A stacked contractive denoising auto-encoder for ECG signal denoising

Artificial Intelligence-Based Atrial Fibrillation Recognition Method for Motion Artifact-Contaminated Electrocardiogram Signals Preprocessed by Adaptive Filtering Algorithm

Atrial fibrillation (AF) is a common arrhythmia, and out-of-hospital, wearable, long-term electrocardiogram (ECG) monitoring can help with the early detection of AF. The presence of a motion artifact (MA) in ECG can significantly affect the characteristics of the ECG signal and hinder early detection of AF. Studies have shown that (a) using reference signals with a strong correlation with MAs in adaptive filtering (ADF) can eliminate MAs from the ECG, and (b) artificial intelligence (AI) algorithms can recognize AF when there is no presence of MAs. However, no literature has been reported on whether ADF can improve the accuracy of AI for recognizing AF in the presence of MAs. Therefore, this paper investigates the accuracy of AI recognition for AF when ECGs are artificially introduced with MAs and processed by ADF. In this study, 13 types of MA signals with different signal-to-noise ratios ranging from +8 dB to -16 dB were artificially added to the AF ECG dataset. Firstly, the accuracy of AF recognition using AI was obtained for a signal with MAs. Secondly, after removing the MAs by ADF, the signal was further identified using AI to obtain the accuracy of the AF recognition. We found that after undergoing ADF, the accuracy of AI recognition for AF improved under all MA intensities, with a maximum improvement of 60%.

A Smartphone-Based M-Health Monitoring System for Arrhythmia Diagnosis

Deep learning technology has been widely adopted in the research of automatic arrhythmia detection. However, there are several limitations in existing diagnostic models, e.g., difficulties in extracting temporal information from long-term ECG signals, a plethora of parameters, and sluggish operation speed. Additionally, the diagnosis performance of arrhythmia is prone to mistakes from signal noise. This paper proposes a smartphone-based m-health system for arrhythmia diagnosis. First, we design a cycle-GAN-based ECG denoising model which takes real-world noise signals as input and aims to produce clean ECG signals. In order to train its two generators and two discriminators simultaneously, we explore an unsupervised pre-training strategy to initialize the generator and accelerate the convergence speed during training. Second, we propose an arrhythmia diagnosis model based on the time convolution network (TCN). This model can identify 34 common arrhythmia events using eight-lead ECG signals, and we deploy such a model on the Android platform to develop an at-home ECG monitoring system. Experimental results have demonstrated that our approach outperforms the existing noise reduction methods and arrhythmia diagnosis models in terms of denoising effect, recognition accuracy, model size, and operation speed, making it more suitable for deployment on mobile devices for m-health monitoring services.

Elimination of Random Mixed Noise in ECG Using Convolutional Denoising Autoencoder With Transformer Encoder

Electrocardiogram (ECG) signals frequently encounter diverse types of noise, such as baseline wander (BW), electrode motion (EM) artifacts, muscle artifact (MA), and others. These noises often occur in combination during the actual data acquisition process, resulting in erroneous or perplexing interpretations for cardiologists. To suppress random mixed noise (RMN) in ECG with less distortion, we propose a Transformer-based Convolutional Denoising AutoEncoder model (TCDAE) in this study. The encoder of TCDAE is composed of three stacked gated convolutional layers and a Transformer encoder block with a point-wise multi-head self-attention module. To obtain minimal distortion in both time and frequency domains, we also propose a frequency weighted Huber loss function in training phase to better approximate the original signals. The TCDAE model is trained and tested on the QT Database (QTDB) and MIT-BIH Noise Stress Test Database (NSTDB), with the training data and testing data coming from different records. All the metrics perform the most robust in overall noise and separate noise intervals for RMN removal compared with the baseline methods. We also conduct generalization tests on the Icentia11k database where the TCDAE outperforms the state-of-the-art models, with a 55% reduction of the false positives in R peak detection after denoising. The TCDAE model approximates the short-term and long-term characteristics of ECG signals and has higher stability even under extreme RMN corruption. The memory consumption and inference speed of TCDAE are also feasible for its deployment in clinical applications.

