Intelligent assessment of atrial fibrillation gradation based on sinus rhythm electrocardiogram and baseline information

BACKGROUND

Atrial fibrillation (AF) is a progressive arrhythmia that significantly affects a patient's quality of life. The 4S-AF scheme is clinically recommended for AF management; however, the evaluation process is complex and time-consuming. This renders its promotion in primary medical institutions challenging. This retrospective study aimed to simplify the evaluation process and present an objective assessment model for AF gradation.

METHODS

In total, 189 12-lead electrocardiogram (ECG) recordings from 64 patients were included in this study. The data were annotated into two groups (mild and severe) according to the 4S-AF scheme. Using a preprocessed ECG during the sinus rhythm (SR), we obtained a synthesized vectorcardiogram (VCG). Subsequently, various features were calculated from both signals, and age, sex, and medical history were included as baseline characteristics. Different machine learning models, including support vector machines, random forests (RF), and logistic regression, were finally tested with a combination of feature selection techniques.

RESULTS

The proposed method demonstrated excellent performance in the classification of AF gradation. With an optimized feature set of VCG and baseline features, the RF model achieved accuracy, sensitivity, and specificity of 83.02 %, 80.56 %, and 88.24 %, respectively, under the inter-patient paradigm.

CONCLUSION

Our results demonstrate the value of physiological signals in AF gradation evaluation, and VCG signals were effective in identifying mild and severe AF. Considering its low computational complexity and high assessment performance, the proposed model is expected to serve as a useful prognostic tool for clinical AF management.

What determines the optimal pharmacological treatment of atrial fibrillation? Insights from in silico trials in 800 virtual atria

The best pharmacological treatment for each atrial fibrillation (AF) patient is unclear. We aim to exploit AF simulations in 800 virtual atria to identify key patient characteristics that guide the optimal selection of anti-arrhythmic drugs. The virtual cohort considered variability in electrophysiology and low voltage areas (LVA) and was developed and validated against experimental and clinical data from ionic currents to ECG. AF sustained in 494 (62%) atria, with large inward rectifier K+ current (IK1 ) and Na+ /K+ pump (INaK ) densities (IK1 0.11 ± 0.03 vs. 0.07 ± 0.03 S mF-1 ; INaK 0.68 ± 0.15 vs. 0.38 ± 26 S mF-1 ; sustained vs. un-sustained AF). In severely remodelled left atrium, with LVA extensions of more than 40% in the posterior wall, higher IK1 (median density 0.12 ± 0.02 S mF-1 ) was required for AF maintenance, and rotors localized in healthy right atrium. For lower LVA extensions, rotors could also anchor to LVA, in atria presenting short refractoriness (median L-type Ca2+ current, ICaL , density 0.08 ± 0.03 S mF-1 ). This atrial refractoriness, modulated by ICaL and fast Na+ current (INa ), determined pharmacological treatment success for both small and large LVA. Vernakalant was effective in atria presenting long refractoriness (median ICaL density 0.13 ± 0.05 S mF-1 ). For short refractoriness, atria with high INa (median density 8.92 ± 2.59 S mF-1 ) responded more favourably to amiodarone than flecainide, and the opposite was found in atria with low INa (median density 5.33 ± 1.41 S mF-1 ). In silico drug trials in 800 human atria identify inward currents as critical for optimal stratification of AF patient to pharmacological treatment and, together with the left atrial LVA extension, for accurately phenotyping AF dynamics. KEY POINTS: Atrial fibrillation (AF) maintenance is facilitated by small L-type Ca2+ current (ICaL ) and large inward rectifier K+ current (IK1 ) and Na+ /K+ pump. In severely remodelled left atrium, with low voltage areas (LVA) covering more than 40% of the posterior wall, sustained AF requires higher IK1 and rotors localize in healthy right atrium. For lower LVA extensions, rotors can also anchor to LVA, if the atria present short refractoriness (low ICaL ) Vernakalant is effective in atria presenting long refractoriness (high ICaL ). For short refractoriness, atria with fast Na+ current (INa ) up-regulation respond more favourably to amiodarone than flecainide, and the opposite is found in atria with low INa . The inward currents (ICaL and INa ) are critical for optimal stratification of AF patient to pharmacological treatment and, together with the left atrial LVA extension, for accurately phenotyping AF dynamics.

