Progressive modification of rotors in persistent atrial fibrillation by stepwise linear ablation.

Introduction

While percutaneous pulmonary vein (PV) isolation has proved to be an effective treatment of paroxysmal atrial fibrillation (AF), radiofrequency ablation of persistent atrial fibrillation (PeAF) has been much less successful. However, improved outcomes are reported for catheter ablation of PeAF guided by real-time dominant frequency analysis and by dynamic 3-dimensional (3D) electroanatomic mapping,3, 4 particularly when this is used to target the origin and trajectory of rotors.5, 6 Here, we present a complete case study of a patient with PeAF who was treated successfully with stepwise linear ablation. A novel frequency analysis approach, in which a wavelet filter and multiscale peak detection methods are combined, was applied retrospectively to electrograms recorded during the procedure. This reveals progressive changes in the number, location, and form of reentrant sources during ablation.

Case report

The patient was a 58-year-old woman with a 7-year history of AF and no prior ablation and classified as having PeAF. A decapolar 6-F electrode catheter (St Jude Medical, St Paul, MN) was positioned in the coronary sinus.3, 4 A 9-F 64-channel electrode balloon was positioned in the middle of the left atrium (LA) via a transseptal approach with the J-tip wire lodged in the left superior PV. An 8-mm-tip deflectable ablation catheter (St Jude Medical) was then introduced into the LA via a second transseptal puncture. Intravenous heparin was administered to maintain an activated clotting time of 300–350 seconds after transseptal puncture. LA endocardial geometry was acquired, and a well-defined linear ablation procedure was followed.3, 4 This involved 2 ablation lesion sets (figure 7 and figure 0 lesions) delivered at upper limits of 60 W and 60°C. A figure 7 lesion set was delivered across the LA roof to the ridge between the left atrial appendage and the left superior PV and then downward to the anterior antrum of the left inferior PV. A second (figure 0) lesion set followed a path from the roof line along the anterolateral junction of the right PVs, and the circuit was closed by extending a line between the right inferior PV and the mitral valve (MV) annulus adjacent to the coronary sinus. Circumferential PV isolation was not performed.3, 4 Noncontact electroanatomic maps were recorded for ~10 seconds at regular specified intervals throughout the ablation process by using an EnSite 3000 system (St Jude Medical; sampling rate 1200 Hz; bandwidth 2–600 Hz; sensitivity 10 µV).

We have exported 2048 3D data cloud coordinates of LA geometry from the EnSite system and created the triangle mesh based on the data cloud by using custom-written MATLAB applications (MathWorks, Inc, Natick, MA). By using the EnSite geometrical data, we estimated that the superior-to-inferior depth of LA geometry was ~38.0 mm, the anterior-to-posterior width was ~41.7 mm, and the lateral-to-lateral length (excluding PV sleeves) was ~50.1 mm. The distance from anywhere in LA geometry excluding the PV sleeves to the center of the EnSite array was <40 mm owing to the position (along the septum-to-ridge direction of the LA) and dimension (45×18 mm2) of the EnSite array. Virtual electrograms from each LA geometry location point were analyzed and linked to their anatomical structure (Figure 1A). Frequency analysis was performed on each recorded electrogram, and key steps involved are illustrated in Figure 1B for a 1-second virtual electrogram segment (panel I) from the posterior inferior septum before ablation.7, 8 The timing and duration of ventricular components were determined, and a tapered attenuation function, which minimized the far-field ventricular signal and preserved low-frequency signal components, was applied within this window (panel II). A continuous wavelet transformation was then applied to the modified electrograms. The first derivative of a Gaussian function was used as the mother wavelet to identify negative deflections across 15 temporal scales (panel III). A filtered derivative signal was then constructed with weighted wavelet components across scales 9–15 (panel IV). Maxima adjacent to suprathreshold derivative gradients were immediately identified as local atrial activation (black circles in panel IV) when ≥70 ms apart. Where maxima satisfied the gradient criterion, but were <70 ms apart, the peak with the highest derivative gradient was identified as local activation and others as fractionation (open circles in panel IV). Regional atrial frequency was estimated from the local activation count during the recorded time interval except the time windows (~120 ms) in which ventricular depolarization occurred. Activation time maps were constructed by tracking the propagation of local activation complexes across the LA surface.

