (x t+k |c t ) The remaining N-1 negative samples are from distribution p (x t+k ) The loss function can be written as (2):
wherein E represents a mathematical expectation operation, f k And (x) represents a function for evaluating the approximation between the true value and the predicted value, as in equation (5).
When the model training is good enough, the probability of correctly distinguishing one positive sample from N-1 negative samples is as shown in equation (3):
where d represents the index of positive samples in sample set X, so when d=i, X i Represents a positive sample, for a single sample p (x i |c t ) Represented by the formula c t From p (x) t+k |c t ) Taking out positive samples x from distribution i Is a function of the probability of (1),is expressed from p (x t+k ) The probability product of taking negative samples from the distribution is that the probability of occurrence of an event is independent of each other, can be directly multiplied, and is +.>Is a constant term, so that the final step equation of (3) is easily obtained, and the molecule +.>The probability of occurrence of 1 positive sample and N-1 negative samples is represented, while the denominator is the probability of occurrence of each of the N samples as positive sample and the other N-1 negative samples is considered.
From the neural probability language model, p (x) t+k |c t ) As in formula (4):
wherein f (x) t+k ,c t ) As shown in formula (5):
wherein,characteristic representation of convolutional neural network output at time t+k, W k Representing a linear matrix.
Equation (5) represents the log bilinear model f (x) t+k ,c t ) Since the optimization targets of equation (3) and equation (4) are the same, f (x) t+k ,c t ) Proportional to
Will f (x) t+k ,c t ) Replaced byBringing into equation (2) reduces the available equation (7):
the summation operation is changed to the desired operation in the formula (7) to obtain the formula (8) because of x j And c t Independent of each other, so p (x) j |c t )＝p(x j ) Obtaining the formula (9). Due to p (x t+k |c t )≥p(x t+k ) Bringing (9) the equation (10) into easy obtainment, and finally bringing (1) the equation (10) into the relation between the loss and the mutual information, wherein minimizing the loss function of the model is equivalent to maximizing x t+k And c t Lower bound of mutual information.
In the embodiment 2, a trained contrast learning SimCLR model is used to process Hilbert spectra and extract Hilbert spectrum characteristics.
In this embodiment 2, the basic framework of the contrast learning SimCLR model includes a data enhancement module, a convolutional neural network, a fully connected network, and a contrast loss function; the data enhancement module is used for converting the picture, including turning, rotating, zooming and random cutting; the convolutional neural network is used for extracting respective characteristic representations of the converted pictures; the fully connected network is used to map the feature representation learned by the convolutional neural network to a one-dimensional feature representation.
In this embodiment 2, the data enhancement module converts the picture, including flipping, rotating, scaling, and random cropping, such as given samples 1 and 2, by converting sample 1 to X1 and X2 and sample 2 to X3 and X4.
The convolutional neural network takes X1, X2, X3 and X4 as input, and extracts the respective characteristic representations of the convolutional neural network to be respectively marked as H1, H2, H3 and H4. Convolutional neural networks contain 33 end-to-end convolutional layers for automatic extraction of features. The calculation formula of each convolution layer is shown in (12):
wherein f (·) represents a nonlinear activation function, N represents the total number of input feature maps, M represents the total number of convolution kernels, X i Represents the ith input feature map, j represents the jth convolution kernel, W ij Weights representing the ith input and jth output convolution kernel, b j Representing the offset of the jth convolution kernel.
The fully connected network learns the characteristic representation H learned by the convolutional neural network i Mapping to one-dimensional feature representation Z i Input into the nonlinear activation function Relu in the activation model,
as shown in formulas (13), (14):
σ i ＝wH i +b (13)
Z i ＝max(0,σ i ) (14)
wherein sigma i Representing the output after passing through the sensor, w and b are weight and bias, respectively, Z i Is the output of the nonlinear activation function.
