Rhythm Classification Research Articles

External validation of a machine learning based classification algorithm for ambulatory heart rhythm diagnostics using smartphone-photoplethysmography - the SMARTBEATS algorithm study

Abstract Introduction Smartphone-photoplethysmography (PPG) can be used for heart rhythm monitoring. According to current guidelines, ECG verification is required for a new diagnosis of atrial fibrillation (AF). Accurate machine learning (ML) based automatic algorithms have the potential to facilitate clinical applications generating large numbers of PPG recordings, by offloading the need for manual over-read. Purpose The aim of this study was to evaluate the diagnostic performance of an automatic ML-based algorithm for heart rhythm diagnostics using smartphone-PPG recorded by patients with AF in an unsupervised ambulatory setting. Methods Patients undergoing direct current cardioversion (DCCV) at a university hospital for treatment of AF or atrial flutter (AFL) were asked to perform one-minute heart rhythm recordings post-cardioversion at least twice daily for 30 days. All participants were provided with an iPhone 7 smartphone running the CORAI Heart Monitor PPG application simultaneously with a single-lead ECG recording (KardiaMobile). The automatic algorithm uses support vector machines (SVM) to classify heart rhythm from smartphone-PPG recordings. The SVM classifier evaluated here was trained on ambulatory PPG recordings made by patients in a separate cardioversion cohort, ensuring complete independence between training and validation sets. PPG recordings in the validation cohort were analyzed by the automatic algorithm and diagnostic performance was calculated by comparing heart rhythm classification output from the SVM classifier to the heart rhythm diagnosis from the simultaneous ECG recordings. ECG recordings were interpreted independently by two cardiologists and were set as gold standard. Results The SVM classifier was trained on data from 153 patients with 11,749 PPG recordings, of which 5,873 (50.0%) were labelled as AF and 5,876 (50.0%) as sinus rhythm (SR), as assessed by manual reading. In the validation cohort, 280 patients, with a median age of 69.0 years (31% women), registered 18,005 simultaneous PPG and ECG recordings. Recordings interpreted as AFL on ECG (2.0%), as having insufficient quality on ECG (4.9%) or PPG (2.8%), and low certainty algorithmic classifications (1.2%) were excluded, leaving 16,057 PPG recordings. Included ECG recordings consisted of 71.2% interpreted as (SR) and 28.8% as AF. Algorithm classification of the PPG recordings in the validation cohort diagnosed AF (sensitivity) in 99.7% and SR (specificity) in 99.7% of the recordings, with an overall accuracy of 99.7%. F1-score was 99.4% and area under the ROC curve (AUC) was 0.999. Conclusion A machine learning based algorithm for automatic heart rhythm diagnostics showed excellent diagnostic performance for smartphone-PPG recordings in an unsupervised ambulatory setting, minimizing the need for ECG verification in cardioversion populations.

Infants' reliance on rhythm to distinguish languages: A critical review.

This article reviews empirical methods and findings on early language discrimination, questioning rhythm-class based hypotheses on language discrimination in infancy, as well as the assumption that early language discrimination is driven primarily (or solely) by temporal prosodic cues. The present work argues that within-rhythm class discrimination which - according to the rhythmic hypothesis - is not applicable very early in life, has not been sufficiently tested with infants under 4months of age, that familiarity with a language is not a prerequisite for its discrimination from another rhythmically similar language, and that the temporal rhythm properties may not universally be the primary cues to language discrimination. Although rhythm taxonomy is now by many understood as outdated, some developmental literature still draws on the assumption that rhythm classification determines infants' language discrimination; other studies consider rhythm along a continuous scale and only a few account for cues to language discrimination other than temporal ones. It is proposed that studies on early language discrimination systematically test the contribution of other than temporal rhythm cues, similarly to recent work on multidimensional psychoacoustic salience in the acquisition of segmental categories.

Evaluating the impacts of digital ECG denoising on the interpretive capabilities of healthcare professionals.