Figure 1

Three-dimensional endocardial surface of the left atrium using coordinates provided by the EnSite system (St Jude Medical) was reconstructed at the anterior view (A), and a typical atrial virtual electrogram was processed by the proposed signal processing tools (B). Wavelet-based peak atrial impulse detection (the far-field ventricular activation complex is indicated in grey). I: Typical virtual electrogram of the patient with persistent atrial fibrillation at the posterior inferior septum. II: The electrogram is attenuated during the ventricular activation time window. III: Wavelet scalogram of panel II, indicating the magnitude and duration of each of the 15 scales. White indicates a negative deflection and black a positive deflection. IV: Derivative signal constructed from 7 lowest wavelet scales. The maxima for which adjacent derivative gradients exceed a preset threshold are indicated by closed circles, with the peak-to-peak distance ≥70 ms. Where peaks are separated by <70 ms, the peak with the highest derivative gradient was identified as local activation (closed circles) and others as fractionation (open circles). ASS = anterior superior septum; LAA = left atrium appendage; LIPV = left inferior pulmonary vein; LSPV = left superior pulmonary vein; MV = mitral valve; PIS = persistent atrial fibrillation; RIPV = right inferior pulmonary vein; RSPV = right superior pulmonary vein.

Before catheter ablation, specific regions on the LA roof and interatrial septum were characterized by high-frequency activity (Figure 2A). Typically activity died out after a few cycles, but then recurred at the proximate regions. Figure 7 lesions altered this pattern, especially along the roof line (Figure 2B). High-frequency activity appeared to become more uniform and was also evident at the junction of the right superior PV and in the inferior septum. Partway through figure 0 lesions, a macroreentrant flutter circuit was established and the corresponding frequency map is shown in Figure 2C. The observation of 64 representative virtual electrograms confirmed a stable 2:1 atrioventricular rhythm with regions of high-amplitude fractionation in the roof and inferior septum indicative of local conduction delays. Activation was initiated in the anterior wall and spread slowly to the roof. Typical virtual electrograms from the site of initial activation (white circle) and the adjacent anterior septum are shown in Figure 2D.

Figure 2

Regional frequency maps rendered on the anterior LA wall immediately before (A) and during (B and C) linear ablation. Intraprocedural AF shown in images C and D presents 2-ms electrogram segments for the highest-frequency region in the inferior interatrial septum (left panel) and the site at which activation originates (right panel; indicated by the circle in image C). Abbreviations as in Figure 1.

We observed transient rotors before and during ablation. These were sustained for 2 cycles at most and originated near sites of high-frequency activity (in the LA roof, the interatrial septum, PV junctions, and the MV annulus). Figure 3A shows a rotor that started near the MV annulus and circulated clockwise on the anterior surface of the LA before ablation. A more short-lived rotor that originated in the interatrial septum after figure 7 lesions is presented in Figure 3B.

Figure 3

Activation time maps rendered on the anterior left atrial wall immediately before linear ablation (A) and after figure 7 lesions (B). These maps show typical reentrant activation patterns that originate adjacent to high-frequency regions. Note that Figures 3A and 3B correspond to Figures 2A and 2B, respectively. Abbreviations as in Figure 1.

Sinus rhythm was established with the completion of figure 0 lesions. Postprocedural AF occurred several seconds later, but sustained sinus rhythm was restored shortly soon after consolidating the “box set” of linear ablation lines. Mapping and ablation took 49 minutes, and the patient was free from AF at 12 months.

