• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Noisy ECG Signal Data Transformation to Augment Classification Accuracy

    2022-08-24 03:26:10IqraAfzalFiazMajeedMuhammadUsmanAliShahzadaKhurramAkberAbidGardeziShafiqAhmadSaadAladyanAlmetwallyMostafaandMuhammadShafiq
    Computers Materials&Continua 2022年5期

    Iqra Afzal,F(xiàn)iaz Majeed,Muhammad Usman Ali,Shahzada Khurram,Akber Abid Gardezi,Shafiq Ahmad,Saad Aladyan,Almetwally M.Mostafa and Muhammad Shafiq

    1Department of Information Technology,University of Gujrat,Gujrat,50700,Pakistan

    2Department of Computer Science,University of Gujrat,Gujrat,50700,Pakistan

    3Faculty of Computing,The Islamia University of Bahawalpur,Bahawalpur,63100,Pakistan

    4Department of Computer Science,COMSATS University Islamabad,Islamabad,Pakistan

    5Industrial Engineering Department,College of Engineering,King Saud University,P.O.Box 800,Riyadh 11421,Saudi Arabia

    6Department of Information Systems,College of Computer and Information Sciences,King Saud University,P.O.Box 800,Riyadh,11421,Saudi Arabia

    7Department of Information and Communication Engineering,Yeungnam University,Gyeongsan,38541,Korea

    Abstract: In this era of electronic health, healthcare data is very important because it contains information about human survival.In addition,the Internet of Things(IoT)revolution has redefined modern healthcare systems and management by providing continuous monitoring.In this case,the data related to the heart is more important and requires proper analysis.For the analysis of heart data, Electrocardiogram (ECG) is used.In this work,machine learning techniques, such as adaptive boosting (AdaBoost) is used for detecting normal sinus rhythm,atrial fibrillation(AF),and noise in ECG signals to improve the classification accuracy.The proposed model uses ECG signals as input and provides results in the form of the presence or absence of disease AF, and classifies other signals as normal, other, or noise.This article derives different features from the signal using Maximal Information Coefficient(MIC)and minimum Redundancy Maximum Relevance(mRMR)technique,and then classifies them based on their attributes.Since the ECG contains some kind of noise and irregular data streams so the purpose of this study is to remove artifacts from the ECG signal by deploying the method of Second-Order-Section (SOS) (filter) and correctly classify them.Several features were extracted to improve the detection of ECG data.Compared with existing methods,this work gives promising results and can help improve the classification accuracy of the ECG signals.

    Keywords: ECG;atrial fibrillation;adaptive boosting;heart rate variability

    1 Introduction

    In order to fulfill the body’s needs such as oxygen supply and nutrition, the heart is the main organ located in the chest and assists the entire circulatory system inside the body.The heart has four chambers.The upper two chambers that are responsible for receiving blood are called atria,and the lower two chambers that discharge blood are called ventricles.In the ordinance that separates the left and right atria and the left and right ventricles, nerves walls called septum are used [1].In the right atrium,there is a node that produces a heartbeat,called the atrial node(or SA node).In the human heart,the process from the end of a heartbeat to the beginning of a new heartbeat is called the cardiac cycle.The cardiac cycle has two phases of contraction or depolarization and diastole or repolarization[2].Systole is a period when the ventricles contract to empty all blood on the other hand,extra-cardiac diastole is part of the cardiac cycle,during which the heart is emptied and refilled with blood during the systolic and repolarization period.Doctors can use electrodes and a suitable device called an electrocardiogram (ECG) to capture the signal produced by the pacemaker.The electrocardiogram is one of the most integrated and important methods for studying the electrical signals produced by the heart.An ECG is a diagnostic tool that can measure and record the electrical activity of the cardiac cycle in impeccable detail.It records the heart’s electrical pulses in vivid graphics.

    Auxiliary stimulation of the nerve impulse of the heart is used to generate an electrocardiogram.ECG is explained by using waves, intervals, and segments.In ECG, a wave can be described as any positive or negative deflection from isoelectric line (a straight line with no positive or negative charges in the ECG to produce deflection), which represent specific electrical events.They are also representatives of different electrophysiological events in the cardiac cycle.A normal ECG consists of P,Q,R,S,and T waves[3].QRS waves collectively made the QRS complex.When the upper chambers of the heart contract they produce P wave.QRS complex refers to the depolarization of the ventricles.Since the myocardium is also repolarized,they induce T waves when the lower chambers of the heart are electrically resting [4].A segment can be defined as the length between two specific events of the cardiac cycle(between depolarization and repolarization.The duration between these events that describes the period of time is called the interval.The interval includes one or more waveforms and one segment.When these waves are generated irregularly,they describe the presence of arrhythmia.

    A condition where the depolarization is very high and the heartbeat is fast or the heartbeat is low and irregular is called arrhythmia.There are two basic types of arrhythmias.Bradycardia is the point where the pulse is too low-Less than 60 beats per minute.In this regard,tachycardia can be identified as a point where the heart pulse is too high-more than 100 heartbeats per minute.If it is more extreme,it can lead to cardiovascular failure or death[5].Heart arrhythmias demonstrate the vast majority of the cardiovascular issues that may prompt chest pain,heart failure,or sudden cardiac death[6].One of the most common tachycardia arrhythmias that widely affect the population is Atrial Fibrillation(AF).AF is detected when there is an irregular flow of ECG waveforms, that is, when atrial contractions experience chaotic electrical impulses(these are waveforms generated in a rapidly uncoordinated and weekly manner).AF is defined as“irregular supraventricular tachycardia(SVT)with no p-wave and duration>30 s”[7].It occurs in 1%to 2%of the total population and can cause heart attacks,heart failure,and other complex heart diseases.

    In the United States,it is estimated that more than 5 million people have AF,and the prevalence of this arrhythmia will increase with age in the next ten years[8].In order to reduce the risk of stroke,patients should be treated with anticoagulants appropriately.ECG is currently the best quality detector for detecting the presence of AF,because it can accurately obtain the electrical activity of the heart.Irregularities in the ECG waveform,such as the absence of P waves and abnormal RR intervals,can detect AF arrhythmia[9].Nowadays,signal processing can basically be found in any ECG evaluation framework, and it has truly verified its importance for completing the improved analysis of a large number of heart diseases [10].Over the years, from the perspective of ECG signal-based analysis,the study of heart beat-to-beat time variation (called heart rate variability (HRV)) has been deeply considered[11].From various electrocardiogram signals,HRV is a feature extracted to better identify cardiac abnormalities and diseases.

