High-density mapping of the average complex interval helps localizing atrial fibrillation drivers and predicts catheter ablation outcomes.

Background: Persistent Atrial Fibrillation (PersAF) electrogram-based ablation is complex, and appropriate identification of atrial substrate is critical. Little is known regarding the value of the Average Complex Interval (ACI) feature for PersAF ablation. Objective: Using the evolution of AF complexity by sequentially computing AF dominant frequency (DF) along the ablation procedure, we sought to evaluate the value of ACI for discriminating active drivers (AD) from bystander zones (BZ), for predicting AF termination during ablation, and for predicting AF recurrence during follow-up. Methods: We included PersAF patients undergoing radiofrequency catheter ablation by pulmonary vein isolation and ablation of atrial substrate identified by Spatiotemporal Dispersion or Complex Fractionated Atrial Electrograms (>70% of recording). Operators were blinded to ACI measurement which was sought for each documented atrial substrate area. AF DF was measured by Independent Component Analysis on 1-minute 12-lead ECGs at baseline and after ablation of each atrial zone. AD were differentiated from BZ either by a significant decrease in DF (>10%), or by AF termination. Arrhythmia recurrence was monitored during follow-up. Results: We analyzed 159 atrial areas (129 treated by radiofrequency during AF) in 29 patients. ACI was shorter in AD than BZ (76.4 ± 13.6 vs. 86.6 ± 20.3 ms; p = 0.0055), and mean ACI of all substrate zones was shorter in patients for whom radiofrequency failed to terminate AF [71.3 (67.5-77.8) vs. 82.4 (74.4-98.5) ms; p = 0.0126]. ACI predicted AD [AUC 0.728 (0.629-0.826)]. An ACI < 70 ms was specific for predicting AD (Sp 0.831, Se 0.526), whereas areas with an ACI > 100 ms had almost no chances of being active in AF maintenance. AF recurrence was associated with more ACI zones with identical shortest value [3.5 (3-4) vs. 1 (0-1) zones; p = 0.021]. In multivariate analysis, ACI < 70 ms predicted AD [OR = 4.02 (1.49-10.84), p = 0.006] and mean ACI > 75 ms predicted AF termination [OR = 9.94 (1.14-86.7), p = 0.038]. Conclusion: ACI helps in identifying AF drivers, and is correlated with AF termination and AF recurrence during follow-up. It can help in establishing an ablation plan, by prioritizing ablation from the shortest to the longest ACI zone.

Fibrillatory Wave Amplitude Evolution during Persistent Atrial Fibrillation Ablation: Implications for Atrial Substrate and Fibrillation Complexity Assessment.

Background. Fibrillatory Wave Amplitude (FWA) has been described as a non-invasive marker of atrial fibrillation (AF) complexity, and it predicts catheter ablation outcome. However, the actual determinants of FWA remain incompletely understood. Objective. To assess the respective implications of anatomical atrial substrate and AF spectral characteristics for FWA. Methods. Persistent AF patients undergoing radiofrequency catheter ablation were included. FWA was measured on 1-min ECG by TQ concatenation in Lead I, V1, V2, and V5 at baseline and immediately before AF termination. FWA evolution during ablation was compared to that of AF dominant frequency (DF) measured by Independent Component Analysis on 12-lead ECG. FWA was compared to the extent of endocardial low-voltage areas (LVA I < 10%; II 10-20%; III 20-30%; IV > 30%), to the surface of healthy left atrial tissue, and to P-wave amplitude in sinus rhythm. The predictive value of FWA for AF recurrence during follow-up was assessed. Results. We included 29 patients. FWA remained stable along ablation procedure with comparable values at baseline and before AF termination (Lead I p = 0.54; V1 p = 0.858; V2 p = 0.215; V5 p = 0.14), whereas DF significantly decreased (5.67 ± 0.68 vs. 4.95 ± 0.58 Hz, p < 0.001). FWA was higher in LVA-I than in LVA-II, -III, and -IV in Lead I and V5 (p = 0.02 and p = 0.01). FWA in V5 was strongly correlated with the surface of healthy left atrial tissue (R = 0.786; p < 0.001). FWA showed moderate to strong correlation to P-wave amplitude in all leads. Finally, FWA did not predict AF recurrence after a follow-up of 23.3 ± 9.8 months. Conclusions. These findings suggest that FWA is unrelated to AF complexity but is mainly determined by the amount of viable atrial myocytes. Therefore, FWA should only be referred as a marker of atrial tissue pathology.