Electrocardiogram (ECG) interpretation is an essential skill across multiple medical disciplines; yet, studies have consistently identified deficiencies in the interpretive performance of healthcare professionals linked to a variety of educational and technological factors. Despite the established correlation between noise interference and erroneous diagnoses, research evaluating the impacts of digital denoising software on clinical ECG interpretation proficiency is lacking. Forty-eight participants from a variety of medical professions and experience levels were prospectively recruited for this study. Participants' capabilities in classifying common cardiac rhythms were evaluated using a sequential blinded and semi-blinded interpretation protocol on a challenging set of single-lead ECG signals (42 × 10 s) pre- and post-denoising with robust, cloud-based ECG processing software. Participants' ECG rhythm interpretation performance was greatest when raw and denoised signals were viewed in a combined format that enabled comparative evaluation. The combined view resulted in a 4.9% increase in mean rhythm classification accuracy (raw: 75.7% ± 14.5% vs. combined: 80.6% ± 12.5%, P = 0.0087), a 6.2% improvement in mean five-point graded confidence score (raw: 4.05 ± 0.58 vs. combined: 4.30 ± 0.48, P < 0.001), and 9.7% reduction in the mean proportion of undiagnosable data (raw: 14.2% ± 8.2% vs. combined: 4.5% ± 2.4%, P < 0.001), relative to raw signals alone. Participants also had a predominantly positive perception of denoising as it related to revealing previously unseen pathologies, improving ECG readability, and reducing time to diagnosis. Our findings have demonstrated that digital denoising software improves the efficacy of rhythm interpretation on single-lead ECGs, particularly when raw and denoised signals are provided in a combined viewing format, warranting further investigation into the impact of such technology on clinical decision-making and patient outcomes.

A circadian behavioral analysis suite for real-time classification of daily rhythms in complex behaviors.

Measuring animal behavior over long timescales has been traditionally limited to behaviors that are easily measurable with real-time sensors. More complex behaviors have been measured over time, but these approaches are considerably more challenging due to the intensive manual effort required for scoring behaviors. Recent advances in machine learning have introduced automated behavior analysis methods, but these often overlook long-term behavioral patterns and struggle with classification in varying environmental conditions. To address this, we developed a pipeline that enables continuous, parallel recording and acquisition of animal behavior for an indefinite duration. As part of this pipeline, we applied a recent breakthrough self-supervised computer vision model to reduce training bias and overfitting and to ensure classification robustness. Our system automatically classifies animal behaviors with a performance approaching that of expert-level human labelers. Critically, classification occurs continuously, across multiple animals, and in real time. As a proof-of-concept, we used our system to record behavior from 97 mice over two weeks to test the hypothesis that sex and estrogen influence circadian rhythms in nine distinct home cage behaviors. We discovered novel sex- and estrogen-dependent differences in circadian properties of several behaviors including digging and nesting rhythms. We present a generalized version of our pipeline and novel classification model, the "circadian behavioral analysis suite," (CBAS) as a user-friendly, open-source software package that allows researchers to automatically acquire and analyze behavioral rhythms with a throughput that rivals sensor-based methods, allowing for the temporal and circadian analysis of behaviors that were previously difficult or impossible to observe.

Diagnostic accuracy of smartwatches for automated atrial fibrillation detection over time: insights from the basel wearable study

