MIT-BIH Normal Sinus Rhythm Database Research Articles

Detection of atrial fibrillation using a nonlinear Lorenz Scattergram and deep learning in primary care

BackgroundAtrial fibrillation (AF) is highly correlated with heart failure, stroke and death. Screening increases AF detection and facilitates the early adoption of comprehensive intervention. Long-term wearable devices have become increasingly popular for AF screening in primary care. However, interpreting data obtained by long-term wearable ECG devices is a problem in primary care. To diagnose the disease quickly and accurately, we aimed to build AF episode detection model based on a nonlinear Lorenz scattergram (LS) and deep learning.MethodsThe MIT-BIH Normal Sinus Rhythm Database, MIT-BIH Arrhythmia Database and the Long-Term AF Database were extracted to construct the MIT-BIH Ambulatory Electrocardiograph (MIT-BIH AE) dataset. We converted the long-term ECG into a two-dimensional LSs. The LSs from MIT-BIH AE dataset was randomly divided into training and internal validation sets in a 9:1 ratio, which was used to develop and internally validated model. We built a MOBILE-SCREEN-AF (MS-AF) dataset from a single-lead wearable ECG device in primary care for external validation. Performance was quantified using a confusion matrix and standard classification metrics.ResultsDuring the evaluation of model performance based on the LS, the sensitivity, specificity and accuracy of the model in diagnosing AF were 0.992, 0.973, and 0.983 in the internal validation set respectively. In the external validation set, these metrics were 0.989, 0.956, and 0.967, respectively. Furthermore, when evaluating the model’s performance based on ECG records in the MS-AF dataset, the sensitivity, specificity and accuracy of model diagnosis paroxysmal AF were 1.000, 0.870 and 0.876 respectively, and 0.927, 1.000 and 0.973 for the persistent AF.ConclusionsThe model based on the nonlinear LS and deep learning has high accuracy, making it promising for AF screening in primary care. It has potential for generalization and practical application.

FPGA implementation of IIR elliptic filters for de-noising ECG signal

De-noising of ECG signal is very much necessary to monitor the heart health and to diagnose disease. This paper presents a real time de-noising system that efficiently wiped out the noises from corrupted ECG signal and improves its SNR which enhances the interpretation of features present in ECG signal. This proposed system is designed with three Infinite Impulse Response (IIR) elliptic filters in cascaded direct form II structure and is implemented on the FPGA platform. Total 376 ECG samples are taken from the open access public database physiobank. All the architectural level FPGA designs have been developed in Xilinx System Generator (XSG) and implemented on Xilinx Zynq 7000 board. Our design utilized only 3.87 % of the total available resource and consumed 0.321 watt on-chip power. Worst Negative Slack (WNS) for the critical path is 2.101 ns. The positive value of WNS signifies our hardware design implementation run successfully by meeting the required time constraint. To judge the performance of the proposed design, the measurement matrices SNR Improvement, RMSE, PRD, SSNR and SINAD are calculated and achieved 81.11% , 88.89% and 85.41% accuracy for ECG recordings of ECG-ID, MIT-BIH Normal Sinus Rhythm and MIT-BIH Arrhythmia database respectively. The novelty of this proposed denoising system is its robustness, which work satisfactorily for Baseline Wander (BW), Electrode Motion (EM), Power Line Interference (PLI), Muscle or Electromyography (EMG) and as well as Additive White Gaussian Noise (AWGN).

Adaptive augmented cubature Kalman filter/smoother for ECG denoising.

