CN113974647A - System and method for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine - Google Patents
System and method for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine Download PDFInfo
- Publication number
- CN113974647A CN113974647A CN202111251594.XA CN202111251594A CN113974647A CN 113974647 A CN113974647 A CN 113974647A CN 202111251594 A CN202111251594 A CN 202111251594A CN 113974647 A CN113974647 A CN 113974647A
- Authority
- CN
- China
- Prior art keywords
- sudden cardiac
- cardiac death
- model
- wave
- sudden
- 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.)
- Granted
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- 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
- A61B5/352—Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
-
- 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
- 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
- A61B5/7253—Details of waveform analysis characterised by using transforms
- A61B5/726—Details of waveform analysis characterised by using transforms using Wavelet 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
-
- 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
-
- 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
- A61B5/7271—Specific aspects of physiological measurement 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
- A61B5/7271—Specific aspects of physiological measurement analysis
- A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
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)
- Cardiology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (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 provides a system and a method for reconstructing sudden cardiac death risk factors by multi-feature quantization and model parameter optimization of a nonlinear support vector machine, which comprises the steps of carrying out data preprocessing on a sudden cardiac death cardiac signal data set and a normal sinus rhythm cardiac signal data set; performing electrocardiographic waveform detection on the processed electrocardiographic data set; extracting the sudden cardiac death risk factor; carrying out characteristic quantization scaling processing on the extracted initial characteristics; a nonlinear support vector machine is used as a verification model of the sudden cardiac death risk factor, and an error penalty parameter C and a nuclear parameter gamma are determined through model parameter optimization; obtaining a sudden cardiac death prediction model through the formulated sudden cardiac death risk factor and the optimized model parameters; the effects of reconstructing and verifying the sudden cardiac death risk factor are achieved, and the method has good guiding significance for researching sudden cardiac death.
Description
Technical Field
The invention belongs to the technical field of machine learning and electrocardiosignal splitting processing, and particularly relates to a system and a method for reconstructing sudden cardiac death risk factors through multi-feature quantization and model parameter optimization of a nonlinear support vector machine.
Background
Sudden Cardiac Death (SCD) refers to unpredictable Sudden Death occurring immediately or within 24 hours. Sudden Cardiac death patients can feel obvious symptoms such as dyspnea, chest pain, palpitation and the like during the common attack period of about 1 hour, and then the Sudden Cardiac Arrest (SCA) phenomenon appears, and if the Sudden Cardiac death patients cannot be effectively rescued within a few minutes, the Sudden Cardiac death patients can be caused. Because sudden cardiac death is paroxysmal, even people without cardiovascular disease history or with slight symptoms can induce sudden cardiac arrest due to poor life style such as smoking and drinking and dangerous inducers such as obesity, hypertension and diabetes, which leads to subconscious loss and sudden cardiac death. Medical imaging and electrocardiography are the main means for clinical risk assessment of sudden cardiac death. Intrinsic activities and changes of the heart can be observed visually through medical images, but because of the limitation of observation time and equipment, heart information cannot be acquired flexibly in real time, and the electrocardiogram is widely applied to heart health monitoring with noninvasive real-time convenience.
The current exploration of sudden cardiac death causes is mainly from the medical point of view, and the specific wave bands in the electrocardiogram are observed in a targeted way based on physiology and anatomy, but the above method is limited by the limitation of human knowledge, so that some hidden pathogenic factors are not paid attention to sometimes. Under the background of the combination of medical work at present, how to utilize the powerful data processing capability of a computer and a related data processing method in the field of artificial intelligence to discover potential hidden factors of sudden cardiac death or discover the influence of each factor on the sudden cardiac death is an urgent problem to be solved.
Although the sudden cardiac death factor is not found to be enough for definite diagnosis or treatment of sudden cardiac death, the screening thereof still has important reference value for relevant research.
Disclosure of Invention
Aiming at the defects and shortcomings in the prior art, the invention aims to provide a system and a method for reconstructing sudden cardiac death risk factors by multi-feature quantization and model parameter optimization of a nonlinear support vector machine, which comprises the steps of carrying out data preprocessing on a sudden cardiac death cardiac signal data set and a normal sinus rhythm cardiac signal data set; performing electrocardiographic waveform detection on the processed electrocardiographic data set; extracting the sudden cardiac death risk factor; carrying out characteristic quantization scaling processing on the extracted initial characteristics; a nonlinear support vector machine is used as a verification model of the sudden cardiac death risk factor, and an error penalty parameter C and a nuclear parameter gamma are determined through model parameter optimization; obtaining a sudden cardiac death prediction model through the formulated sudden cardiac death risk factor and the optimized model parameters; the effects of reconstructing and verifying the sudden cardiac death risk factor are achieved, and the method has good guiding significance for researching sudden cardiac death.