Multichannel high noise level ECG denoising based on adversarial deep learning

This paper proposes a denoising method based on an adversarial deep learning approach for the post-processing of multi-channel fetal electrocardiogram (ECG) signals. As it's well known, noise leads to misinterpretations of fetal ECG signals and thus limits the use of fetal electrocardiography for healthcare applications. Therefore, denoising algorithms are essential for the exploitation of non-invasive fetal ECG. The proposed method is based on the combination of three end-to-end trained sub-networks to convert noisy fetal ECG signals into clean signals. The first two sub-networks are linked by skip connections and form a deep convolutional network that downsamples the noisy signals into a latent representation and subsequently upsamples this latent representation to recover clean signals. The third sub-network aims to boost the decoder sub-network to generate realistic clean signals. Experiments carried out on synthetic and real data showed that the proposed method improved by the signal-to-noise (SNR) of fetal ECG signals with input SNR ranging from [Formula: see text] to 0 dB by an average of 20 dB, and improve fetal signal quality by significantly increasing the number of true detected QRS complexes and halving QRS complex detection errors.

Accelerated model-based iterative reconstruction strategy for sparse-view photoacoustic tomography aided by multi-channel autoencoder priors

Photoacoustic tomography (PAT) commonly works in sparse view due to data acquisition limitations. However, reconstruction suffers from serious deterioration (e.g., severe artifacts) using traditional algorithms under sparse view. Here, a novel accelerated model-based iterative reconstruction strategy for sparse-view PAT aided by multi-channel autoencoder priors was proposed. A multi-channel denoising autoencoder network was designed to learn prior information, which provides constraints for model-based iterative reconstruction. This integration accelerates the iteration process, leading to optimal reconstruction outcomes. The performance of the proposed method was evaluated using blood vessel simulation data and experimental data. The results show that the proposed method can achieve superior sparse-view reconstruction with a significant acceleration of iteration. Notably, the proposed method exhibits excellent performance under extremely sparse condition (e.g., 32 projections) compared with the U-Net method, with an improvement of 48% in PSNR and 12% in SSIM for in vivo experimental data.

Denoising magnetic resonance spectroscopy (MRS) data using stacked autoencoder for improving signal-to-noise ratio and speed of MRS

BACKGROUND

While magnetic resonance imaging (MRI) provides high resolution anatomical images with sharp soft tissue contrast, magnetic resonance spectroscopy (MRS) enables non-invasive detection and measurement of biochemicals and metabolites. However, MRS has low signal-to-noise ratio (SNR) when concentrations of metabolites are in the range of millimolar. Standard approach of using a high number of signal averaging (NSA) to achieve sufficient SNR comes at the cost of a long acquisition time.

PURPOSE

We propose to use deep-learning approaches to denoise MRS data without increasing NSA. This method has potential to reduce the acquisition time as well as improve SNR and quality of spectra, which could enhance the diagnostic value and broaden the clinical applications of MRS.

METHODS

The study was conducted using data collected from the brain spectroscopy phantom and human subjects. We utilized a stack auto-encoder (SAE) network to train deep learning models for denoising low NSA data (NSA = 1, 2, 4, 8, and 16) randomly truncated from high SNR data collected with high NSA (NSA = 192), which were also used to obtain the ground truth. We applied both self-supervised and fully-supervised training approaches and compared their performance of denoising low NSA data based on improvement in SNR. To prevent overfitting, the SAE network was trained in a patch-based manner. We then tested the denoising methods on noise-containing data collected from the phantom and human subjects, including data from brain tumor patients. We evaluated their performance by comparing the SNR levels and mean squared errors (MSEs) calculated for the whole spectra against high SNR "ground truth", as well as the value of chemical shift of N-acetyl-aspartate (NAA) before and after denoising.

