Electrocardiogram Database Research Articles

Express method of biometric person authentication based on one cycle of the ECG signal

The article is devoted to an express method of biometric authentication of a person based on an electrocardiogram (ECG). The method is characterized by high accuracy (efficiency) of authentication of a person based on only one cycle of its ECG. Such characteristics as Accuracy, Balanced Accuracy and F1-score on average are not lower than 96.1% for such binary classifiers as k-Nearest Neighbors, Linear SVM, Decision Tree, Random Forest, Multilayer Perceptron, Adaptive Boosting, Naive Bayes and Statistical Interval Classifier. The research utilized the Combined Measurement of ECG, Breathing, and Seismocardiograms database, which features data from 20 healthy people. A method of constructing confidence intervals for ECG cycles has been developed, which is based on the rhythm-adaptive statistical estimation of the mathematical expectation and the standard deviation of the ECG signal. The method of constructing confidence intervals is based on the functioning of the Statistical Interval Classifier in the system of biometric authentication of a person. The Statistical Interval Classifier has the lowest time computational complexity among the 8 studied classifiers, which justifies its use in portable biometric authentication systems that have negligible computing resources.

Federated Learning with Deep Neural Networks: A Privacy-Preserving Approach to Enhanced ECG Classification.

In response to increasing data privacy regulations, this work examines the use of federated learning for deep residual networks to diagnose cardiac abnormalities from electrocardiogram (ECG) data. This approach allows medical institutions to collaborate without exchanging raw patient data. We utilize the publicly available data from the PhysioNet/Computing in Cardiology Challenge 2021, featuring diverse ECG databases, to compare the classification performance of three federated learning methods against both central training with data sharing and isolated training scenarios. We show that federated learning outperforms ECG classifiers trained in isolation. In particular, our findings demonstrate that a globally trained model fine-tuned to specific local datasets surpasses non-collaborative approaches. This shows that models trained in federation learn general features that can be tailored to specific tasks. Furthermore, federated learning almost matches the performance of central training with data sharing on out-of-distribution data from non-participating institutions. These results highlight the ability of federated learning in developing models that generalize well across diverse patient data, without the need to share data among institutions, thus addressing data privacy concerns.

ECG Biometric Authentication Using Self-Supervised Learning for IoT Edge Sensors.

Wearable Internet of Things (IoT) devices are gaining ground for continuous physiological data acquisition and health monitoring. These physiological signals can be used for security applications to achieve continuous authentication and user convenience due to passive data acquisition. This paper investigates an electrocardiogram (ECG) based biometric user authentication system using features derived from the Convolutional Neural Network (CNN) and self-supervised contrastive learning. Contrastive learning enables us to use large unlabeled datasets to train the model and establish its generalizability. We propose approaches enabling the CNN encoder to extract appropriate features that distinguish the user from other subjects. When evaluated using the PTB ECG database with 290 subjects, the proposed technique achieved an authentication accuracy of 99.15%. To test its generalizability, we applied the model to two new datasets, the MIT-BIH Arrhythmia Database and the ECG-ID Database, achieving over 98.5% accuracy without any modifications. Furthermore, we show that repeating the authentication step three times can increase accuracy to nearly 100% for both PTBDB and ECGIDDB. This paper also presents model optimizations for embedded device deployment, which makes the system more relevant to real-world scenarios. To deploy our model in IoT edge sensors, we optimized the model complexity by applying quantization and pruning. The optimized model achieves 98.67% accuracy on PTBDB, with 0.48% accuracy loss and 62.6% CPU cycles compared to the unoptimized model. An accuracy-vs-time-complexity tradeoff analysis is performed, and results are presented for different optimization levels.

Multi-Label Diagnosis of Arrhythmias Based on a Modified Two-Category Cross-Entropy Loss Function