The invention can specifically adopt the following technical scheme:
a system for reconstructing sudden cardiac death risk factors through multi-feature quantization and model parameter optimization of a nonlinear support vector machine is characterized by comprising the following steps:
the preprocessing module is used for preprocessing the data of the cardiac sudden death cardiac signal data set and the normal sinus rhythm cardiac signal data set;
the waveform detection module is used for performing electrocardiographic waveform detection on the preprocessed electrocardiographic data set to obtain an electrocardiographic waveform for extracting sudden death risk factors;
the initial characteristic acquisition module is used for extracting sudden cardiac death risk factors needing to be confirmed according to the electrocardiographic waveform to obtain electrocardiographic initial characteristics;
the characteristic set construction module is used for processing the electrocardio initial characteristics to obtain a training set;
the characteristic optimizing module is used for obtaining a prediction model of the sudden cardiac death according to the training set through formulated sudden cardiac death risk factors and optimized model parameters, and verifying and recombining the sudden cardiac death risk factors through model feedback; and taking a nonlinear support vector machine as a verification model of the sudden cardiac death risk factor, optimizing model parameters based on a Gaussian kernel function, and determining an error penalty parameter C and a kernel parameter gamma.
Further, the preprocessing module comprises: the device comprises a data cutting sub-module, a frequency band decomposition sub-module, a noise filtering sub-module and a signal reconstruction sub-module.
Further, the waveform detection module detects a QRS complex, compares the sizes of waveform integral areas under fixed-size sliding windows in a retrieval interval based on an RR interval, and detects the tail end of a T wave and the peak of the T wave.
Further, the pair of initial feature acquisition modules includes: potential factors of RR interval, QRS time limit, QTc interval, T peak-to-end interval, Tp-Te/QT index, T wave amplitude and heart rate variability of the electrocardiosignal sequence are extracted.
Further, the feature set construction module calculates the average value, the standard deviation and the approximate entropy of the initial features for constructing a training set.
Further, in the feature set construction module, after constructing the training set, the method further includes the operation of performing feature scaling processing on the initial features: firstly, the difference of dimensions among the features is determined, and then the minimum maximum value scaling or standard scaling is selected to process each feature to obtain a feature value with smaller difference in the same dimension data.
Further, the feature optimizing module utilizes an improved grid search method and a quantum particle swarm optimization algorithm as parameter optimizing algorithms respectively, compares the influence of the two parameter optimizing algorithms on the performance of the sudden cardiac death prediction model, and determines an error punishment parameter C and a nuclear parameter gamma; sensitivity Se, specificity Sp, positive prediction rate Ppv and accuracy Acc are used as indexes for model evaluation.
Further, classifying the electrocardio data 20-30 min, 30-40 min, 40-50 min and 60-70 min before the sudden cardiac death occurs as a sudden cardiac death electrocardio signal data set; the feature optimizing module outputs four models respectively corresponding to four time periods.
And, a method for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of a nonlinear support vector machine, characterized by comprising the following steps:
step S1: carrying out data preprocessing on the sudden cardiac death cardiac signal data set and the normal sinus rhythm cardiac signal data set;
step S2: performing electrocardiographic waveform detection on the electrocardiographic data set processed in the step S1 to obtain an electrocardiographic waveform for extracting sudden death risk factors; detecting a QRS complex based on an R wave peak detection method for constructing an adaptive threshold and a QRS complex detection algorithm of a multi-decision rule, comparing the sizes of waveform integral areas under fixed-size sliding windows in an RR interval-based retrieval interval, and detecting the tail end of a T wave and the peak of the T wave;
step S3: extracting sudden cardiac death risk factors needing to be confirmed to obtain initial electrocardio characteristics, and constructing a characteristic set suitable for model training according to the respective average value, standard deviation and approximate entropy of the initial characteristics;
step S4: performing feature scaling processing on the initial features extracted in step S3;
step S5: taking a nonlinear support vector machine as a verification model of the sudden cardiac death risk factor, optimizing model parameters based on a Gaussian kernel function, and determining an error punishment parameter C and a kernel parameter gamma;
step S6: and obtaining a prediction model of the sudden cardiac death through the formulated sudden cardiac death risk factor and the optimized model parameters, and verifying and recombining the sudden cardiac death risk factor through model feedback.
