Low-complexity detection of atrial fibrillation in continuous long-term monitoring

Associations of Physical Activity and Heart Rate Variability from a Two-Week ECG Monitor with Cognitive Function and Dementia: The ARIC Neurocognitive Study

Low physical activity (PA) measured by accelerometers and low heart rate variability (HRV) measured from short-term ECG recordings are associated with worse cognitive function. Wearable long-term ECG monitors are now widely used, and some devices also include an accelerometer. The objective of this study was to evaluate whether PA or HRV measured from long-term ECG monitors was associated with cognitive function among older adults. A total of 1590 ARIC participants had free-living PA and HRV measured over 14 days using the Zio® XT Patch [aged 72-94 years, 58% female, 32% Black]. Cognitive function was measured by cognitive factor scores and adjudicated dementia or mild cognitive impairment (MCI) status. Adjusted linear or multinomial regression models examined whether higher PA or higher HRV was cross-sectionally associated with higher factor scores or lower odds of MCI/dementia. Each 1-unit increase in the total amount of PA was associated with higher global cognition (β = 0.30, 95% CI: 0.16-0.44) and executive function scores (β = 0.38, 95% CI: 0.22-0.53) and lower odds of MCI (OR = 0.38, 95% CI: 0.22-0.67) or dementia (OR = 0.25, 95% CI: 0.08-0.74). HRV (i.e., SDNN and rMSSD) was not associated with cognitive function. More research is needed to define the role of wearable ECG monitors as a tool for digital phenotyping of dementia.

SEResUTer: a deep learning approach for accurate ECG signal delineation and atrial fibrillation detection

Objective.Accurate detection of electrocardiogram (ECG) waveforms is crucial for computer-aided diagnosis of cardiac abnormalities. This study introduces SEResUTer, an enhanced deep learning model designed for ECG delineation and atrial fibrillation (AF) detection.Approach. Built upon a U-Net architecture, SEResUTer incorporates ResNet modules and Transformer encoders to replace convolution blocks, resulting in improved optimization and encoding capabilities. A novel masking strategy is proposed to handle incomplete expert annotations. The model is trained on the QT database (QTDB) and evaluated on the Lobachevsky University Electrocardiography Database (LUDB) to assess its generalization performance. Additionally, the model's scope is extended to AF detection using the the China Physiological Signal Challenge 2021 (CPSC2021) and the China Physiological Signal Challenge 2018 (CPSC2018) datasets.Main results.The proposed model surpasses existing traditional and deep learning approaches in ECG waveform delineation on the QTDB. It achieves remarkable average F1 scores of 99.14%, 98.48%, and 98.46% for P wave, QRS wave, and T wave delineation, respectively. Moreover, the model demonstrates exceptional generalization ability on the LUDB, achieving average SE, positive prediction rate, and F1 scores of 99.05%, 94.59%, and 94.62%, respectively. By analyzing RR interval differences and the existence of P waves, our method achieves AF identification with 99.20% accuracy on the CPSC2021 test set and demonstrates strong generalization on CPSC2018 dataset.Significance.The proposed approach enables highly accurate ECG waveform delineation and AF detection, facilitating automated analysis of large-scale ECG recordings and improving the diagnosis of cardiac abnormalities.

Six-lead electrocardiography compared to single-lead electrocardiography and photoplethysmography of a wrist-worn device for atrial fibrillation detection controlled by premature atrial or ventricular contractions: six is smarter than one

Background

Smartwatches are commonly capable to record a lead-I-like electrocardiogram (ECG) and perform a photoplethysmography (PPG)-based atrial fibrillation (AF) detection. Wearable technologies repeatedly face the challenge of frequent premature beats, particularly in target populations for screening of AF.

Objective

To investigate the potential diagnostic benefit of six-lead ECG compared to single-lead ECG and PPG-based algorithm for AF detection of the wrist-worn device.

Methods and results

From the database of DoubleCheck-AF 249 adults were enrolled in AF group (n = 121) or control group of SR with frequent premature ventricular (PVCs) or atrial (PACs) contractions (n = 128). Cardiac rhythm was monitored using a wrist-worn device capable of recording continuous PPG and simultaneous intermittent six-lead standard-limb-like ECG. To display a single-lead ECG, the six-lead ECGs were trimmed to lead-I-like ECGs. Two diagnosis-blinded cardiologists evaluated reference, six-lead and single-lead ECGs as "AF", "SR", or "Cannot be concluded". AF detection based on six-lead ECG, single-lead ECG, and PPG yielded a sensitivity of 99.2%, 95.7%, and 94.2%, respectively. The higher number of premature beats per minute was associated with false positive outcomes of single-lead ECG (18.80 vs. 5.40 beats/min, P < 0.01), six-lead ECG (64.3 vs. 5.8 beats/min, P = 0.018), and PPG-based detector (13.20 vs. 5.60 beats/min, P = 0.05). Single-lead ECG required 3.4 times fewer extrasystoles than six-lead ECG to result in a false positive outcome. In a control subgroup of PACs, the specificity of six-lead ECG, single-lead ECG, and PPG dropped to 95%, 83.8%, and 90%, respectively. The diagnostic value of single-lead ECG (AUC 0.898) was inferior to six-lead ECG (AUC 0.971) and PPG-based detector (AUC 0.921). In a control subgroup of PVCs, the specificity of six-lead ECG, single-lead ECG, and PPG was 100%, 96.4%, and 96.6%, respectively. The diagnostic value of single-lead ECG (AUC 0.961) was inferior to six-lead ECG (AUC 0.996) and non-inferior to PPG-based detector (AUC 0.954).

Conclusions

A six-lead wearable-recorded ECG demonstrated the superior diagnostic value of AF detection compared to a single-lead ECG and PPG-based AF detection. The risk of type I error due to the widespread use of smartwatch-enabled single-lead ECGs in populations with frequent premature beats is significant.

