Minimizing discordances in automated classification of fractionated electrograms in human persistent atrial fibrillation

Identification of atrial fibrillation drivers by means of concentric ring electrodes

BACKGROUND AND OBJECTIVE

The prevalence of atrial fibrillation (AF) has tripled in the last 50 years due to population aging. High-frequency (DFdriver) activated atrial regions lead the activation of the rest of the atria, disrupting the propagation wavefront. Fourier based spectral analysis of body surface potential maps have been proposed for DFdriver identification, although these approaches present serious drawbacks due to their limited spectral resolution for short AF epochs and the blurring effect of the volume conductor. Laplacian signals (BC-ECG) from bipolar concentric ring electrodes (CRE) have been shown to outperform the spatial resolution achieved with conventional unipolar recordings. Our aimed was to determine the best DFdriver estimator in endocardial electrograms and to assess the BC-ECG capacity of CRE to quantify AF activity non-invasively.

METHODS

31 AF episodes were simulated using realistic tridimensional models of the atria electrical activity and torso. Periodogram and autoregressive (AR) spectral estimators were computed and the percentile (P90^th, P95^th and P98^th) to impose on the dominant frequencies (DFs) across whole atria to define the best DFdriver estimator evaluated. The identification of DFdriver on DFs from BC-ECG and unipolar surface signals with conventional disc electrodes was compared.

RESULTS

The best DFdriver estimator was P95^th and AR order 100. BC-ECG signals allowed better detection of AF activity than unipolar signals, with a significantly greater percentage of electrode locations in which DFdriver was identified (p-value 0.0095).

CONCLUSIONS

The use of BC-ECG signals for body surface Laplacian potential mapping with CRE could be helpful for better AF diagnosis, prognosis and ablation procedures than those with conventional disk electrodes.

Restitution Slope Affects the Outcome of Dominant Frequency Ablation in Persistent Atrial Fibrillation: CUVIA-AF2 Post-Hoc Analysis Based on Computational Modeling Study

Introduction

Although the dominant frequency (DF) localizes the reentrant drivers and the maximal slope of the action potential duration (APD) restitution curve (Smax) reflects the tendency of the wave-break, their interaction has never been studied. We hypothesized that DF ablation has different effects on atrial fibrillation (AF) depending on Smax.

Methods

We studied the DF and Smax in 25 realistic human persistent AF model samples (68% male, 60 ± 10 years old). Virtual AF was induced by ramp pacing measuring Smax, followed by spatiotemporal DF evaluation for 34 s. We assessed the DF ablation effect depending on Smax in both computational modeling and a previous clinical trial, CUVIA-AF (170 patients with persistent AF, 70.6% male, 60 ± 11 years old).

Results

Mean DF had an inverse relationship with Smax regardless of AF acquisition timing (p < 0.001). Virtual DF ablations increased the defragmentation rate compared to pulmonary vein isolation (PVI) alone (p = 0.015), especially at Smax <1 (61.5 vs. 7.7%, p = 0.011). In post-DF ablation defragmentation episodes, DF was significantly higher (p = 0.002), and Smax was lower (p = 0.003) than in episodes without defragmentation. In the post-hoc analysis of CUVIA-AF2, we replicated the inverse relationship between Smax and DF (r = −0.47, p < 0.001), and we observed better rhythm outcomes of clinical DF ablations in addition to a PVI than of empirical PVI at Smax <1 [hazard ratio 0.45, 95% CI (0.22–0.89), p = 0.022; log-rank p = 0.021] but not at ≥ 1 (log-rank p = 0.177).

Conclusion

We found an inverse relationship between DF and Smax and the outcome of DF ablation after PVI was superior at the condition with Smax <1 in both in-silico and clinical trials.

Unsupervised Classification of Atrial Electrograms for Electroanatomic Mapping of Human Persistent Atrial Fibrillation

OBJECTIVE

Ablation treatment for persistent atrial fibrillation (persAF) remains challenging due to the absence of a 'ground truth' for atrial substrate characterization and the presence of multiple mechanisms driving the arrhythmia. We implemented an unsupervised classification to identify clusters of atrial electrograms (AEGs) with similar patterns, which were then validated by AEG-derived markers.