Model-based Bayesian approaches have been widely applied in Electrocardiogram (ECG) signal processing, where their performances heavily rely on the accurate selection of model parameters, particularly the state and measurement noise covariance matrices. In this study, we introduce an adaptive augmented cubature Kalman filter/smoother (CKF/CKS) for ECG processing, which updates the noise covariance matrices at each time step to accommodate diverse noise types and input signal-to-noise ratios (SNRs). Additionally, we incorporate the dynamic time warping technique to enhance the filter's efficiency in the presence of heart rate variability. Furthermore, we propose a method to significantly reduce the computational complexity required for CKF/CKS implementation in ECG processing. The denoising performance of the proposed filter was compared to those of various nonlinear Kalman-based frameworks involving the Extended Kalman filter/smoother (EKF/EKS), the unscented Kalman filter/smoother (UKF/UKS), and the ensemble Kalman filter (EnKF) that was recently proposed for ECG enhancement. In this study, we conducted a comprehensive evaluation and comparison of the performance of various nonlinear Kalman-based frameworks for ECG signal processing, which have been proposed in recent years. Our assessment was carried out on multiple normal ECG segments extracted from different entries in the MIT-BIH Normal Sinus Rhythm Database (NSRDB). This database provides a diverse set of ECG recordings, allowing us to examine the filters' denoising capabilities across various scenarios. By comparing the performance of these filters on the same dataset, we aimed to provide a thorough analysis and identification of the most effective approach for ECG denoising. Two kinds of noises were introduced to such segments: 1-stationary white Gaussian noise and 2-non-stationary real muscle artifact noise. For evaluation, four comparable measures namely the SNR improvement, PRD, correlation coefficient and MSEWPRD were employed. The findings demonstrated that the suggested algorithm outperforms the EKF/EKS, EnKF/EnKS, UKF/UKS methods in both stationary and nonstationary environments regarding SNR improvement, PRD, correlation coefficient and MSEWPRD metrics.

A portable household detection system based on the combination of bidirectional LSTM and residual block for automatical arrhythmia detection.

Arrhythmia is an important component of cardiovascular disease, and electrocardiogram (ECG) is a method to detect arrhythmia. Arrhythmia detection is often paroxysmal, and ECG signal analysis is time-consuming and expensive. We propose a model and device for convenient monitoring of arrhythmia at any time. This work proposes a model combining residual block and bidirectional long-term short-term memory network (BiLSTM) to detect and classify ECG signals. Residual blocks can extract deep features and avoid performance degradation caused by convolutional networks. Combined with the feature of BiLSTM to strengthen the connection relationship of the local window, it can achieve a better classification and prediction effect. Model optimization experiments were performed on the MIT-BIH Atrial Fibrillation Database (AFDB) and MIT-BIH Arrhythmia Database (MITDB). The accuracy simulation results on both long and short signal was higher than 99 %. To further demonstrate the applicability of the model, validation experiments were conducted on MIT-BIH Normal Sinus Rhythm Database (NSRDB) and the Long-Term AF Database (LTAFDB) datasets, and the related recognition accuracy were 99.830 and 91.252 %, respectively. Additionally, we proposed a portable household detection system including an ECG and a blood pressure detection module. The detection accuracy was higher than 98 % using the collected data as testing set. Hence, we thought our system can be used for practical application.

HGSmark: An efficient ECG watermarking scheme using hunger games search and Bayesian regularization BPNN

The extensive use of telemedicine has led to an exponential increase in the exchange of medical data via insecure networks. For example, diagnosis in telemedicine practice often requires electrocardiograms to be shared via Internet. This raises security and authenticity concerns. In this paper, we have proposed HGSmark—an ECG signal watermarking scheme based on multiple embedding strength (MES). We have used the hunger games search (HGS) algorithm to optimize the MES values locally, while maintaining the imperceptibility-robustness trade-off. However, as the optimization using metaheuristic techniques is time-intensive, we have used HGS only once to train a Bayesian regularization-backpropagation neural network (BR-BPNN) on a set of ECG signals. Subsequently, the trained BR-BPNN is employed to quickly compute the optimum MES values. For watermarking, we have employed a wrapping-based fast discrete curvelet transform (FDCT) and singular value decomposition (SVD) to extract the principal component (PC) coefficients of the ECG signal, and embedded a watermark into PC coefficients using the MES values. We have used PC coefficients for watermarking in order to avoid the problem of false-positive. Further, to enhance security, we have encoded the watermark using the logistic chaotic map. We evaluated the performance of HGSmark on MIT-BIH normal sinus rhythm and MIT-BIH arrhythmia databases. The experimental results show that HGSmark achieved average PSNR and PRD values of 57.725 dB and 0.271, respectively and is robust to common interferences. Moreover, the comparison with the state-of-the-art ECG signal watermarking schemes establishes that HGSmark is superior to others in terms of imperceptibility and robustness.