Generalizable and robust deep learning algorithm for atrial fibrillation diagnosis across geography, ages and sexes

To drive health innovation that meets the needs of all and democratize healthcare, there is a need to assess the generalization performance of deep learning (DL) algorithms across various distribution shifts to ensure that these algorithms are robust. This retrospective study is, to the best of our knowledge, an original attempt to develop and assess the generalization performance of a DL model for AF events detection from long term beat-to-beat intervals across geography, ages and sexes. The new recurrent DL model, denoted ArNet2, is developed on a large retrospective dataset of 2,147 patients totaling 51,386 h obtained from continuous electrocardiogram (ECG). The model's generalization is evaluated on manually annotated test sets from four centers (USA, Israel, Japan and China) totaling 402 patients. The model is further validated on a retrospective dataset of 1,825 consecutives Holter recordings from Israel. The model outperforms benchmark state-of-the-art models and generalized well across geography, ages and sexes. For the task of event detection ArNet2 performance was higher for female than male, higher for young adults (less than 61 years old) than other age groups and across geography. Finally, ArNet2 shows better performance for the test sets from the USA and China. The main finding explaining these variations is an impairment in performance in groups with a higher prevalence of atrial flutter (AFL). Our findings on the relative performance of ArNet2 across groups may have clinical implications on the choice of the preferred AF examination method to use relative to the group of interest.

Influence of Photoplethysmogram Signal Quality on Pulse Arrival Time during Polysomnography

Intervals of low-quality photoplethysmogram (PPG) signals might lead to significant inaccuracies in estimation of pulse arrival time (PAT) during polysomnography (PSG) studies. While PSG is considered to be a "gold standard" test for diagnosing obstructive sleep apnea (OSA), it also enables tracking apnea-related nocturnal blood pressure fluctuations correlated with PAT. Since the electrocardiogram (ECG) is recorded synchronously with the PPG during PSG, it makes sense to use the ECG signal for PPG signal-quality assessment. (1) Objective: to develop a PPG signal-quality assessment algorithm for robust PAT estimation, and investigate the influence of signal quality on PAT during various sleep stages and events such as OSA. (2) Approach: the proposed algorithm uses R and T waves from the ECG to determine approximate locations of PPG pulse onsets. The MESA database of 2055 PSG recordings was used for this study. (3) Results: the proportions of high-quality PPG were significantly lower in apnea-related oxygen desaturation (matched-pairs rc = 0.88 and rc = 0.97, compared to OSA and hypopnea, respectively, when p < 0.001) and arousal (rc = 0.93 and rc = 0.98, when p < 0.001) than in apnea events. The significantly large effect size of interquartile ranges of PAT distributions was between low- and high-quality PPG (p < 0.001, rc = 0.98), and regular and irregular pulse waves (p < 0.001, rc = 0.74), whereas a lower quality of the PPG signal was found to be associated with a higher interquartile range of PAT across all subjects. Suggested PPG signal quality-based PAT evaluation reduced deviations (e.g., rc = 0.97, rc = 0.97, rc = 0.99 in hypopnea, oxygen desaturation, and arousal stages, respectively, when p < 0.001) and allowed obtaining statistically larger differences between different sleep stages and events. (4) Significance: the implemented algorithm has the potential to increase the robustness of PAT estimation in PSG studies related to nocturnal blood pressure monitoring.

Multiscale Encoding of Electrocardiogram Signals with a Residual Network for the Detection of Atrial Fibrillation

Atrial fibrillation (AF) is one of the most common cardiac arrhythmias, and it is an indication of high-risk factors for stroke, myocardial ischemia, and other malignant cardiovascular diseases. Most of the existing AF detection methods typically convert one-dimensional time-series electrocardiogram (ECG) signals into two-dimensional representations to train a deep and complex AF detection system, which results in heavy training computation and high implementation costs. In this paper, a multiscale signal encoding scheme is proposed to improve feature representation and detection performance without the need for using any transformation or handcrafted feature engineering techniques. The proposed scheme uses different kernel sizes to produce the encoded signal by using multiple streams that are passed into a one-dimensional sequence of blocks of a residual convolutional neural network (ResNet) to extract representative features from the input ECG signal. This also allows networks to grow in breadth rather than in depth, thus reducing the computing time by using the parallel processing capability of deep learning networks. We investigated the effects of the use of a different number of streams with different kernel sizes on the performance. Experiments were carried out for a performance evaluation using the publicly available PhysioNet CinC Challenge 2017 dataset. The proposed multiscale encoding scheme outperformed existing deep learning-based methods with an average F₁ score of 98.54%, but with a lower network complexity.

Atrioventricular Synchronization for Detection of Atrial Fibrillation and Flutter in One to Twelve ECG Leads Using a Dense Neural Network Classifier

