RawECGNet: Deep Learning Generalization for Atrial Fibrillation Detection From the Raw ECG.

INTRODUCTION: Deep learning models for detecting episodes of atrial fibrillation (AF) using rhythm information in long-term ambulatory ECG recordings have shown high performance. However, the rhythm-based approach does not take advantage of the morphological information conveyed by the different ECG waveforms, particularly the f-waves. As a result, the performance of such models may be inherently limited. METHODS: To address this limitation, we have developed a deep learning model, named RawECGNet, to detect episodes of AF and atrial flutter (AFl) using the raw, single-lead ECG. We compare the generalization performance of RawECGNet on two external data sets that account for distribution shifts in geography, ethnicity, and lead position. RawECGNet is further benchmarked against a state-of-the-art deep learning model, named ArNet2, which utilizes rhythm information as input. RESULTS: Using RawECGNet, the results for the different leads in the external test sets in terms of the F1 score were 0.91-0.94 in RBDB and 0.93 in SHDB, compared to 0.89-0.91 in RBDB and 0.91 in SHDB for ArNet2. The results highlight RawECGNet as a high-performance, generalizable algorithm for detection of AF and AFl episodes, exploiting information on both rhythm and morphology.

Characterization of Atrial Fibrillation Episode Patterns: A Comparative Study.

OBJECTIVE: The episode patterns of paroxysmal atrial fibrillation (AF) may carry important information on disease progression and complication risk. However, existing studies offer very little insight into to what extent a quantitative characterization of AF patterns can be trusted given the errors in AF detection and various types of shutdown, i.e., poor signal quality and non-wear. This study explores the performance of AF pattern characterizing parameters in the presence of such errors. METHODS: To evaluate the performance of the parameters AF aggregation and AF density, both previously proposed to characterize AF patterns, the two measures mean normalized difference and the intraclass correlation coefficient are used to describe agreement and reliability, respectively. The parameters are studied on two PhysioNet databases with annotated AF episodes, also accounting for shutdowns due to poor signal quality. RESULTS: The agreement is similar for both parameters when computed for detector-based and annotated patterns, which is 0.80 for AF aggregation and 0.85 for AF density. On the other hand, the reliability differs substantially, with 0.96 for AF aggregation but only 0.29 for AF density. This finding suggests that AF aggregation is considerably less sensitive to detection errors. The results from comparing three strategies to handle shutdowns vary considerably, with the strategy that disregards the shutdown from the annotated pattern showing the best agreement and reliability. CONCLUSIONS: Due to its better robustness to detection errors, AF aggregation should be preferred. To further improve performance, future research should put more emphasis on AF pattern characterization.

Can 24 h of ambulatory ECG be used to triage patients to extended monitoring?

BACKGROUND: Access to long-term ambulatory recording to detect atrial fibrillation (AF) is limited for economical and practical reasons. We aimed to determine whether 24 h ECG (24hECG) data can predict AF detection on extended cardiac monitoring. METHODS: We included all US patients from 2020, aged 17-100 years, who were monitored for 2-30 days using the PocketECG device (MEDICALgorithmics), without AF ≥30 s on the first day (n = 18,220, mean age 64.4 years, 42.4% male). The population was randomly split into equal training and testing datasets. A Lasso model was used to predict AF episodes ≥30 s occurring on days 2-30. RESULTS: The final model included maximum heart rate, number of premature atrial complexes (PACs), fastest rate during PAC couplets and triplets, fastest rate during premature ventricular couplets and number of ventricular tachycardia runs ≥4 beats, and had good discrimination (ROC statistic 0.7497, 95% CI 0.7336-0.7659) in the testing dataset. Inclusion of age and sex did not improve discrimination. A model based only on age and sex had substantially poorer discrimination, ROC statistic 0.6542 (95% CI 0.6364-0.6720). The prevalence of observed AF in the testing dataset increased by quintile of predicted risk: 0.4% in Q1, 2.7% in Q2, 6.2% in Q3, 11.4% in Q4, and 15.9% in Q5. In Q1, the negative predictive value for AF was 99.6%. CONCLUSION: By using 24hECG data, long-term monitoring for AF can safely be avoided in 20% of an unselected patient population whereas an overall risk of 9% in the remaining 80% of the population warrants repeated or extended monitoring.