In this embodiment 2, the SimCLR model is trained by self-supervised learning using the contrast loss function, and if N samples are input, 2N sample pairs are output after data enhancement, and then the output obtained after 1 positive sample passes through the fully connected network is recorded as Z i And Z j Then 2N-2 negative samples remain, and the definition formulas of the loss function are shown as (15) and (16):
sim(u,v)＝u T v/||u||||v|| (15)
wherein u, v represent form parameters, u T Represents the transpose of u, τ represents the temperature factor, in order to make exp (sim (z i ,z j ) I tau) tends to 1 faster, equivalent to the loss function reaching a minimum as soon as possible。
Example 3
As shown in fig. 1, in this embodiment 3, there is provided an arrhythmia detection screening system based on contrast learning, the system including: the system comprises an electrocardiosignal acquisition module, an electrocardiosignal preprocessing module, a comparison learning module and an evaluation module, wherein the electrocardiosignal acquisition module is connected with the electrocardiosignal preprocessing module, two paths of data output by the preprocessing module are respectively sent into two corresponding sub-modules in the comparison learning module, and a pre-training model generated after training the comparison learning model is combined with labeled data to train out a clinically required supervised model.
The electrocardiosignal acquisition module is used for filtering and denoising the original electrocardiosignal; the electrocardiosignal preprocessing module is used for carrying out normalization processing on the acquired electrocardiosignals and then outputting one electrocardiosignal time sequence and another electrocardiosignal Hilbert spectrogram; the contrast learning module is composed of a Contrastive Predictive Coding (CPC) module and a Simple Framework for Contrastive Learning of Visual Representations (SimCLR) module and is used for performing feature learning on the untagged electrocardiosignals; the evaluation module is used for evaluating the advantages and disadvantages of the feature extractor trained by the self-supervision model, and adding the labeled data aiming at different disease detection tasks, so that the model can be converged more quickly and the accurate noninvasive arrhythmia disease screening can be ensured.
In this example 3, arrhythmia screening was performed using the system described above, comprising the following specific steps:
step 1: signal acquisition
Firstly, collecting an original electrocardiosignal of a tested person, and removing baseline drift, high-frequency noise interference and 50Hz power frequency interference of the electrocardiosignal by using HHT (high-frequency transform);
step 2: electrocardiosignal preprocessing module
And carrying out normalization processing on the filtered electrocardiosignals by using a maximum and minimum normalization method, outputting time sequence electrocardiosignals, and simultaneously converting the electrocardiosignals into Hilbert spectrograms by using HHT.
Step 3: CPC module feature extraction
Given an electrocardio time sequence of 10 seconds, assuming that the current time is t, c can be represented according to electrocardio signal characteristics before the time t t Predicting an electrocardiograph signal x at time k after t t+k Directly constructing a conditional generative model p (x t+k |c t ) The computation is complex, and the electrocardiosignal context relation is often ignored, so that the effect is not ideal. Therefore, a mutual information formula is introduced as shown in (1):
wherein x is t+k The size (B, C, L) B represents that the batch processing number is generally set to be 32, C represents that the electrocardiograph lead number is defaults to 12, L is a sample point at the t+k time of the electrocardiograph time sequence, and the total length of the electrocardiograph sequence defaults to 1000 (10 seconds electrocardiograph signal sampling frequency is 100 hz). c t The size is (B, L 1 ,H)，L 1 And the characteristic points at the time t after the processing of the cyclic neural network are represented, and H represents the number of hidden layers output by the cyclic neural network. Equation (1) can be considered as the mathematical expectation of a discrete random variable function, and therefore, only the density ratio needs to be calculatedTo calculate the density ratio, a noise contrast estimation algorithm is introduced from p (x t+k |c t ) The data taken from the distribution is recorded as positive samples, the positive samples are the characteristic representation c of the electrocardiosignal before the known t moment t Under the condition, the electrocardiosignals at the moment k after t are sample points of real signals, and the label is 1. From p (x) t+k ) Random taken data in the distribution is noted as negative sample, labeled 0, where p (x t+k ) Is with c t There is no associated noise profile. After obtaining two types of tagged data, a classifier needs to be trained to be able to distinguish between the two types of data.