Further, step S1 is specifically:
step S11: classifying the cardiac sudden death electrocardiogram data according to four time periods of 20-30 min, 30-40 min, 40-50 min and 60-70 min before the sudden cardiac arrest event occurs, and then cutting the cardiac sudden death signal of each time period into data with the time length of 1 min;
step S12: for normal sinus rhythm data, randomly taking a length of 10min for each note of record, and cutting the note into a plurality of lengths of 1 min;
step S13: processing the electrocardiosignals by using discrete wavelet transform and Mallat algorithm to obtain a plurality of layers of electrocardio decomposition signals with different frequency bands;
step S14: carrying out threshold processing on the decomposed signal by using a threshold value and a threshold value function, and filtering the electrocardio noise signal;
step S15: reconstructing the multilayer decomposed signals with the noise filtered out to obtain filtered electrocardiosignals;
step S2 specifically includes:
step S21: the filtered electrocardiosignals pass through a low-pass filter and a high-pass filter, and the two filters form a band-pass filter;
step S22: carrying out point-by-point derivation on each sample point, taking an absolute value, amplifying the slope information of the R wave, and finally carrying out window integration on the signal;
step S23: correcting the R wave determined by the R wave detection rule, and determining the positions of a Q wave crest and an S wave crest based on an RR interval;
step S24: integrating waveforms in a T wave tail end retrieval interval by using a sliding window with a fixed length, comparing an integration result to determine a T wave tail end, and then forwardly retrieving according to the T wave tail end to determine a T wave peak value position;
in step S4, the difference between the dimensions of the features is determined, and the minimum maximum value scaling or the standard scaling is selected to process each feature, so as to obtain a feature value with a small difference between the data in the same dimension;
step S5 specifically includes:
step S51: dividing four data sets corresponding to the four time periods into a training set and a testing set according to a certain proportion, and sending the training set into a support vector machine;
step S52: the method comprises the steps that an improved grid searching method and a quantum particle swarm optimization algorithm are used as parameter optimization algorithms, the influence of the two parameter optimization algorithms on the performance of a sudden cardiac death prediction model is compared, and an error punishment parameter C and a nuclear parameter gamma are determined;
in step S6, four models are output for each of the four time periods, and the sensitivity Se, the specificity Sp, the positive prediction rate Ppv, and the accuracy Acc are used as the indices for model evaluation.
Compared with the prior art, the method and the preferred scheme have the advantages that the difference between the normal sinus rhythm and the electrocardiosignals before sudden cardiac death is quantified, the characteristic set is used for representing the prediction information which is hidden by the electrocardiosignals which look normal before sudden cardiac death, a methodology and a specific machine learning construction scheme are provided for exploring the sudden cardiac death risk factor, and the neglect of the human subjective factors of the possible sudden cardiac death risk factor due to the limitation of the existing knowledge is compensated. In the prior art, basically, under the condition of ensuring higher prediction accuracy, the prediction time is 20 minutes before the sudden cardiac death occurs, the method autonomously establishes and optimizes the model, improves the prediction time to 70 minutes before the sudden cardiac death occurs on the basis of reconstructing and verifying the sudden death risk factor, and guarantees the reliability of the extracted sudden death risk factor based on model training.
Drawings
The invention is described in further detail below with reference to the following figures and detailed description:
fig. 1 is a schematic diagram of a system structure and a work flow according to an embodiment of the present invention.
FIG. 2 is a schematic diagram of a model parameter optimization training process according to an embodiment of the present invention.
Detailed Description
In order to make the features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail as follows:
it should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
As shown in fig. 1, the scheme for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of a nonlinear support vector machine provided by the present invention mainly includes: data preprocessing, waveform detection, initial feature extraction and feature set construction, feature scaling, model parameter optimization model optimization, final verification effect feedback reconstruction risk factor and the like;
the data preprocessing specifically comprises the following steps:
acquiring electrocardiogram data from a Sudden Cardiac Death dynamic electrocardiogram Database (Sudden Cardiac layered Holter Database) and an MIT-BIH Normal Sinus Rhythm Database (MIT-BIH Normal sine Rhythm Database), wherein the Sudden Cardiac Death dynamic electrocardiogram Database has 23 complete Holter records in total and is numbered as 30-52. Each recording contains 2 leads, the sampling rate is 250Hz, and the duration of the recording varies from hours to more than 20 hours. Since the electrocardiographic records numbered 40, 42, 49 do not mark the time at which the cardiac arrest event occurred; recording 35 that the noise interference of the electrocardiosignals is serious, and the waveform detection effect is poor to influence the prediction performance of the model; the electrocardiogram before the sudden cardiac death shows ventricular tachycardia for a long time period, has obvious waveform abnormality and low prediction value, so the 5-stroke electrocardiogram recording is not used in the embodiment. The MIT-BIH normal sinus rhythm database has 18 records, each record also containing 2 leads, with a sampling rate of 128Hz and a recording duration of about 24 hours.
Classifying the cardiac sudden death cardiac electric data according to four time periods of 20-30 min, 30-40 min, 40-50 min and 60-70 min before the sudden cardiac arrest event occurs. Then cutting the electrocardio data of each time period into time length of 1min, and rejecting samples with obvious ventricular arrhythmia or serious noise interference. For normal sinus rhythm data, a length of 10min was randomly taken per note and cut into several 1min lengths.
In the embodiment, based on discrete wavelet transform filtering, a db8 wavelet function is selected according to morphological characteristics of an electrocardiosignal, 8 layers of decomposition layers are selected according to the sampling frequency of an original signal and the distribution frequency band of noise, and then, the original electrocardiosignal is subjected to wavelet decomposition by using a Mallat-based algorithm. And selecting a threshold value and a threshold value function to carry out threshold value processing on the frequency band needing filtering after decomposition, and finally carrying out wavelet inverse decomposition to reconstruct and recover the filtered electrocardiosignal.