A Survey on Physiological Signal-Based Emotion Recognition

Physiological signals are the most reliable form of signals for emotion recognition, as they cannot be controlled deliberately by the subject. Existing review papers on emotion recognition based on physiological signals surveyed only the regular steps involved in the workflow of emotion recognition such as pre-processing, feature extraction, and classification. While these are important steps, such steps are required for any signal processing application. Emotion recognition poses its own set of challenges that are very important to address for a robust system. Thus, to bridge the gap in the existing literature, in this paper, we review the effect of inter-subject data variance on emotion recognition, important data annotation techniques for emotion recognition and their comparison, data pre-processing techniques for each physiological signal, data splitting techniques for improving the generalization of emotion recognition models and different multimodal fusion techniques and their comparison. Finally, we discuss key challenges and future directions in this field.

In-silico drug trials for precision medicine in atrial fibrillation: From ionic mechanisms to electrocardiogram-based predictions in structurally-healthy human atria

Atrial fibrillation (AF) inducibility, sustainability and response to pharmacological treatment of individual patients are expected to be determined by their ionic current properties, especially in structurally-healthy atria. Mechanisms underlying AF and optimal cardioversion are however still unclear. In this study, in-silico drug trials were conducted using a population of human structurally-healthy atria models to 1) identify key ionic current properties determining AF inducibility, maintenance and pharmacological cardioversion, and 2) compare the prognostic value for predicting individual AF cardioversion of ionic current properties and electrocardiogram (ECG) metrics. In the population of structurally-healthy atria, 477 AF episodes were induced in ionic current profiles with both steep action potential duration (APD) restitution (eliciting APD alternans), and high excitability (enabling propagation at fast rates that transformed alternans into discordant). High excitability also favored 211 sustained AF episodes, so its decrease, through prolonged refractoriness, explained pharmacological cardioversion. In-silico trials over 200 AF episodes, 100 ionic profiles and 10 antiarrhythmic compounds were consistent with previous clinical trials, and identified optimal treatments for individual electrophysiological properties of the atria. Algorithms trained on 211 simulated AF episodes exhibited >70% accuracy in predictions of cardioversion for individual treatments using either ionic current profiles or ECG metrics. In structurally-healthy atria, AF inducibility and sustainability are enabled by discordant alternans, under high excitability and steep restitution conditions. Successful pharmacological cardioversion is predicted with 70% accuracy from either ionic or ECG properties, and it is optimal for treatments maximizing refractoriness (thus reducing excitability) for the given ionic current profile of the atria.

Machine learning based potentiating impacts of 12-lead ECG for classifying paroxysmal versus non-paroxysmal atrial fibrillation

Background

Conventional modality requires several days observation by Holter monitor to differentiate atrial fibrillation (AF) between Paroxysmal atrial fibrillation (PAF) and Non-paroxysmal atrial fibrillation (Non-PAF). Rapid and practical differentiating approach is needed.

Objective

To develop a machine learning model that observes 10-s of standard 12-lead electrocardiograph (ECG) for real-time classification of AF between PAF versus Non-PAF.

Methods

In this multicenter, retrospective cohort study, the model training and cross-validation was performed on a dataset consisting of 741 patients enrolled from Severance Hospital, South Korea. For cross-institutional validation, the trained model was applied to an independent data set of 600 patients enrolled from Ewha University Hospital, South Korea. Lasso regression was applied to develop the model.

Results

In the primary analysis, the Area Under the Receiver Operating Characteristic Curve (AUC) on the test set for the model that predicted AF subtype only using ECG was 0.72 (95% CI 0.65–0.80). In the secondary analysis, AUC only using baseline characteristics was 0.53 (95% CI 0.45–0.61), while the model that employed both baseline characteristics and ECG parameters was 0.72 (95% CI 0.65–0.80). Moreover, the model that incorporated baseline characteristics, ECG, and Echocardiographic parameters achieved an AUC of 0.76 (95% CI 0.678–0.855) on the test set.