Abstract Background The first smart-watch (SW) able to record a single lead ECG (SL-ECG) with automated atrial fibrillation (AF) detection algorithm was introduced 2018. Multiple other manufacturers followed. While manufacturers claim to continuously improve their devices’ algorithm, little is known about the impact on diagnostic accuracy on AF detection over time in a real-world cohort of patients. Purpose To assess if the diagnostic accuracy of algorithm-based AF detection of 5 smart devices has improved over time. Methods In this prospective study, consecutive patients undergoing electrophysiological procedures in a tertiary referral center were included between April 2021 to May 2023. Each participant received a 12-lead ECG and 5 SL-ECGs with different smart devices (AliveCor KardiaMobile, Apple Watch 6, Fitbit Sense, Samsung Galaxy Watch 3 and Withings ScanWatch). All SW were continuously updated for newest software or firmware as recommend by the manufacturers. The SW rhythm classification was compared to manual cardiologist-based interpretation of SL-ECGs. The change in number of inconclusive SL-ECGs by manufacturer’s algorithm were investigated over time via Cochran-Armitage test for trend and illustrated over time via logistic regression. Results 247 participants were included (30% female, mean age 66.3 [IQR: 57-74] years) resulting in a total of 1’235 SL-ECGs. Among these, 202 SL-ECGs (16%) were labeled as inconclusive by at least one smart-device. Inconclusive rates for individual SW ranged from 15% to 19%. When comparing the first and last quintile of patients in the study period, the inconclusive rates for the KardiaMobile were 14% vs. 6%, for the Apple Watch 14% vs. 6%, for the Fitbit Sense 20% vs. 14%, for the Samsung Galaxy 18% vs. 10% and for the Withings ScanWatch 22% vs. 14%, respectively. When assessed continuously over time, we observed a trend towards fewer inconclusive tracings, without reaching statistical significance (Figure 1). Conclusion Over 26 months, inconclusive tracings of 5 SW didn’t significantly decrease despite frequent software updates in a real-world patient cohort. Larger, long-term cohorts are needed to assess if automated AF detection has significantly improved over time.Figure 1

Atrial imaging and cardiac Rhythm In cryptogenic Embolic Stroke: the ARIES study

Abstract Background One of the underlying causes of cryptogenic stroke is an unknown cardioembolic etiology, and the most frequent is atrial fibrillation (AF). It is not unusual to find non-sustained (&amp;lt;30 s) "microfibrillatory" atrial tachycardias (nsAT) or frequent atrial ectopic beats (AE) during the work-up but its pathological significance remains uncertain. Purpose Our aim is to analyze the risk of developing AF or nsAT in patients with cryptogenic stroke and compare their atrial function using advanced echocardiography and their clinical outcome during the follow-up. Methods The ARIES study (Atrial imaging and cardiac Rhythm In cryptogenic Embolic Stroke) is an observational, prospective study in patients admitted for recent cryptogenic stroke in a comprehensive Stroke Unit. We analyze the frequency of detecting AF and nsAT in two 30-day ECG records with a wearable device and we compare echocardiographic signs of left atrial (LA) dysfunction according to rhythm classification: AF, high-burden nsAT/AE (&amp;gt;2 nsAT/24 hours or &amp;gt;3.000 AE/24 hours respectively) and normal sinus rhythm (NSR). Finally, we evaluate stroke recurrence and mortality at 12-months and report changes in classification of stroke etiology during the follow-up. Results In total prolonged rhythm study (1st and 2nd 1-month wearable EKG), AF was found in 35 (32.1%) patients, high-burden nsAT/AE in 27 patients (24.8%) and normal rhythm was found in 47 patients (43.1%). Patients with AF compared with patients with NSR presented higher values in LA volume (69.9±21.8 vs 49.4±26.5, p&amp;lt;0.001) , LA index volume (38.8±11.2 vs 27.3±11.8 ml/m2, p&amp;lt;0.001), 3D LA index volume (50.6±17.2 vs 34.0±15.4 ml/m2, p&amp;lt;0.001), 3D-telediastolic LA volume (89.5±27.7 vs 66.8 ±31.9 ml/m2, p=0.002), 3D-telesistolic LA volume (43.6±15.3 vs 25.3±17.4 ml/m2, p &amp;lt;0.001), lower rates of LA ejection fraction (50 ±14.6 vs 62.7±11.8, p=0.001) and LA strain in reservoir (22.0±8.6 vs 30.4±10.5, p&amp;lt;0.001) and contraction (10.5±8.18 vs 17.1±7.5, p&amp;lt;0.001). After multivariable regression analysis (age, HTA and CHASDS2VASC) all the differences remained significative. High-burden nsAT/AE and NSR patients presented significant differences in LA index volume (33.4±12.4 vs 27.3±11.8 ml/m2, p=0.043) and LA strain in reservoir (26.6±12.5) vs 31.0±10.1, p=0.034) that disappeared after multivariate regression analysis. There were no significant differences in ischemic recurrence or mortality between the three groups. Conclusions In patients with cryptogenic stroke, we registered a high rate of AF and high-burden nsAT/AE. AF patients showed notable alterations in LA volume, LA function and reservoir and contraction strain, while patients with high-burden nsAT/AE showed less structural alteration in echocardiography without difference in stroke recurrence among groups.