The waveform detection specifically comprises the following steps:
the QRS wave group and the T wave which are generally concerned by the sudden cardiac death at present are extracted, and meanwhile, the possibility that the waveform which is not generally concerned by potential factors such as J wave, P wave and the like is used as a waveform detection source cannot be ignored.
The QRS complex detection algorithm firstly preprocesses the electrocardiosignals denoised by the discrete wavelet transform, so that the amplitude of the R wave is highlighted, and the amplitudes of other waveforms are reduced. And then the position of the R wave crest is determined according to the R wave detection rule, and then the position of the R wave crest is adjusted and corrected within a range. And finally, determining self-adaptive Q wave and S wave detection windows according to the average RR interval, and determining the positions of Q wave peaks and S wave peaks according to the derivative of sample points in the detection windows.
Preprocessing is carried out before R wave detection, so that the waveform form of the R wave becomes obvious. Firstly, the filtered electrocardiosignals pass through a low-pass filter and a high-pass filter, the two filters form a band-pass filter, the interference of other waveforms such as P waves, T waves and the like is reduced, and the frequency band with concentrated QRS wave group energy is reserved; then, point-by-point derivation is carried out on each sample point and the absolute value is takenAnd amplifying the slope information of the R wave, and finally performing window integration on the signal to smooth the signal. After the wave peak position of the R wave is determined, the average value of RR intervals of the electrocardiosignals is obtained
For the kth R wave peak position RkThe Q wave peak detection window will be defined. Initializing Q waveguide number prevDiffQ, and sequentially calculating derivative diffQ of each sample point in a Q wave detection window from right to left when
Then, this point is regarded as the peak of the Q wave. Defining the S wave peak detection window as
Initializing an S wave guide number prevDiffS, and sequentially calculating derivative diffS of each sample point in an S wave detection window from left to right when
Then, this point is regarded as the S-wave peak.
The specific steps of T wave detection are as follows:
assuming that the sampling frequency of the electrocardiosignal s (k) is fs, the peak R of the ith R waveiAnd the (i + 1) th R wave peak Ri+1Between the end of the T wave and the peak of the T wave, RiAnd Ri+1Is defined as RRi. Defining the T wave end search interval as [ k ]a,kb]，kaAnd kbAre all located at RiAnd Ri+1In the meantime.
For k ═ ka,ka+1,…,kbCalculating the mean value of the corresponding electrocardiosignal amplitudes of p sample points before and after each k value
And the integrated area A of k under a sliding window of length wk：
The size w of the sliding window should not exceed the width of the T wave, otherwise a higher amplitude band than the T wave may be included in the integration range, making the detection of the end of the T wave incorrect. But the width of the T wave is actually unknown, and w is chosen in the patent to be an empirical value 32. The smoothing window parameter p is used for smoothing the signal amplitude of the k point and reducing the noise influence, p is far smaller than w, otherwise, the calculated value is
The signal amplitude of the k point cannot be effectively represented, and p is 4.
Find out [ k ]a,kb]Points k' and k "corresponding to the maximum and minimum areas within the range:
when k' corresponds to area Ak′Area A corresponding to k ″k″Satisfies the following conditions:
the T wave is considered as a biphase T wave, the relatively larger point of k 'and k' is taken as the T wave terminal Tend, otherwise | Ak′Ak and Ak″And taking the k point corresponding to the larger value in the | as the T wave terminal Tend. Where the parameter λ is to ensure that a T-wave, which may be a biphasic T-wave morphology, can be identified. The biphase T wave has a positive T wave crest and a negative T wave crest, and the peak values of the two wave crests have small difference, so the obtained Ak′And Ak″The present embodiment can set λ to 6 within λ times.
In [ k ]a,Tend]And finding the point with the maximum amplitude difference value with the Tend in the range to be the T wave peak Tpeak.
The initial feature extraction and feature set construction specifically comprises the following steps:
the filtered electrocardiosignal is represented by s (x), wherein s (x) has n cardiac cycles and the sampling frequency is fs. Besides the heart rate variability index rMSSD, factors such as RR intervals, QRS time limit, QTc intervals, Tp-Te/QT indexes, T wave amplitude values and the like of different samples have different initial characteristics, and a characteristic set cannot be directly formed for model training. In this embodiment, a feature set suitable for model training is constructed by calculating the average value, standard deviation and approximate entropy of each of the initial features, and finally, elements in the feature set and the initial features corresponding to the elements are determined.
The scaling of the features is specifically as follows:
the features selected in this embodiment are substantially similar to normal distribution, wherein the feature value of the mean value Tamp _ mean of the amplitude of the feature T wave is distributed between [0,3], and the feature value of the standard deviation QRS _ std of the feature QRS time limit is distributed between [0,0.1], and there are differences in dimension, and uniform dimension or normalization processing is required. The features are subjected to standard scaling, and the variation range of each feature after the normalization processing is in the same order of magnitude, so that the support vector machine learning is facilitated. Feature scaling using normalization may be prioritized when the feature distribution is similar to a normal distribution.