ECG Modeling for Simulation of Arrhythmias in Time-Varying Conditions.

The present article proposes an ECG simulator that advances modeling of arrhythmias and noise by introducing time-varying signal characteristics. The simulator is built around a discrete-time Markov chain model for simulating atrial and ventricular arrhythmias of particular relevance when analyzing atrial fibrillation (AF). Each state is associated with statistical information on episode duration and heartbeat characteristics. Statistical, time-varying modeling of muscle noise, motion artifacts, and the influence of respiration is introduced to increase the complexity of simulated ECGs, making the simulator well suited for data augmentation in machine learning. Modeling of how the PQ and QT intervals depend on heart rate is also introduced. The realism of simulated ECGs is assessed by three experienced doctors, showing that simulated ECGs are difficult to distinguish from real ECGs. Simulator usefulness is illustrated in terms of AF detection performance when either simulated or real ECGs are used to train a neural network for signal quality control. The results show that both types of training lead to similar performance.

Atrial fibrillation episode patterns as predictor of clinical outcome of catheter ablation.

Methods for characterization of atrial fibrillation (AF) episode patterns have been introduced without establishing clinical significance. This study investigates, for the first time, whether post-ablation recurrence of AF can be predicted by evaluating episode patterns. The dataset comprises of 54 patients (age 56 ± 11 years; 67% men), with an implantable cardiac monitor, before undergoing the first AF catheter ablation. Two parameters of the alternating bivariate Hawkes model were used to characterize the pattern: AF dominance during the monitoring period (log(mu)) and temporal aggregation of episodes (beta1). Moreover, AF burden and AF density, a parameter characterizing aggregation of AF burden, were studied. The four parameters were computed from an average of 29 AF episodes before ablation. The risk of AF recurrence after catheter ablation using the Hawkes parameters log(mu) and beta1, AF burden, and AF density was evaluated. While the combination of AF burden and AF density is related to a non-significant hazard ratio, the combination of log(mu) and beta1 is related to a hazard ratio of 1.95 (1.03-3.70; p < 0.05). The Hawkes parameters showed increased risk of AF recurrence within 1 year after the procedure for patients with high AF dominance and high episode aggregation and may be used for pre-ablation risk assessment.

Spectral Analysis of Heart Rate Variability in Time-Varying Conditions and in the Presence of Confounding Factors.

The tools for spectrally analyzing heart rate variability (HRV) has in recent years grown considerably, with emphasis on the handling of time-varying conditions and confounding factors. Time-frequency analysis holds since long an important position in HRV analysis, however, this technique cannot alone handle a mean heart rate or a respiratory frequency which vary over time. Overlapping frequency bands represents another critical condition which needs to be dealt with to produce accurate spectral measurements. The present survey offers a comprehensive account of techniques designed to handle such conditions and factors by providing a brief description of the main principles of the different methods. Several methods derive from a mathematical/statistical model, suggesting that the model can be used to simulate data used for performance evaluation. The inclusion of a respiratory signal, whether measured or derived, is another feature of many recent methods, e.g., used to guide the decomposition of the HRV signal so that signals related as well as unrelated to respiration can be analyzed. It is concluded that the development of new approaches to handling time-varying scenarios are warranted, as is benchmarking of performance evaluated in technical as well as in physiological/clinical terms.

A comparative study of the performance of methods for f-wave extraction.

Objective.This study proposes a novel technique for atrial fibrillatory waves (f-waves) extraction and investigates the performance of the proposed method comparing with different f-wave extraction methods.Approach.We propose a novel technique combining a periodic component analysis (PiCA) and echo state network (ESN) for f-waves extraction, denoted PiCA-ESN. PiCA-ESN benefits from the advantages of using both source separation and nonlinear adaptive filtering. PiCA-ESN is evaluated by comparing with other state-of-the-art approaches, which include template subtraction technique based on principal component analysis, spatiotemporal cancellation, nonlinear adaptive filtering using an echo state neural network, and a source separation technique based on PiCA. Quality assessment is performed on a recently published reference database including a large number of simulated ECG signals in atrial fibrillation (AF). The performance of the f-wave extraction methods is evaluated in terms of signal quality metrics (SNR, ΔSNR) and robustness of f-wave features.Main results.The proposed method offers the best signal quality performance, with a ΔSNR of approximately 22 dB across all 8 sets of the reference database, as well as the most robust extraction of f-wave features, with 75% of all estimates of dominant atrial frequency well below 1 Hz.

