Original Electrocardiogram Research Articles

Inferring ECG Waveforms from PPG Signals with a Modified U-Net Neural Network.

There are two widely used methods to measure the cardiac cycle and obtain heart rate measurements: the electrocardiogram (ECG) and the photoplethysmogram (PPG). The sensors used in these methods have gained great popularity in wearable devices, which have extended cardiac monitoring beyond the hospital environment. However, the continuous monitoring of ECG signals via mobile devices is challenging, as it requires users to keep their fingers pressed on the device during data collection, making it unfeasible in the long term. On the other hand, the PPG does not contain this limitation. However, the medical knowledge to diagnose these anomalies from this sign is limited by the need for familiarity, since the ECG is studied and used in the literature as the gold standard. To minimize this problem, this work proposes a method, PPG2ECG, that uses the correlation between the domains of PPG and ECG signals to infer from the PPG signal the waveform of the ECG signal. PPG2ECG consists of mapping between domains by applying a set of convolution filters, learning to transform a PPG input signal into an ECG output signal using a U-net inception neural network architecture. We assessed our proposed method using two evaluation strategies based on personalized and generalized models and achieved mean error values of 0.015 and 0.026, respectively. Our method overcomes the limitations of previous approaches by providing an accurate and feasible method for continuous monitoring of ECG signals through PPG signals. The short distances between the infer-red ECG and the original ECG demonstrate the feasibility and potential of our method to assist in the early identification of heart diseases.

Development and Validation of a Real-Time Service Model for Noise Removal and Arrhythmia Classification Using Electrocardiogram Signals.

Arrhythmias range from mild nuisances to potentially fatal conditions, detectable through electrocardiograms (ECGs). With advancements in wearable technology, ECGs can now be monitored on-the-go, although these devices often capture noisy data, complicating accurate arrhythmia detection. This study aims to create a new deep learning model that utilizes generative adversarial networks (GANs) for effective noise removal and ResNet for precise arrhythmia classification from wearable ECG data. We developed a deep learning model that cleans ECG measurements from wearable devices and detects arrhythmias using refined data. We pretrained our model using the MIT-BIH Arrhythmia and Noise databases. Least squares GANs were used for noise reduction, maintaining the integrity of the original ECG signal, while a residual network classified the type of arrhythmia. After initial training, we applied transfer learning with actual ECG data. Our noise removal model significantly enhanced data clarity, achieving over 30 dB in a signal-to-noise ratio. The arrhythmia detection model was highly accurate, with an F1-score of 99.10% for noise-free data. The developed model is capable of real-time, accurate arrhythmia detection using wearable ECG devices, allowing for immediate patient notification and facilitating timely medical response.

AI-enhanced reconstruction of the 12-lead electrocardiogram via 3-leads with accurate clinical assessment

The 12-lead electrocardiogram (ECG) is an integral component to the diagnosis of a multitude of cardiovascular conditions. It is performed using a complex set of skin surface electrodes, limiting its use outside traditional clinical settings. We developed an artificial intelligence algorithm, trained over 600,000 clinically acquired ECGs, to explore whether fewer leads as input are sufficient to reconstruct a 12-lead ECG. Two limb leads (I and II) and one precordial lead (V3) were required to generate a reconstructed 12-lead ECG highly correlated with the original ECG. An automatic algorithm for detection of ECG features consistent with acute myocardial infarction (MI) performed similarly for original and reconstructed ECGs (AUC = 0.95). When interpreted by cardiologists, reconstructed ECGs achieved an accuracy of 81.4 ± 5.0% in identifying ECG features of ST-segment elevation MI, comparable with the original 12-lead ECGs (accuracy 84.6 ± 4.6%). These results will impact development efforts to innovate ECG acquisition methods with simplified tools in non-specialized settings.