The model parameter optimizing optimization model specifically comprises the following steps:
the feature-scaled feature set is divided into four data sets DS1, DS2, DS3 and DS4, which contain the same normal sinus rhythm cardiac features but different sudden cardiac death cardiac characteristics. The data sets DS1, DS2, DS3 and DS4 respectively comprise the electrocardio characteristics of 20-30 min, 30-40 min, 40-50 min and 60-70 min before sudden cardiac death. The sudden cardiac death cardiac characteristics and the normal sinus rhythm cardiac characteristics in the four data sets are randomly distributed into a training set and a testing set according to a certain proportion, and the recording numbers of the training set and the testing set of the prediction model in the corresponding cardiac database are completely separated.
And selecting a nonlinear support vector machine as a basic model for predicting sudden cardiac death, and selecting a Gaussian kernel function as a kernel function of the support vector machine. There are two parameters for this support vector machine: an error penalty parameter C and a kernel parameter γ. And selecting an improved grid search method and a quantum particle swarm optimization algorithm as parameter optimization algorithms, and comparing the influence of the two parameter optimization algorithms on the performance of the sudden cardiac death prediction model.
And (3) carrying out repeated independent experiments on the support vector machine by applying an improved grid search method, wherein the training set and the test set which are divided in each experiment are different. And setting the search ranges of the parameters C and gamma, wherein the C and the gamma form a grid, and each point in the grid corresponds to a combination of the C and the gamma. And subdividing the training set into a new training set and a new verification set by using a K-fold cross validation method. C corresponding to the model with the highest average classification accuracy*And gamma*The value is used as the optimal parameter combination of the initial search, and the parameter C is set to be [0.1C ] again in a certain step length*,10C*]Sequentially taking values in the range, and setting the parameter gamma to be [0.1 gamma ] in a certain step length*,10γ*]And sequentially taking values in the range, repeating the steps to obtain C and gamma corresponding to the new model with the highest classification accuracy as the optimal parameter combination for fine search, and verifying the two support vector machines corresponding to the optimal parameter combination for preliminary search and the optimal parameter combination for fine search by using a test set.
From the accuracy of the model corresponding to the combination of the two optimal parameters of the initial search and the fine search in the improved grid search method on the test set, it can be seen that the accuracy of the prediction model can be improved by performing the fine search again on the basis of the initial search. In this embodiment, C and γ corresponding to the model with the highest average classification accuracy in the verification set are used as the optimal parameter combination, but the accuracy corresponding to a plurality of C and γ may be the highest in the grid, and at this time, a group with the smallest values of C and γ is selected as the optimal parameter combination by default, but the performance of the support vector machine corresponding to the parameter combination on the test set is not necessarily the best. Thus, the better results performed on the test set in both the preliminary search and the fine search are taken as the overall accuracy of the improved grid search method.
And (3) carrying out repeated independent experiments on the quantum particle swarm optimization algorithm, wherein a training set and a test set used in each experiment correspond to the training set and the test set of the independent repeated experiment in the improved grid search method one by one, and the classification performance of the support vector machine model based on the two parameter optimization algorithms on the test set is transversely compared. Defining a particle group X consisting of N particles, wherein the current position of the particle i at the time t is X _ i (t), the individual optimal position is Pbest _ i (t), the global optimal position of all the particles is Gbest (t), and the global optimal position of the particle group in the whole optimizing process is Gbest. The number of the particles is set, and the number of the optimized parameters determines the dimension of the particles. And defining a fitness function for evaluating whether the current position of the particle is the optimal position, wherein the fitness function is a support vector machine based on a Gaussian kernel, and the smaller the fitness value is, the better the fitness of the particle is, and the closer the fitness function is to the position of the optimal parameter.
The parameter searching process of the quantum particle swarm optimization algorithm is strong in randomness, so that the consumed time is only 1/10 of the grid searching method. In the case of a comparable classification accuracy, sensitivity can be another relatively important indicator of identifying the effects of sudden cardiac death. The sudden cardiac death prediction period selected by the embodiment is 20-70 min before sudden cardiac arrest occurs, so that under the condition of almost poor identification accuracy, the embodiment is more prone to selecting a grid search method with higher sensitivity as a parameter optimization algorithm of the support vector machine. Of course, if the prediction period is within 10min, the quantum-behaved particle swarm optimization algorithm or other less time-consuming parameter optimization algorithm should be considered preferentially, or the search step size in the grid search method should be increased by some.