Training Convolutional Neural Networks on Simulated Photoplethysmography Data: Application to Bradycardia and Tachycardia Detection.

Objective: To develop a method for detection of bradycardia and ventricular tachycardia using the photoplethysmogram (PPG). Approach: The detector is based on a dual-branch convolutional neural network (CNN), whose input is the scalograms of the continuous wavelet transform computed in 5-s segments. Training and validation of the CNN is accomplished using simulated PPG signals generated from RR interval series extracted from public ECG databases. Manually annotated real PPG signals from the PhysioNet/CinC 2015 Challenge Database are used for performance evaluation. The performance is compared to that of a pulse-based reference detector. Results: The sensitivity/specificity were found to be 98.1%/97.9 and 76.6%/96.8% for the CNN-based detector, respectively, whereas the corresponding results for the pulse-based detector were 94.7%/99.8 and 67.1%/93.8%, respectively. Significance: The proposed detector may be useful for continuous, long-term monitoring of bradycardia and tachycardia using wearable devices, e.g., wrist-worn devices, especially in situations where sensitivity is favored over specificity. The study demonstrates that simulated PPG signals are suitable for training and validation of a CNN.

QT Interval Adaptation to Heart Rate Changes in Atrial Fibrillation as a Predictor of Sudden Cardiac Death.

OBJECTIVE: The clinical significance of QT interval adaptation to heart rate changes has been poorly investigated in atrial fibrillation (AF), since QT delineation in the presence of f-waves is challenging. The objective of the present study is to investigate new techniques for QT adaptation estimation in permanent AF. METHODS: A multilead strategy based on periodic component analysis, to emphasize T-wave periodicity, is proposed for QT delineation. QT adaptation is modeled by a linear, time-invariant filter, which describes the dependence between the current QT interval and the preceding RR intervals, followed by a memoryless, nonlinear, function. The QT adaptation time lag is determined from the estimated impulse response. RESULTS: Using simulated ECGs in permanent AF, the transformed lead was found to offer more accurate QT delineation and time lag estimation than did the original ECG leads for a wide range of f-wave amplitudes. In a population with chronic heart failure and permanent AF, the time lag estimated from the transformed lead was found to have the strongest, statistically significant association with sudden cardiac death (SCD) (hazard ratio = 3.49). CONCLUSIONS: Periodic component analysis provides more accurate QT delineation and improves time lag estimation in AF. A prolonged QT adaptation time lag is associated with a high risk for SCD. SIGNIFICANCE: SCD risk markers originally developed for sinus rhythm can also be used in AF, provided that T-wave periodicity is emphasized. The time lag is a potentially useful biomarker for identifying patients at risk for SCD, guiding clinicians in adopting effective therapeutic decisions.

Signal Quality Assessment of a Novel ECG Electrode for Motion Artifact Reduction.

BACKGROUND: The presence of noise is problematic in the analysis and interpretation of the ECG, especially in ambulatory monitoring. Restricting the analysis to high-quality signal segments only comes with the risk of excluding significant arrhythmia episodes. Therefore, the development of novel electrode technology, robust to noise, continues to be warranted. METHODS: The signal quality of a novel wet ECG electrode (Piotrode) is assessed and compared to a commercially available, commonly used electrode (Ambu). The assessment involves indices of QRS detection and atrial fibrillation detection performance, as well as signal quality indices (ensemble standard deviation and time-frequency repeatability), computed from ECGs recorded simultaneously from 20 healthy subjects performing everyday activities. RESULTS: The QRS detection performance using the Piotrode was considerably better than when using the Ambu, especially for running but also for lighter activities. The two signal quality indices demonstrated similar trends: the gap in quality became increasingly larger as the subjects became increasingly more active. CONCLUSIONS: The novel wet ECG electrode produces signals with less motion artifacts, thereby offering the potential to reduce the review burden, and accordingly the cost, associated with ambulatory monitoring.