Comparative analysis of parametric B-spline and Hermite cubic spline based methods for accurate ECG signal modeling

Analyzing Electrocardiogram (ECG) signals is imperative for diagnosing cardiovascular diseases. However, evaluating ECG analysis techniques faces challenges due to noise and artifacts in actual signals. Machine learning for automatic diagnosis encounters data acquisition hurdles due to medical data privacy constraints. Addressing these issues, ECG modeling assumes a crucial role in biomedical and parametric spline-based methods have garnered significant attention for their ability to accurately represent the complex temporal dynamics of ECG signals. This study conducts a comparative analysis of two parametric spline-based methods—B-spline and Hermite cubic spline—for ECG modeling, aiming to identify the most effective approach for accurate and reliable ECG representation. The Hermite cubic spline serves as one of the most effective interpolation methods, while B-spline is an approximation method. The comparative analysis includes both qualitative and quantitative evaluations. Qualitative assessment involves visually inspecting the generated spline-based models, comparing their resemblance to the original ECG signals, and employing power spectrum analysis. Quantitative analysis incorporates metrics such as root mean square error (RMSE), Percentage Root Mean Square Difference (PRD) and cross correlation, offering a more objective measure of the model's performance. Preliminary results indicate promising capabilities for both spline-based methods in representing ECG signals. However, the analysis unveils specific strengths and weaknesses for each method. The B-spline method offers greater flexibility and smoothness, while the cubic spline method demonstrates superior waveform capturing abilities with the preservation of control points, a critical aspect in the medical field. Presented research provides valuable insights for researchers and practitioners in selecting the most appropriate method for their specific ECG modeling requirements. Adjustments to control points and parameterization enable the generation of diverse ECG waveforms, enhancing the versatility of this modeling technique. This approach has the potential to extend its utility to other medical signals, presenting a promising avenue for advancing biomedical research.

Detection of congestive heart failure based on Gramian angular field and two-dimensional symbolic phase permutation entropy

Congestive heart failure (CHF) is a serious threat to human health. Electrocardiogram (ECG) signals have been proven to be useful in the detection of CHF. However, the low amplitude and short duration of the ECG signals, as well as the superimposed noise during the real-time acquisition of the signal, seriously affect the CHF detection. To improve the detection rate of CHF, this paper proposes a congestive heart failure detection method based on Gramian angular field (GAF) and two-dimensional symbolic phase permutation entropy (SPPE2D). The significant advantage of this method is that it reduces the sensitivity to noise, and good performance can be obtained without denoising using raw ECG signals. We segment the original ECG signals into 2 s non-overlapping segments and convert them into images using the GAF method. Then, the SPPE2D algorithm is proposed to measure the complexity between normal sinus rhythm (NSR) and CHF, and analyze the anti-noise performance of the algorithm. Finally, the SPPE2D features of GAF images are computed and input into a support vector machine (SVM) for CHF detection. Classification accuracy on the Massachusetts Institute of Technology − Beth Israel Hospital Normal Sinus Rhythm Database and Beth Israel Deaconess Medical Center Congestive Heart Failure Database is 99.59%, sensitivity is 99.42%, specificity is 99.80%, and F1-score is 99.62%. The accuracy of detecting CHF reach more than 97.75% in the other five CHF databases. The experimental results show that the method based on GAF and SPPE2D can effectively detect CHF by images of ECG signals and has good robustness. CHF can be detected using the 2 s sample lengths of ECG signals recording with high sensitivity, giving clinicians ample time to treat patients with CHF.

Impact of QRS misclassifications on heart-rate-variability parameters (results from the CARLA cohort study).