The verification effect feedback reconstruction risk factor is specifically as follows:
the following four indexes are adopted in the embodiment to evaluate the performance of the sudden death risk factor reconstruction verification model:
sensitivity (Se): the ratio of the number of SCD samples which are correctly predicted by the model to the total number of SCD samples is represented, the identification capability of the model to the sudden cardiac death samples is embodied, and the definition formula is as follows:
specificity (Sp): the ratio of the number of the NSR samples which are predicted correctly by the model to the total number of the NSR samples is represented, the recognition capability of the model to the normal sinus rhythm samples is embodied, and the definition formula is as follows:
positive Predictive rate (Positive Predictive Value, Ppv): the ratio of the number of SCD samples which are predicted correctly by the model to the number of SCD samples which are predicted correctly is expressed, the identification capability of the model to the sudden cardiac death samples is also embodied, and the definition formula is as follows:
accuracy (Accuracy, Acc): the ratio of the number of samples which are correctly predicted to be SCD class and NSR class to the total number of samples is represented, the overall prediction capability of the model is embodied, and the definition formula is as follows:
wherein True Positive (TP) indicates the number of SCD class samples predicted correctly by the model, True Negative (TN) indicates the number of NSR class samples predicted correctly, False Positive (FP) indicates the number of samples predicted as SCD class by the NSR class, and False Negative (FN) indicates the number of samples predicted as NSR class by the SCD class. The data set has balanced samples of sudden cardiac death and normal sinus rhythm, so the accuracy is the evaluation index which most intuitively reflects the overall predictive performance of the model. This embodiment focuses more on the ability to verify sudden death risk factor reconstruction than normal sinus rhythm samples, so sensitivity is another important assessment indicator. Specificity and positive prediction rate are also common indicators for evaluating classifier performance. The verification effect of the model directly reflects the prediction capability of the sudden death risk factor on the sudden death from heart, the model directly serves as feedback to act on the extraction and reconstruction stage of the sudden death risk factor, a set of methodology for verifying the sudden death risk factor is provided for people, and meanwhile the possibility that the extracted factor is effective can be improved by artificially extracting or verifying the reconstruction of the potential sudden death risk factor extracted by an intelligent algorithm.
The above scheme provided by this embodiment can be stored in a computer readable storage medium in a coded form, and implemented in a computer program, and inputs basic parameter information required for calculation through computer hardware, and outputs a calculation result.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, apparatus, 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 of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each flow of the flowcharts, and combinations of flows in the flowcharts, 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.
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.
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.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
The present invention is not limited to the above-mentioned preferred embodiments, and any system and method for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machines in various forms can be derived by anyone with the benefit of the present invention.
Claims (10)
1. A system for reconstructing sudden cardiac death risk factors through multi-feature quantization and model parameter optimization of a nonlinear support vector machine is characterized by comprising the following steps:
the preprocessing module is used for preprocessing the data of the cardiac sudden death cardiac signal data set and the normal sinus rhythm cardiac signal data set;
the waveform detection module is used for performing electrocardiographic waveform detection on the preprocessed electrocardiographic data set to obtain an electrocardiographic waveform for extracting sudden death risk factors;
the initial characteristic acquisition module is used for extracting sudden cardiac death risk factors needing to be confirmed according to the electrocardiographic waveform to obtain electrocardiographic initial characteristics;
the characteristic set construction module is used for processing the electrocardio initial characteristics to obtain a training set;
the characteristic optimizing module is used for obtaining a prediction model of the sudden cardiac death according to the training set through formulated sudden cardiac death risk factors and optimized model parameters, and verifying and recombining the sudden cardiac death risk factors through model feedback; taking a nonlinear support vector machine as a verification model of the sudden cardiac death risk factor, optimizing model parameters based on a Gaussian kernel function, and determining an error penalty parameter C and a kernel parameter
。
2. The system according to claim 1, wherein the system for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine comprises: the preprocessing module comprises: the device comprises a data cutting sub-module, a frequency band decomposition sub-module, a noise filtering sub-module and a signal reconstruction sub-module.
3. The system according to claim 1, wherein the system for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine comprises: the waveform detection module detects QRS wave groups, compares the waveform integral area under a sliding window with a fixed size in a retrieval interval based on an RR interphase, and detects the tail end of a T wave and the wave crest of the T wave.
4. The system according to claim 1, wherein the system for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine comprises: the pair of initial feature acquisition modules comprises: potential factors of RR interval, QRS time limit, QTc interval, T peak-to-end interval, Tp-Te/QT index, T wave amplitude and heart rate variability of the electrocardiosignal sequence are extracted.
5. The system according to claim 1, wherein the system for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine comprises: the characteristic set construction module calculates the average value, the standard deviation and the approximate entropy of the initial characteristics and is used for constructing a training set.
6. The system of claim 5, wherein the system for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine comprises: in the feature set construction module, after constructing the training set, the method further includes the operation of performing feature scaling processing on the initial features: firstly, the difference of dimensions among the features is determined, and then the minimum maximum value scaling or standard scaling is selected to process each feature to obtain a feature value with smaller difference in the same dimension data.
7. The system according to claim 1, wherein the system for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine comprises: the characteristic optimizing module respectively uses an improved grid searching method and a quantum particle swarm optimization algorithm as parameter optimizing algorithms, compares the influence of the two parameter optimizing algorithms on the performance of the sudden cardiac death prediction model, and determines an error punishment parameter C and a nuclear parameter
(ii) a Sensitivity Se, specificity Sp, positive prediction rate Ppv and accuracy Acc are used as indexes for model evaluation.