ECG Identification Based on the Gramian Angular Field and Tested with Individuals in Resting and Activity States

In the last decade, biosignals have attracted the attention of many researchers when designing novel biometrics systems. Many of these works use cardiac signals and their representation as electrocardiograms (ECGs). Nowadays, these solutions are even more realistic since we can acquire reliable ECG records by using wearable devices. This paper moves in that direction and proposes a novel approach for an ECG identification system. For that, we transform the ECG recordings into Gramian Angular Field (GAF) images, a time series encoding technique well-known in other domains but not very common with biosignals. Specifically, the time series is transformed using polar coordinates, and then, the cosine sum of the angles is computed for each pair of points. We present a proof-of-concept identification system built on a tuned VGG19 convolutional neural network using this approach. We confirm our proposal's feasibility through experimentation using two well-known public datasets: MIT-BIH Normal Sinus Rhythm Database (subjects at a resting state) and ECG-GUDB (individuals under four specific activities). In both scenarios, the identification system reaches an accuracy of 91%, and the False Acceptance Rate (FAR) is eight times higher than the False Rejection Rate (FRR).

A New Deep Learning Method with Self-Supervised Learning for Delineation of the Electrocardiogram.

Heartbeat characteristic points are the main features of an electrocardiogram (ECG), which can provide important information for ECG-based cardiac diagnosis. In this manuscript, we propose a self-supervised deep learning framework with modified Densenet to detect ECG characteristic points, including the onset, peak and termination points of P-wave, QRS complex wave and T-wave. We extracted high-level features of ECG heartbeats from the QT Database (QTDB) and two other larger datasets, MIT-BIH Arrhythmia Database (MITDB) and MIT-BIH Normal Sinus Rhythm Database (NSRDB) with no human-annotated labels as pre-training. By applying different transformations to ECG signals, the task of discriminating signals before and after transformation was defined as the pretext task. Subsequently, the convolutional layer was frozen and the weights of the self-supervised network were transferred to the downstream task of characteristic point localizations on heart beats in the QT dataset. Finally, the mean ± standard deviation of the detection errors of our proposed self-supervised learning method in QTDB for detecting the onset, peak, and termination points of P-waves, the onset and termination points of QRS waves, and the peak and termination points of T-waves were -0.24 ± 10.04, -0.48 ± 11.69, -0.28 ± 10.19, -3.72 ± 8.18, -4.12 ± 13.54, -0.68 ± 20.42, and 1.34 ± 21.04. The results show that the deep learning network based on the self-supervised framework constructed in this manuscript can accurately detect the feature points of a heartbeat, laying the foundation for automatic extraction of key information related to ECG-based diagnosis.

ECG enhancement using a modified Bayesian framework and particle swarm optimization

Extended Kalman filter (EKF) is a well-known nonlinear Bayesian framework deployed in various electrocardiogram (ECG) processing applications. However, it is not very effective in removing non-stationary noises, such as muscle artifacts (MA) that are common in ECG recordings. As a model-based framework, this filter relies on its predefined state space model. As a result, EKF fails when ECG dynamics don’t comply its state-space model. In order to solve the above issues, in this paper, we propose to use particle swarm optimization (PSO) to find a better and more accurate state-space model. In addition, in order to improve EKF performance in nonstationary environments, we propose a new measurement model. This Model is modified to include non-Gaussian non-stationary additive and stationary noises. However, in order for EKF to use this model, its equations should be reformulated. Two different approaches are proposed in this paper to reformulate EKF equations; 1-state augmentation and 2- measurement differencing strategies. The proposed formulations for the EKF algorithm in this paper enable it to perform better than standard EKF in removing non-stationary contaminants. The proposed filters also preserve the clinical characteristics of ECG signals better than standard EKF. In order to show the effectiveness of the proposed EKF algorithms, their denoising performances were evaluated on the MIT-BIH Normal Sinus Rhythm Database (NSRDB) in the presence of two different types of non-stationary contaminants; 1- synthetic pink noise and 2- real muscle artifact noise extracted from Physionet noise stress database. The results showed that the proposed modified EKF frameworks significantly outperformed the standard EKF framework in non-stationary environments.