    AF can be detected in two ways by analyzing atrial activity or by examining methods based on ventricular response[12].In the atrial activity analysis method,AF is detected by analyzing the absence of P wave or the presence of fibrillating f wave in the TQ interval.On the other hand,the ventricular response method is based on the analysis of the interbeat time(“RR interval”)of the QRS complex.In the past,many techniques have been applied to predict AF arrhythmia,but this field is still lacking because most of the proposed algorithms and techniques are based on ventricular response analysis methods,which are usually used to clean data by eliminating noise signals.Techniques based on atrial activity analysis are also applied, but their algorithms are not sufficient to deal with noise, because noise is the main factor in the method,so the results provided by these techniques are inefficient and less accurate.The electrocardiogram signal is either affected by external sources or by morphological processes that occur in the human body that cause noise.There are two types of noise in ECG signals.The noise generated by the movement of the patient or the recording cable is called baseline wander(BW) and has a low power frequency.The noise caused by EMG and power line interference refers to Gaussian white noise,which has a normal or high-power frequency[13].In this study,we classify ECG signals as AF and other arrhythmias,and we will apply techniques to remove noise from ECG signals.Since the ECG signal contains noisy data,it may produce unexpected and erroneous results.The data set we used for this research is Physionet/Computing in Cardiology Challenge 2017[14].This research has the following key contributions.

    ·We propose a method to remove noise from ECG signals by using Second-Order Cross-Section(SOS)filter to improve the accuracy of the data set.In this regard,we extracted a set of features,including morphology,statistics,HRV,AF,noise,and other rhythm-related features.

    ·We improve the classification accuracy by implementing a multi-layer model method of binary classification instead of a single-class model, because it provides degrees of freedom in the selection of features and classifiers for each binary classification node.

    ·We verified the accuracy of the proposed technology while analyzing the ECG signal through a large number of medical records.Compared with existing solutions,the proposed technology has achieved gratifying results because the classification accuracy of all categories(F1n,F(xiàn)1a,F(xiàn)1o, and F1p) has been improved, which can help the sensor-based healthcare industry to identify more accurately cardiovascular diseases.

    The rest of the paper is organized as follows:Section 2 briefly introduces related work.Section 3 details the proposed method.Section 4 evaluates the proposed algorithm and discusses the results.Finally,Section 5 concludes the paper by identifying limitations and future work.

    2 Related Works

    This section contains information about the achievements and an overview of ECG analysis and classification.In addition, it covers the different techniques used in ECG analysis.Since the development of this field,a lot of work has been done in various branches and fields of ECG analysis.However,the key artifacts are described below.

    2.1 Techniques for Noise Removal Only

    For electrocardiogram analysis, some studies are only devoted to processing and removing the noise in the electrocardiographic.In [15], the authors proposed the delay error normalized LMS(DENLMS) to remove high frequency and Channel noise.In [16], the authors deployed to replace noise it sets a threshold, and after detecting noise or missing values, replace these values with the appropriate threshold.The results show that the noise replacement technique is more effective than just filtering the noise and also produces more accurate results.This method is used for noise generated in wearable sensors,but the model provides a better alternative instead of removing the noise value.

    2.2 Techniques Based on Inappropriate Features

    Most existing methods (e.g., [1,17–19]) are model to ignore irregular Data flow.These models do not classify the noise classes and just discard the waves with irregular flow,which also reduces the length of the signal.In addition,these technologies only use QRS complexes from various ECG signals.The article proposes an algorithm to identify the RS slope from the QRS complex[20].This algorithm uses five data channels to process different types of signals,even noisy signals.But for preprocessing,they use a very basic filtering method of moving median and mean and Butterworth linear low-pass filter.This article does not classify the signal, and the noise removal technology is also insufficient.Authors of[21]used a powerful method to distinguish between normal and irregular recordings.The proposed model initially loads an ECG file and converts the ECG signal into an envelope diagram.These envelope maps are then used to detect QRS complexes from the original ECG signal.If more than four complexes are detected,these signals are considered to be noisy and the process is terminated.None of these techniques were used to deal with noisy recordings,they were simply rejected.

    2.3 Noise Classification Techniques Without P Wave

    In the past, some studies on ECG analysis used P&T (Pan-Tompkins detector).By using this detector, they retained the R peak, where only the r peak was used to classify the ECG signal, and some of them only used the Rr interval for ECG analysis.In[22,23],the author used a third-order noise removal technique called Butterworth and a median low-pass linear filter.In[24],proposed systems are also applicable to RR but their models are different.These systems have used a combination of three methods,called phasor transform,continuous wavelet transform(CWT),and S transform,with two median filters called wavelet Wiener and Pyramid for noise processing.Paper [25] and [26] use a QRS complex-based model that includes characteristics from the QRS complex,such as skewness and kurtosis, to assign ECG waves.The noise model used in the study [25] is a Butterworth filter,and[26]uses LMS Adaptive filter Least Mean Square(LMS delay error normalization.LMS uses an extended version of DENLMS to remove noise.Paper[6]proposed a new noise removal model based on band-pass filter,segmentation and labeling,and scaling.But this model only extracts features based on r peaks and ensures the validity of the RR interval.Another valuable method in the past is [27],which uses a convolutional neural network (CNN) includes primary and secondary models.Their classification method detects QRS complexes and then uses the signal quality index(SQI)to analyze the signal quality.If the signal category is too low,it marks the signal as a noise segment,otherwise it processes the spectrogram into batch normalization and then performs post-processing, but the technology does not train their model for noise signals.Paper[28]also used SQI technology for noise classification and Savinsky-Golay filter for preprocessing.The main problem with all models proposed in this context is that they only ensure R waves and do not care about keeping other waves,which can lead to misclassification of other categories, because other waves also matter while scrutinizing the ECG.