Artificial Intelligence Interpretation of the Electrocardiogram: A State-of-the-Art Review.

Artificial intelligence (AI) is transforming electrocardiography (ECG) interpretation. AI diagnostics can reach beyond human capabilities, facilitate automated access to nuanced ECG interpretation, and expand the scope of cardiovascular screening in the population. AI can be applied to the standard 12-lead resting ECG and single-lead ECGs in external monitors, implantable devices, and direct-to-consumer smart devices. We summarize the current state of the literature on AI-ECG. Rhythm classification was the first application of AI-ECG. Subsequently, AI-ECG models have been developed for screening structural heart disease including hypertrophic cardiomyopathy, cardiac amyloidosis, aortic stenosis, pulmonary hypertension, and left ventricular systolic dysfunction. Further, AI models can predict future events like development of systolic heart failure and atrial fibrillation. AI-ECG exhibits potential in acute cardiac events and non-cardiac applications, including acute pulmonary embolism, electrolyte abnormalities, monitoring drugs therapy, sleep apnea, and predicting all-cause mortality. Many AI models in the domain of cardiac monitors and smart watches have received Food and Drug Administration (FDA) clearance for rhythm classification, while others for identification of cardiac amyloidosis, pulmonary hypertension and left ventricular dysfunction have received breakthrough device designation. As AI-ECG models continue to be developed, in addition to regulatory oversight and monetization challenges, thoughtful clinical implementation to streamline workflows, avoiding information overload and overwhelming of healthcare systems with false positive results is necessary. Research to demonstrate and validate improvement in healthcare efficiency and improved patient outcomes would be required before widespread adoption of any AI-ECG model.

A Deep Learning Approach for Detecting Atrial Fibrillation using RR Intervals of ECG

Purpose: Atrial Fibrillation (AF) is one of the most common types of heart arrhythmias observed in clinical practice. AF can be detected using an Electrocardiogram (ECG). ECG signals are time-varying and nonlinear in nature. Hence, it is very difficult for a physician to manually perform accurate and rapid classification of different heart rhythms. Materials and Methods: In this paper, we propose a method using Discrete Wavelet Transform (DWT) with db6 as the basis function for denoising ECG signal. Results: The denoised ECG is smoothened using the Savitzky- Golay filter. Deep learning methods, such as a combination of Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) (CNN-LSTM) and ResNet18 are used for the accurate classification of ECG signals using Physionet Challenge 2017 database. Conclusion: With a 10-fold cross-validation method the model provided overall accuracy of 98.25% with the CNN-LSTM classifier.

Real-world validation of smartphone-based photoplethysmography for rate and rhythm monitoring in atrial fibrillation.

Photoplethysmography- (PPG) based smartphone applications facilitate heart rate and rhythm monitoring in patients with paroxysmal and persistent atrial fibrillation (AF). Despite an endorsement from the European Heart Rhythm Association, validation studies in this setting are lacking. Therefore, we evaluated the accuracy of PPG-derived heart rate and rhythm classification in subjects with an established diagnosis of AF in unsupervised real-world conditions. Fifty consecutive patients were enrolled, 4 weeks before undergoing AF ablation. Patients used a handheld single-lead electrocardiography (ECG) device and a fingertip PPG smartphone application to record 3907 heart rhythm measurements twice daily during 8 weeks. The ECG was performed immediately before and after each PPG recording and was given a diagnosis by the majority of three blinded cardiologists. A consistent ECG diagnosis was exhibited along with PPG data of sufficient quality in 3407 measurements. A single measurement exhibited good quality more often with ECG (93.2%) compared to PPG (89.5%; P < 0.001). However, PPG signal quality improved to 96.6% with repeated measurements. Photoplethysmography-based detection of AF demonstrated excellent sensitivity [98.3%; confidence interval (CI): 96.7-99.9%], specificity (99.9%; CI: 99.8-100.0%), positive predictive value (99.6%; CI: 99.1-100.0%), and negative predictive value (99.6%; CI: 99.0-100.0%). Photoplethysmography underestimated the heart rate in AF with 6.6 b.p.m. (95% CI: 5.8 b.p.m. to 7.4 b.p.m.). Bland-Altman analysis revealed increased underestimation in high heart rates. The root mean square error was 11.8 b.p.m. Smartphone applications using PPG can be used to monitor patients with AF in unsupervised real-world conditions. The accuracy of AF detection algorithms in this setting is excellent, but PPG-derived heart rate may tend to underestimate higher heart rates.