Conclusions

Our machine learning model using ECG has potential for automatic differentiation of AF between PAF versus Non-PAF achieving high accuracy. The inclusion of  Echocardiographic parameters further increases model performance. Further studies are needed to clarify the next steps towards clinical translation of the proposed algorithm.

Respiratory Induced Modulation in f-Wave Characteristics During Atrial Fibrillation

The autonomic nervous system (ANS) is an important factor in cardiac arrhythmia, and information about ANS activity during atrial fibrillation (AF) may contribute to personalized treatment. In this study we aim to quantify respiratory modulation in the f-wave frequency trend from resting ECG. First, an f-wave signal is extracted from the ECG by QRST cancelation. Second, an f-wave model is fitted to the f-wave signal to obtain a high resolution f-wave frequency trend and an index for signal quality control ( S ). Third, respiratory modulation in the f-wave frequency trend is extracted by applying a narrow band-pass filter. The center frequency of the band-pass filter is determined by the respiration rate. Respiration rate is estimated from a surrogate respiration signal, obtained from the ECG using homomorphic filtering. Peak conditioned spectral averaging, where spectra of sufficient quality from different leads are averaged, is employed to obtain a robust estimate of the respiration rate. The envelope of the filtered f-wave frequency trend is used to quantify the magnitude of respiratory induced f-wave frequency modulation. The proposed methodology is evaluated using simulated f-wave signals obtained using a sinusoidal harmonic model. Results from simulated signals show that the magnitude of the respiratory modulation is accurately estimated, quantified by an error below 0.01 Hz, if the signal quality is sufficient ( S > 0 . 5 ). The proposed method was applied to analyze ECG data from eight pacemaker patients with permanent AF recorded at baseline, during controlled respiration, and during controlled respiration after injection of atropine, respectively. The magnitude of the respiratory induce f-wave frequency modulation was 0.15 ± 0.01, 0.18 ± 0.02, and 0.17 ± 0.03 Hz during baseline, controlled respiration, and post-atropine, respectively. Our results suggest that parasympathetic regulation affects the magnitude of respiratory induced f-wave frequency modulation.

A Review on the State of the Art in Atrial Fibrillation Detection Enabled by Machine Learning

Atrial Fibrillation (AF) the most commonly occurring type of cardiac arrhythmia is one of the main causes of morbidity and mortality worldwide. The timely diagnosis of AF is an equally important and challenging task because of its asymptomatic and episodic nature. In this paper, state-of-the-art ECG data-based machine learning models and signal processing techniques applied for auto diagnosis of AF are reviewed. Moreover, key biomarkers of AF on ECG and the common methods and equipment used for the collection of ECG data are discussed. Besides that, the modern wearable and implantable ECG sensing technologies used for gathering AF data are presented briefly. In the end, key challenges associated with the development of auto diagnosis solutions of AF are also highlighted. This is the first review paper of its kind that comprehensively presents a discussion on all these aspects related to AF auto-diagnosis in one place. It is observed that there is a dire need for low energy and low cost but accurate auto diagnosis solutions for the proactive management of AF.

Multi-scale Entropy Evaluates the Proarrhythmic Condition of Persistent Atrial Fibrillation Patients Predicting Early Failure of Electrical Cardioversion

Atrial fibrillation (AF) is nowadays the most common cardiac arrhythmia, being associated with an increase in cardiovascular mortality and morbidity. When AF lasts for more than seven days, it is classified as persistent AF and external interventions are required for its termination. A well-established alternative for that purpose is electrical cardioversion (ECV). While ECV is able to initially restore sinus rhythm (SR) in more than 90% of patients, rates of AF recurrence as high as 20-30% have been found after only a few weeks of follow-up. Hence, new methods for evaluating the proarrhythmic condition of a patient before the intervention can serve as efficient predictors about the high risk of early failure of ECV, thus facilitating optimal management of AF patients. Among the wide variety of predictors that have been proposed to date, those based on estimating organization of the fibrillatory (f-) waves from the surface electrocardiogram (ECG) have reported very promising results. However, the existing methods are based on traditional entropy measures, which only assess a single time scale and often are unable to fully characterize the dynamics generated by highly complex systems, such as the heart during AF. The present work then explores whether a multi-scale entropy (MSE) analysis of the f-waves may provide early prediction of AF recurrence after ECV. In addition to the common MSE, two improved versions have also been analyzed, composite MSE (CMSE) and refined MSE (RMSE). When analyzing 70 patients under ECV, of which 31 maintained SR and 39 relapsed to AF after a four week follow-up, the three methods provided similar performance. However, RMSE reported a slightly better discriminant ability of 86%, thus improving the other multi-scale-based outcomes by 3-9% and other previously proposed predictors of ECV by 15-30%. This outcome suggests that investigation of dynamics at large time scales yields novel insights about the underlying complex processes generating f-waves, which could provide individual proarrhythmic condition estimation, thus improving preoperative predictions of ECV early failure.