Discussion

3D electroanatomic mapping over extended periods is beneficial for an accurate dynamic analysis of AF. This is difficult to achieve with point-to-point contact mapping systems because electrical activity in AF is nonstationary. However, 3D mapping can be carried out using atrial electrograms recorded simultaneously with multielectrode intracardiac basket catheters. The CONFIRM trial5, 6 demonstrates the potential of this approach, but it requires electrodes to be in contact with the endocardial surface of the atria, which is not always possible. An alternative method is to deploy noncontact arrays and use inverse mapping to reconstruct 3D potential distributions.3, 4

In this case study, simultaneous virtual electrograms at multiple atria regions have been recorded at intervals throughout a successful PeAF ablation procedure. These were analyzed subsequently using novel wavelet-based signal analysis tools that enable regional activation times and frequencies to be determined. The findings reinforce the utility of real-time 3D electroanatomic mapping for identifying nonstationary reentrant sources during AF ablation and demonstrate that noncontact intracardiac mapping systems can be used for this purpose. A potential advantage of our approach is that it can be used to identify regions of unstable rotor activity.

A problem with virtual electrograms reconstructed with noncontact methods is that they are unipolar signals. They have greater spatiotemporal complexity than do bipolar contact recordings and can be more difficult to interpret, particularly during PeAF. We have adapted a wavelet-based technique reported by Houben et al to detect local activation times in the presence of far-field potentials. This has enabled us to track reentrant activation and to determine local activation frequency more directly than is possible with dominant frequency analysis.2, 10

In this work, we have demonstrated that originally and subsequently identified sources are located primarily in the atrial roof and septum. The first roof line ablation was ~10 mm from the closest high-frequency activity (Figure 2A). After the completion of figure 7 lesions, high-frequency activity in the roof was shifted by >20 mm from the roof line (Figure 2B) and disappeared after subsequent ablation. Figure 0 lesions (Figures 2A–2C) were ~10 mm away from the high-frequency region in the septum and sources in this region disappeared after the completion of lines, adjacent to the MV, atrial floor, and left lateral wall. In this patient, reentrant sources were associated with regions of high-frequency activity and these regions were also characterized by temporally dispersed fractionation during intra- and postprocedural flutter. Reentrant sources and focal firing were consistent with those reported elsewhere, but were not sustained continuously. We postulate that the linear ablation approach has worked here by isolating microreentrant sources and interrupting intra- and postprocedural macroentrant circuits rather than by targeting rotors directly.3, 4, 5, 6 Furthermore, we observed that occasionally only some of the rotors disappeared owing to collision with fibrillatory conduction and then reappeared at a nearby location. Nevertheless, this case indicates that electrical rotors can be nonstationary during AF ablation and highlights the need for real-time 3D global mapping systems and analysis tools that can track their trajectories throughout ablation procedures. The present clinical practice is unique in that linear ablation was performed as a primary therapy rather than as an adjunct to PV isolation. Our clinical study suggests that the linear ablation approach without PV isolation could achieve high ablation outcomes and reduce the probability of PV stenosis or atrial-esophageal fistula since neither PVs nor the posterior wall was targeted.3, 4 Also, the systematic approach used here resulted in termination of AF using less lesion lines, potentially simplifying the procedure.

There are some limitations of this study. First, atrial electrograms recorded in the LA from 1 patient were analyzed; in the future, we will include more patients and biatrial data. Second, only activation maps were used instead of more robust phase analysis to characterize rotors, even though the work by Narayan et al concluded that both approaches led to similar qualitative results. Furthermore, in this study we directly processed unipolar electrograms which is a challenge to apply the existing phase analysis approach due to specific morphology of the atrial signals. Lastly, it is known that virtual electrograms in AF exhibit morphological artifact when compared with contact bipolar electrograms, particularly in some regions of the larger atria where the distance from the center of the mapping probe is >40 mm.

KEY TEACHING POINTS

• A novel frequency analysis approach is demonstrated to effectively process atrial unipolar electrograms acquired by a noncontact balloon. • Frequency analysis reveals progressive changes in the number, location, and form of reentrant sources during linear ablation. • 3D panoramic electroanatomic mapping is crucial for accurate dynamic description of atrial fibrillation and guidance of catheter ablation.