A comprehensive review on efficient artificial intelligence models for classification of abnormal cardiac rhythms using electrocardiograms

Deep learning has made many advances in data classification using electrocardiogram (ECG) waveforms. Over the past decade, data science research has focused on developing artificial intelligence (AI) based models that can analyze ECG waveforms to identify and classify abnormal cardiac rhythms accurately. However, the primary drawback of the current AI models is that most of these models are heavy, computationally intensive, and inefficient in terms of cost for real-time implementation. In this review, we first discuss the current state-of-the-art AI models utilized for ECG-based cardiac rhythm classification. Next, we present some of the upcoming modeling methodologies which have the potential to perform real-time implementation of AI-based heart rhythm diagnosis. These models hold significant promise in being lightweight and computationally efficient without compromising the accuracy. Contemporary models predominantly utilize 12-lead ECG for cardiac rhythm classification and cardiovascular status prediction, increasing the computational burden and making real-time implementation challenging. We also summarize research studies evaluating the potential of efficient data setups to reduce the number of ECG leads without affecting classification accuracy. Lastly, we present future perspectives on AI's utility in precision medicine by providing opportunities for accurate prediction and diagnostics of cardiovascular status in patients.

Rhythm Research in Interactive System Design: A Literature Review

Rhythm can create a sense of structure or order, making it easier for users to understand and navigate interactive systems, thereby making user engagement more enjoyable and motivating. Recently, designers are increasingly interested in integrating rhythm into the design of interactive systems. However, the existing literature on this topic is fragmented, making it difficult for designers to find comprehensive guidance. This paper presents the first literature review of rhythm research for designing an interactive system based on the PRISMA method. Through an analysis of 59 papers based on the critical review method (e.g., research through design), we identify six research genres and provide a map to help future researchers locate their research position. Because there is currently no clear classification framework to guide designers in integrating rhythm into interactive systems, we propose a classification framework that includes six design qualities: three human rhythm-driven qualities and three computer rhythm-driven qualities based on the literature review. This classification framework provides a foundation for designers to regulate rhythm in interactive systems. In addition, we discuss emerging issues related to rhythm genres and rhythm classification framework and design qualities. We reflect on the effectiveness of our proposed framework by revisiting previous works on designing interactive systems, and provide potential directions for future research in this area. Overall, this paper aims to provide comprehensive guidance for designers who wish to integrate rhythm into the design of interactive systems, and to advance the understanding of the role of rhythm in interactive system design.