On the fractal geometry of different heart rhythms

In this study, we explore the possibility that ECG recordings belong to class of multifractal process for which a large number of scaling exponents are required to characterize their scaling structures. We use the BIDMC Congestive Heart Failure database including long term ECG recordings from 11 men, aged 22 to 71, and 4 women, aged 54 to 63 with severe congestive heart failure and the MIT-BIH Arrhythmia database that contains 48 half-hour excerpts of two-channel ambulatory ECG recordings, obtained from 47 subjects studied by the BIH Arrhythmia Laboratory between 1975 and 1979. We compare these two chronic heart diseases with the control people in the MIT-BIH Normal Sinus Rhythm database which includes 18 long-term ECG recordings of 5 men, aged 26 to 45, and 13 women, aged 20 to 50 without significant arrhythmia. The vibration analysis such as power spectral densities (PSD) analysis has been performed for differentiating the time series. Multifractal dynamics of heartbeat interval signals have been assessed by multifractal spectrum analysis to differentiate normal signals with arrhythmia and severe congestive heart failure patients. We apply Higuchi algorithm to find the fractal complexity of each cardiac rhythm and then compare the signals for different time intervals. According to our analysis, we investigate that neither the power spectral densities nor the Higuchi algorithm to find the fractal dimension alone were sufficient to separate different classes of patients and healthy people. However, when multifractal analysis and scaling exponent were used as a classifier, the three classes were well separated. In addition, multifractal analysis revealed that we have a narrow range of exponents for arrhythmia and congestive heart failure subjects and as a result, a clear loss of multifractality for them. It was of great significance to show that we have a narrower range of exponents for arrhythmia subject compared to congestive heart failure subject which is useful to recognize these two classes of patients with heart disease. Our findings provide a comprehensive framework for diagnostic and classifying different patients with cardiac disease such as arrhythmia and congestive heart failure and differentiate them with normal people without heart disease which is crucial in finding the best diagnostic and controlling strategy in fight against chronic heart disease.

Optimising and validating deep learning approaches for diagnosing atrial fibrillation from few-lead ambulatory electrocardiogram signals

Abstract Funding Acknowledgements Type of funding sources: None. Background Deep learning (DL) has received much attention as a solution for automatically diagnosing atrial fibrillation (AF) from raw ECG signals. However, few studies exist to investigate how DL approaches can be optimally configured and whether their diagnostic performance is externally validated. Purpose To explore how signal-related parameter tuning affects the ability of DL approaches to diagnose AF and validate the optimal approach internally and externally. Methods We applied two dedicated DL models (InceptionTime and MINIROCKET) on a set of 7,966 AF and non-AF (normal or with other abnormalities) ambulatory ECG samples, originating from the MIT-BIH AF, MIT-BIH Normal Sinus Rhythm and Long Term AF databases. We tested the effect of different sample lengths (30sec (s), 10s, 30/10s -30s with a "sliding window" of 10s-), sampling frequencies (200, 100, 50 Hz) and lead numbers (two-, single-), and the role of denoising (Discrete Wavelet Transformation, no denoising) on the ability to diagnose AF, by measuring ROC AUC and sensitivity (SEN) after repeated model training and testing. Under the optimal configuration, we trained 10 replicas of both models on 90% of the data and tested their performance on the remaining 10% (internal validation). Finally, we applied both pre-trained models on a separate dataset (MIT-BIH Arrhythmia) to determine their external validity. Results Although the diagnostic performance did not differ between 30s and 10s signals, the 30/10s setting displayed significantly higher median AUC (0.98) and sensitivity (97.3%, p<0.05 for all comparisons). Signals sampled at 50Hz performed poorer (AUC=0.88, SEN=79.9%) than those at 100Hz (AUC=0.92, SEN=88.7%) and 200Hz (AUC=0.93, SEN=89.2%), although this difference slightly failed to reach statistical significance. Despite denoised signals showing a higher median AUC (0.95 vs. 0.92) and sensitivity (92.8% vs. 88.7%), the difference was not found significant. Similarly, two-lead signals performed better than single-lead ones (AUC=0.92 vs. 0.9 and SEN=88.7% vs. 84.1%, respectively), but without crossing the significance threshold. The internal validation with denoised, 30/10s, two-lead signals, at 100Hz, yielded similarly high performance metrics for both InceptionTime and MINIROCKET (AUC=0.98, SEN=96.9% and AUC=0.98, SEN=97.4%, respectively). In contrast, the performance on the external set dropped significantly (AUC=0.79, SEN=81.4% and AUC=0.72, SEN=83.7%, respectively, p<0.001 for all comparisons). Conclusions Both DL approaches can effectively detect AF in ambulatory ECG signals, with only 3 out of 100 cases missed, designating their promising utility as screening tools for automated AF detection. While optimising tunable parameters can enhance the internal performance of such efforts, their external validation is necessary to establish their robustness "in the wild", since their performance on "unseen" data can be, similarly to our case, notably lower.