Data Augmentation for Automatic Identification of Spatiotemporal Dispersion Electrograms in Persistent Atrial Fibrillation Ablation Using Machine Learning.

Catheter ablation is increasingly used to treat atrial fibrillation (AF), the most common sustained cardiac arrhythmia encountered in clinical practice. A recent breakthrough finding in AF ablation consists in identifying ablation sites based on their spatiotemporal dispersion (STD). STD stands for a delay of the cardiac activation observed in intracardiac electrograms (EGMs) across contiguous leads. In practice, interventional cardiologists localize STD sites visually using the PentaRay multipolar mapping catheter. This work aims at automatically characterizing STD by classifying EGM data into STD vs. non STD groups using machine learning (ML) techniques. A dataset of 23082 multichannel EGM recordings acquired by the PentaRay coming from 16 persistent AF patients is included in this study. A major problem hampering the classification performance lies in the highly imbalanced dataset ratio. We suggest to tackle data imbalance using adapted data augmentation techniques including 1) undersampling 2) oversampling 3) lead shift 4) time reversing and 5) time shift. These tools are designed to preserve the integrity of the cardiac data and are validated by a partner cardiologist. They provide enhancement in classification performance in terms of sensitivity, which increases from 50% to 80% while maintaining accuracy and AUC around 90% with oversampling. Bootstrapping is applied to check the variability of the trained classifiers.Clinical relevance-The machine learning techniques developed in this contribution are expected to aid cardiologists in performing patient-tailored catheter ablation procedures for treating persistent AF.

Association between circadian Holter ECG changes and sudden cardiac death in patients with Chagas heart disease.

OBJECTIVE: Chagas disease (ChD) is a parasitic illness, largely spread over South America. ChD usually causes progressive myocardium damage, either by direct parasite action or through autoimmune response. Sudden cardiac death (SCD) is prevalent in the early disease stages, being associated with a high variety of ectopic cardiac beats. This study aims at applying heart rate variability (HRV) and heart rate turbulence (HRT) techniques over Holter electrocardiogram (ECG) records to investigate the association with SCD in Chagas heart disease (ChHD). APPROACH: From a retrospective evaluation of a local database, the Holter records from 78 outpatients (34 female) were divided into groups: SCD deaths (20) and alive patients (56). To consider circadian autonomic changes, the analysis was performed in three periods: (a) entire 24 h record, (b) 12 h daylight period, and (c) the remaining 12 h including night rest. Eight variables were extracted using HRV and HRT approaches from each record and analysed together with the left ventricular ejection fraction (LVEF) estimated by echocardiography. MAIN RESULTS: The set of parameters was reduced by both the forward- and backward-stepwise approach and classification was performed using the k-nearest neighbours method and a leave-one-out cross-validation in a set of ten bootstrap trials, where SCD data were randomly taken and repositioned to balance the groups. The best 24 h model predicted SCD with 89.9% ± 0.9% accuracy using three HRV variables. The use of 12 h segments increased the accuracy up to 91.0% ± 1.2% in a model with the standard deviation parameter measured during the day (SDNNday) and night (SDNNnight). Although considered as playing a major role in SCD, LVEF did not show an association with SCD in this sample. SIGNIFICANCE: The degree of HRV and its circadian changes are associated with SCD in ChHD patients.

Unsupervised Common Spatial Patterns.

The common spatial pattern (CSP) method is a dimensionality reduction technique widely used in brain-computer interface (BCI) systems. In the two-class CSP problem, training data are linearly projected onto directions maximizing or minimizing the variance ratio between the two classes. The present contribution proves that kurtosis maximization performs CSP in an unsupervised manner, i.e., with no need for labeled data, when the classes follow Gaussian or elliptically symmetric distributions. Numerical analyses on synthetic and real data validate these findings in various experimental conditions, and demonstrate the interest of the proposed unsupervised approach.