Smart-device ECGs and inconclusive results: impact of ECG anomalies

Abstract Background Evermore smart-devices are capable of recording single-lead (SL) electrocardiograms (ECGs). Each recording comes with an automated rhythm interpretation usually consisting of sinus rhythm (SR), atrial fibrillation (AF) or inconclusive. The impact of common ECG anomalies possibly leading to an inconclusive interpretation by the devices algorithm is largely unknown. Purpose to assess the impact of baseline ECG anomalies on automated smart-device rhythm classification. Methods In this prospective study, consecutive hospitalized patients undergoing electrophysiological procedures in a tertiary referral center were included. Each participant obtained a 12-lead ECG followed by SL-ECGs with 5 different smart-devices (AliveCor KardiaMobile, Apple Watch 6, Fitbit Sense, Samsung Galaxy Watch3 and Withings ScanWatch). The smart-device generated rhythm determination was compared to manual interpretation. ECG anomalies were assessed by 2 independent cardiologists and included atrial Flutter (AFlut), pacing and conduction delay consisting of AV-block or left/right bundle branch block, artifacts and ectopy. Inconclusive tracings were labelled false-positive if the goldstandard was SR in the 12-lead ECG and false-negative if the goldstandard was AF in the 12-lead ECG. Relative Risk (RR) for anomalies to result as inconclusive diagnosis by the smart-device was calculated. Results 245 participants were included (34% female, mean age 64±13 years) resulting in a total of 1’165 recorded SL-ECGs. Among these, 247 SL-ECGs (21%) were labeled as inconclusive by at least one smart-device (Figure 1). Overall, 834 ECG anomalies were present. 169 inconclusive SL-ECGs were labelled false-positive with 192 anomalies present. Among these, 34 (18%) had ventricular pacing, RR 2.9 (CI 2.2-3.8), 52 (27%) showed conduction delay, RR 1.5 (CI 1.1-2), 44 (23 %) had ectopy, RR 2.3 (CI 1.7-3), and 62 (32%) some sort of artifacts, RR 1.9 (CI 1.5-2.5). Among 78 inconclusive SL-ECGs with false-negative results, 84 anomalies were present. 18 (21%) were in AFlut, RR 1.5 (CI 0.94-2.3), 12 (14%) had ventricular pacing RR, 4.2 (CI 3.1-5.6), 14 (17%) showed conduction delay, RR 1.3 (CI 0.8-2.1), 9 (11%) had ectopy, RR 0.99 (CI 0.54-1.8), and 31 (37%) some sort of artifacts, RR 1.7 (CI 1.2-2.5). There were no differences between assessed smart-devices. Conclusion Underlying ECG anomalies are common in inconclusive SL-ECGs and impact automated rhythm classification.Figure 1

Optimization of Arrhythmia-based ECG-lead Selection for Computer-interpreted Heart Rhythm Classification.

The 12-lead ECG only has 8 independent ECG leads, which leads to diagnostic redundancy when using all 12 leads for heart arrhythmias classification. We have previously developed a deep learning (DL)-based computer-interpreted ECG (CIE) approach to identify an optimal 4-lead ECG subset for classifying heart arrhythmias. However, the clinical diagnostic criteria of cardiac arrhythmia types are often lead-specific, so this study is going to explore the selection of arrhythmia-based ECG-lead subsets rather than one general optimal ECG-lead subset, which could improve the classification performance for the CIE. The DL-based CIE model previously developed was used to learn 4 common types of heart arrhythmias (LBBB, RBBB, AF, and I-AVB) for identifying corresponding optimal ECG-lead subsets. A public dataset that splits into training (approx. 70%), validation (approx. 15%), and test (approx. 15%) sets from the PhysioNet Cardiology Challenge 2020 was used to explore the study. The results demonstrated that the DL-based CIE model identified an optimal ECG-lead subset for each arrhythmia: I, II, aVR, aVL, V1, V3, and V5 for I-AVB; I, II, aVR, and V3 for AF; I, II, aVR, aVF, V1, V3, and V4 for LBBB; and I, II, III, aVR, V1, V4, and V6 for RBBB. For each arrhythmia classification, the DL-based CIE model using the optimal ECG-lead subset significantly outperformed the model using the full 12-lead ECG set on the validation set and on the external test dataset.The results support the hypothesis that using an optimal ECG-lead subset instead of the full 12-lead ECG set can improve the classification performance of a specific arrhythmia when using the DL-based CIE approach.Clinical Relevance- Using an arrhythmia-based optimal ECG-lead subset, the classification performance of a deep-learning-based model can be achieved without loss of accuracy in comparison with the full 12-lead set (p<0.05).

Development of an AI based automated analysis of pediatric Apple Watch iECGs.