RESULTS

With the SAE model, the SNR of low NSA data (NSA = 1) obtained from the phantom increased by 28.5% and the MSE decreased by 42.9%. For low NSA data of the human parietal and temporal lobes, the SNR increased by 32.9% and the MSE decreased by 63.1%. In all cases, the chemical shift of NAA in the denoised spectra closely matched with the high SNR spectra without significant distortion to the spectra after denoising. Furthermore, the denoising performance of the SAE model was more effective in denoising spectra with higher noise levels.

CONCLUSIONS

The reported SAE denoising method is a model-free approach to enhance the SNR of MRS data collected with low NSA. With the denoising capability, it is possible to acquire MRS data with a few NSA, shortening the scan time while maintaining adequate spectroscopic information for detecting and quantifying the metabolites of interest. This approach has the potential to improve the efficiency and effectiveness of clinical MRS data acquisition by reducing the scan time and increasing the quality of spectroscopic data.

A novel transformer-based ECG dimensionality reduction stacked auto-encoders for arrhythmia beat detection

BACKGROUND

Electrocardiogram (ECG) is a powerful tool for studying cardiac activity and diagnosing various cardiovascular diseases, including arrhythmia. While machine learning and deep learning algorithms have been applied to ECG interpretation, there is still room for improvement. For instance, the commonly used Recurrent Neural Networks (RNNs), reply on its previous state to update and is therefore ineffective for parallel computing. RNN also struggles to efficiently address the issue of long-distance reliance.

PURPOSE

To reduce computational complexity by dimensionality reduction of ECG signals we constructed a Stacked Auto-encoders model using Transformer for ECG-based arrhythmia detection. And overcome the challenges of long-term dependencies and limited parallelizability in traditional RNNs when applied to ECG signal processing.

METHODS

In this paper, a Transformer-Based ECG Dimensionality Reduction Stacked Auto-encoders model is proposed for ECG-based arrhythmia detection. The transformer is used to encode ECG signals into a feature matrix, which is then dimensionally reduced using unsupervised greedy training through the four linear layers. This resulted in a low-dimensional representation of ECG features, which are subsequently classified using support vector machines (SVM) to minimize overfitting.

RESULTS

The proposed method is benchmarked on the MIT-BIH Arrhythmia database. In the 10-fold cross validation of beat-based arrhythmia detection, the average accuracy, sensitivity, specificity and F1 score of the proposed method are 99.83%, 98.84%, 99.84% and 99.13%, respectively, for the record-based arrhythmia detection which refers to the approach where the training and testing sets use ECG data from independent recorded patients are 88.10%, 49.79%, 91.56% and 39.95%, respectively.

CONCLUSIONS

Compared to other existing ECG-based arrhythmia detection methods, our proposed approach exhibits improved detection accuracy and stronger generalization for arrhythmia beats. Additionally, the use of the record-based data division method makes our approach more suitable for clinical practice.

A denoising method of ECG signal based on variational autoencoder and masked convolution

Wearable electrocardiogram (ECG) equipment can realize continuous monitoring of cardiovascular diseases, but these devices are more susceptible to interference from various noises, which will seriously reduce the diagnostic correctness. In this work, a novel noise reduction model for ECG signals is proposed based on variational autoencoder and masked convolution. The variational Bayesian inference is conducted to capture the global features of the ECG signals by encouraging the approximate posterior of the latent variables to fit the prior distribution, and we use the skip connection and feature concatenation to realize the information interaction across the channels. To strengthen the connection of local features of the ECG signals, the masked convolution module is used to extract local feature information, which supplement the global features and the noise reduction performance of whole model can be greatly improved. Experiments are carried out on the MIT-BIH arrythmia database, and the results display that the performance metrics of signal-to-noise ratio (SNR) and root mean square error (RMSE) are significantly improved compared with other approaches while causing less signal distortion.