This study investigates the use of atrioventricular (AV) synchronization as an important diagnostic criterion for atrial fibrillation and flutter (AF) using one to twelve ECG leads. Heart rate, lead-specific AV conduction time, and P-/f-wave amplitude were evaluated by three representative ECG metrics (mean value, standard deviation), namely RR-interval (RRi-mean, RRi-std), PQ-interval (PQi-mean, PQI-std), and PQ-amplitude (PQa-mean, PQa-std), in 71,545 standard 12-lead ECG records from the six largest PhysioNet CinC Challenge 2021 databases. Two rhythm classes were considered (AF, non-AF), randomly assigning records into training (70%), validation (20%), and test (10%) datasets. In a grid search of 19, 55, and 83 dense neural network (DenseNet) architectures and five independent training runs, we optimized models for one-lead, six-lead (chest or limb), and twelve-lead input features. Lead-set performance and SHapley Additive exPlanations (SHAP) input feature importance were evaluated on the test set. Optimal DenseNet architectures with the number of neurons in sequential [1st, 2nd, 3rd] hidden layers were assessed for sensitivity and specificity: DenseNet [16,16,0] with primary leads (I or II) had 87.9-88.3 and 90.5-91.5%; DenseNet [32,32,32] with six limb leads had 90.7 and 94.2%; DenseNet [32,32,4] with six chest leads had 92.1 and 93.2%; and DenseNet [128,8,8] with all 12 leads had 91.8 and 95.8%, indicating sensitivity and specificity values, respectively. Mean SHAP values on the entire test set highlighted the importance of RRi-mean (100%), RR-std (84%), and atrial synchronization (40-60%) for the PQa-mean (aVR, I), PQi-std (V2, aVF, II), and PQi-mean (aVL, aVR). Our focus on finding the strongest AV synchronization predictors of AF in 12-lead ECGs would lead to a comprehensive understanding of the decision-making process in advanced neural network classifiers. DenseNet self-learned to rely on a few ECG behavioral characteristics: first, characteristics usually associated with AF conduction such as rapid heart rate, enhanced heart rate variability, and large PQ-interval deviation in V2 and inferior leads (aVF, II); second, characteristics related to a typical P-wave pattern in sinus rhythm, which is best distinguished from AF by the earliest negative P-peak deflection of the right atrium in the lead (aVR) and late positive left atrial deflection in lateral leads (I, aVL). Our results on lead-selection and feature-selection practices for AF detection should be considered for one- to twelve-lead ECG signal processing settings, particularly those measuring heart rate, AV conduction times, and P-/f-wave amplitudes. Performances are limited to the AF diagnostic potential of these three metrics. SHAP value importance can be used in combination with a human expert's ECG interpretation to change the focus from a broad observation of 12-lead ECG morphology to focusing on the few AV synchronization findings strongly predictive of AF or non-AF arrhythmias. Our results are representative of AV synchronization findings across a broad taxonomy of cardiac arrhythmias in large 12-lead ECG databases.

Accurate detection of atrial fibrillation events with R-R intervals from ECG signals

Atrial fibrillation (AF) is a typical category of arrhythmia. Clinical diagnosis of AF is based on the detection of abnormal R-R intervals (RRIs) with an electrocardiogram (ECG). Previous studies considered this detection problem as a classification problem and focused on extracting a number of features. In this study we demonstrate that instead of using any specific numerical characteristic as the input feature, the probability density of RRIs from ECG conserves comprehensive statistical information; hence, is a natural and efficient input feature for AF detection. Incorporated with a support vector machine as the classifier, results on the MIT-BIH database indicates that the proposed method is a simple and accurate approach for AF detection in terms of accuracy, sensitivity, and specificity.

Automatic Atrial Fibrillation Arrhythmia Detection Using Univariate and Multivariate Data

Atrial fibrillation (AF) is still a major cause of disease morbidity and mortality, making its early diagnosis desirable and urging researchers to develop efficient methods devoted to automatic AF detection. Till now, the analysis of Holter-ECG recordings remains the gold-standard technique to screen AF. This is usually achieved by studying either RR interval time series analysis, P-wave detection or combinations of both morphological characteristics. After extraction and selection of meaningful features, each of the AF detection methods might be conducted through univariate and multivariate data analysis. Many of these automatic techniques have been proposed over the last years. This work presents an overview of research studies of AF detection based on RR interval time series. The aim of this paper is to provide the scientific community and newcomers to the field of AF screening with a resource that presents introductory concepts, clinical features, and a literature review that describes the techniques that are mostly followed when RR interval time series are used for accurate detection of AF.

Deepaware: A hybrid deep learning and context-aware heuristics-based model for atrial fibrillation detection

BACKGROUND

State-of-the-art automatic atrial fibrillation (AF) detection models trained on RR-interval (RRI) features generally produce high performance on standard benchmark electrocardiogram (ECG) AF datasets. These models, however, result in a significantly high false positive rates (FPRs) when applied on ECG data collected under free-living ambulatory conditions and in the presence of non-AF arrhythmias.

METHOD

This paper proposes DeepAware, a novel hybrid model combining deep learning (DL) and context-aware heuristics (CAH), which reduces the FPR effectively and improves the AF detection performance on participant-operated ambulatory ECG from free-living conditions. It exploits the RRI and P-wave features, as well as the contextual features from the ambulatory ECG.

RESULTS

DeepAware is shown to be very generalizable and superior to the state-of-the-art models when applied on unseen benchmark ECG AF datasets. Most importantly, the model is able to detect AF efficiently when applied on participant-operated ambulatory ECG recordings from free-living conditions and has achieved a sensitivity (Se), specificity (Sp), and accuracy (Acc) of 97.94%, 98.39%, 98.06%, respectively. Results also demonstrate the effect of atrial activity analysis (via P-waves detection) and CAH in reducing the FPR over the RRI features-based AF detection model.

CONCLUSIONS

The proposed DeepAware model can substantially reduce the physician's workload of manually reviewing the false positives (FPs) and facilitate long-term ambulatory monitoring for early detection of AF.

High Specificity Wearable Device With Photoplethysmography and Six-Lead Electrocardiography for Atrial Fibrillation Detection Challenged by Frequent Premature Contractions: DoubleCheck-AF

Background

Consumer smartwatches have gained attention as mobile health (mHealth) tools able to detect atrial fibrillation (AF) using photoplethysmography (PPG) or a short strip of electrocardiogram (ECG). PPG has limited accuracy due to the movement artifacts, whereas ECG cannot be used continuously, is usually displayed as a single-lead signal and is limited in asymptomatic cases.

Objective

DoubleCheck-AF is a validation study of a wrist-worn device dedicated to providing both continuous PPG-based rhythm monitoring and instant 6-lead ECG with no wires. We evaluated its ability to differentiate between AF and sinus rhythm (SR) with particular emphasis on the challenge of frequent premature beats.

Methods and Results

We performed a prospective, non-randomized study of 344 participants including 121 patients in AF. To challenge the specificity of the device two control groups were selected: 95 patients in stable SR and 128 patients in SR with frequent premature ventricular or atrial contractions (PVCs/PACs). All ECG tracings were labeled by two independent diagnosis-blinded cardiologists as “AF,” “SR” or “Cannot be concluded.” In case of disagreement, a third cardiologist was consulted. A simultaneously recorded ECG of Holter monitor served as a reference. It revealed a high burden of ectopy in the corresponding control group: 6.2 PVCs/PACs per minute, bigeminy/trigeminy episodes in 24.2% (31/128) and runs of ≥3 beats in 9.4% (12/128) of patients. AF detection with PPG-based algorithm, ECG of the wearable and combination of both yielded sensitivity and specificity of 94.2 and 96.9%; 99.2 and 99.1%; 94.2 and 99.6%, respectively. All seven false-positive PPG-based cases were from the frequent PVCs/PACs group compared to none from the stable SR group (P < 0.001). In the majority of these cases (6/7) cardiologists were able to correct the diagnosis to SR with the help of the ECG of the device (P = 0.012).