METHODS

956 bipolar AEGs were collected from 11 persAF patients. CARTO variables (Biosense Webster; ICL, ACI and SCI) were used to create a 3D space, and subsequently used to perform an unsupervised classification with k-means. The characteristics of the identified groups were investigated using nine AEG-derived markers: sample entropy (SampEn), dominant frequency, organization index (OI), determinism, laminarity, recurrence rate (RR), peak-to-peak (PP) amplitude, cycle length (CL), and wave similarity (WS).

RESULTS

Five AEG classes with distinct characteristics were identified (F = 582, P<0.0001). The presence of fractionation increased from class 1 to 5, as reflected by the nine markers. Class 1 (25%) included organized AEGs with high WS, determinism, laminarity, and RR, and low SampEn. Class 5 (20%) comprised fractionated AEGs with in low WS, OI, determinism, laminarity, and RR, and in high SampEn. Classes 2 (12%), 3 (13%) and 4 (30%) suggested different degrees of AEG organization.

CONCLUSIONS

Our results expand and reinterpret the criteria used for automated AEG classification. The nine markers highlighted electrophysiological differences among the five classes found by the k-means, which could provide a more complete characterization of persAF substrate during ablation target identification in future clinical studies.

Short-Time Estimation of Fractionation in Atrial Fibrillation with Coarse-Grained Correlation Dimension for Mapping the Atrial Substrate

Atrial fibrillation (AF) is currently the most common cardiac arrhythmia, with catheter ablation (CA) of the pulmonary veins (PV) being its first line therapy. Ablation of complex fractionated atrial electrograms (CFAEs) outside the PVs has demonstrated improved long-term results, but their identification requires a reliable electrogram (EGM) fractionation estimator. This study proposes a technique aimed to assist CA procedures under real-time settings. The method has been tested on three groups of recordings: Group 1 consisted of 24 highly representative EGMs, eight of each belonging to a different AF Type. Group 2 contained the entire dataset of 119 EGMs, whereas Group 3 contained 20 pseudo-real EGMs of the special Type IV AF. Coarse-grained correlation dimension (CGCD) was computed at epochs of 1 s duration, obtaining a classification accuracy of 100% in Group 1 and 84.0–85.7% in Group 2, using 10-fold cross-validation. The receiver operating characteristics (ROC) analysis for highly fractionated EGMs, showed 100% specificity and sensitivity in Group 1 and 87.5% specificity and 93.6% sensitivity in Group 2. In addition, 100% of the pseudo-real EGMs were correctly identified as Type IV AF. This method can consistently express the fractionation level of AF EGMs and provides better performance than previous works. Its ability to compute fractionation in short-time can agilely detect sudden changes of AF Types and could be used for mapping the atrial substrate, thus assisting CA procedures under real-time settings for atrial substrate modification.

Recurrence Quantification Analysis for Investigating Atrial Fibrillation Dynamics in a Heterogeneous Simulation Setup

The outcomes of ablation targeting either reentry activations or fractionated activity during persistent atrial fibrillation (AF) therapy remain suboptimal due to, among others, the intricate underlying AF dynamics. In the present work, we sought to investigate such AF dynamics in a heterogeneous simulation setup using recurrence quantification analysis (RQA). AF was simulated in a spherical model of the left atrium, from which 412 unipolar atrial electrograms (AEGs) were extracted (2 s duration; 5 mm spacing). The phase was calculated using the Hilbert transform, followed by the identification of points of singularity (PS). Three regions were defined according to the occurrence of PSs: 1) no rotors; 2) transient rotors and; 3) long-standing rotors. Bipolar AEGs (1114) were calculated from pairs of unipolar nodes and bandpass filtered (30-300 Hz). The CARTO criterion (Biosense Webster) was used for AEGs classification (normal vs. fractionated). RQA attributes were calculated from the filtered bipolar AEGs: determinism (DET); recurrence rate (RR); laminarity (LAM). Sample entropy (SampEn) and dominant frequency (DF) were also calculated from the AEGs. Regions with longstanding rotors have shown significantly lower RQA attributes and SampEn when compared to the other regions, suggesting a higher irregular behaviour (P≤0.01 for all cases). Normal and fractionated AEGs were found in all regions (respectively; Region 1: 387 vs. 15; Region 2: 221 vs. 13; Region 3: 415 vs. 63). Region 1 vs. Region 3 have shown significant differences in normal AEGs (P≤0.0001 for all RQA attributes and SampEn), and significant differences in fractionated AEGs for LAM, RR and SampEn (P=0.0071, P=0.0221 and P=0.0086, respectively). Our results suggest the co-existence of normal and fractionated AEGs within long-standing rotors. RQA has unveiled distinct dynamic patterns-irrespective of AEGs classification-related to regularity structures and their nonstationary behaviour in a rigorous deterministic context.