Classifier loss function construction, assuming the set x= { X for a given positive and negative sample 1 ,x 2 ,...x N A positive sample x t+k From distribution p (x t+k |c t ) The rest N-1 negativeThe samples are from distribution p (x t+k ) The loss function can be written as (2):
when the model training is good enough, the probability of correctly distinguishing one positive sample from N-1 negative samples is as shown in equation (3):
d represents the index of positive samples in sample set X, so when d=i, X i Represents a positive sample, for a single sample p (x i |c t ) Represented by the formula c t From p (x) t+k |c t ) Taking out positive samples x from distribution i Is a function of the probability of (1),is expressed from p (x t+k ) The probability product of taking negative samples from the distribution is that the probability of occurrence of an event is independent of each other, can be directly multiplied, and is +.>Is a constant term, so that the final step equation of (3) is easily obtained, and the molecule +.>The probability of occurrence of 1 positive sample and N-1 negative samples is represented, while the denominator is the probability of occurrence of each of the N samples as positive sample and the other N-1 negative samples is considered.
From the neural probability language model, p (x) t+k |c t ) As in formula (4):
wherein f (x) t+k ,c t ) As shown in formula (5):
equation (5) represents the log bilinear model f (x) t+k ,c t ) Since the optimization targets of equation (3) and equation (4) are the same, f (x) t+k ,c t ) Proportional toWill f (x) t+k ,c t ) Replaced by->Bringing into equation (2) reduces the available equation (7): />
The summation operation is changed to the desired operation in the formula (7) to obtain the formula (8) because of x j And c t Independent of each other, so p (x) j |c t )＝p(x j ) Obtaining the formula (9). Due to p (x t+k |c t )≥p(x t+k ) Bringing (9) the equation (10) into easy obtainment, and finally bringing (1) the equation (10) into the relation between the loss and the mutual information, wherein minimizing the loss function of the model is equivalent to maximizing x t+k And c t Lower bound of mutual information.
As shown in fig. 2, in the CPC module, the convolutional neural network is used to input electrocardiographic data x at time t in 10 seconds electrocardiographic signals t Feature extraction is performed to obtain z t Generating a representation c of the fusion history information by using an autoregressive model (a cyclic neural network) to represent the characteristics at and before the time t t 。c t Can be combined with z t+k (k∈[1,2,3,4]) Form a positive sample, and the negative sample is composed of c t And feature representation of arbitrarily sampled sequence pointsComposition is prepared.
Step 4: simCLR module feature extraction
The SimCLR module is basically illustrated in FIG. 5 and includes a data enhancement module, a convolutional neural network, a fully-connected network, and a contrast loss function.
The data enhancement module converts the picture, including flipping, rotating, scaling, random cropping, giving samples 1 and 2, converting sample 1 as shown in fig. 3 to X1 and X2, and converting sample 2 as shown in fig. 4 to X3 and X4 via the data enhancement module.
The convolutional neural network takes X1, X2, X3 and X4 as input, and extracts the respective characteristic representations of the convolutional neural network to be respectively marked as H1, H2, H3 and H4. Convolutional neural networks contain 33 end-to-end convolutional layers for automatic extraction of features. The calculation formula of each convolution layer is shown in (12):
wherein f (·) represents a nonlinear activation function, N represents the total number of input feature maps, M represents the total number of convolution kernels, X i Represents the ith input feature map, j represents the jth convolution kernel, W ij Weights representing the ith input and jth output convolution kernel, b j Representing the offset of the jth convolution kernel.
(3) The fully connected network learns the characteristic representation H learned by the convolutional neural network i Mapping to one-dimensional feature representation Z i The non-linear activation function Relu input to the activation model is as shown in formulas (13), (14):
σ i ＝wH i +b (13)
Z i ＝max(0,σ i ) (14)
wherein sigma i Representing the output after passing through the sensor, w and b are weight and bias, respectively, Z i Is the output of the nonlinear activation function.