8. The system according to claim 1, wherein the system for reconstructing sudden cardiac death risk factor by multi-feature quantization and model parameter optimization of nonlinear support vector machine comprises: classifying the electrocardio data 20-30 min, 30-40 min, 40-50 min and 60-70 min before the sudden cardiac death occurs as a sudden cardiac death electrocardio signal data set; the feature optimizing module outputs four models respectively corresponding to four time periods.
9. A method for reconstructing sudden cardiac death risk factors through multi-feature quantization and model parameter optimization of a nonlinear support vector machine is characterized by comprising the following steps:
step S1: carrying out data preprocessing on the sudden cardiac death cardiac signal data set and the normal sinus rhythm cardiac signal data set;
step S2: performing electrocardiographic waveform detection on the electrocardiographic data set processed in the step S1 to obtain an electrocardiographic waveform for extracting sudden death risk factors; detecting a QRS complex based on an R wave peak detection method for constructing an adaptive threshold and a QRS complex detection algorithm of a multi-decision rule, comparing the sizes of waveform integral areas under fixed-size sliding windows in an RR interval-based retrieval interval, and detecting the tail end of a T wave and the peak of the T wave;
step S3: extracting sudden cardiac death risk factors needing to be confirmed to obtain initial electrocardio characteristics, and constructing a characteristic set suitable for model training according to the respective average value, standard deviation and approximate entropy of the initial characteristics;
step S4: performing feature scaling processing on the initial features extracted in step S3;
step S5: taking a nonlinear support vector machine as a verification model of the sudden cardiac death risk factor, optimizing model parameters based on a Gaussian kernel function, and determining an error penalty parameter C and a kernel parameter
；
Step S6: and obtaining a prediction model of the sudden cardiac death through the formulated sudden cardiac death risk factor and the optimized model parameters, and verifying and recombining the sudden cardiac death risk factor through model feedback.
10. The method of claim 9, wherein the method comprises the steps of:
step S1 specifically includes:
step S11: classifying the cardiac sudden death electrocardiogram data according to four time periods of 20-30 min, 30-40 min, 40-50 min and 60-70 min before the sudden cardiac arrest event occurs, and then cutting the cardiac sudden death signal of each time period into data with the time length of 1 min;
step S12: for normal sinus rhythm data, randomly taking a length of 10min for each note of record, and cutting the note into a plurality of lengths of 1 min;
step S13: processing the electrocardiosignals by using discrete wavelet transform and Mallat algorithm to obtain a plurality of layers of electrocardio decomposition signals with different frequency bands;
step S14: carrying out threshold processing on the decomposed signal by using a threshold value and a threshold value function, and filtering the electrocardio noise signal;
step S15: reconstructing the multilayer decomposed signals with the noise filtered out to obtain filtered electrocardiosignals;
step S2 specifically includes:
step S21: the filtered electrocardiosignals pass through a low-pass filter and a high-pass filter, and the two filters form a band-pass filter;
step S22: carrying out point-by-point derivation on each sample point, taking an absolute value, amplifying the slope information of the R wave, and finally carrying out window integration on the signal;
step S23: correcting the R wave determined by the R wave detection rule, and determining the positions of a Q wave crest and an S wave crest based on an RR interval;
step S24: integrating waveforms in a T wave tail end retrieval interval by using a sliding window with a fixed length, comparing an integration result to determine a T wave tail end, and then forwardly retrieving according to the T wave tail end to determine a T wave peak value position;
in step S4, the difference between the dimensions of the features is determined, and the minimum maximum value scaling or the standard scaling is selected to process each feature, so as to obtain a feature value with a small difference between the data in the same dimension;
step S5 specifically includes:
step S51: dividing four data sets corresponding to the four time periods into a training set and a testing set according to a certain proportion, and sending the training set into a support vector machine;
step S52: an improved grid search method and a quantum particle swarm optimization algorithm are used as parameter optimization algorithms, the influence of the two parameter optimization algorithms on the performance of the sudden cardiac death prediction model is compared, and an error punishment parameter C and a nuclear parameter are determined
；
In step S6, four models are output for each of the four time periods, and the sensitivity Se, the specificity Sp, the positive prediction rate Ppv, and the accuracy Acc are used as the indices for model evaluation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111251594.XA CN113974647B (en) | 2021-10-26 | 2021-10-26 | System and method for reconstructing sudden cardiac death risk factor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111251594.XA CN113974647B (en) | 2021-10-26 | 2021-10-26 | System and method for reconstructing sudden cardiac death risk factor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113974647A true CN113974647A (en) | 2022-01-28 |