The 12-lead resting electrocardiogram (ECG) is commonly used in hospitals to assess heart health. The ECG can reflect a variety of cardiac abnormalities, requiring multi-label classification. However, the diagnosis results in previous studies have been imprecise. For example, in some previous studies, some cardiac abnormalities that cannot coexist often appeared in the diagnostic results. In this work, we explore how to realize the effective multi-label diagnosis of ECG signals and prevent the prediction of cardiac arrhythmias that cannot coexist. In this work, a multi-label classification method based on a convolutional neural network (CNN), long short-term memory (LSTM), and an attention mechanism is presented for the multi-label diagnosis of cardiac arrhythmia using resting ECGs. In addition, this work proposes a modified two-category cross-entropy loss function by introducing a regularization term to avoid the existence of arrhythmias that cannot coexist. The effectiveness of the modified cross-entropy loss function is validated using a 12-lead resting ECG database collected by our team. Using traditional and modified cross-entropy loss functions, three deep learning methods are employed to classify six types of ECG signals. Experimental results show the modified cross-entropy loss function greatly reduces the number of non-coexisting label pairs while maintaining prediction accuracy. Deep learning methods are effective in the multi-label diagnosis of ECG signals, and diagnostic efficiency can be improved by using the modified cross-entropy loss function. In addition, the modified cross-entropy loss function helps prevent diagnostic models from outputting two arrhythmias that cannot coexist, further reducing the false positive rate of non-coexisting arrhythmic diseases, thereby demonstrating the potential value of the modified loss function in clinical applications.

Inter-patient ECG arrhythmia heartbeat classification network based on multiscale convolution and FCBA

Cardiac arrhythmias that can lead to sudden cardiac death are common. Electrocardiograms (ECGs) offer valuable information about cardiac status and play a crucial role in evaluating patients with arrhythmia in clinical diagnosis. The development of machine learning technologies has made the computer-aided diagnosis of cardiac arrhythmias possible, which can improve the efficiency and quality of medical treatment. In this study, a new method for the automatic classification of heartbeats is developed. This paper presents a novel approach to detect arrhythmias in the inter-patient paradigm using a convolutional neural network (CNN) that integrates multiscale convolutional blocks, frequency convolutional block attention (FCBA) modules, and RR interval features. The proposed method also includes oversampling for the heartbeats of the minority class and the addition of random noise to increase training samples and alleviate the problem of data imbalance. We used the MIT-BIH-AR arrhythmia database , which is a globally recognized ECG database, to evaluate the classification performance of the proposed model in this paper. Based on the experiments, the sensitivities of the normal beats (N), supraventricular ectopic beats (SVEB), and ventricular ectopic beats (VEB) were 96.9%, 89.3%, and 93.3%, respectively. In the inter-patient heartbeat classification paradigm, the overall accuracy of the proposed method is 95.60% in classifying heartbeats. The study shows that the proposed scheme outperforms other published schemes in terms of classification results.

A Deep Multi-Task Learning Approach for Bioelectrical Signal Analysis

Deep learning is a promising technique for bioelectrical signal analysis, as it can automatically discover hidden features from raw data without substantial domain knowledge. However, training a deep neural network requires a vast amount of labeled samples. Additionally, a well-trained model may be sensitive to the study object, and its performance may deteriorate sharply when transferred to other study objects. We propose a deep multi-task learning approach for bioelectrical signal analysis to address these issues. Explicitly, we define two distinct scenarios, the consistent source-target scenario and the inconsistent source-target scenario based on the motivation and purpose of the tasks. For each scenario, we present methods to decompose the original task and dataset into multiple subtasks and sub-datasets. Correspondingly, we design the generic deep parameter-sharing neural networks to solve the multi-task learning problem and illustrate the details of implementation with one-dimension convolutional neural networks (1D CNN), vanilla recurrent neural networks (RNN), recurrent neural networks with long short-term memory units (LSTM), and recurrent neural networks with gated recurrent units (GRU). In these two scenarios, we conducted extensive experiments on four electrocardiogram (ECG) databases. The results demonstrate the benefits of our approach, showing that our proposed method can improve the accuracy of ECG data analysis (up to 5.2%) in the MIT-BIH arrhythmia database.

Optimization of CNN using modified Honey Badger Algorithm for Sleep Apnea detection

Sleep Apnea (SA) is the most prevalent breathing sleep problem, and if left untreated, it can lead to catastrophic neurological and cardiovascular illnesses. Conventionally, polysomnography (PSG) is used to diagnose SA. Nonetheless, this approach necessitates several electrodes, cables, and a professional to oversee the experiment. A promising alternative is using a single-channel signal for SA diagnosis, with the electrocardiogram (ECG) signal being among the most relevant and easily recordable. Recently, a convolutional neural network (CNN) has been used to extract efficient features from training data instead of manually selecting characteristics from ECG. However, selecting the best hyperparameter values for CNN can be challenging due to the vast number of possibilities. To address this, we propose a modified Honey Badger Algorithm (MHBA) combined with three improvement initiatives: quasi-opposition learning, arbitrary weighting agent, and adaptive mutation method. Our approach is evaluated on the Physionet Apnea ECG database, consisting of 70 single-lead ECG recordings annotated by qualified medical professionals. The experiments show that the MHBA outperforms traditional CNN and machine learning methods with an accuracy of 91.3%, AUC of 97.5%, specificity of 93.6%, and sensitivity of 90.1%. Our results demonstrate the effectiveness of the MHBA for SA detection.