Electrocardiogram Biometrics Using Transformer's Self-Attention Mechanism for Sequence Pair Feature Extractor and Flexible Enrollment Scope Identification.

The existing electrocardiogram (ECG) biometrics do not perform well when ECG changes after the enrollment phase because the feature extraction is not able to relate ECG collected during enrollment and ECG collected during classification. In this research, we propose the sequence pair feature extractor, inspired by Bidirectional Encoder Representations from Transformers (BERT)’s sentence pair task, to obtain a dynamic representation of a pair of ECGs. We also propose using the self-attention mechanism of the transformer to draw an inter-identity relationship when performing ECG identification tasks. The model was trained once with datasets built from 10 ECG databases, and then, it was applied to six other ECG databases without retraining. We emphasize the significance of the time separation between enrollment and classification when presenting the results. The model scored 96.20%, 100.0%, 99.91%, 96.09%, 96.35%, and 98.10% identification accuracy on MIT-BIH Atrial Fibrillation Database (AFDB), Combined measurement of ECG, Breathing and Seismocardiograms (CEBSDB), MIT-BIH Normal Sinus Rhythm Database (NSRDB), MIT-BIH ST Change Database (STDB), ECG-ID Database (ECGIDDB), and PTB Diagnostic ECG Database (PTBDB), respectively, over a short time separation. The model scored 92.70% and 64.16% identification accuracy on ECGIDDB and PTBDB, respectively, over a long time separation, which is a significant improvement compared to state-of-the-art methods.

ECGsound for human identification

Novel biometric systems have emerged in recent years as an alternative or complement to traditional identification systems based on passwords (something you know) or tokens (something you have). In this sense, biopotentials signals such as electrocardiograms (cardiac signal) or electroencephalograms (brain signals) have attracted many researchers’ attention. This work proposes an innovative identification technique based on electrocardiograms (ECGs) and musical features (e.g., dynamics, rhythm or timbre) commonly used to characterise audio files. In a nutshell, after pre-processing ECG recordings, we transform them into audio wave files, split them into segments, extract features into five musical dimensions and finally fed a classifier with these instances. The proposal’s workability is confirmed by experimentation using the MIT-BIH Normal Sinus Rhythm Database with 18 subjects and offering an accuracy of 96.6 and a low error rate with FAR and FRR 0.002 and 0.004, respectively.