Heart rate variability (HRV), an important marker of autonomic nervous system activity, is usually determined from electrocardiogram (ECG) recordings corrected for extrasystoles and artifacts. Especially in large population-based studies, computer-based algorithms are used to determine RR intervals. The Modular ECG Analysis System MEANS is a widely used tool, especially in large studies. The aim of this study was therefore to evaluate MEANS for its ability to detect non-sinus ECG beats and artifacts and to compare HRV parameters in relation to ECG processing. Additionally, we analyzed how ECG processing affects the statistical association of HRV with cardiovascular disease (CVD) risk factors. 20-min ECGs from 1,674 subjects of the population-based CARLA study were available for HRV analysis. All ECGs were processed with the ECG computer program MEANS. A reference standard was established by experienced clinicians who visually inspected the MEANS-processed ECGs and reclassified beats if necessary. HRV parameters were calculated for 5-minute segments selected from the original 20-minute ECG. The effects of misclassified typified normal beats on i) HRV calculation and ii) the associations of CVD risk factors (sex, age, diabetes, myocardial infarction) with HRV were modeled using linear regression. Compared to the reference standard, MEANS correctly classified 99% of all beats. The averaged sensitivity of MEANS across all ECGs to detect non-sinus beats was 76% [95% CI: 74.1;78.5], but for supraventricular extrasystoles detection sensitivity dropped to 38% [95% CI: 36.8;38.5]. Time-domain parameters were less affected by false sinus beats than frequency parameters. Compared to the reference standard, MEANS resulted in a higher SDNN on average (mean absolute difference 1.4ms [95% CI: 1.0;1.7], relative 4.9%). Other HRV parameters were also overestimated as well (between 6.5 and 29%). The effect estimates for the association of CVD risk factors with HRV did not differ between the editing methods. We have shown that the use of the automated MEANS algorithm may lead to an overestimation of HRV due to the misclassification of non-sinus beats, especially in frequency domain parameters. However, in population-based studies, this has no effect on the observed associations of HRV with risk factors, and therefore an automated ECG analyzing algorithm as MEANS can be recommended here for the determination of HRV parameters.

Detection of Cardiovascular Disease Using ECG Images in Machine Learning and Deep Learning

One of the most prominent tools for detecting cardiovascular problems is the electrocardiogram (ECG). The electrocardiogram (ECG or EKG) is a diagnostic tool that is used to routinely assess the electrical and muscular functions of the heart. Even though it is a comparatively simple test to perform, the interpretation of the ECG charts requires considerable amounts of training. Till recently, the majority of ECG records were kept on paper. Thus, manually examining and re-examining the ECG paper records often can be a time-consuming and daunting process. If we digitize such paper ECG records, we can perform automated diagnosis and analysis. The main goal of this project is to use machine learning to convert ECG paper records into a 1-D signal. This can be achieved by extracting the P, QRS, and T waves that exist in ECG signals to demonstrate the electrical activity of the heart using various techniques. The techniques include splitting the original ECG report into 13 Leads, extracting and converting into the signal, smoothing, converting them to binary images using threshold and scaling. Post-feature-extraction, dimension reduction techniques like Principal Component Analysis are applied to understand the data. Multiple classifiers like k-nearest neighbors (KNN), Logistic Regression, Support Vector Machine (SVM), and Voting Based Ensemble Classifier are implemented, and based on the acceptable criteria on the accuracy, precision, recall, f1-score, and support, the model will be finalized. This final model will aid in the diagnosing of cardiac diseases, to detect whether a patient has/had Myocardial Infarction, Abnormal Heartbeat, or the patient is hale and healthy by inferring the ECG reports

Universal 12-lead ECG representation for signal denoising and cardiovascular disease detection by fusing generative and contrastive learning