Automated Arrhythmia Classification Using Farmland Fertility Algorithm with Hybrid Deep Learning Model on Internet of Things Environment.

In recent years, the rapid progress of Internet of Things (IoT) solutions has offered an immense opportunity for the collection and dissemination of health records in a central data platform. Electrocardiogram (ECG), a fast, easy, and non-invasive method, is generally employed in the evaluation of heart conditions that lead to heart ailments and the identification of heart diseases. The deployment of IoT devices for arrhythmia classification offers many benefits such as remote patient care, continuous monitoring, and early recognition of abnormal heart rhythms. However, it is challenging to diagnose and manually classify arrhythmia as the manual diagnosis of ECG signals is a time-consuming process. Therefore, the current article presents the automated arrhythmia classification using the Farmland Fertility Algorithm with Hybrid Deep Learning (AAC-FFAHDL) approach in the IoT platform. The proposed AAC-FFAHDL system exploits the hyperparameter-tuned DL model for ECG signal analysis, thereby diagnosing arrhythmia. In order to accomplish this, the AAC-FFAHDL technique initially performs data pre-processing to scale the input signals into a uniform format. Further, the AAC-FFAHDL technique uses the HDL approach for detection and classification of arrhythmia. In order to improve the classification and detection performance of the HDL approach, the AAC-FFAHDL technique involves an FFA-based hyperparameter tuning process. The proposed AAC-FFAHDL approach was validated through simulation using the benchmark ECG database. The comparative experimental analysis outcomes confirmed that the AAC-FFAHDL system achieves promising performance compared with other models under different evaluation measures.

Quantum conditional generative adversarial network based on patch method for abnormal electrocardiogram generation

To address the scarcity and class imbalance of abnormal electrocardiogram (ECG) databases, which are crucial in AI-driven diagnostic tools for potential cardiovascular disease detection, this study proposes a novel quantum conditional generative adversarial algorithm (QCGAN-ECG) for generating abnormal ECG signals. The QCGAN-ECG constructs a quantum generator based on patch method. In this method, each sub-generator generates distinct features of abnormal heartbeats in different segments. This patch-based generative algorithm conserves quantum resources and makes QCGAN-ECG practical for near-term quantum devices. Additionally, QCGAN-ECG introduces quantum registers as control conditions. It encodes information about the types and probability distributions of abnormal heartbeats into quantum registers, rendering the entire generative process controllable. Simulation experiments on Pennylane demonstrated that the QCGAN-ECG could generate completely abnormal heartbeats with an average accuracy of 88.8%. Moreover, the QCGAN-ECG can accurately fit the probability distribution of various abnormal ECG data. In the anti-noise experiments, the QCGAN-ECG showcased outstanding robustness across various levels of quantum noise interference. These results demonstrate the effectiveness and potential applicability of the QCGAN-ECG for generating abnormal ECG signals, which will further promote the development of AI-driven cardiac disease diagnosis systems. The source code is available at github.com/VanSWK/QCGAN_ECG.

Arrhythmia classification detection based on multiple electrocardiograms databases.

According to the World Health Organization, cardiovascular diseases are the leading cause of deaths globally. Electrocardiogram (ECG) is a non-invasive approach for detecting heart diseases and reducing the risk of heart disease-related death. However, there are limited numbers of ECG samples and imbalance distribution for existing ECG databases. It is difficult to train practical and efficient neural networks. Based on the analysis and research of many existing ECG databases, this paper conduct an in-depth study on three fine-labeled ECG databases, to extract heartbeats, unify the sampling frequency, and propose a self-processing method of heartbeats, and finally form a unified ECG arrhythmia classification database, noted as Hercules-3. It is separated into training sets (80%) and testing sets (the remaining 20%). In order to verify its capabilities, we have trained a 16-classification fully connected neural network based on Hercules-3 and it achieves an accuracy rate of up to 98.67%. Compared with other data processing, our proposed method improves classification recall by at least 6%, classification accuracy by at least 4%, and F1-score by at least 7%.