Deep learning-based protoacoustic signal denoising for proton range verification

Proton therapy is a type of radiation therapy that can provide better dose distribution compared to photon therapy by delivering most of the energy at the end of range, which is called the Bragg peak (BP). The protoacoustic technique was developed to determine the BP locationsin vivo, but it requires a large dose delivery to the tissue to obtain a high number of signal averaging (NSA) to achieve a sufficient signal-to-noise ratio (SNR), which is not suitable for clinical use. A novel deep learning-based technique has been proposed to denoise acoustic signals and reduce BP range uncertainty with much lower doses. Three accelerometers were placed on the distal surface of a cylindrical polyethylene (PE) phantom to collect protoacoustic signals. In total, 512 raw signals were collected at each device. Device-specific stack autoencoder (SAE) denoising models were trained to denoise the noise-containing input signals, which were generated by averaging only 1, 2, 4, 8, 16, or 24 raw signals (low NSA signals), while the clean signals were obtained by averaging 192 raw signals (high NSA). Both supervised and unsupervised training strategies were employed, and the evaluation of the models was based on mean squared error (MSE), SNR, and BP range uncertainty. Overall, the supervised SAEs outperformed the unsupervised SAEs in BP range verification. For the high accuracy detector, it achieved a BP range uncertainty of 0.20 ± 3.44 mm by averaging over 8 raw signals, while for the other two low accuracy detectors, they achieved the BP uncertainty of 1.44 ± 6.45 mm and -0.23 ± 4.88 mm by averaging 16 raw signals, respectively. This deep learning-based denoising method has shown promising results in enhancing the SNR of protoacoustic measurements and improving the accuracy in BP range verification. It greatly reduces the dose and time for potential clinical applications.

Hierarchical deep learning with Generative Adversarial Network for automatic cardiac diagnosis from ECG signals

Cardiac disease is the leading cause of death in the US. Accurate heart disease detection is critical to timely medical treatment to save patients' lives. Routine use of the electrocardiogram (ECG) is the most common method for physicians to assess the cardiac electrical activities and detect possible abnormal conditions. Fully utilizing the ECG data for reliable heart disease detection depends on developing effective analytical models. In this paper, we propose a two-level hierarchical deep learning framework with Generative Adversarial Network (GAN) for ECG signal analysis. The first-level model is composed of a Memory-Augmented Deep AutoEncoder with GAN (MadeGAN), which aims to differentiate abnormal signals from normal ECGs for anomaly detection. The second-level learning aims at robust multi-class classification for different arrhythmia identification, which is achieved by integrating the transfer learning technique to transfer knowledge from the first-level learning with the multi-branching architecture to handle the data-lacking and imbalanced data issues. We evaluate the performance of the proposed framework using real-world ECG data from the MIT-BIH arrhythmia database. Experimental results show that our proposed model outperforms existing methods that are commonly used in current practice.

A Reparameterization Multifeature Fusion CNN for Arrhythmia Heartbeats Classification

Aiming at arrhythmia heartbeats classification, a novel multifeature fusion deep learning-based method is proposed. The stationary wavelet transforms (SWT) and RR interval features are firstly extracted. Based on the traditional one-dimensional convolutional neural network (1D-CNN), a parallel multibranch convolutional network is designed for training. The subband of SWT is input into the multiscale 1D-CNN separately. The output fused with RR interval features are fed to the fully connected layer for classification. To achieve the lightweight network while maintaining the powerful inference capability of the multibranch structure, the redundant branches of the network are removed by reparameterization. Experimental results and analysis show that it outperforms existing methods by many in arrhythmic heartbeat classification.