Spectral decomposition of heart rate variability using generalized harmonic analysis

In this paper, a method for decomposing harmonics in the spectrum of heart rate variability (HRV) using generalized harmonic analysis (GHA) is introduced. First, a simulated RR interval signal generated by an integral pulse frequency modulation model was decomposed spectra by GHA, Pisarenko harmonic decomposition (PHD), and autoregressive (AR) spectral decomposition model. The spectral profiles were obtained by the GHA, PHD, and AR methods for various numbers of extracted sinusoids and model orders from 1 to 48. The spectral profiles of GHA were the most stable. Of the power values of the sinusoids extracted by each method, it was clear that the power values estimated by GHA were approximately equal to the mean square value and closer than that obtained using the PHD or a fast Fourier transform (FFT). Second, a comparison of the power of the low-frequency (LF) and high-frequency (HF) band components reveals that the values obtained by GHA are similar to those obtained by FFT for the analysis using a real ECG signal. Third, Bland-Altman analysis reveals that LF and HF band power value calculated by the GHA are compatible with ones by the Lomb-Scargle periodogram using MIT-BIH normal sinus rhythm database from short term recordings of 30 min. These results suggest that GHA is a useful tool for calculating the power of a component in the spectral analysis of HRV. The limitations of spectral decomposition by GHA are still discussed.

CSNet: A deep learning approach for ECG compressed sensing

Remote electrocardiogram (ECG) monitoring plays a very important role in the prevention and treatment of cardiovascular diseases. However, the current long-term ECG monitoring generates a large amount of data, which puts great pressure on the bandwidth and transmission systems. Compressed sensing (CS) has great attraction for resource-limited wearable devices used in remote ECG monitoring. Traditional CS reconstruction algorithms often require complex signal processing and prior knowledge, and the reconstruction process is time-consuming, which limits the application of CS in remote ECG monitoring systems. This paper proposes a non-iterative fast reconstruction algorithm based on CS and deep learning, which combines convolutional neural network (CNN) and long short-term memory (LSTM) to directly learn the mapping relationship between the rising dimension signal of the measurements and the original signal, and is validated in the MIT-BIH Arrhythmia Database (MITDB). The experimental results show that the reconstruction error of this method is lower than that of traditional algorithms including basis pursuit (BP), orthogonal matching pursuit (OMP), bound-optimization-based block sparse Bayesian learning (BSBL-BO) and rotate-singular value decomposition + basis pursuit (R-SVD+BP) at high compression ratios (CRs, when CR⩾50%). At the same time, for a 30-min ECG signal, only about 0.12 s is needed to complete the reconstruction, which is at least 45 times faster than the traditional algorithms, which is enough to support real-time applications. In addition, the proposed method is also validated in MIT-BIH Normal Sinus Rhythm Database (NSRDB), MIT-BIH Atrial Fibrillation Database (AFDB) and European ST-T Database (EDB), and the reconstructed signals of NSRDB and AFDB meet the clinical requirements at CR⩽70%, and the reconstructed signals of EDB meet the clinical requirements at CR⩽90%.

AN EFFICIENT ALGORITHM FOR R PEAKS DETECTION OF ELECTROCARDIOGRAM SIGNALS

Efficient [Formula: see text] peaks detection is the key to the accurate analysis of electrocardiogram (ECG) signals which is a benefit to the early detection of cardiovascular diseases. In recent years, many effective [Formula: see text] peaks detection methods have been proposed, however, the false detection rate is relatively high when the noisy ECG signal is involved. Based on the property of MTEO that it could enhance the features of signal, a novel [Formula: see text] peaks detection algorithm is proposed in this paper to deal with ECG signals with low SNR. The algorithm includes two stages. In the first stage, a band-pass filter is used for eliminating noise, then the first-order forward differentiation and MTEO are used to transform the ECG signals, at last, the output of MTEO is smoothed with a Moving Averaging filter. In the second stage, the adaptive thresholds method and efficient decision rules are applied to detect the true [Formula: see text] peaks. The efficiency and robustness of the proposed method are substantiated on MIT-BIH Arrhythmia Database (MITDB), Fantasia Database and MIT-BIH Normal Sinus Rhythm Database. The testing of the proposed method on the MITDB showed the following results: Sensitivity [Formula: see text], Positive predictivity [Formula: see text] and Accuracy [Formula: see text]. On Fantasia Database involvement, [Formula: see text], [Formula: see text] and [Formula: see text]. On MIT-BIH Normal Sinus Rhythm Database involvement, [Formula: see text], [Formula: see text] and [Formula: see text]. Compared with other [Formula: see text] peaks detection methods, the proposed algorithm is simple, efficient and robust.