ElectroCardioGuard: Preventing patient misidentification in electrocardiogram databases through neural networks

Electrocardiograms (ECGs) are commonly used by cardiologists to detect heart-related pathological conditions. Reliable collections of ECGs are crucial for precise diagnosis. However, in clinical practice, the assignment of captured ECG recordings to incorrect patients can occur inadvertently. In collaboration with a clinical and research facility which recognized this challenge and reached out to us, we present a study that addresses this issue. In this work, we propose a small and efficient neural-network based model for determining whether two ECGs originate from the same patient. Our model demonstrates great generalization capabilities and achieves state-of-the-art performance in gallery-probe patient identification on PTB-XL while utilizing 760x fewer parameters. Furthermore, we present a technique leveraging our model for detection of recording-assignment mistakes, showcasing its applicability in a realistic scenario. Finally, we evaluate our model on a newly collected ECG dataset specifically curated for this study, and make it public for the research community.

Increased ventricular ectopy precedes Torsades de Pointes in patients with prolonged QT.

Torsades de Pointes (TdP) is a potentially lethal ventricular tachydysrhythmia. Prolonged heartrate corrected QT interval (QTc) predicts TdP; however, with poor specificity. We performed this study to identify other predictors of TdP among patients with prolonged QTc. We performed a retrospective case control study with 2:1 matching at an urban academic hospital. We searched our hospital electrocardiogram (ECG) database for tracings with heartrate ≤ 60, QTc≥500, and QRS<120, followed by a natural language search for electronic records with "Torsades," "polymorphic VT," or similar to identify TdP cases from 2005 to 19. We identified controls from a similar ECG database search matching for QTc, heartrate, age, and sex. We compared cardiologic and historical factors, medications, laboratory values, and ECG measurements including ectopy using univariate statistics. For those cases with saved telemetry strips that included preceding beats or TdP onset, we compared ectopy and TdP onset characteristics between the ECG and telemetry strips using mixed linear modeling. Seventy-five cases including 50 with telemetry strips and 150 controls were included. Historical, pharmacologic, laboratory, and cardiologic testing results were similar between cases and controls. The proportion of telemetry tracings with premature ventricular contractions (PVC's) preceding TdP was 0.78 compared to 0.16 for case ECG's (difference 0.62(95%CI 0.44-0.75)) and 0.10 for control ECGs (difference 0.68(95%CI 0.56-0.80)). Average telemetry heartrate was 72 and QTc 549 immediately preceding TdP, similar to the ECG values. Clinical factors don't differentiate patients with long QTc who develop TdP, however, an increase in PVC's in patients with prolonged QTc may usefully predict imminent TdP.

1DIEN: Cross-session Electrocardiogram Authentication Using 1D Integrated EfficientNet

The potential of using electrocardiogram (ECG), an important physiological signal for humans, as a new biometric trait has been demonstrated, and ongoing efforts have focused on utilizing deep learning (e.g., 2D neural networks) to improve authentication accuracy (with some efficiency tradeoffs). In most of the existing ECG-based authentication approaches, the ECG recordings for enrollment and testing are collected within short intervals (e.g., within an hour). However, since ECG biometrics change over time, this design may decrease authentication accuracy when ECG recordings are collected weeks or even months prior. In this article, we propose 1D Integrated EfficientNet (1DIEN) to achieve cross-session ECG authentication. We adopt 1D neural networks as a lightweight alternative to 2D neural networks, and a voting scheme is designed to reduce variance and improve general authentication performance. We use three public ECG databases (i.e., an inter-session database, a mixed-session database, and an intra-session database) to evaluate our proposed 1DIEN under different authentication scenarios. The experimental results show that our approach achieves satisfactory performance for ECG authentication at a 3-month interval and is suitable for practical applications.