Conclusions

This is the first wearable combining PPG-based AF detection algorithm for screening of AF together with an instant 6-lead ECG with no wires for manual rhythm confirmation. The system maintained high specificity despite a remarkable amount of frequent single or multiple premature contractions.

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.

Application of Dense Neural Networks for Detection of Atrial Fibrillation and Ranking of Augmented ECG Feature Set

Considering the significant burden to patients and healthcare systems globally related to atrial fibrillation (AF) complications, the early AF diagnosis is of crucial importance. In the view of prominent perspectives for fast and accurate point-of-care arrhythmia detection, our study optimizes an artificial neural network (NN) classifier and ranks the importance of enhanced 137 diagnostic ECG features computed from time and frequency ECG signal representations of short single-lead strips available in 2017 Physionet/CinC Challenge database. Based on hyperparameters' grid search of densely connected NN layers, we derive the optimal topology with three layers and 128, 32, 4 neurons per layer (DenseNet-3@128-32-4), which presents maximal F1-scores for classification of Normal rhythms (0.883, 5076 strips), AF (0.825, 758 strips), Other rhythms (0.705, 2415 strips), Noise (0.618, 279 strips) and total F1 relevant to the CinC Challenge of 0.804, derived by five-fold cross-validation. DenseNet-3@128-32-4 performs equally well with 137 to 32 features and presents tolerable reduction by about 0.03 to 0.06 points for limited input sets, including 8 and 16 features, respectively. The feature reduction is linked to effective application of a comprehensive method for computation of the feature map importance based on the weights of the activated neurons through the total path from input to specific output in DenseNet. The detailed analysis of 20 top-ranked ECG features with greatest importance to the detection of each rhythm and overall of all rhythms reveals DenseNet decision-making process, noticeably corresponding to the cardiologists' diagnostic point of view.

Detection of Brief Episodes of Atrial Fibrillation Based on Electrocardiomatrix and Convolutional Neural Network

Background: Brief episodes of atrial fibrillation (AF) may evolve into longer AF episodes increasing the chances of thrombus formation, stroke, and death. Classical methods for AF detection investigate rhythm irregularity or P-wave absence in the ECG, while deep learning approaches profit from the availability of annotated ECG databases to learn discriminatory features linked to different diagnosis. However, some deep learning approaches do not provide analysis of the features used for classification. This paper introduces a convolutional neural network (CNN) approach for automatic detection of brief AF episodes based on electrocardiomatrix-images (ECM-images) aiming to link deep learning to features with clinical meaning. Materials and Methods: The CNN is trained using two databases: the Long-Term Atrial Fibrillation and the MIT-BIH Normal Sinus Rhythm, and tested on three databases: the MIT-BIH Atrial Fibrillation, the MIT-BIH Arrhythmia, and the Monzino-AF. Detection of AF is done using a sliding window of 10 beats plus 3 s. Performance is quantified using both standard classification metrics and the EC57 standard for arrhythmia detection. Layer-wise relevance propagation analysis was applied to link the decisions made by the CNN to clinical characteristics in the ECG. Results: For all three testing databases, episode sensitivity was greater than 80.22, 89.66, and 97.45% for AF episodes shorter than 15, 30 s, and for all episodes, respectively. Conclusions: Rhythm and morphological characteristics of the electrocardiogram can be learned by a CNN from ECM-images for the detection of brief episodes of AF.

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.

Automated Arrhythmia Detection Based on RR Intervals

Abnormal heart rhythms, also known as arrhythmias, can be life-threatening. AFIB and AFL are examples of arrhythmia that affect a growing number of patients. This paper describes a method that can support clinicians during arrhythmia diagnosis. We propose a deep learning algorithm to discriminate AFIB, AFL, and NSR RR interval signals. The algorithm was designed with data from 4051 subjects. With 10-fold cross-validation, the algorithm achieved the following results: ACC = 99.98%, SEN = 100.00%, and SPE = 99.94%. These results are significant because they show that it is possible to automate arrhythmia detection in RR interval signals. Such a detection method makes economic sense because RR interval signals are cost-effective to measure, communicate, and process. Having such a cost-effective solution might lead to widespread long-term monitoring, which can help detecting arrhythmia earlier. Detection can lead to treatment, which improves outcomes for patients.

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.

Poincaré Plot Image and Rhythm-Specific Atlas for Atrial Bigeminy and Atrial Fibrillation Detection

A detector based only on RR intervals capable of classifying other tachyarrhythmias in addition to atrial fibrillation (AF) could improve cardiac monitoring. In this paper a new classification method based in a 2D non-linear RRI dynamics representation is presented. For this aim, the concepts of Poincaré Images and Atlases are introduced. Three cardiac rhythms were targeted: Normal sinus rhythm (NSR), AF and atrial bigeminy (AB). Three Physionet open source databases were used. Poincaré Images were generated for all signals using different Poincaré plot configurations: RR, dRR and RRdRR. The study was computed for different time window lengths and bin sizes. For each rhythm, the Poincaré Images of the 80% of that rhythm's patients were used to create a reference image, a Poincaré Atlas. The remaining 20% were used as test set and classified into one of the three rhythms using normalized mutual information and 2D correlation. The process was iterated in a tenfold cross-validation and patient-wise dataset division. Sensitivity results obtained for RRdRR configuration and bin size 40 ms, for a 60 s time window were 94.35% ±3.68, 82.07% ±9.18 and 88.86% ±12.79 with a specificity of 85.52% ±7.46, 95.91% ±3.14, 96.10% ±2.25 for AF, NSR and AB respectively. Results suggest that a rhythms general RRI pattern may be captured using Poincaré Atlases and that these can be used to classify other signal segments using Poincaré Images. In contrast with other studies, the former method could be generalized to more cardiac rhythms and does not depend on rhythm-specific thresholds.