Changes in f-wave characteristics during cryoballoon catheter ablation

OBJECTIVE

Changes in ECG-derived parameters are studied in atrial fibrillation (AF) patients undergoing cryoballoon catheter ablation.

APPROACH

Parameters characterizing f-wave frequency, morphology by phase dispersion, and amplitude are estimated using a model-based statistical approach. These parameters are studied before, during, and after ablation, as well as for AF type (paroxysmal/persistent). Seventy-seven (49/28 paroxysmal/persistent) AF patients undergoing de novo catheter ablation are included in the study, out of which 31 (16/15 paroxysmal/persistent) were in AF during the whole procedure. A signal quality index (SQI) is used to identify analyzable segments.

MAIN RESULTS

f-wave frequency decreased significantly during ablation (p  =  0.001), in particular after ablation of the inferior right pulmonary vein (p  <  0.05). Frequency and phase dispersion differed significantly between paroxysmal and persistent AF (p  =  0.001 and p  <  0.05, respectively).

SIGNIFICANCE

This study demonstrates that a decrease in f-wave frequency can be distinguished during catheter ablation. The use of an SQI ensures reliable analysis and produces results significantly different from those obtained without an SQI.

Classification of persistent and long-standing persistent atrial fibrillation by means of surface electrocardiograms

Atrial fibrillation, which is the most common cardiac arrhythmia, is typically classified into four clinical subtypes: paroxysmal, persistent, long-standing persistent and permanent. The ability to distinguish between them is of crucial significance in choosing the most suitable therapy for each patient. Nevertheless, classification is currently established once the natural history of the arrhythmia has been disclosed as it is not possible to make an early differentiation. This paper presents a novel method to discriminate persistent and long-standing atrial fibrillation patients by means of a time-frequency analysis of the surface electrocardiogram. Classification results provide approximately 75% accuracy when evaluating ECGs of consecutive unselected patients from a tertiary center and higher than 80% when patients are not under antiarrhythmic treatment or do not have structural heart disease (76% sensitivity and 88% specificity). Moreover, to our knowledge, this is the first study that discriminates between persistent and long-standing persistent subtypes in a heterogeneous population sample and without discontinuing antiarrhythmic therapy to patients. Thus, it can help clinicians to address the most suitable therapeutic approach for each patient.

Preoperative study of the surface ECG for the prognosis of atrial fibrillation maze surgery outcome at discharge

The Cox-maze surgery is an effective procedure for terminating atrial fibrillation (AF) in patients requiring open-heart surgery associated with another heart disease. After the intervention, regardless of the patient's rhythm, all are treated with oral anticoagulants and antiarrhythmic drugs prior to discharge. Furthermore, patients maintaining AF before discharge could also be treated with electrical cardioversion (ECV). In view of this, a preoperative prognosis of the patient's rhythm at discharge would be helpful for optimizing drug therapy planning as well as for advancing ECV therapy. This work analyzes 30 preoperative electrocardiograms (ECGs) from patients suffering from AF in order to predict the Cox-maze surgery outcome at discharge. Two different characteristics of the AF pattern have been studied. On the one hand, the atrial activity (AA) organization, which provides information about the number of propagating wavelets in the atria, was investigated. AA organization has been successfully used in previous studies related to spontaneous reversion of paroxysmal AF and to the outcome of ECV. To assess organization, the dominant atrial frequency (DAF) and sample entropy (SampEn) have been computed. On the other hand, the second characteristic studied was the fibrillatory wave (f-wave) amplitude, which has been demonstrated to be a valuable indicator of the Cox-maze surgery outcome in previous studies. Moreover, this parameter has been obtained through a new methodology, based on computing the f-wave average power (fWP). Finally, all the computed indices were combined in a decision tree in order to improve prediction capability. Results for the DAF yielded a sensitivity (Se), a specificity (Sp) and an accuracy (Acc) of 61.54%, 82.35% and 73.33%, respectively. For SampEn the values were 69.23%, 76.00% and 73.33%, respectively, and for fWP they were 92.31%, 82.35% and 86.67%, respectively. Finally, the decision tree combining the three parameters analyzed improved the preoperative prognosis of the Cox-maze outcome with values of Se, Sp and Acc of 100%, 82.35% and 90%, respectively. As a consequence, the analysis of parameters related to the f-wave pattern, extracted from the preoperative ECG, has provided a considerable ability to predict the outcome of AF Cox-maze surgery at discharge.