Ultra-rapid high-density mapping system with the phase singularity technique is feasible in identifying rotors and focal sources and predicting AF termination

INTRODUCTION

Phase singularity (PS) mapping provides additional insight into the AF mechanism and is accurate in identifying rotors. The study aimed to evaluate the feasibility of PS mapping in identifying AF rotors using data obtained from an automatic ultra-rapid high-resolution mapping system with a high-density mini-basket catheter.

METHODS

Twenty-three pigs underwent rapid right atrial (RA) pacing (RAP 480 bpm) for 5 weeks before the experiment. During AF, RA endocardial automatic continuous mappings with a mini-basket catheter were generated using an automatic ultra-rapid mapping system. Both fractionation mapping and waveform similarity measurements using a PS mapping algorithm were applied on the same recording signals to localize substrates maintaining AF.

RESULTS

Seventeen (74%) pigs developed sustained AF after RAP. Three were excluded because of periprocedural ventricular arrhythmia and corrupted digital data. RA fractionation maps were acquired with 6.17 ± 4.29 minutes mean acquisition time, 13768 ± 12698 acquisition points mapped during AF from 581 ± 387 beats. Fractionation mapping identified extensively distributed (66.7%) RA complex fractionated atrial electrogram (CFAE), whereas the nonlinear analysis identified high similarity index (SI > 0.7) parts in limited areas (23.7%). There was an average of 1.67 ± 0.87 SI sites with 0.43 ± 0.76 rotor/focal source/chamber. AF termination occurred in 11/16 (68.75%) AF events in 14 pigs during ablation targeting max CFAE. There was a higher incidence of rotor/focal source at AF termination sites compared with non-AF termination sites (54.5% vs 0%, P = 0.011).

CONCLUSIONS

The data obtained from ultra-rapid high-density automatic mapping is feasible and effective in identifying AF rotors/focal sources using PS technique, and those critical substrates were closely related to AF procedural termination.

Systematic differences of non-invasive dominant frequency estimation compared to invasive dominant frequency estimation in atrial fibrillation

Non-invasive analysis of atrial fibrillation (AF) using body surface mapping (BSM) has gained significant interest, with attempts at interpreting atrial spectro-temporal parameters from body surface signals. As these body surface signals could be affected by properties of the torso volume conductor, this interpretation is not always straightforward. This paper highlights the volume conductor effects and influences of the algorithm parameters for identifying the dominant frequency (DF) from cardiac signals collected simultaneously on the torso and atrial surface. Bi-atrial virtual electrograms (VEGMs) and BSMs were recorded simultaneously for 5 min from 10 patients undergoing ablation for persistent AF. Frequency analysis was performed on 4 s segments. DF was defined as the frequency with highest power between 4 and 10 Hz with and without applying organization index (OI) thresholds. The volume conductor effect was assessed by analyzing the highest DF (HDF) difference of each VEGM HDF against its BSM counterpart. Significant differences in HDF values between intra-cardiac and torso signals could be observed, independent of OI threshold. This difference increases with increasing endocardial HDF (BSM-VEGM median difference from -0.13 Hz for VEGM HDF at 6.25 ± 0.25 Hz to -4.24 Hz at 9.75 ± 0.25 Hz), thereby confirming the theory of the volume conductor effect in real-life situations. Applying an OI threshold strongly affected the BSM HDF area size and location and atrial HDF area location. These results suggest that volume conductor and measurement algorithm effects must be considered for appropriate clinical interpretation.