Variations of Time Irreversibility of Heart Rate Variability Under Normobaric Hypoxic Exposure.

In the field of biomedicine, time irreversibility is used to describe how imbalanced and asymmetric biological signals are. As an important feature of signals, the direction of time is always ignored. To find out the variation regularity of time irreversibility of heart rate variability (HRV) in the initial stage of hypoxic exposure, the present study implemented 2 h acute normobaric hypoxic exposure on six young subjects who have no plateau or hypoxia experiences; oxygen concentration was set as 12.9%. Electrocardiogram (ECG) signals were recorded in the whole process and RR interval sequences were extracted. Mathematical operations were executed to transform the difference of adjacent RR intervals into proportion and distance with delay time to conduct time irreversibility analysis of HRV. The same calculating method was implemented on six items randomly picked out from the MIT-BIH normal sinus rhythm database as a control group. Results show that variation of time irreversibility of HRV in a hypoxic environment is different from that in a normoxic environment, time irreversibility indices of a hypoxic group decreases continually at a delay time of 1 and 2, and indices curves of time irreversibility gradually tend to be steady and gather with each other at a delay time of 3 or 4. The control group shows no consistent tendency no matter what the delay time is in the range of 1–4. Our study indicates that in short-time hypoxic exposure, as hypoxic time goes by, regulation of the cardiovascular autonomic nervous system weakens; regulation times and intensity of sympathetic and parasympathetic nerves tend to be equal.

Atrial fibrillation detection with and without atrial activity analysis using lead-I mobile ECG technology

ObjectiveThis paper presents the performance of an atrial fibrillation (AF) detection algorithm in a lead-I mobile ECG setting. Our aim is to demonstrate that false positive cases occurring due to non-AF arrhythmias can be significantly reduced by taking atrial activity into account in addition to heart rhythm. MethodsAF detection was carried out in two ways: only by heart rhythm assessment based on Poincaré plot of RR intervals (A1), and by also performing P wave analysis on the average majority cycle (A2). ResultsThe algorithm was tested on the PhysioNet MIT-BIH Atrial Fibrillation Database, Long-Term AF Database, MIT-BIH Arrhythmia Database, and MIT-BIH Normal Sinus Rhythm Database. A1 produced an average sensitivity (Se) of 97.64 % and specificity (Sp) of 93.06 %. A2 resulted Se = 96.52 % and Sp = 99.12 %. Additionally, clinical data acquisition was performed by the Sanatmetal WIWE mobile ECG system. On the latter records, Se = 100 % and Sp = 95.31 % were obtained with A1, while A2 resulted Se = 100 % and Sp = 100 %. ConclusionAtrial activity analysis significantly increased Sp in records with frequent premature beats (by almost 12 %) and marked sinus arrhythmia (by 7.5 %). Based on our results and data from the literature, we suggest that AF detection methods should be tested more carefully on non-AF arrhythmia cases. Applied in WIWE, the A2 version of our algorithm outperforms the AF detectors of current single-channel mobile ECG systems. SignificanceThe results contribute to the significant reduction of false positive AF detections in lead-I mobile ECG technology.

Denoising of Electrocardiogram Signal Using S-Transform Based Time–Frequency Filtering Approach

Electrocardiogram (ECG) signals are damaged by various types of noise during acquisition and transmission which may mislead the analysis. In this paper, an automated denoising technique based on time–frequency filtering approach is proposed. The S-transform based time–frequency method with morphological processing is employed to visualize the spectrum of the ECG signal. The time–frequency plane is surface fitted to estimate the noise and then a threshold is used to eliminate it. The proposed method has been assessed with numerous abnormal and normal ECG signals selected from the MIT-BIH normal sinus rhythm database. Several noises with varying signal-to-noise ratio are considered for the simulation study. The results showed that the proposed technique is superior to the existing wavelet-based approach. It significantly reduces the mean square error, percentage root mean square difference and improves the signal-to-noise ratio (SNR). Moreover, at lower SNR condition, the proposed approach efficiently suppresses the noise. In the proposed approach, the requirement of the reference signal is eliminated; and at the same time, the structural information is preserved in the denoised signal.