| CN113974647B CN113974647B (en) | 2023-08-04 |
Family
ID=79742028
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111251594.XA Active CN113974647B (en) | 2021-10-26 | 2021-10-26 | System and method for reconstructing sudden cardiac death risk factor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113974647B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110190593A1 (en) * | 2009-12-31 | 2011-08-04 | Cerner Innovation, Inc. | Computerized Systems and Methods for Stability-Theoretic Prediction and Prevention of Falls |
| CN104107042A (en) * | 2014-07-10 | 2014-10-22 | 杭州电子科技大学 | Electromyographic signal gait recognition method based on particle swarm optimization and support vector machine |
| CN106344006A (en) * | 2016-11-03 | 2017-01-25 | 太原理工大学 | J wave detection method based on pole symmetrical mode decomposition and support vector machine |
| US20200069205A1 (en) * | 2018-08-30 | 2020-03-05 | Tata Consultancy Services Limited | Non-invasive detection of coronary heart disease from short single-lead ecg |
-
2021
- 2021-10-26 CN CN202111251594.XA patent/CN113974647B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110190593A1 (en) * | 2009-12-31 | 2011-08-04 | Cerner Innovation, Inc. | Computerized Systems and Methods for Stability-Theoretic Prediction and Prevention of Falls |
| CN104107042A (en) * | 2014-07-10 | 2014-10-22 | 杭州电子科技大学 | Electromyographic signal gait recognition method based on particle swarm optimization and support vector machine |
| CN106344006A (en) * | 2016-11-03 | 2017-01-25 | 太原理工大学 | J wave detection method based on pole symmetrical mode decomposition and support vector machine |
| US20200069205A1 (en) * | 2018-08-30 | 2020-03-05 | Tata Consultancy Services Limited | Non-invasive detection of coronary heart disease from short single-lead ecg |
Non-Patent Citations (6)
| Title |
|---|
| 刘红夺儿;朱逸;湛萍;王志刚;彭屹;: "短时非线性方法用于心率变异性分析", 中国生物医学工程学报, no. 02, pages 229 - 234 * |
| 尹蕾蕾;宋涛;张英涛;王秀杰;承恒达;曲秀芬;: "基于动态心电的支持向量机模型对心肌梗死后心脏性死亡的预警研究", 中国心脏起搏与心电生理杂志, no. 05, pages 408 - 412 * |
| 张清丽;苏士美;尹咪咪;张建华;刘莹;: "基于心电特征参数的心肌梗死疾病辅助诊断模型的建立", 郑州大学学报(医学版), vol. 52, no. 02, pages 151 - 153 * |
| 张爱华;师文韬;: "基于心电信号循环平稳特征的心脏性猝死识别研究", 中国医疗器械杂志, no. 05, pages 322 - 326 * |
| 李文;王兴浩;何云霄;罗学科;: "基于支持向量机的水质浊度补偿研究", 计算机测量与控制, vol. 28, no. 06, pages 140 - 143 * |
| 王新康 等: "基于自适应向量机检测睡眠呼吸暂停综合征的最优特征组合筛选", 《中国医药导报》, vol. 16, no. 12, pages 165 - 168 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113974647B (en) | 2023-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sasikala et al. | Robust r peak and qrs detection in electrocardiogram using wavelet transform | |
| Rakshit et al. | An efficient wavelet-based automated R-peaks detection method using Hilbert transform | |
| US9050007B1 (en) | Extraction of cardiac signal data | |
| US8204583B2 (en) | System for seizure monitoring and detection | |
| EP2759065B1 (en) | Physiological signal denoising | |
| DİKER et al. | A diagnostic model for identification of myocardial infarction from electrocardiography signals | |
| Kelwade et al. | Prediction of cardiac arrhythmia using artificial neural network | |
| CN114052744B (en) | Electrocardiosignal classification method based on impulse neural network | |
| CN111358459A (en) | Arrhythmia identification method, device, equipment and storage medium | |
| Kshirsagar et al. | Classification of ECG-signals using artificial neural networks | |
| Raj et al. | Automated recognition of cardiac arrhythmias using sparse decomposition over composite dictionary | |
| Shi et al. | Automated atrial fibrillation detection based on feature fusion using discriminant canonical correlation analysis | |
| Mahesh et al. | ECG arrhythmia classification based on logistic model tree | |
| Krak et al. | Electrocardiogram classification using wavelet transformations | |
| Papadogiorgaki et al. | Heart rate classification using ECG signal processing and machine learning methods | |
| Vimal et al. | Random forest classifier based ECG arrhythmia classification | |
| Chiu et al. | Get rid of the beat in mobile EEG applications: A framework towards automated cardiogenic artifact detection and removal in single-channel EEG | |
| CN113974647B (en) | System and method for reconstructing sudden cardiac death risk factor | |
| Ouelli et al. | Multivariate autoregressive modeling for cardiac arrhythmia classification using multilayer perceptron neural networks | |
| Celin et al. | A novel method for ECG classification using polynomial based curve fitting | |
| Rodrigues et al. | The issue of automatic classification of heartbeats | |
| Sanamdikar et al. | Analysis of several characteristics of ECG signal for cardiac arrhythmia detection | |
| Jayagopi et al. | On the classification of arrhythmia using supplementary features from tetrolet transforms. | |
| Talatov et al. | Methodology for processing and analysis of diagnostic indicators electrocardiogram based on Labview | |
| Murthy et al. | Design and implementation of hybrid techniques and DA-based reconfigurable FIR filter design for noise removal in EEG signals on FPGA |
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 |