With the wide use of electrocardiogram (ECG) technology, more and more ECGs have been collected and stored. However, ECG labeling is costly and laborious, the utilization of unlabeled ECG is still a critical challenge. Self-supervised learning (SSL) is a way to deal with this problem, which can learn representations from unlabeled ECG and use them for downstream tasks. In this study, a novel SSL model that fuses generative learning (denoising autoencoder, DAE) and contrastive learning (CL) is proposed to learn robust representations from unlabeled ECG for downstream denoising and classification. The 12-lead ECGs are separated into one-dimensional single-lead ECGs and ECG-specific noises are added to the original ECGs as the data augmentation method. To improve the model's denoising and feature extraction abilities, the reconstruction loss and contrastive loss are combined during the pretraining phase. In the downstream classification task, a multi-branch network is used to enhance the correlation between different ECG leads. As a result, it improves the denoising performance over the standard DAE by an average of 5.20%. In the classification tasks for the Chapman, PTB-XL, and CPSC2018 databases, compared to existing work, the accuracies in linear probe are increased by at least 2.49%, 1.09%, and 2.61%, respectively. In addition, an SoC (system-on-a-chip) based heterogeneous deployment scheme is designed. It is 13.69–14.26 times faster in inference than pure software deployment and enables real-time detection of ECG signals. The proposed method provides a good way to utilize unlabeled ECG, and the designed deployment scheme has great potential for medical applications.

Sleep apnea detection from single-lead electrocardiogram signals using effective deep-shallow fusion network

Objective. Explore a network architecture that can efficiently perform single-lead electrocardiogram (ECG) sleep apnea (SA) detection by utilizing the beneficial information of extended ECG segments and reducing the impact of their noisy information. Approach. We propose an effective deep-shallow fusion network (EDSFnet). The deeper residual network is used to extract high-level features with stronger semantics and less noise from the original ECG segments. The shallower convolutional neural network is used to extract lower-level features with higher resolution containing more detailed neighborhood information from the extended ECG segments. These two types of features are then fused using Effective Channel Attention, implementing automatic weight assignment to take advantage of their complementary nature. Main results. The performance of EDSFnet is evaluated on the Apnea-ECG dataset and the FAH-ECG dataset. In the Apnea-ECG dataset with 35 subjects as the training set and 35 subjects as the test set, the accuracy of EDSFnet was 92.6% and 100% for per-segment and per-recording test, respectively. In the FAH-ECG dataset with 348 subjects as the training set and 88 subjects as the test set, the accuracy of EDSFnet was 89.0% and 93.2% for per-segment and per-recording test, respectively. EDSFnet has achieved state-of-the-art results in both experiments using the publicly available Apnea-ECG dataset and subject-independent experiments using the FAH-ECG clinical dataset. Significance. The success of EDSFnet in handling SA detection underlines its robustness and adaptability. By achieving superior results across different datasets, EDSFnet offers promise in advancing the cost-effective and efficient detection of SA through single-lead ECG, reducing the burden on patients and healthcare systems alike.

Semi-Supervised Domain Adaptation for Individual Identification from Electrocardiogram Signals

When acquiring electrocardiogram (ECG) signals, the placement of electrode patches is crucial for acquiring electrocardiographic signals. Constant displacement positions are essential for ensuring the consistency of the ECG signal when used for individual identification. However, achieving constant placement of ECG electrode patches in every trial for data acquisition is challenging. This is because different individuals may attach patches, and even when the same person attaches them, it may be difficult to specify the exact position. Therefore, gathering ECG data from various locations is necessary. However, this process requires a substantial amount of labor and time, owing to the requirement for multiple attempts. Nonetheless, persisting with these efforts enables the endurance of some ECG differences. To reduce labor and time, we propose a semi-supervised domain adaptation for individual identification using ECG signals. The method operates with a full set of original ECG signals and a small set of ECG signals from different placements to account for the differences between the signals in the generative adversarial network (CycleGAN). Specifically, to train the CycleGAN, the ECG signals were transformed into time–frequency representations, and the trained generator was used to generate ECG signals to expand the small set of ECG signals from different placements. Subsequently, both the original and generated signals were used to train the classifier for individual identification. This scenario can also be applied to the classification of ECG signals from different sensors. The PTB-ECG dataset was used for this experiment. We found that the proposed method showed higher accuracy than when only the original ECG signals were used for the training classifier.

Detection of driver drowsiness level using a hybrid learning model based on ECG signals.