RPCA-based detection and quantification of motion artifacts in ECG signals

In this paper, a recursive principal component analysis (RPCA)-based algorithm is applied for detecting and quantifying the motion artifact episodes encountered in an ECG signal. The motion artifact signal is synthesized by low-pass filtering a random noise signal with different spectral ranges of LPF (low pass filter): 0-5 Hz, 0-10 Hz, 0-15 Hz and 0-20 Hz. Further, the analysis of the algorithm is carried out for different values of SNR levels and forgetting factors (α) of an RPCA algorithm. The algorithm derives an error signal, wherever a motion artifact episode (noise) is present in the entire ECG signal with 100% accuracy. The RPCA error magnitude is almost zero for the clean signal portion and considerably high wherever the motion artifacts (noisy episodes) are encountered in the ECG signals. Further, the general trend of the algorithm is to produce a smaller magnitude of error for higher SNR (i.e. low level of noise) and vice versa. The quantification of the RPCA algorithm has been made by applying it over 25 ECG data-sets of different morphologies and genres with three different values of SNRs for each forgetting factor and for each of four spectral ranges.

Prevalence and Predisposing Factors of Atrial Fibrillation in a Multi-Ethnic Society: The Impact of Racial Differences in Bahrain

Background

The prevalence and epidemiological data of atrial fibrillation (AF) among multi-ethnic populations is less well studied worldwide.

Aim

Evaluation of the prevalence and predisposing factors of AF in patients who were admitted to acute medical emergencies (ER) in Bahrain over the period of one year.

Methods

Two hundred and fifty three patients with onset of AF were studied. The mean difference of biochemical data and clinical characteristics between Middle Eastern (ME) and sub continental (SC) patients was evaluated. The odds ratio of different predisposing factors for the development of clinical events in AF patients was assessed using multiple logistic regression analysis.

Results

Out of 7,450 patients that were admitted to ER over one year, 253 had AF based on twelve leads Electrocardiogram (ECG), with prevalence of 3.4%. In the whole study, the mean age was 59.45 ± 18.27 years, with 164 (65%) male. There were 150 ME patients (59%), and 107 (41%) SC, 55 (22%) were Indian (IND) and 48 (19%) were South Asian (SA). In the whole study clinical presentation was of 48% for palpitation, pulmonary edema was of 14%, angina pectoris on rest of 12%, 10% had embolic phenomena, 6% had dizziness, and 7% were asymptomatic. The odds ratio of different variables for occurrence of clinical events in the study was positive of 2.2 for history of hypertension, 1.8 for sickle cell disease, 1.2 for high body mass index (BMI) >30, 1.1 for mitral valve disease. The ME patients, compared with SC, were older, had significantly higher body mass index, higher history of rheumatic valve disease, sickle cell disease with high level of uric acid and lower hemoglobin. The history of hypertension, DM and smoking was higher among the SC patients. The rate of thyroid disease was equal in both groups.

Conclusion

The prevalence of atrial fibrillation was 3.4% with male predominance of 65%. Patients of sub continental origin were younger with a significantly high history of hypertension and ischemic heart disease. The patients of Middle Eastern origin had significantly high rate of rheumatic heart disease, and sickle cell disease. The history of hypertension was the most important independent clinical predictor of adverse events in patients presented with AF.