Identification of Transient Noise to Reduce False Detections in Screening for Atrial Fibrillation.

Screening for atrial fibrillation (AF) with a handheld device for recording the ECG is becoming increasingly popular. The poorer signal quality of such ECGs may lead to false detection of AF, often caused by transient noise. Consequently, the need for expert review in AF screening can become extensive. A convolutional neural network (CNN) is proposed for transient noise identification in AF detection. The network is trained using the events produced by a QRS detector, classified into either true beat detections or false detections. The CNN and a low-complexity AF detector are trained and tested using the StrokeStop I database, containing 30-s ECGs from mass screening for AF in the elderly population. Performance evaluation of the CNN-based quality control using a subset of the database resulted in sensitivity, specificity, and accuracy of 96.4, 96.9, and 96.9%, respectively. By inserting the CNN before the AF detector, the false AF detections were reduced by 22.5% without any loss in sensitivity. The results show that the number of recordings calling for expert review can be significantly reduced thanks to the identification of transient noise. The reduction of false AF detections is directly linked to the time and cost spent on expert review.

Considerations on Performance Evaluation of Atrial Fibrillation Detectors.

OBJECTIVE: A large number of atrial fibrillation (AF) detectors have been published in recent years, signifying that the comparison of detector performance plays a central role, though not always consistent. The aim of this study is to shed needed light on aspects crucial to the evaluation of detection performance. METHODS: Three types of AF detector, using either information on rhythm, rhythm and morphology, or segments of ECG samples, are implemented and studied on both real and simulated ECG signals. The properties of different performance measures are investigated, for example, in relation to dataset imbalance. RESULTS: The results show that performance can differ considerably depending on the way detector output is compared to database annotations, i.e., beat-to-beat, segment-to-segment, or episode-to-episode comparison. Moreover, depending on the type of detector, the results substantiate that physiological and technical factors, e.g., changes in ECG morphology, rate of atrial premature beats, and noise level, can have a considerable influence on performance. CONCLUSION: The present study demonstrates overall strengths and weaknesses of different types of detector, highlights challenges in AF detection, and proposes five recommendations on how to handle data and characterize performance.

Modeling and Estimation of Temporal Episode Patterns in Paroxysmal Atrial Fibrillation.

OBJECTIVE: The present study proposes a model-based, statistical approach to characterizing episode patterns in paroxysmal atrial fibrillation (AF). Thanks to the rapid advancement of noninvasive monitoring technology, the proposed approach should become increasingly relevant in clinical practice. METHODS: History-dependent point process modeling is employed to characterize AF episode patterns, using a novel alternating, bivariate Hawkes self-exciting model. In addition, a modified version of a recently proposed statistical model to simulate AF progression throughout a lifetime is considered, involving non-Markovian rhythm switching and survival functions. For each model, the maximum likelihood estimator is derived and used to find the model parameters from observed data. RESULTS: Using three databases with a total of 59 long-term ECG recordings, the goodness-of-fit analysis demonstrates that the proposed alternating, bivariate Hawkes model fits SR-to-AF transitions in 40 recordings and AF-to-SR transitions in 51; the corresponding numbers for the AF model with non-Markovian rhythm switching are 40 and 11, respectively. Moreover, the results indicate that the model parameters related to AF episode clustering, i.e., aggregation of temporal AF episodes, provide information complementary to the well-known clinical parameter AF burden. CONCLUSION: Point process modeling provides a detailed characterization of the occurrence pattern of AF episodes that may improve the understanding of arrhythmia progression.

Restoration of the electrocardiogram during mechanical cardiopulmonary resuscitation.