Hybrid Decision Support to Monitor Atrial Fibrillation for Stroke Prevention

In this paper, we discuss hybrid decision support to monitor atrial fibrillation for stroke prevention. Hybrid decision support takes the form of human experts and machine algorithms working cooperatively on a diagnosis. The link to stroke prevention comes from the fact that patients with Atrial Fibrillation (AF) have a fivefold increased stroke risk. Early diagnosis, which leads to adequate AF treatment, can decrease the stroke risk by 66% and thereby prevent stroke. The monitoring service is based on Heart Rate (HR) measurements. The resulting signals are communicated and stored with Internet of Things (IoT) technology. A Deep Learning (DL) algorithm automatically estimates the AF probability. Based on this technology, we can offer four distinct services to healthcare providers: (1) universal access to patient data; (2) automated AF detection and alarm; (3) physician support; and (4) feedback channels. These four services create an environment where physicians can work symbiotically with machine algorithms to establish and communicate a high quality AF diagnosis.

Remote Atrial Fibrillation Burden Estimation Using Deep Recurrent Neural Network

OBJECTIVE

The atrial fibrillation burden (AFB) is defined as the percentage of time spent in atrial fibrillation (AF) over a long enough monitoring period. Recent research has suggested the added prognostic value of using the AFB compared to a binary diagnosis. We evaluate, for the first time, the ability to estimate the AFB over long-term continuous recordings, using a deep recurrent neutral network (DRNN) approach.

METHODS

The models were developed and evaluated on a large database of p = 2,891 patients, totaling t = 68,800 hours of continuous electrocardiography (ECG) recordings from the University of Virginia. Specifically, 24h beat-to-beat time series were obtained from a single portable ECG channel. The network, denoted ArNet, was benchmarked against a gradient boosting (XGB) model, trained on 21 features including the coefficient of sample entropy (CosEn) and AFEvidence that is derived from the number of irregular points revealed by the Lorenz plot. The generalizations of ArNet and XGB were also evaluated on the independent PhysioNet LTAF test database.

RESULTS

the absolute AF burden estimation error [Formula: see text], median and interquartile, on the test set, was 1.2 (0.1-6.7) for ArNet and 2.8 (0.9-11.7) for XGB for AF individuals. Generalization results on LTAF were consistent with [Formula: see text] of 2.7 (1.1-14.7) for ArNet and 3.6 (1.0-16.7) for XGB.

CONCLUSION

This research demonstrates the feasibility of AFB estimation from 24h beat-to-beat interval time series utilizing DRNNs.

SIGNIFICANCE

The novel data-driven approach enables robust remote diagnosis and phenotyping of AF.

HAN-ECG: An interpretable atrial fibrillation detection model using hierarchical attention networks

Atrial fibrillation (AF) is one of the most prevalent cardiac arrhythmias that affects the lives of many people around the world and is associated with a five-fold increased risk of stroke and mortality. Like other problems in the healthcare domain, artificial intelligence (AI)-based models have been used to detect AF from patients' ECG signals. The cardiologist level performance in detecting this arrhythmia is often achieved by deep learning-based methods, however, they suffer from the lack of interpretability. In other words, these approaches are unable to explain the reasons behind their decisions. The lack of interpretability is a common challenge toward a wide application of machine learning (ML)-based approaches in the healthcare which limits the trust of clinicians in such methods. To address this challenge, we propose HAN-ECG, an interpretable bidirectional-recurrent-neural-network-based approach for the AF detection task. The HAN-ECG employs three attention mechanism levels to provide a multi-resolution analysis of the patterns in ECG leading to AF. The detected patterns by this hierarchical attention model facilitate the interpretation of the neural network decision process in identifying the patterns in the signal which contributed the most to the final detection. Experimental results on two AF databases demonstrate that our proposed model performs better than the existing algorithms. Visualization of these attention layers illustrates that our proposed model decides upon the important waves and heartbeats which are clinically meaningful in the detection task (e.g., absence of P-waves, and irregular R-R intervals for the AF detection task).

A Deep Learning Approach for Featureless Robust Quality Assessment of Intermittent Atrial Fibrillation Recordings from Portable and Wearable Devices

Atrial fibrillation (AF) is the most common heart rhythm disturbance in clinical practice. It often starts with asymptomatic and very short episodes, which are extremely difficult to detect without long-term monitoring of the patient’s electrocardiogram (ECG). Although recent portable and wearable devices may become very useful in this context, they often record ECG signals strongly corrupted with noise and artifacts. This impairs automatized ulterior analyses that could only be conducted reliably through a previous stage of automatic identification of high-quality ECG intervals. So far, a variety of techniques for ECG quality assessment have been proposed, but poor performances have been reported on recordings from patients with AF. This work introduces a novel deep learning-based algorithm to robustly identify high-quality ECG segments within the challenging environment of single-lead recordings alternating sinus rhythm, AF episodes and other rhythms. The method is based on the high learning capability of a convolutional neural network, which has been trained with 2-D images obtained when turning ECG signals into wavelet scalograms. For its validation, almost 100,000 ECG segments from three different databases have been analyzed during 500 learning-testing iterations, thus involving more than 320,000 ECGs analyzed in total. The obtained results have revealed a discriminant ability to detect high-quality and discard low-quality ECG excerpts of about 93%, only misclassifying around 5% of clean AF segments as noisy ones. In addition, the method has also been able to deal with raw ECG recordings, without requiring signal preprocessing or feature extraction as previous stages. Consequently, it is particularly suitable for portable and wearable devices embedding, facilitating early detection of AF as well as other automatized diagnostic facilities by reliably providing high-quality ECG excerpts to further processing stages.