    2.4 P Wave-Based Noise Classification Methods

    Researchers are also committed to extracting features of QRS complex and P wave.Paper[29]uses classical analysis to explore the rules of P wave and ventricular utilization accumulation RR responds at intervals.For preprocessing,they use existing methods along with MATLAB.Similarly[30]also use atrial activity and ventricular response for classification.The author performed wavelet multifractal and symbol analysis to obtain a regular ECG waveform.For ventricular analysis, they used the old P&T detector technology.Paper [31] proposed a technique that follows the same strategy, which is suitable for atrial and ventricular response characteristics.The moving average filters are used to detect baseline drift and low-frequency noise signals for de-noising,and they use wavelet transform methods to cut off high-frequency irregularities.The authors of[32]also experienced a different method.They extract different features and apply R peak detector by setting thresholds on the basis of these peaks.They further detect all other waves if any Peaks do not match their set threshold they mark them as noisy.Paper[33]also suggested that morphology and HRV characteristics are most important when analyzing ECG.If RR interval is greater than 2 s, this study marks the waveforms as noisy.In the context of HRV feature-based analysis,paper[34]describes a QRS detection algorithm called“gqrs”to generate HRV signals.For noise processing,this work uses the“double median filter”and the“taut string linear approximation method”.These technologies respectively eliminate baseline drift and reduce the influence of abnormal signals.Paper[35]use two-layer binary classification and a simple spectrogram-based method to identify noise and extract ventricles,atria,and HRV and morphological features.This model marked RR intervals with high spectral power above 50 Hz as noisy and rejected them before further processing.All these studies have some shortcomings,for example[31]and[30]only use P&T for QRS, which alone is not enough for better analysis and classification of ECG.Similarly[32]detect all waves based on R,which leads to misclassification and inadequate results.

    3 Proposed Methodology

    This section contains information about the tools, methods, and materials used to conduct the research.In addition,it also details the system processes for performing data analysis and operations,including data collection,data processing and extraction.

    3.1 Data Set for ECG Signal

    The data set used in our research comes from[14].The data set contains 8,528 different single-lead ECG records.In these 8,528 records,5154 are normal types,771 are AF,2557 are other types,and 46 are noisy.The frequency used to sample these records is 300 Hz.These signals are based on binary data,where time is measured in seconds on the x-axis and amplitude is measured in milliamps on the y-axis.The normal category contains data of normal ECG waveforms,AF contains data about patients with atrial fibrillation disease,and the noise rhythm contains very irregular data that cannot provide any information.Fig.1 shows the proposed methodology,which includes multiple steps detailed below.

    3.2 Preprocessing

    Preprocessing the electrocardiogram is a technique to remove different kinds of artifacts from the electrocardiogram.Past research on P&T algorithm[25]and spectrogram technique[35]are not enough,because spectrogram represents time as a sine wave that is not localized in time and does not represent abrupt changes.Therefore,in this study,we use Second-Order Cross Section Filter(SOS).This SOS contains high-pass and low-pass filter orders.It performs band-pass filtering by removing the low-scale coefficient of the low-frequency signal, and eliminates the high-scale of the waveform with too low frequency.In order to further remove the noise in the signal, it will create a waveform closer to the ECG signal, thereby preserving the QRS and P and T waves.In addition, the signal length remains the same,and a well-reconstructed signal is generated.Fig.2 elaborates the process of preprocessing.

    Figure 1:Architecture of the proposed methodology

    Figure 2:Transformation phases of the ECG signal

    3.3 Identification of ECG Waves

    This module is used to detect ECG waves including P,Q,R,S and T.Pan&Tompkins algorithm is used for detecting and identifying R waves.After detecting the R wave, we identified other waves on the basis of it.The P wave is complementary,and the R wave is detected as the maximum value,which is called the maximum value because the maximum value is higher than the two edges of the R peak.The salient minimum,called the minimum of the R wave,is detected as the Q wave because it is the most prominent low on both sides.After this extreme value,the R peak is used to detect the S wave and T wave.Since extreme values also have maximum and minimum points on both sides, the minimum value in this recognized extreme value is called an S wave with high amplitude.On the other hand,the maximum point of this extreme value is detected as T wave.

    3.4 Extraction of Features

    Since we have discussed that the purpose of this study is to classify ECG signals into one of four categories,in the direction of achieving this goal,we extract features from data signals.After extracting the features,we store them in cells,which help us to classify the signal.According to the characteristics,these features are divided into the following categories.

    3.4.1 Morphological Features

    3.4.2 State of the Art AF Features

    AF represents irregular ventricular response,so it can be detected as irregular R peaks in ECG,and it does not have P waves.This module detects different features from existing technologies,including AFEvidence, raw count, irregular evidence, pace count, density evidence, anisotropy evidence,and AFEvidence from RR interval[36].More features are extracted in this context,some of which are based on sample entropy,while others are related to RR and delta RR intervals.The features extracted from these are their coefficients and variation.

    3.4.3 Time Domain Features of HRV

    In ECG, we have a variety of heart cycles, so we can measure their distance, which leads to a heartbeat in HRV.HRV features include features about intervals,such as standard deviation(SDSN),SDSD(standard deviation of successive differences)features.We extract features in this context based on RMSSD (Root Mean Square of Continuous Differences).nn50 The number of consecutive NN pairs with a difference of more than 50 ms pnn50 NN50 divided by the ratio of the total number of 50 ms nn20 Paired continuous NNs with a difference of more than 20 milliseconds pnn20[37].

    3.4.4 Statistical Characteristics of RR Interval

    This module is about extracting the characteristics of the RR interval statistical measurement.These statistical measures include the average,median,maximum,and minimum of the RR interval.These features also include irregular intervals called RR interval skewness and features including large intervals called kurtosis.Some other extracted features are variance,RR interval range,and dRR(delta RR)interval.

    3.4.5 Features from Frequency Domain

    How beautiful it looks with its tongue stretched out like that! It was a long night, but the Snow-man did not find it so; there he stood, wrapt in his pleasant thoughts, and they froze, so that he cracked

    In the context of visualizing HRV features, Poincare plotting technology was sued.For this purpose, Poincare features are tested, which represent how the prediction of the next RR interval is predicted by the previous interval.In addition,the frequency analysis of the ECG signal is also carried out in this section.For this frequency analysis,a technique called Time Fourier Transform STFT is used.