Fatigue has a considerable impact on the driver's vehicle and even the driver's own operating ability. An intelligent algorithm is proposed for the problem that it is difficult to classify the degree of drowsiness generated by the driver during the driving process. By studying the driver's electrocardiogram (ECG) during driving, two models were established to jointly classify the ECG signals as awake, stress, and fatigue or drowsiness states for drowsiness levels. Firstly, the deep learning method was used to establish the model_1 to predict the drowsiness of the original ECG, and model_2 was developed using the combination of principal component analysis (PCA) and weighted K-nearest neighbor (WKNN) algorithm to classify the heart rate variability characteristics. Then, the drowsiness prediction results of the two models were weighted according to certain rules, and the hybrid learning model combining dilated convolution and bidirectional long short-term memory network with PCA and WKNN algorithm was established, and the mixed model was denoted as DiCNN-BiLSTM and PCA-WKNN (DBPW). Finally, the validity of the DBPW model was verified by simulation of the public database. The experimental results show that the average accuracy, sensitivity and F1 score of the test model in the dataset containing multiple drivers are 98.79, 98.81, and 98.79 % respectively, and the recognition accuracy for drowsiness or drowsiness state is 99.33 %. Using the proposed algorithm, it is possible to identify driver anomalies and provide new ideas for the development of intelligent vehicles.

Inter-patient ECG heartbeat classification for arrhythmia classification: a new approach of multi-layer perceptron with weight capsule and sequence-to-sequence combination.

Objective: The objective of this research is to construct a method to alleviate the problem of sample imbalance in classification, especially for arrhythmia classification. This approach can improve the performance of the model without using data enhancement. Methods: In this study, we have developed a new Multi-layer Perceptron (MLP) block and have used a Weight Capsule (WCapsule) network with MLP combined with sequence-to-sequence (Seq2Seq) network to classify arrhythmias. Our work is based on the MIT-BIH arrhythmia database, the original electrocardiogram (ECG) data is classified according to the criteria recommended by the American Association for Medical Instrumentation (AAMI). Also, our method's performance is further evaluated. Results: The proposed model is evaluated using the inter-patient paradigm. Our proposed method shows an accuracy (ACC) of 99.88% under sample imbalance. For Class N, sensitivity (SEN) is 99.79%, positive predictive value (PPV) is 99.90%, and specificity (SPEC) is 99.19%. For Class S, SEN is 97.66%, PPV is 96.14%, and SPEC is 99.85%. For Class V, SEN is 99.97%, PPV is 99.07%, and SPEC is 99.94%. For Class F, SEN is 97.94%, PPV is 98.70%, and SPEC is 99.99%. When using only half of the training sample, our method shows that the SEN of Class N and V is 0.97% and 5.27% higher than the traditional machine learning algorithm. Conclusion: The proposed method combines MLP, weight capsule network with Seq2seq network, effectively addresses the problem of sample imbalance in arrhythmia classification, and produces good performance. Our method also shows promising potential in less samples.

An error-bounded median filter for correcting ECG baseline wander.

The baseline wander (BLW) in electrocardiogram (ECG) is a common disturbance that has a significant influence on the ECG wave pattern recognition. Many methods, such as IIR filter, mean filter, etc., can be used to correct BLW; However, most of them work on the original ECG signals. Compressed ECG data are economic for data storage and transmission, and if the baseline correction can be processed on them, it will be more efficient than we decompress them first and then do such correction. In this paper, we propose a new type of median filter CM_Filter, which works on the synopses of straight lines achieved from ECG by piecewise linear approximation (PLA) under maximum error bound. In CM_Filter, a heuristic strategy "Quick-Finding" is deduced by a property of straight lines in order to get the quality-assured median values from the synopses. The extended experimental tests demonstrate that the proposed filter is very efficient in execution time, and effective for correcting both slow and abrupt ECG baseline wander.