A Review of Atrial Fibrillation Detection Methods as a Service

Atrial Fibrillation (AF) is a common heart arrhythmia that often goes undetected, and even if it is detected, managing the condition may be challenging. In this paper, we review how the RR interval and Electrocardiogram (ECG) signals, incorporated into a monitoring system, can be useful to track AF events. Were such an automated system to be implemented, it could be used to help manage AF and thereby reduce patient morbidity and mortality. The main impetus behind the idea of developing a service is that a greater data volume analyzed can lead to better patient outcomes. Based on the literature review, which we present herein, we introduce the methods that can be used to detect AF efficiently and automatically via the RR interval and ECG signals. A cardiovascular disease monitoring service that incorporates one or multiple of these detection methods could extend event observation to all times, and could therefore become useful to establish any AF occurrence. The development of an automated and efficient method that monitors AF in real time would likely become a key component for meeting public health goals regarding the reduction of fatalities caused by the disease. Yet, at present, significant technological and regulatory obstacles remain, which prevent the development of any proposed system. Establishment of the scientific foundation for monitoring is important to provide effective service to patients and healthcare professionals.

Detection of Atrial Fibrillation Episodes in Long-Term Heart Rhythm Signals Using a Support Vector Machine

Atrial fibrillation (AF) is a serious heart arrhythmia leading to a significant increase of the risk for occurrence of ischemic stroke. Clinically, the AF episode is recognized in an electrocardiogram. However, detection of asymptomatic AF, which requires a long-term monitoring, is more efficient when based on irregularity of beat-to-beat intervals estimated by the heart rate (HR) features. Automated classification of heartbeats into AF and non-AF by means of the Lagrangian Support Vector Machine has been proposed. The classifier input vector consisted of sixteen features, including four coefficients very sensitive to beat-to-beat heart changes, taken from the fetal heart rate analysis in perinatal medicine. Effectiveness of the proposed classifier has been verified on the MIT-BIH Atrial Fibrillation Database. Designing of the LSVM classifier using very large number of feature vectors requires extreme computational efforts. Therefore, an original approach has been proposed to determine a training set of the smallest possible size that still would guarantee a high quality of AF detection. It enables to obtain satisfactory results using only 1.39% of all heartbeats as the training data. Post-processing stage based on aggregation of classified heartbeats into AF episodes has been applied to provide more reliable information on patient risk. Results obtained during the testing phase showed the sensitivity of 98.94%, positive predictive value of 98.39%, and classification accuracy of 98.86%.

Modeling Framework for the Generation of Synthetic RR Series during Atrial Arrhythmias

We introduced a modeling framework for the generation of realistic ventricular interval (RR) series to be used in the validation of atrial arrhythmia detection algorithms. The framework included three previously proposed models, which reproduced the specific variability properties of RR series in normal sinus rhythm, atrial flutter (AFL) and atrial fibrillation (AF). Transitions between the three rhythms were governed by a three-state continuous-time Markov chain model, which could be tuned to obtain arrhythmic episodes of the requested length. As a representative application, the modeling framework was used to generate a database of RR series for the validation of a previously proposed AF detection algorithm, which was based on RR pattern similarity. The validation showed the deterioration of detector performance in presence of simulated AFL episodes. Thanks to the detailed reproduction of the specific features of the two most common atrial arrhythmias, our modeling framework may constitute a novel tool for the assessment and comparison of detection algorithm performance.

ECGNET: Learning where to attend for detection of atrial fibrillation with deep visual attention

The complexity of the patterns associated with atrial fibrillation (AF) and the high level of noise affecting these patterns have significantly limited the application of current signal processing and shallow machine learning approaches to accurately detect this condition. Deep neural networks have shown to be very powerful to learn the non-linear patterns in various problems such as computer vision tasks. While deep learning approaches have been utilized to learn complex patterns related to the presence of AF in electrocardiogram (ECG) signals, they can considerably benefit from knowing which parts of the signal is more important to focus on during learning. In this paper, we introduce a two-channel deep neural network to more accurately detect the presence of AF in the ECG signals. The first channel takes in an ECG signal and automatically learns where to attend for detection of AF. The second channel simultaneously takes in the same ECG signal to consider all features of the entire signal. Besides improving detection accuracy, this model can guide the physicians via visualization that what parts of the given ECG signal are important to attend while trying to detect atrial fibrillation. The experimental results confirm that the proposed model significantly improves the performance of AF detection on well-known MIT-BIH AF database with 5-s ECG segments (achieved a sensitivity of 99.53%, specificity of 99.26% and accuracy of 99.40%).

Detection of atrial fibrillation using a wrist-worn device

OBJECTIVE

This study proposes an algorithm for the detection of atrial fibrillation (AF), designed to operate on extended photoplethysmographic (PPG) signals recorded using a wrist-worn device of own design.

APPROACH

Robustness against false alarms is achieved by means of signal quality assessment and different techniques for suppression of ectopic beats, bigeminy, and respiratory sinus arrhythmia. The decision logic is based on our previously proposed RR interval-based AF detector, but modified to account for differences between interbeat intervals in the ECG and the PPG. The detector is evaluated on simulated PPG signals as well as on clinical PPG signals recorded during cardiac rehabilitation after myocardial infarction.

MAIN RESULTS

Analysis of the clinical signals showed that 1.5 false alarms were on average produced per day with a sensitivity of 72.0% and a specificity of 99.7% when 89.2% of the database was available for analysis, whereas as many as 15 when the RR interval-based AF detector, boosted by accelerometer information for signal quality assessment, was used. However, a sensitivity of 97.2% and a specificity of 99.6% were achieved when increasing the demands on signal quality so that 50% was available for analysis.

SIGNIFICANCE

The proposed detector offers promising performance and is particularly well-suited for implementation in low-power wearable devices, e.g. wrist-worn devices, with significance in mass screening applications.

Automatic, electrocardiographic-based detection of autonomic arousals and their association with cortical arousals, leg movements, and respiratory events in sleep

Study Objectives

The current definition of sleep arousals neglects to address the diversity of arousals and their systemic cohesion. Autonomic arousals (AA) are autonomic activations often associated with cortical arousals (CA), but they may also occur in relation to a respiratory event, a leg movement event or spontaneously, without any other physiological associations. AA should be acknowledged as essential events to understand and explore the systemic implications of arousals.