    3.4.6 Features from Delta(RR)and RR

    In this module, the probability density estimates of RR and incremental RR intervals are extracted.It includes skewness, kurtosis, and other characteristics.Other characteristics include maximum distance, minimum distance, and length, or the number of peaks between RR and delta RR.

    3.4.7 Features for Other Rhythm

    To describe categories other than AF, normal and noisy features are used, such as irregular R peaks,abnormal duration of R peaks,slope,segmentation,increased and abnormal heart rate.

    3.4.8 Features Regarding the Noisy Rhythm

    Any abnormality of the ECG peak and any sudden changes in structure reveal noisy segments.Therefore,in order to detect the features of these segments,the features including the amplitude and variance based on the QRS complex and other peaks are extracted from the frequency domain and the time domain.Because of this amplitude and variance, the noise segment can be detected more accurately, because the noise segment is nothing more than the change in the peak amplitude and variance in the ECG.

    3.5 Features Selection

    Feature selection is more important because all results are based on the selection of extracted features, which can eliminate irrelevant features, and can discard duplicate features and store them in only one file.we used well-known tools to select the most relevant features,namely the maximum information coefficient (MIC) and the minimum redundancy and maximum correlation (mRMR)techniques[35].

    3.6 Classification

    In this study, we use AdaBoost (Adaptive Boost) for classification.This is a machine learning integration technology based on boosting.Based on the work of AdaBoost, a similar task was performed in our research.Classification is performed on different layers by using different AdaBoost models.Different models of the classifier take different metrics on the data and train on these data to obtain better results from weak classifiers and at the same time enhance poor classification results.This method will generate a robust classifier by introducing inadequate,unskilled and weak classifiers.Due to its user-friendly interface and multiple built-in data analysis technologies,all feature extraction and classification are performed in MATLAB.In order to obtain the classification results and check the performance according to accuracy,F(xiàn)1 for each of the four types is defined in the following.In this regard,F(xiàn)1 for normal rhythm can be written as,

    F1 for AF rhythm is calculated as follows,

    F1 for other rhythms is given by,

    and,F(xiàn)1 for noisy signal is defined as follows,

    For F1n in Eq.(1),Nn is the number of normal instances classified as normal by both the reference and prediction classifier,is the total number of instances in the reference category,andn is the sum of all instances classified as normal by the classifier.For AF in Eq.(2),A is the total amount of data represented by the reference classifier as AF, anda is the total amount of classified data predicted as AF by the proposed classification method.For Eq.(3),Oo in the other rhythm formula section has the number of instances predicted as‘other’class by the referenced and proposed classifier.represents the aggregated instance of the other class by the reference classifier,ando has the total amount of data predicted by the classifier to be the other class.In F1p in Eq.(4),Pp is the number of Noisy instances that are predicted by the reference and working classifiers and labelled as Noisy.On the other hand,has the total number of instances classified as Noisy by the reference classification,andrepresents the total number of instances predicted by the working classifier.

    4 Results and Discussions

    This section provides detailed information about the experiment and interpretation of the results.In the experimental analysis, we used Matlab to analyze the ECG data through signal processing because it is the most important technique for analyzing ECG waveforms because ECG waves are sometimes in a continuous and episodic rhythm with the help of signal processing.In this regard,we perform feature extraction after performing preprocessing to clean the ECG signal.In past research,the features of ECG classification were to extract general types of features, such as RS slope, QRS complex, and heart rate.These features are not enough when we need to divide the signal into four categories.We, therefore, extracted some other features to support “Other” and “noisy” classes to improve the accuracy.The classification is performed after obtaining the numerical characteristics of the features.In order to obtain more accurate results, we used AdaBoost (Adaptive Boost) with a multi-layer method for classification.

    4.1 Preprocessing

    Preprocessing experiments are performed to obtain pre-processed ECG signals and to remove baseline drift, other high-frequency and low-frequency artifacts.The ECG waveform has sporadic data streams, which contain noise fragments (only the coefficients of extremely high frequency and extremely low frequency),which cannot be classified more accurately,and often produce wrong results,so this step in our experiment helps to improve result.As a result of this, we get bpfecg (bandpass filtered electrocardiogram).After getting bpfecg,because the ECG does not have all the artifacts,we get the filtered form of the ECG signal as shown in Fig.3.We observed that the ECG we get is in a state in which the lower scale coefficients of the high-frequency signal and the higher scale coefficients of the waveform with too low frequency are eliminated.In addition,the waveform closest to the ECG signal is created,all waveforms are preserved,and the signal length remains the same.

    Figure 3:Filtered ECG signal after noise removal

    4.2 P,Q,R,S,and T Detection

    In order to obtain the PQRST wave,we follow the maximum and minimum method discussed in the methodology which leads us to identify these indicators and then detect all waves from the ECG signal.Fig.4 represents the indexes we get in the form of P,Q and S.After these values and indexes are displayed in the graph,the P,Q,R,S and T points are detected and identified in the ECG as flaunted in Fig.5.If the ECG can be distinguished effectively,the system can predict the result more accurately.In the existing techniques,only the three-wave QRS complex is the center wave,and the classification is based on this complex.This is not enough to get better results, because all other waves will also affect the output.Therefore,we aim and hit all waves to obtain more accurate results.

    4.3 Feature Extraction

    In Fig.6,we have obtained some RR and deltaRR(dRR)interval values since features such as R wave and its interval play an important role in this classification.After outputting the form of feat1,the values of AF evidence (AFEv), the coefficient of variation of RR (CVrr), and the coefficient of variation of dRR(CVdrr)are observed as-11.0000,0.0443,and 0.0315,respectively.

    Figure 4:ECG index:(a)P wave index,(b)Q wave index,(c)S wave index

    These statistical measurements also help us in the classification of noisy categories and all other categories.Through these extracted features, we can locate the position of the ECG wave, we can detect the location where the P wave does not exist,or we can say that P is zero as expressed in Fig.7 to detect a specific category.In order to detect the regular and irregular flow of ECG waves,we need to extract the PP and RR indexes so that the system can accurately make decisions on noise and noisy categories.This can be achieved by using the PP and RR index features interpreted in Fig.8.It will help the system understand how the duration between R streams exists and how its frequency and other data streams change before and after R.All the features extracted in the experiment use MIC and mRMR technology to help us extract the most relevant features without any redundancy.Figs.6,7 and 8 show the statistical measures of ECG streams,which are of great significance when classifying ECG because they extract these features to predict the skewness and kurtosis and most other relevant and basic features.We train our model to produce better results with more enormous features.