MODELING AND SYNTHESIS OF THE WISTAR RAT ELECTROCARDIOGRAM

In this work, we present a synthesizer of electrocardiographic signals of the experimental model of the Wistar rat. Firstly, we modeled the P, R, and S waves using Gaussian functions, while for the T-wave, we used a Gumbel function. We validated the proposed model using a Wistar rat electrocardiogram database, which contains 41 animals with 62,907 beats. Model global performance was CC (correlation [Formula: see text], and RE (relative [Formula: see text]. We also propose a transformation that allows the aforementioned model to be built from the standard ECG parameters, such as RR interval, P wave duration, PR interval, RS interval, QT interval, and P, R, S, and T wave amplitudes. In all cases, the correlation coefficients obtained were greater than 0.966, while the mean of the absolute errors of the parameters did not exceed 6[Formula: see text]ms and 30[Formula: see text][Formula: see text]V. Also, we have developed software that integrates the proposed model and the transformation. We used a system of coupled differential equations to synthesize several beats based on one cardiac beat pattern with the possibility of entering variability in the ECG parameters and adding noise to give realism to the electrocardiographic signals. In conclusion, we developed a synthesizer of Wistar rats ECG useful for several applications, such as delineation and segmentation, training and testing of automatic algorithms to detect and monitor, as well as novel filtering methods of these types of ECG signals.

FGSQA-Net: A Weakly Supervised Approach to Fine-Grained Electrocardiogram Signal Quality Assessment.

Due to the lack of fine-grained labels, current research can only evaluate the signal quality at a coarse scale. This article proposes a weakly supervised fine-grained electrocardiogram (ECG) signal quality assessment method, which can produce continuous segment-level quality scores with only coarse labels. A novel network architecture, i.e. FGSQA-Net, is developed for signal quality assessment, which consists of a feature shrinking module and a feature aggregation module. Multiple feature shrinking blocks, which combine residual CNN block and max pooling layer, are stacked to produce a feature map corresponding to continuous segments along the spatial dimension. Segment-level quality scores are obtained by feature aggregation along the channel dimension. The proposed method was evaluated on two real-world ECG databases and one synthetic dataset. Our method produced an average AUC value of 0.975, which outperforms the state-of-the-art beat-by-beat quality assessment method. The results are visualized for 12-lead and single-lead signals over a granularity from 0.64 to 1.7 seconds, demonstrating that high-quality and low-quality segments can be effectively distinguished at a fine scale. FGSQA-Net is flexible and effective for fine-grained quality assessment for various ECG recordings and is suitable for ECG monitoring using wearable devices. This is the first study on fine-grained ECG quality assessment using weak labels and can be generalized to similar tasks for other physiological signals.

Residual neural networks based on empirical mode decomposition for mitral regurgitation prediction

A multi-modal residual neural network based on empirical mode decomposition (EMD) was proposed in this work and used for screening patients with mitral regurgitation (MR). EMD was used to decompose the synchronous electrocardiogram (ECG) and phonocardiogram (PCG) signals, and then the component with the highest degree of correlation with the original signals was selected for reconstruction, and then the reconstructed signals were converted into images by gramian angular difference field (GADF). Finally, the residual neural network was used to get the prediction results. In the present work, we established a database called Synchronized ECG and PCG Database for Patients with Mitral Regurgitation (SEP-MRDb) consisting of 1046 synchronous ECG and PCG recordings from patients with MR (n = 56) and without MR (n = 990). The experimental results showed that the accuracy rate of the proposed model was 96.90%, the precision was 97.10%, and recall was 97.10%, which can effectively capture the combined information of ECG and PCG signals and realize the recognition of MR.

Domain and Patient Adversarial Multi-Task Learning for Arrhythmia Classification.

Manual screening of electrocardiograms (ECGs) for heart arrhythmias by clinicians is time-consuming and labor-intensive. A machine learning model for the automated diagnosis of heart arrhythmia from ECG signals can facilitate improved diagnosis, greater accessibility and earlier intervention for patients. The potential of such models is limited however by the small size of clinical datasets available for training. Methods that can be trained with multiple datasets to classify heart arrhythmia are needed to overcome this problem.In this paper, we propose using adversarial multi-task learning (AMTL) to extract domain and patient invariant features from two electrocardiogram databases. We further investigated the influence of beat segmentation location and beat normalization on domain invariance. Our proposed methods were tested on the MIT-BIH Arrhythmia and the St Petersburg INCART 12-lead Arrhythmia Databases. The domain adversarial models achieved a higher accuracy and average F1 score than their counterparts without domain adversarial learning. In particular, the patient and domain adversarial model improved the F1 scores on the two tested databases from 70% and 74% to 77% each.Clinical Relevance-This establishes that adversarial multitask learning with multiple datasets and multiple adversarial tasks can improve the F1 score of arrhythmia classification.