Characterization of human persistent atrial fibrillation electrograms using recurrence quantification analysis

Atrial fibrillation (AF) is regarded as a complex arrhythmia, with one or more co-existing mechanisms, resulting in an intricate structure of atrial activations. Fractionated atrial electrograms (AEGs) were thought to represent arrhythmogenic tissue and hence have been suggested as targets for radiofrequency ablation. However, current methods for ablation target identification have resulted in suboptimal outcomes for persistent AF (persAF) treatment, possibly due to the complex spatiotemporal dynamics of these mechanisms. In the present work, we sought to characterize the dynamics of atrial tissue activations from AEGs collected during persAF using recurrence plots (RPs) and recurrence quantification analysis (RQA). 797 bipolar AEGs were collected from 18 persAF patients undergoing pulmonary vein isolation (PVI). Automated AEG classification (normal vs. fractionated) was performed using the CARTO criteria (Biosense Webster). For each AEG, RPs were evaluated in a phase space estimated following Takens' theorem. Seven RQA variables were obtained from the RPs: recurrence rate; determinism; average diagonal line length; Shannon entropy of diagonal length distribution; laminarity; trapping time; and Shannon entropy of vertical length distribution. The results show that the RQA variables were significantly affected by PVI, and that the variables were effective in discriminating normal vs. fractionated AEGs. Additionally, diagonal structures associated with deterministic behavior were still present in the RPs from fractionated AEGs, leading to a high residual determinism, which could be related to unstable periodic orbits and suggesting a possible chaotic behavior. Therefore, these results contribute to a nonlinear perspective of the spatiotemporal dynamics of persAF.

Sample Entropy Analysis of Noisy Atrial Electrograms during Atrial Fibrillation

Most cardiac arrhythmias can be classified as atrial flutter, focal atrial tachycardia, or atrial fibrillation. They have been usually treated using drugs, but catheter ablation has proven more effective. This is an invasive method devised to destroy the heart tissue that disturbs correct heart rhythm. In order to accurately localise the focus of this disturbance, the acquisition and processing of atrial electrograms form the usual mapping technique. They can be single potentials, double potentials, or complex fractionated atrial electrogram (CFAE) potentials, and last ones are the most effective targets for ablation. The electrophysiological substrate is then localised by a suitable signal processing method. Sample Entropy is a statistic scarcely applied to electrograms but can arguably become a powerful tool to analyse these time series, supported by its results in other similar biomedical applications. However, the lack of an analysis of its dependence on the perturbations usually found in electrogram data, such as missing samples or spikes, is even more marked. This paper applied SampEn to the segmentation between non-CFAE and CFAE records and assessed its class segmentation power loss at different levels of these perturbations. The results confirmed that SampEn was able to significantly distinguish between non-CFAE and CFAE records, even under very unfavourable conditions, such as 50% of missing data or 10% of spikes.

Atrial Electrogram Fractionation Distribution before and after Pulmonary Vein Isolation in Human Persistent Atrial Fibrillation-A Retrospective Multivariate Statistical Analysis

Purpose: Complex fractionated atrial electrograms (CFAE)-guided ablation after pulmonary vein isolation (PVI) has been used for persistent atrial fibrillation (persAF) therapy. This strategy has shown suboptimal outcomes due to, among other factors, undetected changes in the atrial tissue following PVI. In the present work, we investigate CFAE distribution before and after PVI in patients with persAF using a multivariate statistical model.

Methods: 207 pairs of atrial electrograms (AEGs) were collected before and after PVI respectively, from corresponding LA regions in 18 persAF patients. Twelve attributes were measured from the AEGs, before and after PVI. Statistical models based on multivariate analysis of variance (MANOVA) and linear discriminant analysis (LDA) have been used to characterize the atrial regions and AEGs.