Washable and Flexible Screen-Printed Ag/AgCl Electrode on Textiles for ECG Monitoring.

Electrocardiogram (ECG) electrodes are important sensors for detecting heart disease whose performance determines the validity and accuracy of the collected original ECG signals. Due to the large drawbacks (e.g., allergy, shelf life) of traditional commercial gel electrodes, textile electrodes receive widespread attention for their excellent comfortability and breathability. This work demonstrated a dry electrode for ECG monitoring fabricated by screen printing silver/silver chloride (Ag/AgCl) conductive ink on ordinary polyester fabric. The results show that the screen-printed textile electrodes have good and stable electrical and electrochemical properties and excellent ECG signal acquisition performance. Furthermore, the resistance of the screen-printed textile electrode is maintained within 0.5 Ω/cm after 5000 bending cycles or 20 washing and drying cycles, exhibiting excellent flexibility and durability. This research provides favorable support for the design and preparation of flexible and wearable electrophysiological sensing platforms.

Extraction and Digitization of ECG Signals from Standard Clinical Portable Document Format Files for the Principal Component Analysis of T-wave Morphology.

T-wave analysis from standard electrocardiogram (ECG) remains one of the most available clinical and research methods for evaluating myocardial repolarization. T-wave morphology was recently evaluated to aid with diagnosis and characterization of diastolic dysfunction. Unfortunately, PDF stored ECG datasets limit additional numerical post-processing of ECG waveforms. In this study, we apply a simple custom process pipeline to extract and re-digitize T-wave signals and subject them to principal component analysis (PCA) to define primary T-wave shape variations. We propose simple pre-processing and digitization algorithms programmable as a MATLAB tool using standard thresholding functions without the need for advanced signal analysis. To validate digitized datasets, we compared clinically standard measurements in 20 different ECGs with the original ECG machine interpreted values as a gold standard. Afterwards, we analyzed 212 individual ECGs for T-wave shape analysis using PCA. The re-digitized signal was shown to preserve the original information as evidenced by excellent agreement between original - machine interpreted and re-digitized clinical variables including heart rate: bias ~ 1bpm (95% CI: -1.0 to 3.5), QT interval: bias ~ 0.000 ms (95% CI: -0.012 to 0.012), PR interval: bias = -0.015 ms (95% CI: -0.015 to 0.003), and QRS duration: bias = -0.001 ms (95% CI: -0.007 to 0.006). PCA revealed that the first principal component universally modulates the T-wave height or amount of repolarization voltage regardless of the investigated ECG lead. The second and third principal components described variation in the T-wave peak onset and the T-wave peak morphology, respectively. This study presents a straightforward method for re-digitizing ECGs stored in the PDF format utilized in many academic electronic medical record systems. This process can yield re-digitized lead specific signals which can be retrospectively analyzed using advanced custom post-processing numerical analysis independent of commercially available platforms.

Security enhancement of the access control scheme in IoMT applications based on fuzzy logic processing and lightweight encryption