OBJECTIVE: An artefact-free electrocardiogram (ECG) is essential during cardiac arrest to decide therapy such as defibrillation. Mechanical cardiopulmonary resuscitation (CPR) devices cause movement artefacts that alter the ECG. This study analyzes the effectiveness of mechanical CPR artefact suppression filters to restore clinically relevant ECG information. APPROACH: In total, 495 10 s ECGs were used, of which 165 were in ventricular fibrillation (VF), 165 in organized rhythms (OR) and 165 contained mechanical CPR artefacts recorded during asystole. CPR artefacts and rhythms were mixed at controlled signal-to-noise ratios (SNRs), ranging from -20 dB to 10 dB. Mechanical artefacts were removed using least mean squares (LMS), recursive least squares (RLS) and Kalman filters. Performance was evaluated by comparing the clean and the restored ECGs in terms of restored SNR, correlation-based similarity measures, and clinically relevant features: QRS detection performance for OR, and dominant frequency, mean amplitude and waveform irregularity for VF. For each filter, a shock/no-shock support vector machine algorithm based on multiresolution analysis of the restored ECG was designed, and evaluated in terms of sensitivity (Se) and specificity (Sp). MAIN RESULTS: The RLS filter produced the largest correlation coefficient (0.80), the largest average increase in SNR (9.5 dB), and the best QRS detection performance. The LMS filter best restored VF with errors of 10.3% in dominant frequency, 18.1% in amplitude and 11.8% in waveform irregularity. The Se/Sp of the diagnosis of the restored ECG were 95.1/94.5% using the RLS filter and 97.0/91.4% using the LMS filter. SIGNIFICANCE: Suitable filter configurations to restore ECG waveforms during mechanical CPR have been determined, allowing reliable clinical decisions without interrupting mechanical CPR therapy.

A Comparative Study of ECG-derived Respiration in Ambulatory Monitoring using the Single-lead ECG.

Cardiorespiratory monitoring is crucial for the diagnosis and management of multiple conditions such as stress and sleep disorders. Therefore, the development of ambulatory systems providing continuous, comfortable, and inexpensive means for monitoring represents an important research topic. Several techniques have been proposed in the literature to derive respiratory information from the ECG signal. Ten methods to compute single-lead ECG-derived respiration (EDR) were compared under multiple conditions, including different recording systems, baseline wander, normal and abnormal breathing patterns, changes in breathing rate, noise, and artifacts. Respiratory rates, wave morphology, and cardiorespiratory information were derived from the ECG and compared to those extracted from a reference respiratory signal. Three datasets were considered for analysis, involving a total 59 482 one-min, single-lead ECG segments recorded from 156 subjects. The results indicate that the methods based on QRS slopes outperform the other methods. This result is particularly interesting since simplicity is crucial for the development of ECG-based ambulatory systems.

Short-term reproducibility of parameters characterizing atrial fibrillatory waves.

OBJECTIVE: To study reproducibility of f-wave parameters in terms of inter- and intrapatient variation. APPROACH: Five parameters are investigated: dominant atrial frequency (DAF), f-wave amplitude, phase dispersion, spectral organization, and spatiotemporal variability. For each parameter, the variance ratio R, defined as the ratio between inter- and intrapatient variance, is computed; a larger R corresponds to better stability and reproducibility. The study population consists of 20 high-quality ECGs recorded from patients with atrial fibrillation (11/9 paroxysmal/persistent). MAIN RESULTS: The well-established parameters DAF and f-wave amplitude were associated with considerably larger R-values (13.1 and 21.0, respectively) than phase dispersion (2.4), spectral organization (2.4), and spatiotemporal variability (2.7). The use of an adaptive harmonic frequency tracker to estimate the DAF resulted in a larger R (13.1) than did block-based maximum likelihood estimation (6.3). SIGNIFICANCE: This study demonstrates a noticeable difference in reproducibility among f-wave parameters, a result which should be taken into account when performing f-wave analysis.

Detection of Needle Dislodgement Using Extracorporeal Pressure Signals: A Feasibility Study.

Venous needle dislodgement (VND) during dialysis is a rarely occurring adverse event, which becomes life-threatening if not handled promptly. Because the standard venous pressure alarm, implemented in most dialysis machines, has low sensitivity, a novel approach using extracted cardiac information to detect needle dislodgement is proposed. Four features are extracted from the arterial and venous pressure signals of the dialysis machine, characterizing the mean venous pressure, the venous cardiac pulse pressure, the time delay, and the correlation between the two pressure signals. The features serve as input to a support vector machine (SVM), which determines whether dislodgement has occurred. The SVM is first trained on a set of laboratory data, and then tested on another set of laboratory data as well as on a small data set from clinical hemodialysis sessions. The results show that dislodgement can be detected after 12-17 s, corresponding to 24-143 ml blood loss. The standard venous pressure alarm used in clinical routine only detects 50% of the VNDs, whereas the novel method detects all VNDs and has a false alarm rate of 0.12 per hour, provided that the amplitude of the extracted cardiac pressure signal exceeds 1 mmHg. The results are promising; however, the method needs to be tested on a larger set of clinical data to better establish its performance.