A lightweight U-net for ECG denoising using knowledge distillation

Objective.Electrocardiogram (ECG) signals are easily polluted by various noises which are likely to have adverse effects on subsequent interpretations. Research on model lightweighting can promote the practical application of deep learning-based ECG denoising methods in real-time processing.Approach.Firstly, grouped convolution and conventional convolution are combined to replace the continuous conventional convolution in the model, and the depthwise convolution with stride is used to compress the feature map in the encoder modules. Secondly, additional identity connections and a local maximum and minimum enhancement module are designed, which can retain the detailed information and characteristic waveform in the ECG waveform while effectively denoising. Finally, we develop knowledge distillation in the experiments, which further improves the ECG denoising performance without increasing the model complexity. The ground-truth ECG is from The China Physiological Signal Challenge (CPSC) 2018, and the noise signal is from the MIT-BIH Noise Stress Test Database (NSTDB). We evaluate denoising performance using the signal-to-noise ratio (SNR), the root mean square error (RMSE) and the Pearson correlation coefficient (P). We use the floating point of operations (FLOPs) and parameters to calculate computational complexity.Main Results.Different data generation processes are used to conduct experiments: group 1, group 2 and group 3. The results show that the proposed model (U_Lde-net) can improve SNRs by 10.30 dB, 12.16 dB and 12.61 dB; reduce RMSEs by 9.88 × 10^-2, 20.63 × 10^-2and 15.25 × 10^-2; and increasePs by 14.77 × 10^-2, 27.74 × 10^-2and 21.32 × 10^-2. Moreover, the denoising performance after knowledge distillation is further improved. The U_Lde-net has parameters of 6.9 K and FLOPs of 6.6 M, which are much smaller than the compared models.Significance.We designed a lightweight model, but also retain adequate ECG denoising performance. We believe that this method can be successfully applied to practical applications under time or memory limits.

Improving the Accuracy of R-Peak Detection in a Wearable Armband Device for Daily Life Electrocardiogram Monitoring Using a Deep Convolutional Denoising Encoder-Decoder Network

Continuous long-term heart rate (HR) monitoring using wearable devices is desirable to aid in the diagnosis of many health-related conditions. Recently, we have developed an armband device that does not use obstructive leads, has dry electrodes which are convenient for long-term electrocardiogram (ECG) recording, and has been shown to be an effective alternate approach for continuous ECG monitoring. However, motion artifacts (MA) due to electromyogram (EMG) contractions are acknowledged as the major challenge of an armband. In this study, we used a deep convolutional neural network denoising encoder-decoder (CNNDED) to enhance the accuracy of R-peak detection in MA-corrupted ECG recordings obtained by an armband device. We collected simultaneous 24-hour ECG recordings using both the armband device and a Holter monitor on 10 subjects. Each 10-sec ECG segment was converted to a time-frequency representation and subsequently used as the input to CNNDED. During the training process, the model learned to accentuate the location of R peaks by amplifying their values in each ECG beat and suppressing the remaining waveforms. For the training output, the model used the R-peak location information from the simultaneously collected Holter ECG data, which were considered as the reference. The performance of CNNDED was evaluated on an independent test data set using the standard performance metrics. The mean relative error of the estimated HR with respect to the Holter data was 17.5 and 7.3 beats/min, pre- and post-CNNDED, respectively. The mean relative difference of the root mean square of successive difference values were 0.23 and 0.06 before and after applying CNNDED, respectively. Although further study is needed, the current preliminary results suggest that CNNDED can improve detection of R peaks even when they are completely buried in the presence of EMG artifacts.

An ECG Signal Denoising Method Using Conditional Generative Adversarial Net

In this paper, a novel denoising method for electrocardiogram (ECG) signal is proposed to improve performance and availability under multiple noise cases. The method is based on the framework of conditional generative adversarial network (CGAN), and we improved the CGAN framework for ECG denoising. The proposed framework consists of two networks: a generator that is composed of the optimized convolutional auto-encoder (CAE) and a discriminator that is composed of four convolution layers and one full connection layer. As the convolutional layers of CAE can preserve spatial locality and the neighborhood relations in the latent higher-level feature representations of ECG signal, and the skip connection facilitates the gradient propagation in the denoising training process, the trained denoising model has good performance and generalization ability. The extensive experimental results on MIT-BIH databases show that for single noise and mixed noises, the average signal-to-noise ratio (SNR) of denoised ECG signal is above 39dB, and it is better than that of the state-of-the-art methods. Furthermore, the denoised classification results of four cardiac diseases show that the average accuracy increased above 32% under multiple noises under SNR=0dB. So, the proposed method can remove noise effectively as well as keep the details of the features of ECG signals.