Results: PVI significantly reduced CFAEs in the LA (70 vs. 40%; P < 0.0001). Four types of LA regions were identified, based on the AEGs characteristics: (i) fractionated before PVI that remained fractionated after PVI (31% of the collected points); (ii) fractionated that converted to normal (39%); (iii) normal prior to PVI that became fractionated (9%) and; (iv) normal that remained normal (21%). Individually, the attributes failed to distinguish these LA regions, but multivariate statistical models were effective in their discrimination (P < 0.0001).

Conclusion: Our results have unveiled that there are LA regions resistant to PVI, while others are affected by it. Although, traditional methods were unable to identify these different regions, the proposed multivariate statistical model discriminated LA regions resistant to PVI from those affected by it without prior ablation information.

The temporal behavior and consistency of bipolar atrial electrograms in human persistent atrial fibrillation

The unstable temporal behavior of atrial electrical activity during persistent atrial fibrillation (persAF) might influence ablation target identification, which could explain the conflicting persAF ablation outcomes in previous studies. We sought to investigate the temporal behavior and consistency of atrial electrogram (AEG) fractionation using different segment lengths. Seven hundred ninety-seven bipolar AEGs were collected with three segment lengths (2.5, 5,and 8 s) from 18 patients undergoing persAF ablation. The AEGs with 8-s duration were divided into three 2.5-s consecutive segments. AEG fractionation classification was applied off-line to all cases following the CARTO criteria; 43% of the AEGs remained fractionated for the three consecutive AEG segments, while nearly 30% were temporally unstable. AEG classification within the consecutive segments had moderate correlation (segment 1 vs 2: Spearman’s correlation ρ = 0.74, kappa score κ = 0.62; segment 1 vs 3: ρ = 0.726, κ = 0.62; segment 2 vs 3: ρ = 0.75, κ = 0.68). AEG classifications were more similar between AEGs with 5 and 8 s (ρ = 0.96, κ = 0.87) than 2.5 versus 5 s (ρ = 0.93, κ = 0.84) and 2.5 versus 8 s (ρ = 0.90, κ = 0.78). Our results show that the CARTO criteria should be revisited and consider recording duration longer than 2.5 s for consistent ablation target identification in persAF.

An interactive platform to guide catheter ablation in human persistent atrial fibrillation using dominant frequency, organization and phase mapping

BACKGROUND AND OBJECTIVE

Optimal targets for persistent atrial fibrillation (persAF) ablation are still debated. Atrial regions hosting high dominant frequency (HDF) are believed to participate in the initiation and maintenance of persAF and hence are potential targets for ablation, while rotor ablation has shown promising initial results. Currently, no commercially available system offers the capability to automatically identify both these phenomena. This paper describes an integrated 3D software platform combining the mapping of both frequency spectrum and phase from atrial electrograms (AEGs) to help guide persAF ablation in clinical cardiac electrophysiological studies.

METHODS

30s of 2048 non-contact AEGs (EnSite Array, St. Jude Medical) were collected and analyzed per patient. After QRST removal, the AEGs were divided into 4s windows with a 50% overlap. Fast Fourier transform was used for DF identification. HDF areas were identified as the maximum DF to 0.25Hz below that, and their centers of gravity (CGs) were used to track their spatiotemporal movement. Spectral organization measurements were estimated. Hilbert transform was used to calculate instantaneous phase.

RESULTS

The system was successfully used to guide catheter ablation for 10 persAF patients. The mean processing time was 10.4 ± 1.5min, which is adequate comparing to the normal electrophysiological (EP) procedure time (120∼180min).

CONCLUSIONS

A customized software platform capable of measuring different forms of spatiotemporal AEG analysis was implemented and used in clinical environment to guide persAF ablation. The modular nature of the platform will help electrophysiological studies in understanding of the underlying AF mechanisms.