(4) Contrast loss function for training self-supervision model, falseIf N samples are input, 2N sample pairs are output after data enhancement, and then the output obtained after 1 positive sample passes through (3) is recorded as Z i And Z j Then 2N-2 negative samples remain, and the definition formulas of the loss function are shown as (15) and (16):
sim(u,v)＝u T v/||u||||v|| (15)
step 5: the evaluator evaluates the prediction result
After 3 and 4 self-supervision training models, two pre-training models can be obtained, at the moment, parameters of the pre-training models are fixed, a classifier is connected according to the requirement of a downstream task, data with labels are input, supervision training is carried out, and at the moment, the accuracy of the models is checked by comparing the prediction results output by the models with real labels.
Example 4
Embodiment 4 of the present invention provides a non-transitory computer-readable storage medium storing computer instructions which, when executed by a processor, implement an arrhythmia detection method as described above, the method comprising:
acquiring an original electrocardiosignal to be detected;
preprocessing an original electrocardiosignal to respectively obtain a time sequence electrocardiosignal and a Hilbert spectrogram;
based on the time sequence electrocardiosignal and the Hilbert spectrogram, respectively extracting the time sequence electrocardiosignal characteristic and the Hilbert spectrogram characteristic;
processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
Example 5
Embodiment 5 of the present invention provides a computer program (product) comprising a computer program for implementing a method of arrhythmia detection as described above when run on one or more processors, the method comprising:
acquiring an original electrocardiosignal to be detected;
preprocessing an original electrocardiosignal to respectively obtain a time sequence electrocardiosignal and a Hilbert spectrogram;
based on the time sequence electrocardiosignal and the Hilbert spectrogram, respectively extracting the time sequence electrocardiosignal characteristic and the Hilbert spectrogram characteristic;
processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
Example 6
Embodiment 6 of the present invention provides an electronic device, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and wherein the computer program is stored in the memory, said processor executing the computer program stored in said memory when the electronic device is running, to cause the electronic device to execute instructions for implementing a method for arrhythmia detection as described above, the method comprising:
acquiring an original electrocardiosignal to be detected;
preprocessing an original electrocardiosignal to respectively obtain a time sequence electrocardiosignal and a Hilbert spectrogram;
based on the time sequence electrocardiosignal and the Hilbert spectrogram, respectively extracting the time sequence electrocardiosignal characteristic and the Hilbert spectrogram characteristic;
processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention 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 invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations 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.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it should be understood that various changes and modifications could be made by one skilled in the art without the need for inventive faculty, which would fall within the scope of the invention.
Claims (9)
1. A method of arrhythmia detection comprising:
acquiring an original electrocardiosignal to be detected;
preprocessing an original electrocardiosignal to respectively obtain a time sequence electrocardiosignal and a Hilbert spectrogram;
based on the time sequence electrocardiosignal and the Hilbert spectrogram, respectively extracting the time sequence electrocardiosignal characteristic and the Hilbert spectrogram characteristic;
processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not;
based on the positive sample and the negative sample, the contrast predictive coding CPC model is trained, and the trained contrast predictive coding CPC model is used for processing the time sequence electrocardiosignal and extracting the time sequence electrocardiosignal characteristics.
2. The method of claim 1, wherein acquiring the raw electrocardiographic signal to be detected comprises: and acquiring an original electrocardiosignal of a tested person, removing baseline drift and high-frequency noise interference by using an HHT band-pass filter, and removing power frequency interference by using a trap to obtain the original electrocardiosignal to be detected.
3. The arrhythmia detection method according to claim 1, wherein the raw electrocardiograph signals are normalized by using a maximum and minimum normalization method to obtain time-series electrocardiograph signals; the electrocardiosignals were converted to Hilbert spectra using HHT.
4. The method according to claim 1, wherein the characteristic of a certain moment in time-series electrocardiograph signals is extracted by using a convolutional neural network; generating fusion history information by using a cyclic neural network to generate electrocardiosignal characteristics before the certain moment; the fusion history information and the predicted electrocardiosignal characteristic after a certain moment form a positive sample, and the fusion history information and the characteristic representation of any sampled sequence point form a negative sample.