The Apple Watch valuably records event-based electrocardiograms (iECG) in children, as shown in recent studies by Paech et al. In contrast to adults, though, the automatic heart rhythm classification of the Apple Watch did not provide satisfactory results in children. Therefore, ECG analysis is limited to interpretation by a pediatric cardiologist. To surmount this difficulty, an artificial intelligence (AI) based algorithm for the automatic interpretation of pediatric Apple Watch iECGs was developed in this study. A first AI-based algorithm was designed and trained based on prerecorded and manually classified i.e., labeled iECGs. Afterward the algorithm was evaluated in a prospectively recruited cohort of children at the Leipzig Heart Center. iECG evaluation by the algorithm was compared to the 12-lead-ECG evaluation by a pediatric cardiologist (gold standard). The outcomes were then used to calculate the sensitivity and specificity of the Apple Software and the self-developed AI. The main features of the newly developed AI algorithm and the rapid development cycle are presented. Forty-eight pediatric patients were enrolled in this study. The AI reached a specificity of 96.7% and a sensitivity of 66.7% for classifying a normal sinus rhythm. The current study presents a first AI-based algorithm for the automatic heart rhythm classification of pediatric iECGs, and therefore provides the basis for further development of the AI-based iECG analysis in children as soon as more training data are available. More training in the AI algorithm is inevitable to enable the AI-based iECG analysis to work as a medical tool in complex patients.

Method for Solving Difficulties in Rhythm Classification Caused by Few Samples and Similar Characteristics in Electrocardiograms.

A method for accurately analyzing electrocardiograms (ECGs), which are obtained from electrical signals generated by cardiac activity, is essential in heart disease diagnosis. However, rhythms are typically obtained with relatively few data samples and similar characteristics, making them difficult to classify. To solve these issues, we proposed a novel method that distinguishes a given ECG rhythm using a beat score map (BSM) image. Through the proposed method, the associations between beats and previously used features, such as the R-R interval, were considered. Rhythm classification was implemented by training a convolutional neural network model and using transfer learning with the created BSM image. As a result, the proposed method for ECG rhythms with small data samples showed significant results. It also showed good performance in differentiating atrial fibrillation (AFIB) and atrial flutter (AFL) rhythms, which are difficult to distinguish due to their similar characteristics. The performance for rhythms with a small number of samples of the proposed method is 20% better than an existing method. In addition, the performance based on the F-1 score for classifying AFIB and AFL of the proposed method is 30% better than the existing method. This study solved the previous limitations caused by small sample numbers and similar rhythms.

Functional timing or rhythmical timing, or both? A corpus study of English and Mandarin duration.

It has been long held that languages of the world are divided into rhythm classes so that they are either stress-timed, syllable-timed or mora-timed. It is also known for a long time that duration serves various informational functions in speech. But it is unclear whether these two kinds of uses of duration are complementary to each other, or they are actually one and the same. There has been much empirical research that raises questions about the rhythm class hypothesis due to lack of evidence of the suggested isochrony in any language. Yet the alleged cross-language rhythm classification is still widely taken for granted and continues to be researched. Here we conducted a corpus study of English, an archetype of a stress-timed language, and Mandarin, an alleged syllable-timed language, to look for evidence of at least a tendency toward isochrony when much of the informational use of duration is controlled for. We examined the relationship between segment and syllable duration and the relationship of syllable and phrase duration in the two languages. The results show that in English syllables are largely incompressible to allow stress-timing because segment duration is inflexible to allow variable syllable duration beyond its functional use. Surprisingly, Mandarin does show a small tendency toward both equal syllable duration and equal phrase duration. Additionally, the duration of pre-boundary syllables in English increases linearly with break index, whereas in Mandarin, the duration increase stops after break index 2, which is accompanied by the insertion of silent pauses. We conclude, therefore, timing and duration in speech are predominantly used for encoding information rather being controlled by a rhythmic principle, and the residual equal-duration tendency in the two languages examined here show exactly the opposite patterns from the predictions of the rhythm class hypothesis.

Tracking the Rhythm: Pansori Rhythm Segmentation and Classification Methods and Datasets

This paper presents two methods to understand the rhythmic patterns of the voice in Korean traditional music called Pansori. We used semantic segmentation and classification-based structural analysis methods to segment the seven rhythmic categories of Pansori. We propose two datasets; one is for rhythm classification and one is for segmentation. Two classification and two segmentation neural networks are trained and tested in an end-to-end manner. The standard HR network and DeepLabV3+ network are used for rhythm segmentation. A modified HR network and a novel GlocalMuseNet are used for the classification of music rhythm. The GlocalMuseNet outperforms the HR network for Pansori rhythm classification. A novel segmentation model (a modified HR network) is proposed for Pansori rhythm segmentation. The results show that the DeepLabV3+ network is superior to the HR network. The classifier networks are used for time-varying rhythm classification that behaves as the segmentation using overlapping window frames in a spectral representation of audio. Semantic segmentation using the DeepLabV3+ and the HR network shows better results than the classification-based structural analysis methods used in this work; however, the annotation process is relatively time-consuming and costly.