    4.4 Classification

    Experiments are performed with the training data and the verification data given by the challenge to check the verification of the system.By using the formula, we have F1n=0.9764 on the test data and a score of 0.99 on the training data.In addition, the overall accuracy of F1 is 0.9613.We score F1a=0.9592,which ranks higher compared to other technologies.Our proposed work score of F1o=0.9482.The classification accuracy of our proposed work on noise segment is F1p=0.8571.The confusion matrix of the training set and the validation set are shown in Fig.9.The highlighted cells(in the Normal,AF,Other,Noisy)row and column represent data that our classifier has correctly predicted.

    Figure 5:P,Q,R,S and T waves

    Figure 6:RR Interval

    Figure 7:Zero position of P wave

    Figure 8:PP and RR index

    Figure 9:Confusion matrix obtained on validation and training sets(a)Validation set(b)Training set

    Fig.10 shows the comparison between the proposed technique and other techniques in terms of classification accuracy of F1n,F(xiàn)1a,F(xiàn)1o,and F1p results.We observe that our technique is superior to other techniques because the classification accuracy of F1n, F1a, F1o, and F1p is higher.This is because our method uses SOS filters to remove noise from ECG signals and extract many features,including morphology, statistics, HRV, AF, noise, and other rhythm-related features.This provides evidence for our framework to clean up the signals in the ECG structure that are more likely to be wavelets,carefully detect all waves present in the ECG,and extract classification features that do not have any redundancy.It turns out that our framework ranks higher compared to other technologies,and it also shows that our multi-layer model for binary classification provides better results compared to the single-layer model.Most of the top-ranked techniques (see Fig.10) are considered for this comparison.The simulation results prove that our proposed work ranks high in the classification of all categories.

    Figure 10:Archived accuracy of the proposed technique compared with different techniques

    5 Conclusions

    Heart disease is a very common health disorder in every society,and its diagnosis or testing plays an important role in saving the lives of millions of people around the world.Atrial fibrillation is one of the most important cardiovascular diseases and has received great attention from computer science and engineering disciplines to automatically detect it with the help of different algorithms.However,this is a challenging problem due to identification issues and low accuracy.In this paper,we proposed an effective method to extract the most relevant and novel features from the medical data of binary ECG signal.These features are morphological,state-of-the-art AF,time domain and HRV features,including SDSN and SDSDS.The other features extracted are the statistical features of the RR interval, the features from the frequency domain, and the features from the probability density estimation of delta(RR)and RR.These derived features are then selected through cascaded feature selection to classify the ECG signals of the normal,AF,other,or noise categories that form the specific instance of the data set.Extensive experiments have been conducted with publicly available ECG waveform data sets.Compared with the previous literature,the classification accuracy of the proposed technique shows improved results in all categories including F1n,F(xiàn)1a,F(xiàn)1o,and F1p.This work will also provide strong support to the sensor-based healthcare industry,which will focus on continuous patient monitoring systems in the future.This work can also be extended to the identification and classification of many other heart diseases, especially the irregular streams associated with non-AF arrhythmias.

    Acknowledgement:The authors extend their appreciation to the Deanship of Scientific Research at King Saud University for funding this work through research group No(RG-1441-425).

    Funding Statement:This work was supported by the Deanship of Scientific Research at King Saud University through research group No(RG-1441-425).

    Conflicts of Interest:The authors declare that they have no conflicts of interest to report regarding the present study.