Methods

We developed an automatic AA detection algorithm based on intelligent feature selection and advanced machine learning using the electrocardiogram. The model was trained and tested with respect to CA systematically scored in 258 (181 training size/77 test size) polysomnographic recordings from the Wisconsin Sleep Cohort.

Results

A precision value of 0.72 and a sensitivity of 0.63 were achieved when evaluated with respect to CA. Further analysis indicated that 81% of the non-CA-associated AAs were associated with leg movement (38%) or respiratory (43%) events.

Conclusions

The presented algorithm shows good performance when considering that more than 80% of the false positives (FP) found by the detection algorithm appeared in relation to either leg movement or respiratory events. This indicates that most FP constitute autonomic activations that are indistinguishable from those with cortical cohesion. The proposed algorithm provides an automatic system trained in a clinical environment, which can be utilized to analyze the systemic and clinical impacts of arousals.

A novel wavelet-based filtering strategy to remove powerline interference from electrocardiograms with atrial fibrillation

OBJECTIVE

The electrocardiogram (ECG) is currently the most widely used recording to diagnose cardiac disorders, including the most common supraventricular arrhythmia, such as atrial fibrillation (AF). However, different types of electrical disturbances, in which power-line interference (PLI) is a major problem, can mask and distort the original ECG morphology. This is a significant issue in the context of AF, because accurate characterization of fibrillatory waves (f-waves) is unavoidably required to improve current knowledge about its mechanisms. This work introduces a new algorithm able to reduce high levels of PLI and preserve, simultaneously, the original ECG morphology.

APPROACH

The method is based on stationary wavelet transform shrinking and makes use of a new thresholding function designed to work successfully in a wide variety of scenarios. In fact, it has been validated in a general context with 48 ECG recordings obtained from pathological and non-pathological conditions, as well as in the particular context of AF, where 380 synthesized and 20 long-term real ECG recordings were analyzed.

MAIN RESULTS

In both situations, the algorithm has reported a notably better performance than common methods designed for the same purpose. Moreover, its effectiveness has proven to be optimal for dealing with ECG recordings affected by AF, since f-waves remained almost intact after removing very high levels of noise.

SIGNIFICANCE

The proposed algorithm may facilitate a reliable characterization of the f-waves, preventing them from not being masked by the PLI nor distorted by an unsuitable filtering applied to ECG recordings with AF.

ECG signal classification for the detection of cardiac arrhythmias using a convolutional recurrent neural network

OBJECTIVE

The electrocardiogram (ECG) provides an effective, non-invasive approach for clinical diagnosis in patients with cardiac diseases such as atrial fibrillation (AF). AF is the most common cardiac rhythm disturbance and affects ~2% of the general population in industrialized countries. Automatic AF detection in clinics remains a challenging task due to the high inter-patient variability of ECGs, and unsatisfactory existing approaches for AF diagnosis (e.g. atrial or ventricular activity-based analyses).

APPROACH

We have developed RhythmNet, a 21-layer 1D convolutional recurrent neural network, trained using 8528 single-lead ECG recordings from the 2017 PhysioNet/Computing in Cardiology (CinC) Challenge, to classify ECGs of different rhythms including AF automatically. Our RhythmNet architecture contained 16 convolutions to extract features directly from raw ECG waveforms, followed by three recurrent layers to process ECGs of varying lengths and to detect arrhythmia events in long recordings. Large 15  ×  1 convolutional filters were used to effectively learn the detailed variations of the signal within small time-frames such as the P-waves and QRS complexes. We employed residual connections throughout RhythmNet, along with batch-normalization and rectified linear activation units to improve convergence during training.

MAIN RESULTS

We evaluated our algorithm on 3658 testing data and obtained an F 1 accuracy of 82% for classifying sinus rhythm, AF, and other arrhythmias. RhythmNet was also ranked 5th in the 2017 CinC Challenge.

SIGNIFICANCE

Potentially, our approach could aid AF diagnosis in clinics and be used for patient self-monitoring to improve the early detection and effective treatment of AF.

Accurate detection of atrial fibrillation and atrial flutter using the electrocardiomatrix technique

BACKGROUND

Atrial fibrillation (AFIB) and atrial flutter (AFL) are two common cardiac arrhythmias that predispose patients to serious medical conditions. There is a need to accurately detect these arrhythmias to prevent diseases and reduce mortality. Apart from accurately detecting these arrhythmias, it is also important to distinguish between AFIB and AFL due to differing clinical treatments.

METHODS

In this study, we applied a new technology, the electrocardiomatrix (ECM) invented in our lab, in detecting AFIB and AFL in human patients. ECM converts 2D ECG signals into a 3D color matrix, which renders arrhythmia detection intuitive, fast, and accurate. Using ECM, we analyzed the ECG signals from the online MIT-BIH Atrial Fibrillation Database (PhysioNet), and compared our ECM-based results to manual annotations based on ECG by physicians.

RESULTS

Results demonstrate that ECM and PhysioNet annotations of AFIB and AFL agree more than 99% of the time. The sensitivities of the ECM for AFIB and AFL detection were 99.2% and 98.0%, respectively, and the specificities of the ECM for AFIB and AFL were both at 99.8% and 99.8%.

CONCLUSIONS

This study demonstrates that ECM is a reliable method for accurate identification of AFIB and AFL.

A comparison of entropy approaches for AF discrimination

OBJECTIVE

This study focuses on the comparison of single entropy measures for ventricular response analysis-based AF detection.