Optimized Solutions of Electrocardiogram Lead and Segment Selection for Cardiovascular Disease Diagnostics.

Most of the existing multi-lead electrocardiogram (ECG) detection methods are based on all 12 leads, which undoubtedly results in a large amount of calculation and is not suitable for the application in portable ECG detection systems. Moreover, the influence of different lead and heartbeat segment lengths on the detection is not clear. In this paper, a novel Genetic Algorithm-based ECG Leads and Segment Length Optimization (GA-LSLO) framework is proposed, aiming to automatically select the appropriate leads and input ECG length to achieve optimized cardiovascular disease detection. GA-LSLO extracts the features of each lead under different heartbeat segment lengths through the convolutional neural network and uses the genetic algorithm to automatically select the optimal combination of ECG leads and segment length. In addition, the lead attention module (LAM) is proposed to weight the features of the selected leads, which improves the accuracy of cardiac disease detection. The algorithm is validated on the ECG data from the Huangpu Branch of Shanghai Ninth People's Hospital (defined as the SH database) and the open-source Physikalisch-Technische Bundesanstalt diagnostic ECG database (PTB database). The accuracy for detection of arrhythmia and myocardial infarction under the inter-patient paradigm is 99.65% (95% confidence interval: 99.20-99.76%) and 97.62% (95% confidence interval: 96.80-98.16%), respectively. In addition, ECG detection devices are designed using Raspberry Pi, which verifies the convenience of hardware implementation of the algorithm. In conclusion, the proposed method achieves good cardiovascular disease detection performance. It selects the ECG leads and heartbeat segment length with the lowest algorithm complexity while ensuring classification accuracy, which is suitable for portable ECG detection devices.

Machine learning models for automated interpretation of 12-lead electrocardiographic signals: a narrative review of techniques, challenges, achievements and clinical relevance

Background and Objective: Novel advances in machine learning (ML) and its subfield, deep learning (DL), as well as the recent release of large-scale electrocardiogram (ECG) databases, have driven a sharp increase in research related to automated ECG interpretation. This review aims to summarize the recent ML approaches for automatically interpreting standard 12-lead ECG signals.

PO-05-229 DEEP LEARNING CLASSIFICATION OF PAROXYSMAL SUPRAVENTRICULAR TACHYCARDIA TYPES USING SINUS RHYTHM ELECTROCARDIOGRAMS

Only sinus rhythm electrocardiograms (ECGs) are often available in patients with paroxysmal supraventricular tachycardia (PSVT). However, sinus rhythm ECGs have a limited value in classifying PSVT types. We aimed to investigate the deep learning approach of classifying Atrioventricular Nodal Reentry Tachycardia (AVNRT) from concealed Atrioventricular Reentry Tachycardia (AVRT) using sinus rhythm ECGs. We collected the patients with AVNRT or conceald AVRT, and their 12-lead ECGs with sinus rhythm. The diagnosis of each patient was validated with an electrophysiology study. ResNet-34 was used for the deep learning model, and it was trained to classify sinus rhythm ECGs with underlying AVNRT and conceald AVRT. Due to the limited training dataset, transfer learning was performed by pre-training of an open-source ECG database (PhysioNet/CinC Challenge 2021). Ten-fold cross-validation was used to validate deep learning performance. Integrated gradient analysis was used to visualize which ECG segment is important for the model to classify the two arrhythmias. A total of 696 patients with AVNRT and 305 patients with conceald AVRT were analyzed. Compared to the AVNRT group, the conceald AVRT group was significantly younger and had a higher male proportion; age 48 vs. 51 years (p<0.001); male 63.3 vs. 34.3% (p<0.001). After the pre-training, the deep learning performance was significantly improved; AUROC 0.73 (95% CI 0.69-0.76) vs. 0.67 (0.62-0.72) for with and without pre-training; p<0.001 (Figure A). Deep learning showed modest performance for classifying the two arrhythmias; AUPRC, F1-score, sensitivity, and specificity were 0.84 (0.81-0.87), 0.75 (0.70-0.80), 0.68 (0.61-0.75), and 0.71 (0.65-0.77), respectively. Deep learning focused on PR intervals and T waves for identifying AVNRT, and P-waves for concealed AVRT (Figure B). Deep learning classification of AVNRT and conceald AVRT was feasible with sinus rhythm ECGs.