    国产成人影院久久av| 日本 av在线| 亚洲国产欧美网| 久久午夜综合久久蜜桃| 最近最新免费中文字幕在线| 99在线人妻在线中文字幕| 国产欧美日韩一区二区三| 欧美+亚洲+日韩+国产| 搞女人的毛片| 精品久久久久久久人妻蜜臀av| 国产av又大| 亚洲精品色激情综合| 久久天堂一区二区三区四区| 精品久久久久久,| 人妻久久中文字幕网| 亚洲av电影在线进入| 在线观看一区二区三区| 老汉色∧v一级毛片| 久久久久久久久免费视频了| 欧美人与性动交α欧美精品济南到| 极品教师在线免费播放| 亚洲成人国产一区在线观看| 久久欧美精品欧美久久欧美| 国产精品一区二区精品视频观看| 午夜福利在线观看吧| 在线永久观看黄色视频| 美女国产高潮福利片在线看| 成人18禁高潮啪啪吃奶动态图| 国产色视频综合| 亚洲人成77777在线视频| 在线观看一区二区三区| 悠悠久久av| 一本大道久久a久久精品| 中文在线观看免费www的网站 | www日本黄色视频网| 久久青草综合色| 69av精品久久久久久| 叶爱在线成人免费视频播放| 老熟妇乱子伦视频在线观看| 美女扒开内裤让男人捅视频| 久久九九热精品免费| 在线观看午夜福利视频| 丝袜人妻中文字幕| 男女做爰动态图高潮gif福利片| 91九色精品人成在线观看| 国产激情偷乱视频一区二区| svipshipincom国产片| 精品久久久久久久毛片微露脸| 亚洲专区中文字幕在线| 久久久久久久久免费视频了| 日本在线视频免费播放| 久久草成人影院| 国产主播在线观看一区二区| 黑人巨大精品欧美一区二区mp4| 欧美中文日本在线观看视频| av超薄肉色丝袜交足视频| 国产成人精品久久二区二区91| 搡老妇女老女人老熟妇| 长腿黑丝高跟| 他把我摸到了高潮在线观看| 国产爱豆传媒在线观看 | 日本三级黄在线观看| 精品日产1卡2卡| www.www免费av| 午夜久久久在线观看| 波多野结衣av一区二区av| 成人国产一区最新在线观看| 成人欧美大片| 久久久国产成人精品二区| 热99re8久久精品国产| av超薄肉色丝袜交足视频| 一夜夜www| 亚洲av电影在线进入| 国产精品日韩av在线免费观看| 一级a爱片免费观看的视频| 久久香蕉精品热| 中文字幕av电影在线播放| √禁漫天堂资源中文www| 黄色片一级片一级黄色片| 这个男人来自地球电影免费观看| 国产亚洲精品久久久久5区| 深夜精品福利| 国产在线观看jvid| 欧美一区二区精品小视频在线| 激情在线观看视频在线高清| 久久国产亚洲av麻豆专区| 伦理电影免费视频| 久久久久久亚洲精品国产蜜桃av| 99在线视频只有这里精品首页| 亚洲国产欧美一区二区综合| 久久久精品欧美日韩精品| 99精品欧美一区二区三区四区| 精品国产一区二区三区四区第35| 国产精品久久久av美女十八| 老司机在亚洲福利影院| 真人做人爱边吃奶动态| 成人一区二区视频在线观看| 三级毛片av免费| 天堂动漫精品| ponron亚洲| 亚洲专区中文字幕在线| 国产精品影院久久| 国产亚洲av高清不卡| 国产伦在线观看视频一区| 欧美性长视频在线观看| 少妇 在线观看| 人人妻人人看人人澡| 1024手机看黄色片| 露出奶头的视频| 亚洲精品久久成人aⅴ小说| avwww免费| 黄片小视频在线播放| 嫩草影视91久久| 热re99久久国产66热| 精品久久久久久久久久免费视频| 国产单亲对白刺激| 欧美一级毛片孕妇| 最新美女视频免费是黄的| 99精品在免费线老司机午夜| 十分钟在线观看高清视频www| 18禁黄网站禁片午夜丰满| 99久久国产精品久久久| 午夜精品在线福利| 久99久视频精品免费| aaaaa片日本免费| 欧美色欧美亚洲另类二区| 青草久久国产| 国产亚洲欧美在线一区二区| 午夜福利欧美成人| 国产精品久久久av美女十八| 久久婷婷人人爽人人干人人爱| 麻豆久久精品国产亚洲av| 妹子高潮喷水视频| 免费在线观看影片大全网站| 日韩欧美 国产精品| 妹子高潮喷水视频| 久久国产精品影院| 国产精品免费视频内射| 美国免费a级毛片| 国产激情欧美一区二区| 高潮久久久久久久久久久不卡| 两个人看的免费小视频| 丰满人妻熟妇乱又伦精品不卡| 欧美中文综合在线视频| 婷婷丁香在线五月| 久久精品国产99精品国产亚洲性色| 精品无人区乱码1区二区| 精品一区二区三区四区五区乱码| 中文字幕精品亚洲无线码一区 | 99在线视频只有这里精品首页| 制服诱惑二区| 国产av又大| 麻豆成人午夜福利视频| av福利片在线| 午夜老司机福利片| 国产日本99.免费观看| 日本撒尿小便嘘嘘汇集6| 嫁个100分男人电影在线观看| 满18在线观看网站| 亚洲七黄色美女视频| 国产黄色小视频在线观看| 亚洲精品中文字幕一二三四区| 精品久久久久久久毛片微露脸| 老汉色av国产亚洲站长工具| 搡老妇女老女人老熟妇| 91国产中文字幕| 久久久久久大精品| 丰满的人妻完整版| 男女午夜视频在线观看| 久久婷婷人人爽人人干人人爱| 久久久久久大精品| 日本 欧美在线| 1024视频免费在线观看| 母亲3免费完整高清在线观看| 在线观看日韩欧美| 国产激情久久老熟女| 黑人欧美特级aaaaaa片| 欧美中文日本在线观看视频| 老司机午夜十八禁免费视频| 无人区码免费观看不卡| 欧美色欧美亚洲另类二区| 中文字幕最新亚洲高清| 国产伦在线观看视频一区| 一二三四在线观看免费中文在| 在线十欧美十亚洲十日本专区| av免费在线观看网站| 国产亚洲av高清不卡| 999久久久国产精品视频| 久久久国产欧美日韩av| 亚洲成av片中文字幕在线观看| 久久精品国产亚洲av高清一级| 桃红色精品国产亚洲av| 欧美成狂野欧美在线观看| 精品一区二区三区四区五区乱码| 精品久久蜜臀av无| 久久精品国产99精品国产亚洲性色| 一本精品99久久精品77| 欧美乱码精品一区二区三区| 国产一级毛片七仙女欲春2 | 久久欧美精品欧美久久欧美| 白带黄色成豆腐渣| 成人特级黄色片久久久久久久| 