Respiratory Waveform Estimation From Multiple Accelerometers: An Optimal Sensor Number and Placement Analysis.

Respiratory patterns are commonly measured to monitor and diagnose cardiovascular, metabolic, and sleep disorders. Electronic devices such as masks used to record respiratory waveforms usually require medical staff support and obstruct the patients' breathing, causing discomfort. New techniques are being investigated to overcome such limitations. An emerging approach involves accelerometers to estimate the respiratory waveform based on chest motion. However, most of the existing techniques employ a single accelerometer placed on an arbitrary thorax position. The present work investigates the use and optimal placement of multiple accelerometers located on the thorax and the abdomen. The study population is composed of 30 healthy volunteers in three different postures. By means of a custom-made microcontrolled system, data are acquired from an array of ten accelerometers located on predefined positions and a pneumotachograph used as reference. The best sensor locations are identified by optimal linear reconstruction of the reference waveform from the accelerometer data in the minimum mean square error sense. The analysis shows that right-hand side locations contribute more often to optimal respiratory waveform estimates, a sound finding given that the right lung has a larger volume than the left lung. In addition, we show that the respiratory waveform can be blindly extracted from the recorded accelerometer data by means of independent component analysis. In conclusion, linear processing of multiple accelerometers in optimal positions can successfully recover respiratory information in clinical settings, where the use of masks may be contraindicated.

Time-varying assessment of heart rate variability parameters using respiratory information.

Analysis of heart rate variability (HRV) is commonly used for characterization of autonomic nervous system. As high frequency (HF, known as the respiratory-related) component of HR, overlaps with the typical low frequency (LF) band when the respiratory rate is low, a reference signal for HF variations would help in better discriminating the LF and HF components of HR. The present study proposes a model for time-varying separation of HRV components as well as estimation of HRV parameters using respiration information. An autoregressive moving average with exogenous input (ARMAX) model of HRV is considered with a parametrically modeled respiration signal as the input. The model parameters are estimated using smoothed extended Kalman filtering. Results for different synthetic data show that our proposed joint model outperforms the classical AR modeling in estimation of HRV parameters especially in the case of low respiration rate. In addition, the possibility of using pulse transit time (PTT) and the amplitude of photoplethysmogram (PPGamp) as surrogates of the input respiratory signal has been investigated. To this end, electrocardiogram (ECG), PPG and respiration have been recorded from 21 healthy subjects (10 males and 11 females, mean age 27.5 ± 4.1) during normal and deep respiration. Results show that indeed PTT and PPGamp offer good potential to be used as references for respiratory-related variations of HR, thus avoiding additional devices for recording respiration.

Spectral and spatiotemporal variability ECG parameters linked to catheter ablation outcome in persistent atrial fibrillation.

With the increasing prevalence of atrial fibrillation (AF), there is a strong clinical interest in determining whether a patient suffering from persistent AF will benefit from catheter ablation (CA) therapy at long term. This work presents several regression models based on noninvasive measures automatically computed from the standard 12-lead electrocardiogram (ECG) such as AF dominant frequency (DF), spectral concentration and spatiotemporal variability (STV). Sixty-two AF patients referred to CA were enrolled in this study. Forty-seven of them had no recurrence after CA during an average follow-up of 14 ± 8 months. The ECG features were extracted from an ECG recorded before the CA intervention and they were combined by means of logistic regression. The combination of DF and STV values from different precordial leads reached AUC = 0.939, outperforming the best results by using only one kind of features, such as DF (AUC = 0.801), and yielding a global accuracy of 93.5% for discriminating the best long-term responders to CA. These results point out the need to take into consideration the spatial variation of spectral ECG parameters to build predictive models dealing with AF.

On the Link Between L1-PCA and ICA.