Does Additional Electrogram-Guided Ablation After Linear Ablation Reduce Recurrence After Catheter Ablation for Longstanding Persistent Atrial Fibrillation? A Prospective Randomized Study

Background

Although circumferential pulmonary vein isolation (CPVI) catheter ablation may not be sufficient for long‐standing persistent atrial fibrillation (L‐PeAF), it is not clear which ablation strategy is beneficial in addition to CPVI. We sought to investigate whether additional complex fractionated atrial electrogram (CFAE)‐guided ablation improves clinical outcomes in L‐PeAF patients who exhibit continuous atrial fibrillation (AF) after CPVI and linear ablation (Line).

Methods and Results

This study enrolled 137 L‐PeAF patients (71.4% male, 61.6±10.9 years old) who underwent radiofrequency catheter ablation. We conducted CPVI+Line based on the Dallas lesion set (posterior box+anterior line) after baseline CFAE mapping in all patients. If AF was defragmented (terminated or changed to atrial tachycardia), the procedure was stopped (AF‐Defrag group, n=29). If AF was maintained after CPVI+Line, we mapped the CFAE again and randomly assigned the patient to the CPVI+Line group (n=54) or the additional CFAE ablation group (CPVI+Line+CFAE group, n=54). L‐PeAF was defragmented during CPVI+Line in 21.2% of patients (29/137, AF‐Defrag group). The mean CFAE cycle length was prolonged (P<0.001), and CFAE area (CFAE cycle length <120 milliseconds) was reduced (P<0.001) after CPVI+Line in the remaining patients. Procedure time was longer in the CPVI+Line+CFAE group than the CPVI+Line group (P=0.023), but procedure‐related complication rates did not vary. During 22.3±13.2 months of follow‐up, the clinical recurrence rates were 17.2% in the AF‐Defrag group, 18.5% in the CPVI+Line group, and 32.1% in the CPVI+Line+CFAE group (log rank, P=0.166).

Conclusions

Although CPVI+Line reduces and localizes CFAE area, additional CFAE ablation after CPVI+Line does not improve the clinical outcomes of catheter ablation in patients with L‐PeAF.

Semi-supervised clustering of fractionated electrograms for electroanatomical atrial mapping

BACKGROUND

Electrogram-guided ablation procedures have been proposed as an alternative strategy consisting of either mapping and ablating focal sources or targeting complex fractionated electrograms in atrial fibrillation (AF). However, the incomplete understanding of the mechanism of AF makes difficult the decision of detecting the target sites. To date, feature extraction from electrograms is carried out mostly based on the time-domain morphology analysis and non-linear features. However, their combination has been reported to achieve better performance. Besides, most of the inferring approaches applied for identifying the levels of fractionation are supervised, which lack of an objective description of fractionation. This aspect complicates their application on EGM-guided ablation procedures.

METHODS

This work proposes a semi-supervised clustering method of four levels of fractionation. In particular, we make use of the spectral clustering that groups a set of widely used features extracted from atrial electrograms. We also introduce a new atrial-deflection-based feature to quantify the fractionated activity. Further, based on the sequential forward selection, we find the optimal subset that provides the highest performance in terms of the cluster validation. The method is tested on external validation of a labeled database. The generalization ability of the proposed training approach is tested to aid semi-supervised learning on unlabeled dataset associated with anatomical information recorded from three patients.

RESULTS

A joint set of four extracted features, based on two time-domain morphology analysis and two non-linear dynamics, are selected. To discriminate between four considered levels of fractionation, validation on a labeled database performs a suitable accuracy (77.6 %). Results show a congruence value of internal validation index among tested patients that is enough to reconstruct the patterns over the atria to located critical sites with the benefit of avoiding previous manual classification of AF types.

CONCLUSIONS

To the best knowledge of the authors, this is the first work reporting semi-supervised clustering for distinguishing patterns in fractionated electrograms. The proposed methodology provides high performance for the detection of unknown patterns associated with critical EGM morphologies. Particularly, obtained results of semi-supervised training show the advantage of demanding fewer labeled data and less training time without significantly compromising accuracy. This paper introduces a new method, providing an objective scheme that enables electro-physiologist to recognize the diverse EGM morphologies reliably.