Security of Internet-of-Medical-Things (IoMT) networks has evolved as a vital issue in recent years. The IoMT networks are designed to link patients with caregivers. All reports, data, and medical signals are transferred over these networks. Hence, these networks require robust and secure access strategies for patients to send their vital data or reports. Hence, hacking of these networks may lead to harmful effects on patients. One of the vulnerable points to hacking in these networks is the access point. Access to these networks could be performed with biometrics. The popular biometric traits for this purpose are biomedical signals such as Electrocardiogram (ECG) signals, as they are continuously monitored and measured for patients. A common thread between all biometric authentication systems is the possibility of losing the biometric traits forever if hacking attempts manage to concur the biometric template storage. A new trend in the field of biometric authentication is to avoid the utilization of original biometrics in access control processes. A possible alternative is to use cancelable biometrics instead. Cancelable biometrics can be generated through encryption schemes or non-invertible transforms. This paper adopts both strategies in a unified framework for cancelable ECG signal recognition that can be used in the access step of IoMT networks. The proposed framework begins with applying a non-invertible transformation on the ECG signals through fuzzy logic to change the dynamic range of the signals. As this process is non-invertible in nature, it prevents the recovery of the original ECG signals from the processed versions, which is the main target of cancelable biometric systems. After that, lightweight encryption through XOR operation with user-specific patterns is implemented. Here, the high complexity of full encryption schemes that need a large processing burden is eliminated. The addition of the encryption stage enhances the security of cancelable biometric traits, allowing a hybrid nature of the proposed cancelable biometric framework through the merging of non-invertible transforms and encryption algorithms. Moreover, an FPGA hardware implementation is introduced for real implementation of the proposed ECG-based cancelable biometric recognition framework. This hardware can accompany the user to allow access of the IoMT network when requested. Experimental results show a promising performance of the proposed framework with a large Area under the Receiver Operating Characteristic curve (AROC) of 99.5% and an Equal Error Rate (EER) of 0.058%.

DCDA-Net: Dual-convolutional dual-attention network for obstructive sleep apnea diagnosis from single-lead electrocardiograms

Obstructive sleep apnea (OSA) is a breathing-related chronic disease in which the soft palate and tongue collapse and block the upper airway for at least 10 s during sleep. It can lead to many heart diseases such as hypertension, myocardial infarction, and coronary heart syndrome if not detected early. Artificial intelligence has facilitated the diagnosis of many diseases in healthcare. Polysomnography is a widely used but unpleasant, time-consuming, technically demanding, and financially expensive procedure to detect OSA. Some previous methods have detected OSA using time-domain information from an electrocardiogram (ECG), whereas others have used frequency-domain information. The limitations of these two approaches can be handled using the data’s time–frequency representation. Nevertheless, there is room for enhancing the detection accuracy of OSA using the time–frequency representation approach. Therefore, we propose a novel technique that takes the ECG signal and detects R-peaks from the QRS complexes. Afterward, we interpolate those R-peaks by linear interpolation and get an interpolated-R signal. Then we magnify the interpolated-R signal corresponding to the apnea and normal frequency ranges. After magnification in the time domain, we transformed the magnified version into a scalogram. We also transformed the original one-minute ECG signal into a spectrogram after denoising. Overall, we used ECG signals to generate scalograms and spectrograms for 2 dimensional convolutional neural network (2D CNN) to classify obstructive sleep apnea. For apnea classification, we proposed a dual convolutional dual attention network (DCDA-Net) that includes a dual convolutionally modified inception module, a spatial attention module, and a channel attention module. Finally, we apply a support vector machine to the probability scores obtained from DCDA-Net based on the scalogram and spectrogram. Extensive experimental results using the open PhysioNet apnea ECG dataset confirm the effectiveness of our method in terms of accuracy and F1 score of 98% and 97.5%, respectively, which outperforms state-of-the-art methods.

Multimodal multi-instance learning for long-term ECG classification

Recently, deep learning-based models have been widely used for electrocardiogram (ECG) classification tasks. Most ECG signals are long-term time series that contain a large number of sample points. However, existing deep learning-based models resize or crop the original long-term ECG signal due to the limitation of input size and hardware, which results in information loss. To address this issue, a multimodal multi-instance learning neural network (MAMIL) is proposed for long-term ECG classification. The proposed MAMIL has three major components. First, the original ECG signal and Gramian Angular Field (GAF) image converted from the ECG signal are utilized as multimodal inputs, which enables the model to learn complementary information between different modalities. Second, multi-instance learning (MIL) is introduced to avoid information loss. Each long-term ECG signal and GAF image are treated as bags, and each heartbeat from a long-term ECG signal and each patch from a GAF image are treated as instances. Convolutional neural networks (CNNs) are utilized to extract instance features from different modalities. Third, a novel attention mechanism-based feature fusion method is proposed to aggregate the instance features from multiple modalities to obtain the bag feature for final classification. Our feature fusion method adopts pooling to obtain positive instances, which can effectively eliminate redundant information and achieve low computational complexity. The proposed MAMIL is evaluated on both intrapatient and interpatient patterns of two commonly used ECG datasets. Experimental results show that our model not only outperforms common deep learning-based models, but also outperforms previous MIL-based models.