APPROACH

To enhance the performance of entropy-based AF detectors, we developed a normalized fuzzy entropy, [Formula: see text], a novel metric that (1) uses a fuzzy function to determine vector similarity, (2) replaces probability estimation with density estimation for entropy approximation, (3) utilizes a flexible distance threshold parameter, and (4) adjusts for heart rate by subtracting the natural log value of the mean RR interval. An AF detector based on [Formula: see text] was trained using the MIT-BIH atrial fibrillation (AF) database, and tested on the MIT-BIH normal sinus rhythm (NSR) and MIT-BIH arrhythmia databases. The [Formula: see text]-based AF detector was compared to AF detectors based on three other entropy measures: sample entropy ([Formula: see text]), fuzzy measure entropy ([Formula: see text]) and coefficient of sample entropy ([Formula: see text]), over three standard window sizes.

MAIN RESULTS

To classify AF and non-AF rhythms, [Formula: see text] achieved the highest area under receiver operating characteristic curve (AUC) values of 92.72%, 95.27% and 96.76% for 12-, 30- and 60-beat window lengths respectively. This was higher than the performance of the next best technique, [Formula: see text], over all windows sizes, which provided respective AUCs of 91.12%, 91.86% and 90.55%. [Formula: see text] and [Formula: see text] resulted in lower AUCs (below 90%) over all window sizes. [Formula: see text] also provided superior performance for all other tested statistics, including the Youden index, sensitivity, specificity, accuracy, positive predictivity and negative predictivity. In conclusion, we show that [Formula: see text] can be used to accurately identify AF from RR interval time series. Furthermore, longer window lengths (up to one minute) increase the performance of all entropy-based AF detectors under evaluation except the [Formula: see text]-based method.

SIGNIFICANCE

Our results demonstrate that the new developed normalized fuzzy entropy is an accurate measure for detecting AF.

Model-Based Assessment of f-Wave Signal Quality in Patients With Atrial Fibrillation

OBJECTIVE

The detection and analysis of atrial fibrillation (AF) in the ECG is greatly influenced by signal quality. The present study proposes and evaluates a model-based f-wave signal quality index (SQI), denoted , for use in the QRST-cancelled ECG signal.

METHODS

is computed using a harmonic f-wave model, allowing for variation in frequency and amplitude. The properties of  are evaluated on both f-waves and P-waves using 378 12-lead ECGs, 1875 single-lead ECGs, and simulated signals.

RESULTS

decreases monotonically when noise is added to f-wave signals, even for noise which overlaps spectrally with f-waves. Moreover, is shown to be closely associated with the accuracy of AF frequency estimation, where implies accurate estimation. When  is used as a measure of f-wave presence, AF detection performance improves: the sensitivity increases from 97.0% to 98.1% and the specificity increases from 97.4% to 97.8% when compared to the reference detector.

CONCLUSION

The proposed SQI represents a novel approach to assessing f-wave signal quality, as well as to determining whether f-waves are present.

SIGNIFICANCE

The use of  improves the detection of AF and benefits the analysis of noisy ECGs.

Electrocardiogram modeling during paroxysmal atrial fibrillation: application to the detection of brief episodes

OBJECTIVE

A model for simulating multi-lead ECG signals during paroxysmal atrial fibrillation (AF) is proposed.

SIGNIFICANCE

The model is of particular significance when evaluating detection performance in the presence of brief AF episodes, especially since annotated databases with such episodes are lacking.

APPROACH

The proposed model accounts for important characteristics such as switching between sinus rhythm and AF, varying P-wave morphology, repetition rate of f-waves, presence of atrial premature beats, and various types of noise.

MAIN RESULTS

Two expert cardiologists assessed the realism of simulated signals relative to real ECG signals, both in sinus rhythm and AF. The cardiologists identified the correct rhythm in all cases, and considered two-thirds of the simulated signals as realistic. The proposed model was also investigated by evaluating the performance of two AF detectors which explored either rhythm only or both rhythm and morphology. The results show that detection performance is strongly dependent on AF episode duration, and, consequently, demonstrate that the model can play a significant role in the investigation of detector properties.

Modeling of the photoplethysmogram during atrial fibrillation

A phenomenological model for simulating the photoplethysmogram (PPG) during atrial fibrillation (AF) is proposed. The simulated PPG is solely based on RR interval information, and, therefore, any annotated ECG database can be used to model sinus rhythm, AF, or rhythms with premature beats. A PPG pulse is modeled by a linear combination of a log-normal and two Gaussian waveforms. The model PPG is obtained by placing individual pulses according to the RR intervals so that a connected signal is created. The model is evaluated on synchronously recorded ECG and PPG signals from the MIMIC and the University of Queensland Vital Signs Dataset databases. The results show that the model PPG signals closely resemble real signal for sinus rhythm, premature beats, as well as for AF. The model is used to study the performance of a low-complexity RR interval-based AF detector on simulated PPG signals with five different pulse types generated using the MIT-BIH AF database at signal-to-noise ratios (SNRs) from 0 to 30dB. PPGs composed of pulses with a dicrotic notch tend to increase the rate of false alarms, especially at lower SNRs. The model is capable of generating simulated PPG signals from RR interval series with sinus rhythm, AF, and premature beats. Considering the lack of annotated, public PPG databases with arrhythmias, the simulation of realistic PPG signals based on annotated ECG signals is expected to facilitate the development and testing of PPG-specific AF detectors.

Automated detection of atrial fibrillation using R-R intervals and multivariate-based classification

Automated detection of AF from the electrocardiogram (ECG) still remains a challenge. In this study, we investigated two multivariate-based classification techniques, Random Forests (RF) and k-nearest neighbor (k-nn), for improved automated detection of AF from the ECG. We have compiled a new database from ECG data taken from existing sources. R-R intervals were then analyzed using four previously described R-R irregularity measurements: (1) the coefficient of sample entropy (CoSEn), (2) the coefficient of variance (CV), (3) root mean square of the successive differences (RMSSD), and (4) median absolute deviation (MAD). Using outputs from all four R-R irregularity measurements, RF and k-nn models were trained. RF classification improved AF detection over CoSEn with overall specificity of 80.1% vs. 98.3% and positive predictive value of 51.8% vs. 92.1% with a reduction in sensitivity, 97.6% vs. 92.8%. k-nn also improved specificity and PPV over CoSEn; however, the sensitivity of this approach was considerably reduced (68.0%).