5. The method according to claim 4, wherein the electrocardiosignal characteristics after a certain time are predicted by introducing a mutual information formula based on an electrocardiosignal context relation based on an electrocardiosignal characteristic representation before the certain time.
6. The method for detecting arrhythmia according to claim 1, wherein a trained contrast learning SimCLR model is used to process Hilbert spectra and extract Hilbert spectral features.
7. The method of claim 6, wherein the basic framework of contrast learning SimCLR model includes a data enhancement module, a convolutional neural network, a fully connected network, and a contrast loss function; the data enhancement module is used for converting the picture, including turning, rotating, zooming and random cutting; the convolutional neural network is used for extracting respective characteristic representations of the converted pictures; the fully connected network is used to map the feature representation learned by the convolutional neural network to a one-dimensional feature representation.
8. An arrhythmia detection system, comprising:
the acquisition module is used for acquiring an original electrocardiosignal to be detected;
the preprocessing module is used for preprocessing the original electrocardiosignals to respectively obtain time sequence electrocardiosignals and Hilbert spectrograms;
the extraction module is used for respectively extracting the time sequence electrocardiosignal characteristics and the Hilbert spectrum characteristics based on the time sequence electrocardiosignal and the Hilbert spectrum;
the detection module is used for processing the sequential electrocardiosignal characteristics and the Hilbert spectrum characteristics by using a trained detection model to obtain a final detection result; the detection result comprises whether the original electrocardiosignal to be detected is an arrhythmia signal or not;
based on the positive sample and the negative sample, the contrast predictive coding CPC model is trained, and the trained contrast predictive coding CPC model is used for processing the time sequence electrocardiosignal and extracting the time sequence electrocardiosignal characteristics.
9. An electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and wherein the computer program is stored in the memory, which processor, when the electronic device is running, executes the computer program stored in the memory to cause the electronic device to execute instructions implementing the arrhythmia detection method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110983506.9A CN113598784B (en) | 2021-08-25 | 2021-08-25 | Arrhythmia detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110983506.9A CN113598784B (en) | 2021-08-25 | 2021-08-25 | Arrhythmia detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113598784A CN113598784A (en) | 2021-11-05 |
| CN113598784B true CN113598784B (en) | 2024-04-09 |
Family
ID=78341997
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110983506.9A Active CN113598784B (en) | 2021-08-25 | 2021-08-25 | Arrhythmia detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113598784B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115294342B (en) * | 2022-09-26 | 2023-02-28 | 荣耀终端有限公司 | Image processing method and related device |
| CN118285809B (en) * | 2024-03-21 | 2024-09-03 | 济南大学 | Multi-feature auxiliary diagnosis method based on double-lead ECG signal, terminal and readable storage medium |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107811626A (en) * | 2017-09-10 | 2018-03-20 | 天津大学 | A kind of arrhythmia classification method based on one-dimensional convolutional neural networks and S-transformation |
| CN109497991A (en) * | 2019-01-04 | 2019-03-22 | 济南汇医融工科技有限公司 | Heart failure detection device based on machine learning |
| CN109745033A (en) * | 2018-12-25 | 2019-05-14 | 东南大学 | Dynamic electrocardiogram method for evaluating quality based on time-frequency two-dimensional image and machine learning |
| CN109953755A (en) * | 2019-03-15 | 2019-07-02 | 度特斯(大连)实业有限公司 | A kind of extracting method and device of electrocardial vector data characteristics |