An ECG Stitching Scheme for Driver Arrhythmia Classification Based on Deep Learning

This study proposes an electrocardiogram (ECG) signal stitching scheme to detect arrhythmias in drivers during driving. When the ECG is measured through the steering wheel during driving, the data are always exposed to noise caused by vehicle vibrations, bumpy road conditions, and the driver's steering wheel gripping force. The proposed scheme extracts stable ECG signals and transforms them into full 10 s ECG signals to classify arrhythmias using convolutional neural networks (CNN). Before the ECG stitching algorithm is applied, data preprocessing is performed. To extract the cycle from the collected ECG data, the R peaks are found and the TP interval segmentation is applied. An abnormal P peak is very difficult to find. Therefore, this study also introduces a P peak estimation method. Finally, 4 × 2.5 s ECG segments are collected. To classify arrhythmias with stitched ECG data, each time series' ECG signal is transformed via the continuous wavelet transform (CWT) and short-time Fourier transform (STFT), and transfer learning is performed for classification using CNNs. Finally, the parameters of the networks that provide the best performance are investigated. According to the classification accuracy, GoogleNet with the CWT image set shows the best results. The classification accuracy is 82.39% for the stitched ECG data, while it is 88.99% for the original ECG data.

Artificial intelligence-enabled electrocardiographic screening for left ventricular systolic dysfunction and mortality risk prediction.

Left ventricular systolic dysfunction (LVSD) characterized by a reduced left ventricular ejection fraction (LVEF) is associated with adverse patient outcomes. We aimed to build a deep neural network (DNN)-based model using standard 12-lead electrocardiogram (ECG) to screen for LVSD and stratify patient prognosis. This retrospective chart review study was conducted using data from consecutive adults who underwent ECG examinations at Chang Gung Memorial Hospital in Taiwan between October 2007 and December 2019. DNN models were developed to recognize LVSD, defined as LVEF <40%, using original ECG signals or transformed images from 190,359 patients with paired ECG and echocardiogram within 14 days. The 190,359 patients were divided into a training set of 133,225 and a validation set of 57,134. The accuracy of recognizing LVSD and subsequent mortality predictions were tested using ECGs from 190,316 patients with paired data. Of these 190,316 patients, we further selected 49,564 patients with multiple echocardiographic data to predict LVSD incidence. We additionally used data from 1,194,982 patients who underwent ECG only to assess mortality prognostication. External validation was performed using data of 91,425 patients from Tri-Service General Hospital, Taiwan. The mean age of patients in the testing dataset was 63.7 ± 16.3 years (46.3% women), and 8,216 patients (4.3%) had LVSD. The median follow-up period was 3.9 years (interquartile range 1.5-7.9 years). The area under the receiver-operating characteristic curve (AUROC), sensitivity, and specificity of the signal-based DNN (DNN-signal) to identify LVSD were 0.95, 0.91, and 0.86, respectively. DNN signal-predicted LVSD was associated with age- and sex-adjusted hazard ratios (HRs) of 2.57 (95% confidence interval [CI], 2.53-2.62) for all-cause mortality and 6.09 (5.83-6.37) for cardiovascular mortality. In patients with multiple echocardiograms, a positive DNN prediction in patients with preserved LVEF was associated with an adjusted HR (95% CI) of 8.33 (7.71 to 9.00) for incident LVSD. Signal- and image-based DNNs performed equally well in the primary and additional datasets. Using DNNs, ECG becomes a low-cost, clinically feasible tool to screen LVSD and facilitate accurate prognostication.