Principal component analysis (PCA) based on L1-norm maximization is an emerging technique that has drawn growing interest in the signal processing and machine learning research communities, especially due to its robustness to outliers. The present work proves that L1-norm PCA can perform independent component analysis (ICA) under the whitening assumption. However, when the source probability distributions fulfil certain conditions, the L1-norm criterion needs to be minimized rather than maximized, which can be accomplished by simple modifications on existing optimal algorithms for L1-PCA. If the sources have symmetric distributions, we show in addition that L1-PCA is linked to kurtosis optimization. A number of numerical experiments illustrate the theoretical results and analyze the comparative performance of different algorithms for ICA via L1-PCA. Although our analysis is asymptotic in the sample size, this equivalence opens interesting new perspectives for performing ICA using optimal algorithms for L1-PCA with guaranteed global convergence while inheriting the increased robustness to outliers of the L1-norm criterion.

Non-invasive prediction of catheter ablation outcome in persistent atrial fibrillation by fibrillatory wave amplitude computation in multiple electrocardiogram leads.

BACKGROUND: Catheter ablation (CA) of persistent atrial fibrillation (AF) is challenging, and reported results are capable of improvement. A better patient selection for the procedure could enhance its success rate while avoiding the risks associated with ablation, especially for patients with low odds of favorable outcome. CA outcome can be predicted non-invasively by atrial fibrillatory wave (f-wave) amplitude, but previous works focused mostly on manual measures in single electrocardiogram (ECG) leads only. AIM: To assess the long-term prediction ability of f-wave amplitude when computed in multiple ECG leads. METHODS: Sixty-two patients with persistent AF (52 men; mean age 61.5±10.4years) referred for CA were enrolled. A standard 1-minute 12-lead ECG was acquired before the ablation procedure for each patient. F-wave amplitudes in different ECG leads were computed by a non-invasive signal processing algorithm, and combined into a mutivariate prediction model based on logistic regression. RESULTS: During an average follow-up of 13.9±8.3months, 47 patients had no AF recurrence after ablation. A lead selection approach relying on the Wald index pointed to I, V1, V2 and V5 as the most relevant ECG leads to predict jointly CA outcome using f-wave amplitudes, reaching an area under the curve of 0.854, and improving on single-lead amplitude-based predictors. CONCLUSION: Analysing the f-wave amplitude in several ECG leads simultaneously can significantly improve CA long-term outcome prediction in persistent AF compared with predictors based on single-lead measures.

Atrial signal extraction in atrial fibrillation electrocardiograms using a tensor decomposition approach.

Atrial fibrillation (AF) is the most common cardiac arrhythmia encountered in clinical practice and remains a major challenge in cardiology. The noninvasive analysis of AF usually requires the estimation of the atrial activity (AA) signal in surface electrocardiogram (ECG) recordings. The present contribution puts forward a tensor decomposition approach for noninvasive AA extraction in AF ECG recordings. As opposed to the matrix approach, tensor decompositions are generally unique under mild conditions and have the potential to perform source separation in scenarios with a limited number of electrodes. An experimental study on a synthethic signal model and a real AF ECG recording evaluates the performance of the so-called block term tensor decomposition approach as compared to matrix techniques such as principal component analysis and independent component analysis.

Empirical mode decomposition of multiple ECG leads for catheter ablation long-term outcome prediction in persistent atrial fibrillation.

Predictive models arouse increasing interest in clinical practice, not only to improve successful intervention rates but also to extract information of diverse physiological disorders. This is the case of persistent atrial fibrillation (AF), the most common cardiac arrhythmia in adults. Currently, catheter ablation (CA) is one of the preferred therapies to face this disease. However, selecting the best responders to CA by standard noninvasive techniques such as the electrocardiogram (ECG) remains a challenge. This work presents different predictive models for determining long-term CA outcome based on the dominant frequency (DF) of atrial activity measured in the ECG. The ensemble empirical mode decomposition (EEMD) is employed to obtain the intrinsic mode functions (IMFs) composing the ECG signal in each lead. The IMF DFs computed in multiple leads are then combined into a logistic regression (LR) model. The IMF DF features are discriminant enough to reach 79% accuracy for long-term CA outcome prediction, outperforming other methods based on DF computation. Our study shows EEMD as a valuable alternative to extract clinically relevant spectral information from AF ECGs and confirms the advantage of LR to build multivariate predictive models as compared with univariate analysis.