On the use of singular value decomposition for QRS detection and ECG denoising

Abstract QRS detection is a pre-processing step to detect the heartbeat in an electrocardiogram (ECG) for subsequent rhythm classification. However, measured ECG waveforms may differ as a result of intrinsic variability or due to artefacts or noise. If the signals are distorted, then this often leads to difficulties in QRS detection. Of course, a high QRS detection performance is an important part of an ECG analysis algorithm, and furthermore, it must work even for highly noisy signals. Singular value decompositon (SVD) is the factorization of a matrix into the product three matrices. SVD allows us to find important components of data and, thus, can be used for dimension reduction or denoising. We introduce SVD based methods for QRS detection and ECG denoising, especially for short unknown signal segments, and show application results.

A review of arrhythmia detection based on electrocardiogram with artificial intelligence

Introduction With the widespread availability of portable electrocardiogram (ECG) devices, there will be a surge in ECG diagnoses. Traditional computer-aided diagnosis of arrhythmia mainly relies on the rules of medical knowledge, which are insufficient due to the limitations of data quality and human expert knowledge. The research of arrhythmia detection methods based on artificial intelligence (AI) techniques can assist physicians in high-precision arrhythmia diagnosis. AI algorithms can also be embedded in smart ECG devices to help more people perform early screening for arrhythmia. Areas covered The primary objective of this paper is to describe the application of AI methods in the process of arrhythmia detection. Meanwhile, the advantages and limitations of various approaches in different applications are summarized to provide guidance and reference for future research work. Expert opinion Machine learning (ML) and deep learning (DL) algorithms can be more effectively employed to handle ECG signal denoising and quality assessment, wave detection and delineation, and arrhythmia classification problems. The DL approach can automatically learn deep representation features and temporal features of the ECG signal for heartbeat or rhythm classification. The application of AI methods for arrhythmia detection systems will significantly relieve the pressure on physicians to analyze ECGs.

A method for continuous rhythm classification and early detection of ventricular fibrillation during CPR

AimWe developed a method which continuously classifies the ECG rhythm during CPR in order to guide clinical care. MethodsWe conducted a retrospective study of 432 patients treated following out-of-hospital cardiac arrest. Continuous ECG sequences from two-minute CPR cycles were extracted from defibrillator recordings and further divided into five-second clips. We developed an algorithm using wavelet analysis, hidden semi-Markov modeling, and random forest classification. The algorithm classifies individual clips as asystole, organized rhythm, ventricular fibrillation, or Inconclusive while integrating information from previous clips. Classifications were compared to manual annotations to estimate accuracy in an independent validation dataset. Continuous sequences were classified as shockable, non-shockable, or Inconclusive; classifications were used to compute shock sensitivity and specificity. ResultsOf 432 patient-cases, 290 were used for development and 142 for validation. In the 12,294 validation ECG clips during CPR, accuracies were 0.88 (95% CI 0.85–0.91) for asystole, 0.98 (95% CI 0.98–0.99) for organized rhythm, and 0.97 (95% CI 0.96–0.97) for ventricular fibrillation, with 43% classified as Inconclusive. Of 457 continuous sequences, shock sensitivity was 0.90 (95% CI 0.86–0.93), shock specificity was 0.98 (95% CI 0.93–0.99), and 7% were Inconclusive. Median delay to ventricular fibrillation recognition was 10 (IQR 5–32) seconds. ConclusionA novel algorithm continuously classified the primary resuscitation rhythms—asystole, organized rhythms, and ventricular fibrillation—with 88–98% accuracy, enabling accurate shock advisory guidance during most two-minute CPR cycles. Additional investigation is required to understand how algorithm implementation could affect rescuer actions and clinical outcomes.