ECG-Derived Respiratory Rate in Atrial Fibrillation.

OBJECTIVE: The present study addresses the problem of estimating the respiratory rate from the morphological ECG variations in the presence of atrial fibrillatory waves (f-waves). The significance of performing f-wave suppression before respiratory rate estimation is investigated. METHODS: The performance of a novel approach to ECG-derived respiration, named "slope range" (SR) and designed particularly for operation in atrial fibrillation (AF), is compared to that of two well-known methods based on either R-wave angle (RA) or QRS loop rotation angle (LA). A novel rule is proposed for spectral peak selection in respiratory rate estimation. The suppression of f-waves is accomplished using signal- and noise-dependent QRS weighted averaging. The performance evaluation embraces real as well as simulated ECG signals acquired from patients with persistent AF; the estimation error of the respiratory rate is determined for both types of signals. RESULTS: Using real ECG signals and reference respiratory signals, rate estimation without f-wave suppression resulted in a median error of 0.015 ± 0.021 Hz and 0.019 ± 0.025 Hz for SR and RA, respectively, whereas LA with f-wave suppression resulted in 0.034 ± 0.039 Hz. Using simulated signals, the results also demonstrate that f-wave suppression is superfluous for SR and RA, whereas it is essential for LA. CONCLUSION: The results show that SR offers the best performance as well as computational simplicity since f-wave suppression is not needed. SIGNIFICANCE: The respiratory rate can be robustly estimated from the ECG in the presence of AF.

Reference database and performance evaluation of methods for extraction of atrial fibrillatory waves in the ECG.

OBJECTIVE: This study proposes a reference database, composed of a large number of simulated ECG signals in atrial fibrillation (AF), for investigating the performance of methods for extraction of atrial fibrillatory waves (f-waves). APPROACH: The simulated signals are produced using a recently published and validated model of 12-lead ECGs in AF. The database is composed of eight signal sets together accounting for a wide range of characteristics known to represent major challenges in f-wave extraction, including high heart rates, high morphological QRST variability, and the presence of ventricular premature beats. Each set contains 30 5 min signals with different f-wave amplitudes. The database is used for the purpose of investigating the statistical association between different indices, designed for use with either real or simulated signals. MAIN RESULTS: Using the database, available at the PhysioNet repository of physiological signals, the performance indices unnormalized ventricular residue (uVR), designed for real signals, and the root mean square error, designed for simulated signals, were found to exhibit the strongest association, leading to the recommendation that uVR should be used when characterizing performance in real signals. SIGNIFICANCE: The proposed database facilitates comparison of the performance of different f-wave extraction methods and makes it possible to express performance in terms of the error between simulated and extracted f-wave signals.

A New Method for Continuous Relative Blood Volume and Plasma Sodium Concentration Estimation During Hemodialysis.

OBJECTIVE: Non-invasive sensing and reliable estimation of physiological parameters are important features of hemodialysis machines, especially for therapy customization (biofeedback). In this paper, we present a new method for joint estimation of two important hemodialysis-related physiological parameters-relative blood volume and plasma sodium concentration. METHODS: Our method makes use of a non-invasive sensor setup and a mathematical estimator. The estimator, based on the Kalman filter, allows merging data from multiple sensors, newly designed as well as onboard, with modeling knowledge about the hemodialysis process. The system was validated on in vitro hemodialysis sessions using bovine blood. RESULTS: The estimation error we obtained (0.97 ± 0.73% on relative blood volume and 0.47 ± 0.19 mM on plasmatic sodium) proved to be comparable with that of the reference data for both parameters-the system is sufficiently accurate to be relevant in a clinical context. CONCLUSION: Our system has the potential to provide accurate and important information on the state of a patient undergoing hemodialysis, while only low-cost modifications to the existing onboard sensors are required. SIGNIFICANCE: Through improved knowledge of blood parameters during hemodialysis, our method will allow better patient monitoring and therapy customization in hemodialysis.