Two-stage ECG signal denoising based on deep convolutional network

Background.An electrocardiogram (ECG) is an effective and non-invasive indicator for the detection and prevention of arrhythmia. ECG signals are susceptible to noise contamination, which can lead to errors in ECG interpretation. Therefore, ECG pretreatment is important for accurate analysis.Methods.The ECG data used are from CPSC2018, and the noise signal is from MIT-BIH Noise Stress Test Database. In the experiment, the signal-to-noise ratio (SNR), the root mean square error (RMSE), and the correlation coefficientPare used to evaluate the performance of the network. The method proposed is divided into two stages. In the first stage, a U_de-net model is designed for ECG signal denoising to eliminate noise. The DR-net model in the second stage is used to reconstruct the ECG signal and to correct the waveform distortion caused by noise removal in the first stage. In this paper, the U_de-net and the DR-net are constructed by the convolution method to achieve end-to-end mapping from noisy ECG signals to clean ECG signals.Result.In SNR, RMSE andPindicators, U_de-net + DR-net proposed in this paper can achieve the best performance compared with the other five schemes (FCN, U-net etc). In the three data sets, SNR can be increased by 11.61 dB, 13.71 dB and 14.40 dB and RMSE can be reduced by 10.46 × 10^-2, 21.55 × 10^-2and 15.98 × 10^-2.Conclusions.Despite the contradictory results, the proposed two-stages method can achieve both the elimination of noise and the preservation of effective details to a large extent of the signals. The proposed method has good application prospects in clinical practice.

Applications of Machine Learning in Ambulatory ECG

The ambulatory ECG (AECG) is an important diagnostic tool for many heart electrophysiology-related cases. AECG covers a wide spectrum of devices and applications. At the core of these devices and applications are the algorithms responsible for signal conditioning, ECG beat detection and classification, and event detections. Over the years, there has been huge progress for algorithm development and implementation thanks to great efforts by researchers, engineers, and physicians, alongside the rapid development of electronics and signal processing, especially machine learning (ML). The current efforts and progress in machine learning fields are unprecedented, and many of these ML algorithms have also been successfully applied to AECG applications. This review covers some key AECG applications of ML algorithms. However, instead of doing a general review of ML algorithms, we are focusing on the central tasks of AECG and discussing what ML can bring to solve the key challenges AECG is facing. The center tasks of AECG signal processing listed in the review include signal preprocessing, beat detection and classification, event detection, and event prediction. Each AECG device/system might have different portions and forms of those signal components depending on its application and the target, but these are the topics most relevant and of greatest concern to the people working in this area.

A New ECG Denoising Framework Using Generative Adversarial Network

This paper presents a novel Electrocardiogram (ECG) denoising approach based on the generative adversarial network (GAN). Noise is often associated with the ECG signal recording process. Denoising is central to most of the ECG signal processing tasks. The current ECG denoising techniques are based on the time domain signal decomposition methods. These methods use some kind of thresholding and filtering approaches. In our proposed technique, convolutional neural network (CNN) based GAN model is effectively trained for ECG noise filtering. In contrast to existing techniques, we performed end-to-end GAN model training using the clean and noisy ECG signals. MIT-BIH Arrhythmia database is used for all the qualitative and quantitative analyses. The improved ECG denoising performance open the door for further exploration of GAN based ECG denoising approach.

Using the Redundant Convolutional Encoder-Decoder to Denoise QRS Complexes in ECG Signals Recorded with an Armband Wearable Device