Classification of human electrocardiograms by multi-layer convolutional neural network and hyperparameter optimization

A multi-layer convolutional neural network (MCNN) with hyperparameter optimization (HyperMCNN) is proposed for classifying human electrocardiograms (ECGs). For performance tests of the HyperMCNN, ECG recordings for patients with cardiac arrhythmia (ARR), congestive heart failure (CHF), and normal sinus rhythm (NSR) were obtained from three PhysioNet databases: MIT-BIH Arrhythmia Database, BIDMC Congestive Heart Failure Database, and MIT-BIH Normal Sinus Rhythm Database, respectively. The MCNN hyperparameters in convolutional layers included number of filters, filter size, padding, and filter stride. The hyperparameters in max-pooling layers were pooling size and pooling stride. Gradient method was also a hyperparameter used to train the MCNN model. Uniform experimental design approach was used to optimize the hyperparameter combination for the MCNN. In performance tests, the resulting 16-layer CNN with an appropriate hyperparameter combination (16-layer HyperMCNN) was used to distinguish among ARR, CHF, and NSR. The experimental results showed that the average correct rate and standard deviation obtained by the 16-layer HyperMCNN were superior to those obtained by a 16-layer CNN with a hyperparameter combination given by Matlab examples. Furthermore, in terms of performance in distinguishing among ARR, CHF, and NSR, the 16-layer HyperMCNN was superior to the 25-layer AlexNet, which was the neural network that had the best image identification performance in the ImageNet Large Scale Visual Recognition Challenge in 2012.

WITHDRAWN: Early detection of cardiovascular diseases using deep convolutional neural network & fourier wavelet transform

CVD (Cardiovascular) Diseases is leading cause of human deaths globally. The increasing threats of CVD can be early detected with various medical tests including electrocardiogram (ECG), and also 2D Echo, Stress Test. As ECG is non-invasive, clinical therapeutic agent, so with the help of an ECG signal, early detection of CVD is possible and proper medication can be provided for human life. All these signals from different equipment can, however, be non-stationary and repetitive which takes more time to process and exhausting for physical examination. Moreover, Heart Signal from ECG machine is not a stationary indicator, the discrepancies might not be repeated and may demonstrate up at various periods, hence there is need to adopt computer aided model for fast and accurate prognosis of CVD’s. Also now a day’s various smart devices are available in market which captures the heart rate through pulse detection and also needs fast processing of that heart rate signals for accurately identifying the ECG Signal Wavelengths which further helps in identification of CVD’s. Through this study, we are proposing CVD detection by using Deep Neural Network (DNN) with the help of HRV (Heart Rate Variability). The experimental result had shown that we had achieved 99% accuracy by using our DNN model. Earlier Mental stress detection methods using machine learning methodologies like Support Vector Machine (SVM), Decision Tree (DT) or Random Forest (RF) which gives the accuracy in the range of 80%-85%. In comparison to that we had achieved the highest accuracy rate because we had used Fourier wavelet transform mechanism on ECG signal to detect the variance in HRV (Heart Rate Variability). We had also used 2 additional intermediately layer for training the DNN model and implemented four-fold average accuracy and precision which had given us better result. We had used the combination of MIT-BIH Arrhythmia Database; BIDMC Congestive Heart Failure Database & MIT-BIH Normal Sinus Rhythm Database ECG Dataset. This dataset was resampled with rate of 128 Hz and data length was truncated to a common length of 65,536 samples. This study thus demonstrated the improvement of DNN on CVD detection with the highest accuracy of deep learning model as of now.