| CN110090012A (en) * | 2019-03-15 | 2019-08-06 | 上海图灵医疗科技有限公司 | A kind of human body diseases detection method and testing product based on machine learning |
| CN110537910A (en) * | 2019-09-18 | 2019-12-06 | 山东大学 | coronary heart disease noninvasive screening system based on electrocardio and heart sound signal joint analysis |
| CN110974172A (en) * | 2019-12-13 | 2020-04-10 | 北京理工大学 | Real-time physiological parameter measuring system |
| CN111067505A (en) * | 2019-12-25 | 2020-04-28 | 北京多唯阳光科技有限责任公司 | Arrhythmia detection method, device and storage medium |
| CN111449644A (en) * | 2020-03-19 | 2020-07-28 | 复旦大学 | Bioelectricity signal classification method based on time-frequency transformation and data enhancement technology |
| CN111588370A (en) * | 2020-05-18 | 2020-08-28 | 胡硕 | Heart signal integrated intelligent acquisition sensing system terminal, method and cloud control platform |
| CN112329609A (en) * | 2020-11-03 | 2021-02-05 | 山东大学 | Feature fusion transfer learning arrhythmia classification system based on 2D heart beat |
| CN112426160A (en) * | 2020-11-30 | 2021-03-02 | 贵州省人民医院 | Electrocardiosignal type identification method and device |
| CN112617849A (en) * | 2020-12-31 | 2021-04-09 | 山西三友和智慧信息技术股份有限公司 | Atrial fibrillation detection and classification method based on CNN + LSTM |
| CN112674779A (en) * | 2020-12-25 | 2021-04-20 | 西华大学 | Automatic sleep staging method based on ECG (electrocardiogram) signals |
| CN112932498A (en) * | 2021-01-29 | 2021-06-11 | 山东大学 | T wave morphology classification system with strong generalization capability based on deep learning |
| CN113116300A (en) * | 2021-03-12 | 2021-07-16 | 复旦大学 | Physiological signal classification method based on model fusion |
| CN113288156A (en) * | 2021-05-18 | 2021-08-24 | 浙江大学 | Method for generating electrocardiographic data of any lead visual angle |
| WO2023025650A1 (en) * | 2021-08-26 | 2023-03-02 | Nemo Healthcare B.V. | Fetal heart rate signal processing |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017084546A1 (en) * | 2015-11-17 | 2017-05-26 | 安徽华米信息科技有限公司 | Wearable device-based user interest information determination method, device and wearable device |
| CN111183424B (en) * | 2017-08-30 | 2024-04-12 | 深圳市长桑技术有限公司 | System and method for identifying users |
| US11311201B2 (en) * | 2018-11-02 | 2022-04-26 | Samsung Electronics Co., Ltd. | Feature selection for cardiac arrhythmia classification and screening |
| US20220261651A1 (en) * | 2021-02-08 | 2022-08-18 | Salesforce.Com, Inc. | Systems and methods for self supervised multi-view representation learning for time series |
-
2021
- 2021-08-25 CN CN202110983506.9A patent/CN113598784B/en active Active
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107811626A (en) * | 2017-09-10 | 2018-03-20 | 天津大学 | A kind of arrhythmia classification method based on one-dimensional convolutional neural networks and S-transformation |
| CN109745033A (en) * | 2018-12-25 | 2019-05-14 | 东南大学 | Dynamic electrocardiogram method for evaluating quality based on time-frequency two-dimensional image and machine learning |
| CN109497991A (en) * | 2019-01-04 | 2019-03-22 | 济南汇医融工科技有限公司 | Heart failure detection device based on machine learning |
| CN109953755A (en) * | 2019-03-15 | 2019-07-02 | 度特斯(大连)实业有限公司 | A kind of extracting method and device of electrocardial vector data characteristics |
| CN110090012A (en) * | 2019-03-15 | 2019-08-06 | 上海图灵医疗科技有限公司 | A kind of human body diseases detection method and testing product based on machine learning |
| CN110537910A (en) * | 2019-09-18 | 2019-12-06 | 山东大学 | coronary heart disease noninvasive screening system based on electrocardio and heart sound signal joint analysis |
| CN110974172A (en) * | 2019-12-13 | 2020-04-10 | 北京理工大学 | Real-time physiological parameter measuring system |
| CN111067505A (en) * | 2019-12-25 | 2020-04-28 | 北京多唯阳光科技有限责任公司 | Arrhythmia detection method, device and storage medium |
| CN111449644A (en) * | 2020-03-19 | 2020-07-28 | 复旦大学 | Bioelectricity signal classification method based on time-frequency transformation and data enhancement technology |