Noninvasive prediction of catheter ablation acute outcome in persistent atrial fibrillation based on logistic regression of ECG fibrillatory wave amplitude and spatio-temporal variability.

Catheter ablation (CA) is increasingly employed to treat persistent atrial fibrillation (AF), yet assessment of procedural AF termination is still a subject of debate in the medical community. This has motivated the development of different criteria based on the standard electrocardiogram (ECG) to characterize ablation immediate effectiveness. However, most of conventional descriptors are merely computed in one ECG lead, thus neglecting significant information provided by the other leads. The present study proposes a novel predictor of CA outcome by exploiting a subset of the 12 leads in the standard ECG. Our method predicts the need for electrical cardioversion subsequent to CA by suitably combining two sets of multilead features, namely, a measure of fibrillatory wave amplitude and an index of AF spatio-temporal variability per lead. These features are obtained on a reduced-rank approximation determined by principal component analysis emphasizing the highest-variance components in the multilead atrial activity signal, and are then combined by logistic regression. On a database of over 50 persistent AF patients, our method provides reliable predictive measures and proves more robust and informative than classical AF descriptors.

Exploiting intracardiac and surface recording modalities for atrial signal extraction in atrial fibrillation.

During atrial fibrillation (AF), atrial activity (AA) on the surface ECG consists of a pattern of quasi-periodic oscillations (f-waves), which are related to the electrical activation of the atrial substrate. However, to date no direct comparison between the extracted f-wave pattern in surface recordings and specific activation sites within the atria has been carried out. In the present study, one reference intracardiac modality consisting of a bipolar electrogram (EGM) recorded from the left atrial appendage (LAA) is exploited for the first time to guide the extraction of LAA electrical activity from standard 12-lead ECG recordings. A periodic component analysis (πCA) technique is employed for this task. The performance of the proposed multimodal extraction technique is compared to that obtained employing a noninvasive, fully blind approach, namely, independent component analysis (ICA). On a database of 31 AF patients, results suggest that the estimation of LAA activity is indeed possible, even though its contribution to the ECG total power is relatively low. Interestingly, ICA seems to provide a slightly better estimation of LAA activation rate, expressed in terms of dominant frequency (DF). On the other hand, the multimodal invasive approach performs better QRST complex suppression and provides AA waveforms with narrower spectra.

Spatial variability of the 12-lead surface ECG as a tool for noninvasive prediction of catheter ablation outcome in persistent atrial fibrillation.

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia encountered in clinical practice. Radiofrequency catheter ablation (CA) is increasingly employed to treat this disease, yet the selection of persistent AF patients who will benefit from this treatment remains a challenging task. Several parameters of the surface electrocardiogram (ECG) have been analyzed in previous works to predict AF termination by CA, such as fibrillatory wave (f-wave) amplitude. However, they are usually manually computed and only a subset of electrodes is inspected. In this study, a novel perspective of the role of f-wave amplitude as a potential noninvasive predictor of CA outcome is adopted by exploring ECG interlead spatial variability. An automatic procedure for atrial amplitude computation based on cubic Hermite interpolation is first proposed. To describe the global f-wave peak-to-peak amplitude distribution, signal contributions from multiple leads are then combined by condensing the most representative features of the atrial signal in a reduced-rank approximation based on principal component analysis (PCA). We show that exploiting ECG spatial diversity by means of this PCA-based multilead approach does not only increase the robustness to electrode selection, but also substantially improves the predictive power of the amplitude parameter.

Multidimensional characterization of fibrillatory wave amplitude on surface ECG to describe catheter ablation impact on persistent atrial fibrillation.