Long-term electrocardiogram (ECG) recordings while performing normal daily routines are often corrupted with motion artifacts, which in turn, can result in the incorrect calculation of heart rates. Heart rates are important clinical information, as they can be used for analysis of heart-rate variability and detection of cardiac arrhythmias. In this study, we present an algorithm for denoising ECG signals acquired with a wearable armband device. The armband was worn on the upper left arm by one male participant, and we simultaneously recorded three ECG channels for 24 h. We extracted 10-s sequences from armband recordings corrupted with added noise and motion artifacts. Denoising was performed using the redundant convolutional encoder-decoder (R-CED), a fully convolutional network. We measured the performance by detecting R-peaks in clean, noisy, and denoised sequences and by calculating signal quality indices: signal-to-noise ratio (SNR), ratio of power, and cross-correlation with respect to the clean sequences. The percent of correctly detected R-peaks in denoised sequences was higher than in sequences corrupted with either added noise (70-100% vs. 34-97%) or motion artifacts (91.86% vs. 61.16%). There was notable improvement in SNR values after denoising for signals with noise added (7-19 dB), and when sequences were corrupted with motion artifacts (0.39 dB). The ratio of power for noisy sequences was significantly lower when compared to both clean and denoised sequences. Similarly, cross-correlation between noisy and clean sequences was significantly lower than between denoised and clean sequences. Moreover, we tested our denoising algorithm on 60-s sequences extracted from recordings from the Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) arrhythmia database and obtained improvement in SNR values of 7.08 ± 0.25 dB (mean ± standard deviation (sd)). These results from a diverse set of data suggest that the proposed denoising algorithm improves the quality of the signal and can potentially be applied to most ECG measurement devices.

Opportunities and challenges of deep learning methods for electrocardiogram data: A systematic review

BACKGROUND

The electrocardiogram (ECG) is one of the most commonly used diagnostic tools in medicine and healthcare. Deep learning methods have achieved promising results on predictive healthcare tasks using ECG signals.

OBJECTIVE

This paper presents a systematic review of deep learning methods for ECG data from both modeling and application perspectives.

METHODS

We extracted papers that applied deep learning (deep neural network) models to ECG data that were published between January 1st of 2010 and February 29th of 2020 from Google Scholar, PubMed, and the Digital Bibliography & Library Project. We then analyzed each article according to three factors: tasks, models, and data. Finally, we discuss open challenges and unsolved problems in this area.

RESULTS

The total number of papers extracted was 191. Among these papers, 108 were published after 2019. Different deep learning architectures have been used in various ECG analytics tasks, such as disease detection/classification, annotation/localization, sleep staging, biometric human identification, and denoising.

CONCLUSION

The number of works on deep learning for ECG data has grown explosively in recent years. Such works have achieved accuracy comparable to that of traditional feature-based approaches and ensembles of multiple approaches can achieve even better results. Specifically, we found that a hybrid architecture of a convolutional neural network and recurrent neural network ensemble using expert features yields the best results. However, there are some new challenges and problems related to interpretability, scalability, and efficiency that must be addressed. Furthermore, it is also worth investigating new applications from the perspectives of datasets and methods.

SIGNIFICANCE

This paper summarizes existing deep learning research using ECG data from multiple perspectives and highlights existing challenges and problems to identify potential future research directions.

Deep Learning-Based Stacked Denoising and Autoencoder for ECG Heartbeat Classification

The electrocardiogram (ECG) is a widely used, noninvasive test for analyzing arrhythmia. However, the ECG signal is prone to contamination by different kinds of noise. Such noise may cause deformation on the ECG heartbeat waveform, leading to cardiologists’ mislabeling or misinterpreting heartbeats due to varying types of artifacts and interference. To address this problem, some previous studies propose a computerized technique based on machine learning (ML) to distinguish between normal and abnormal heartbeats. Unfortunately, ML works on a handcrafted, feature-based approach and lacks feature representation. To overcome such drawbacks, deep learning (DL) is proposed in the pre-training and fine-tuning phases to produce an automated feature representation for multi-class classification of arrhythmia conditions. In the pre-training phase, stacked denoising autoencoders (DAEs) and autoencoders (AEs) are used for feature learning; in the fine-tuning phase, deep neural networks (DNNs) are implemented as a classifier. To the best of our knowledge, this research is the first to implement stacked autoencoders by using DAEs and AEs for feature learning in DL. Physionet’s well-known MIT-BIH Arrhythmia Database, as well as the MIT-BIH Noise Stress Test Database (NSTDB). Only four records are used from the NSTDB dataset: 118 24 dB, 118 −6 dB, 119 24 dB, and 119 −6 dB, with two levels of signal-to-noise ratio (SNRs) at 24 dB and −6 dB. In the validation process, six models are compared to select the best DL model. For all fine-tuned hyperparameters, the best model of ECG heartbeat classification achieves an accuracy, sensitivity, specificity, precision, and F1-score of 99.34%, 93.83%, 99.57%, 89.81%, and 91.44%, respectively. As the results demonstrate, the proposed DL model can extract high-level features not only from the training data but also from unseen data. Such a model has good application prospects in clinical practice.