亚洲美女黄片视频| av免费在线观看网站| 成人国产综合亚洲| 亚洲国产高清在线一区二区三 | 国产主播在线观看一区二区| 深夜精品福利| 极品教师在线免费播放| 日韩视频一区二区在线观看| 久久中文字幕一级| 国产亚洲av嫩草精品影院| 免费看十八禁软件| a级毛片在线看网站| a级毛片在线看网站| 精品欧美国产一区二区三| bbb黄色大片| 少妇的丰满在线观看| 亚洲国产精品久久男人天堂| 精品一区二区三区av网在线观看| 精品国产国语对白av| 麻豆一二三区av精品| 亚洲av成人一区二区三| 亚洲熟女毛片儿| www日本黄色视频网| 啦啦啦免费观看视频1| 午夜福利在线在线| 成人三级黄色视频| 欧美午夜高清在线| 91av网站免费观看| 午夜老司机福利片| 国产精品永久免费网站| 在线播放国产精品三级| 成人欧美大片| 一边摸一边做爽爽视频免费| 国产精品香港三级国产av潘金莲| 国产亚洲欧美在线一区二区| 丝袜美腿诱惑在线| 岛国视频午夜一区免费看| 91麻豆精品激情在线观看国产| 国产亚洲欧美在线一区二区| 丝袜人妻中文字幕| 女人被狂操c到高潮| 国产精品 欧美亚洲| 人妻久久中文字幕网| 国产精品一区二区免费欧美| 香蕉久久夜色| 一区二区三区高清视频在线| 午夜福利18| 色综合站精品国产| 久9热在线精品视频| 久久久久精品国产欧美久久久| 在线播放国产精品三级| 国产激情欧美一区二区| 国产乱人伦免费视频| 黄色 视频免费看| 亚洲第一av免费看| 成在线人永久免费视频| 在线观看日韩欧美| 欧美一级a爱片免费观看看 | 午夜福利18| 美女免费视频网站| 韩国精品一区二区三区| 国产成人系列免费观看| 国产精品爽爽va在线观看网站 | 国产高清有码在线观看视频 | 婷婷精品国产亚洲av在线| 黄片小视频在线播放| 性色av乱码一区二区三区2| 午夜精品在线福利| 欧美另类亚洲清纯唯美| 午夜日韩欧美国产| 国产视频一区二区在线看| 嫩草影视91久久| 一区二区日韩欧美中文字幕| 免费女性裸体啪啪无遮挡网站| 国产精华一区二区三区| 日韩有码中文字幕| 高清在线国产一区| 国产成人精品久久二区二区免费| 日本成人三级电影网站| 亚洲美女黄片视频| 99在线人妻在线中文字幕| 亚洲五月天丁香| 亚洲国产欧美一区二区综合| 亚洲色图 男人天堂 中文字幕| 午夜久久久久精精品| 白带黄色成豆腐渣| 999精品在线视频| 国产黄片美女视频| 亚洲熟女毛片儿| 一a级毛片在线观看| 亚洲九九香蕉| 成人国语在线视频| 无限看片的www在线观看| 欧美+亚洲+日韩+国产| 亚洲激情在线av| 在线观看午夜福利视频| 高清在线国产一区| 国产视频一区二区在线看| 黄色女人牲交| or卡值多少钱| 久久天躁狠狠躁夜夜2o2o| 日本一区二区免费在线视频| 国产又黄又爽又无遮挡在线| 中文亚洲av片在线观看爽| 精品乱码久久久久久99久播| 一区二区三区精品91| 久久人妻av系列| 亚洲精品av麻豆狂野| 一区二区三区激情视频| 日韩欧美一区二区三区在线观看| 叶爱在线成人免费视频播放| 久久久精品国产亚洲av高清涩受| 淫妇啪啪啪对白视频| 亚洲中文字幕一区二区三区有码在线看 | 在线观看www视频免费| 国产精品久久久av美女十八| 国产亚洲精品久久久久5区| 亚洲成人久久爱视频| 久久伊人香网站| 又紧又爽又黄一区二区| 欧美激情 高清一区二区三区| 91国产中文字幕| 欧美日韩亚洲国产一区二区在线观看| 精品午夜福利视频在线观看一区| 亚洲 欧美一区二区三区| 国产黄片美女视频| 日本 av在线| 禁无遮挡网站| 国产精品久久久久久人妻精品电影| 精品熟女少妇八av免费久了| xxx96com| 欧美三级亚洲精品| 亚洲,欧美精品.| 午夜免费激情av| 最新美女视频免费是黄的| 精品少妇一区二区三区视频日本电影| 久久精品aⅴ一区二区三区四区| 日日干狠狠操夜夜爽| 在线av久久热| 久久热在线av| 亚洲精品在线美女| 黄片小视频在线播放| 欧美在线一区亚洲| 中文字幕精品亚洲无线码一区 | 国产成+人综合+亚洲专区| 亚洲精品在线美女| 日韩欧美三级三区| 日韩大码丰满熟妇| 国产成人啪精品午夜网站| 亚洲最大成人中文| 美女大奶头视频| 免费观看人在逋| 成人国语在线视频| 国产伦在线观看视频一区| 国产男靠女视频免费网站| 90打野战视频偷拍视频| 亚洲人成伊人成综合网2020| 中文字幕高清在线视频| 久久精品人妻少妇| 欧美日韩中文字幕国产精品一区二区三区| 麻豆国产av国片精品| 亚洲欧美日韩高清在线视频| 国产99白浆流出| www.999成人在线观看| 久久香蕉激情| 色综合欧美亚洲国产小说| 国产熟女xx| 每晚都被弄得嗷嗷叫到高潮| 欧美日韩中文字幕国产精品一区二区三区| aaaaa片日本免费| 脱女人内裤的视频| 婷婷亚洲欧美| 国产爱豆传媒在线观看 | 搞女人的毛片| 12—13女人毛片做爰片一| 听说在线观看完整版免费高清| 欧美成人免费av一区二区三区| 一二三四在线观看免费中文在| 色综合亚洲欧美另类图片| 禁无遮挡网站| 国产激情欧美一区二区| 丁香欧美五月| 视频在线观看一区二区三区| 免费在线观看日本一区| 日韩欧美国产一区二区入口| 久久久久久大精品| 操出白浆在线播放| 精品第一国产精品| 亚洲成人免费电影在线观看| 香蕉av资源在线| 久久午夜亚洲精品久久| 成人欧美大片| 精华霜和精华液先用哪个| 18美女黄网站色大片免费观看| 老司机福利观看| 精品卡一卡二卡四卡免费| 丁香欧美五月| 99riav亚洲国产免费| 黄色毛片三级朝国网站| 亚洲国产精品sss在线观看| 日韩av在线大香蕉| 免费观看精品视频网站| 午夜影院日韩av| 18美女黄网站色大片免费观看| 免费在线观看日本一区| 国产野战对白在线观看| 亚洲性夜色夜夜综合| 亚洲国产中文字幕在线视频| 国产午夜福利久久久久久| 在线天堂中文资源库| av片东京热男人的天堂| 美国免费a级毛片| 亚洲激情在线av| 