Radiofrequency catheter ablation (CA) is increasingly employed to treat persistent atrial fibrillation (AF). Nevertheless, its success is not always guaranteed, as selection of patients who could positively respond to this therapy does not rely on systematic criteria and still remains an open issue. Moreover, very little is known about the quantitative effects of this treatment over AF electrophysiology, so their quantitative evaluation is not a trivial task. In this contribution, ablation impact is quantified by a descriptor of fibrillatory wave (f-wave) amplitude, so far regarded as a predictor of short-term CA outcome. By means of principal component analysis (PCA), surface electrocardiogram (ECG) spatial diversity is exploited and contributions from all leads are combined to describe average f-wave peak-to-peak amplitude, whose value is automatically computed by an algorithm based on cubic spline interpolation. Our work demonstrates how CA influences f-wave amplitude during the procedure as quantified by ECG inter-lead spatial variability. In addition, we show how such variations depend on procedural outcome and the duration of the postoperative blanking period.

Predicting catheter ablation outcome in persistent atrial fibrillation using atrial dominant frequency and related spectral features.

Pre-procedural atrial fibrillation dominant frequency (AFDF) has been reported to play a role as a predictor of catheter ablation (CA) outcome for the treatment of persistent atrial fibrillation (AF). The present study analyzes some spectral features of the atrial signal aimed at evaluating the quality of surface AFDF estimation and discusses their predictive power. First, automated extraction of surface atrial activity (AA) on pre-procedural 12-lead ECG recordings is performed by means of an independent component analysis (ICA) method. AFDF is then estimated by means of short-time Fourier analysis of the extracted atrial sources and simultaneous endocardial electrograms (EGM) used as reference. On a database of 20 patients in persistent AF undergoing CA, AFDF does not appear to play a role as a predictor of CA outcome at follow-up, neither on ECG nor on EGM recordings. The quality of surface AFDF estimation is assessed by means of the correlation coefficient r between surface and EGM AFDF, as well as the spectral concentration (SC) of the estimated atrial signal. It is shown that the quality of surface AFDF estimation is significantly lower for non-terminating CA procedures, both in terms of r and SC. The latter, in particular, seems to play a significant role in distinguishing terminating from non-terminating CA procedures and therefore in the non-invasive prediction of CA outcome.

Topographic imaging of the atrial electrical activity during atrial fibrillation for the analysis of uniform distributions of the surface electrical potentials.

Atrial fibrillation (AF) is a progressive arrhythmia which causes time dependent impairing of the cardiac muscle. This makes that proper therapeutic interventions depend on the degree of AF progression, i.e., on the temporal decrease of the organization of the electrical patterns observed during AF. Standard effective treatments are still lacking nowadays, and this calls for suitable noninvasive analysis of AF. In this sense, an appropriate therapy relies on the knowledge of AF characteristics, as its degree of organization. To this purpose, fast and accurate imaging of cardiac electrical activity can be helpful. Relying on the results of previous work on noninvasive assessment of the complexity of AF, we put forward a method to obtain visual maps of the topographic projection of the main atrial activity (AA) component given by principal component analysis, which is shown to provide detailed information about AA potential pattern distributions on the body surface. Different AA potential pattern distributions can then be identified, depending on the underlying degree of AF organization. An automated way to assess AF organization degree is then proposed, based on topographic projections. Similarities with previous studies suggest its usefulness for determining uniform distributions in the activation patterns on the body surface.

Non-invasive prediction of catheter ablation outcome in persistent atrial fibrillation by exploiting the spatial diversity of surface ECG.

Atrial fibrillation (AF) is the most common cardiac arrhythmia encountered in clinical practice. Radiofrequency catheter ablation (CA) is becoming one of the most widely employed therapies. Yet selection of patients who will benefit from this treatment remains a challenging task. Previous works have examined several electrocardiogram (ECG) parameters as potential predictors of CA success, such as fibrillatory wave (f-wave) amplitude. However, they require a manual computation and consider only a subset of electrodes, so inter-lead spatial variability of the 12-lead ECG is not fully exploited. The present study puts forward an automatic procedure for f-wave amplitude computation to non-invasively predict CA outcome. An extension of this quantitative measure to the whole set of leads is also proposed, based on Principal Component Analysis (PCA). We show that exploiting the spatial diversity present in the surface ECG not only improves the robustness to electrode selection but also increases the predictive power of the amplitude parameter.