| CN111588370A (en) * | 2020-05-18 | 2020-08-28 | 胡硕 | Heart signal integrated intelligent acquisition sensing system terminal, method and cloud control platform |
| CN112329609A (en) * | 2020-11-03 | 2021-02-05 | 山东大学 | Feature fusion transfer learning arrhythmia classification system based on 2D heart beat |
| CN112426160A (en) * | 2020-11-30 | 2021-03-02 | 贵州省人民医院 | Electrocardiosignal type identification method and device |
| CN112674779A (en) * | 2020-12-25 | 2021-04-20 | 西华大学 | Automatic sleep staging method based on ECG (electrocardiogram) signals |
| CN112617849A (en) * | 2020-12-31 | 2021-04-09 | 山西三友和智慧信息技术股份有限公司 | Atrial fibrillation detection and classification method based on CNN + LSTM |
| CN112932498A (en) * | 2021-01-29 | 2021-06-11 | 山东大学 | T wave morphology classification system with strong generalization capability based on deep learning |
| CN113116300A (en) * | 2021-03-12 | 2021-07-16 | 复旦大学 | Physiological signal classification method based on model fusion |
| CN113288156A (en) * | 2021-05-18 | 2021-08-24 | 浙江大学 | Method for generating electrocardiographic data of any lead visual angle |
| WO2023025650A1 (en) * | 2021-08-26 | 2023-03-02 | Nemo Healthcare B.V. | Fetal heart rate signal processing |
Non-Patent Citations (1)
| Title |
|---|
| 基于机器学习方法的心电信号分类研究;章洁;武东;;科技风;20190610(第16期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113598784A (en) | 2021-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113598784B (en) | Arrhythmia detection method and system | |
| Singh et al. | Classification of unsegmented heart sound recording using KNN classifier | |
| CN111759345B (en) | Heart valve abnormality analysis method, system and device based on convolutional neural network | |
| Abdalla et al. | ECG arrhythmia classification using artificial intelligence and nonlinear and nonstationary decomposition | |
| CN112036467B (en) | Abnormal heart sound identification method and device based on multi-scale attention neural network | |
| CN108511055B (en) | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules | |
| CN111449644A (en) | Bioelectricity signal classification method based on time-frequency transformation and data enhancement technology | |
| CN110705456A (en) | Micro motor abnormity detection method based on transfer learning | |
| CN111956214A (en) | QRS wave automatic detection method based on U-net end-to-end neural network model | |
| CN111626114A (en) | Electrocardiosignal arrhythmia classification system based on convolutional neural network | |
| CN113076878B (en) | Constitution identification method based on attention mechanism convolution network structure | |
| CN112426160A (en) | Electrocardiosignal type identification method and device | |
| CN115281676B (en) | Fatigue detection method based on GRU neural network and ECG signal | |
| CN113349793A (en) | Arrhythmia classification method based on layered attention model | |
| CN118035926B (en) | Model training and water detection method and system based on multivariate data diffusion | |
| CN113796873B (en) | Wearable dynamic electrocardiosignal classification method and system | |
| Jasim | ECG signal classification based on deep learning by using convolutional neural network (CNN) | |
| Tiwari et al. | Heartbeat sound classification using Mel-frequency cepstral coefficients and deep convolutional neural network | |
| Zhang et al. | Research on lung sound classification model based on dual-channel CNN-LSTM algorithm | |
| CN116584902A (en) | Heart sound classification device based on feature optimization and visualization | |
| CN116340812A (en) | Transformer partial discharge fault mode identification method and system | |
| CN116491956A (en) | Abnormal electroencephalogram automatic detection method based on multi-domain feature fusion | |
| Mitrokhin et al. | Deep learning approach for QRS wave detection in ECG monitoring | |
| Raghav et al. | Fractal feature based ECG arrhythmia classification | |
| CN116712089B (en) | Epileptiform discharge enriching epileptiform interval and method for predicting focus |
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 |