午夜激情福利司机影院| 白带黄色成豆腐渣| 欧美激情 高清一区二区三区| cao死你这个sao货| 亚洲人成伊人成综合网2020| 免费一级毛片在线播放高清视频| 久久久久久免费高清国产稀缺| 免费看美女性在线毛片视频| 成人18禁高潮啪啪吃奶动态图| 成人av一区二区三区在线看| 亚洲熟妇中文字幕五十中出| 日韩有码中文字幕| 超碰成人久久| 国产私拍福利视频在线观看| 一进一出抽搐动态| 曰老女人黄片| 99在线人妻在线中文字幕| 国产精品亚洲av一区麻豆| 变态另类成人亚洲欧美熟女| 日本精品一区二区三区蜜桃| 久久久精品国产亚洲av高清涩受| 女人爽到高潮嗷嗷叫在线视频| 久久狼人影院| 欧美日韩福利视频一区二区| 久久精品亚洲精品国产色婷小说| 亚洲国产看品久久| 久久九九热精品免费| 色综合婷婷激情| 日韩精品免费视频一区二区三区| 成人手机av| 国产亚洲精品一区二区www| 久久精品aⅴ一区二区三区四区| 午夜精品在线福利| 欧美乱码精品一区二区三区| 久久精品aⅴ一区二区三区四区| 真人一进一出gif抽搐免费| 嫩草影视91久久| 亚洲精品国产精品久久久不卡| 成年女人毛片免费观看观看9| 欧美成人午夜精品| 精品久久久久久久毛片微露脸| 午夜久久久在线观看| 一进一出抽搐gif免费好疼| 国产黄色小视频在线观看| 久久人妻av系列| 欧美日本亚洲视频在线播放| 18美女黄网站色大片免费观看| 国产激情偷乱视频一区二区| 久久精品aⅴ一区二区三区四区| 国产欧美日韩一区二区精品| 老司机深夜福利视频在线观看| 国产精品二区激情视频| 精品久久蜜臀av无| 国产一区二区三区视频了| 69av精品久久久久久| 色播在线永久视频| 草草在线视频免费看| 国产极品粉嫩免费观看在线| 老司机在亚洲福利影院| 亚洲av日韩精品久久久久久密| 正在播放国产对白刺激| 搡老岳熟女国产| 18禁黄网站禁片午夜丰满| 亚洲人成网站在线播放欧美日韩| 一夜夜www| 男男h啪啪无遮挡| 男女做爰动态图高潮gif福利片| 女警被强在线播放| 中文亚洲av片在线观看爽| 性色av乱码一区二区三区2| 免费一级毛片在线播放高清视频| 免费av毛片视频| 狂野欧美激情性xxxx| ponron亚洲| 一个人观看的视频www高清免费观看 | 久9热在线精品视频| 国产黄片美女视频| 国内毛片毛片毛片毛片毛片| 欧美最黄视频在线播放免费| 精品国产美女av久久久久小说| 免费在线观看成人毛片| 国产精品98久久久久久宅男小说| 露出奶头的视频| 我的亚洲天堂| 午夜福利一区二区在线看| 精品一区二区三区四区五区乱码| 亚洲精品国产区一区二| 一夜夜www| 国产成人精品久久二区二区免费| 十八禁网站免费在线| 国产人伦9x9x在线观看| 啦啦啦免费观看视频1| 97超级碰碰碰精品色视频在线观看| 日本黄色视频三级网站网址| 99国产精品一区二区蜜桃av| 亚洲av片天天在线观看| 哪里可以看免费的av片| 国内精品久久久久精免费| 露出奶头的视频| 精品第一国产精品| 一区二区日韩欧美中文字幕| 99精品久久久久人妻精品| 少妇 在线观看| 两个人视频免费观看高清| 午夜日韩欧美国产| 老汉色av国产亚洲站长工具| 亚洲天堂国产精品一区在线| 一区福利在线观看| 国产精品亚洲av一区麻豆| 嫩草影视91久久| 给我免费播放毛片高清在线观看| 成人国产综合亚洲| 俄罗斯特黄特色一大片| 精品高清国产在线一区| 日韩三级视频一区二区三区| 亚洲在线自拍视频| 黄片小视频在线播放| 欧洲精品卡2卡3卡4卡5卡区| 国产又色又爽无遮挡免费看| 麻豆成人午夜福利视频| 白带黄色成豆腐渣| 欧美日韩乱码在线| 黑人操中国人逼视频| АⅤ资源中文在线天堂| 黑人欧美特级aaaaaa片| 一进一出好大好爽视频| 法律面前人人平等表现在哪些方面| 亚洲国产精品sss在线观看| 亚洲欧美激情综合另类| 日本免费一区二区三区高清不卡| 亚洲成av人片免费观看| 亚洲人成网站在线播放欧美日韩| 欧美zozozo另类| 免费在线观看视频国产中文字幕亚洲| 露出奶头的视频| 国产在线精品亚洲第一网站| 免费av毛片视频| 丰满人妻熟妇乱又伦精品不卡| 成人三级黄色视频| 免费在线观看成人毛片| 午夜视频精品福利| 男人操女人黄网站| 亚洲三区欧美一区| 1024视频免费在线观看| 手机成人av网站| 制服丝袜大香蕉在线| 亚洲av成人一区二区三| 久久久久久国产a免费观看| 日韩大尺度精品在线看网址| 亚洲成人国产一区在线观看| 高清在线国产一区| 国产又色又爽无遮挡免费看| 桃色一区二区三区在线观看| 日韩欧美国产一区二区入口| 一进一出抽搐动态| 好男人电影高清在线观看| 亚洲一卡2卡3卡4卡5卡精品中文| 身体一侧抽搐| 久久久国产成人免费| 老熟妇仑乱视频hdxx| 午夜免费激情av| 777久久人妻少妇嫩草av网站| 精品免费久久久久久久清纯| av电影中文网址| 叶爱在线成人免费视频播放| 一本大道久久a久久精品| 18禁美女被吸乳视频| 欧美精品啪啪一区二区三区| 成人午夜高清在线视频 | 国产精品二区激情视频| 最近最新中文字幕大全电影3 | 国产高清有码在线观看视频 | 午夜久久久在线观看| 国产免费av片在线观看野外av| 国产久久久一区二区三区| 一区二区日韩欧美中文字幕| 国内久久婷婷六月综合欲色啪| 国产高清激情床上av| svipshipincom国产片| 久热这里只有精品99| 日本三级黄在线观看| 欧美黑人精品巨大| 日本在线视频免费播放| 国产私拍福利视频在线观看| 亚洲人成网站高清观看| 国产av不卡久久| 男男h啪啪无遮挡| 香蕉丝袜av| 美女国产高潮福利片在线看| 国产精品久久久久久人妻精品电影| 成年版毛片免费区| 久久精品91无色码中文字幕| 午夜两性在线视频| 久9热在线精品视频| 国产精品久久久久久亚洲av鲁大| 亚洲成a人片在线一区二区| 亚洲自拍偷在线| 午夜久久久在线观看| 国产av一区二区精品久久| 观看免费一级毛片|