Automated beat-wise arrhythmia diagnosis using modified U-net on extended electrocardiographic recordings with heterogeneous arrhythmia types

Abnormality of the cardiac conduction system can induce arrhythmia - abnormal heart rhythm - that can frequently lead to other cardiac diseases and complications, and are sometimes life-threatening. These conduction system perturbations can manifest as morphological changes on the surface electrocardiographic (ECG) signal. Assessment of these morphological changes can be challenging and time-consuming, as ECG signal features are often low in amplitude and subtle. The main aim of this study is to develop an automated computer aided diagnostic (CAD) system that can expedite the process of arrhythmia diagnosis, as an aid to clinicians to provide appropriate and timely intervention to patients. We propose an autoencoder of ECG signals that can diagnose normal sinus beats, atrial premature beats (APB), premature ventricular contractions (PVC), left bundle branch block (LBBB) and right bundle branch block (RBBB). Apart from the first, the rest are morphological beat-to-beat elements that characterize and constitute complex arrhythmia. The novelty of this work lies in how we modified the U-net model to perform beat-wise analysis on heterogeneously segmented ECGs of variable lengths derived from the MIT-BIH arrhythmia database. The proposed system has demonstrated self-learning ability in generating class activations maps, and these generated maps faithfully reflect the cardiac conditions in each ECG cardiac cycle. It has attained a high classification accuracy of 97.32% in diagnosing cardiac conditions, and 99.3% for R peak detection using a ten-fold cross validation strategy. Our developed model can help physicians to screen ECG accurately, potentially resulting in timely intervention of patients with arrhythmia.

Deep Learning on 1-D Biosignals: a Taxonomy-based Survey

OBJECTIVES

 Deep learning models such as convolutional neural networks (CNNs) have been applied successfully to medical imaging, but biomedical signal analysis has yet to fully benefit from this novel approach. Our survey aims at (i) reviewing deep learning techniques for biosignal analysis in computer- aided diagnosis; and (ii) deriving a taxonomy for organizing the growing number of applications in the field.

METHODS

 A comprehensive literature research was performed using PubMed, Scopus, and ACM. Deep learning models were classified with respect to the (i) origin, (ii) dimension, and (iii) type of the biosignal as input to the deep learning model; (iv) the goal of the application; (v) the size and (vi) type of ground truth data; (vii) the type and (viii) schedule of learning the network; and (ix) the topology of the model.

RESULTS

 Between January 2010 and December 2017, a total 71 papers were published on the topic. The majority (n = 36) of papers are on electrocariography (ECG) signals. Most applications (n = 25) aim at detection of patterns, while only a few (n = 6) at predection of events. Out of 36 ECG-based works, many (n = 17) relate to multi-lead ECG. Other biosignals that have been identified in the survey are electromyography, phonocardiography, photoplethysmography, electrooculography, continuous glucose monitoring, acoustic respiratory signal, blood pressure, and electrodermal activity signal, while ballistocardiography or seismocardiography have yet to be analyzed using deep learning techniques. In supervised and unsupervised applications, CNNs and restricted Boltzmann machines are the most and least frequently used, (n = 34) and (n = 15), respectively.

CONCLUSION

 Our key-code classification of relevant papers was used to cluster the approaches that have been published to date and demonstrated a large variability of research with respect to data, application, and network topology. Future research is expected to focus on the standardization of deep learning architectures and on the optimization of the network parameters to increase performance and robustness. Furthermore, application-driven approaches and updated training data from mobile recordings are needed.