A Novel Tool for the Identification and Characterization of Repetitive Patterns in High-Density Contact Mapping of Atrial Fibrillation

Abstract

Introduction: Electrical contact mapping provides a detailed view of conduction patterns in the atria during atrial fibrillation (AF). Identification of repetitive wave front propagation mechanisms potentially initiating or sustaining AF might provide more insights into temporal and spatial distribution of candidate AF mechanism and identify targets for catheter ablation. We developed a novel tool based on recurrence plots to automatically identify and characterize repetitive conduction patterns in high-density contact mapping of AF.

Materials and methods: Recurrence plots were constructed by first transforming atrial electrograms recorded by a multi-electrode array to activation-phase signals and then quantifying the degree of similarity between snapshots of the activation-phase in the electrode array. An AF cycle length dependent distance threshold was applied to discriminate between repetitive and non-repetitive snapshots. Intervals containing repetitive conduction patterns were detected in a recurrence plot as regions with a high recurrence rate. Intervals that contained similar repetitive patterns were then grouped into clusters. To demonstrate the ability to detect and quantify the incidence, duration and size of repetitive patterns, the tool was applied to left and right atrial recordings in a goat model of different duration of persistent AF [3 weeks AF (3 wkAF, n = 8) and 22 weeks AF (22 wkAF, n = 8)], using a 249-electrode mapping array (2.4 mm inter-electrode distance).

Results: Recurrence plots identified frequent recurrences of activation patterns in all recordings and indicated a strong correlation between recurrence plot threshold and AF cycle length. Prolonged AF duration was associated with shorter repetitive pattern duration [mean maximum duration 3 wkAF: 74 cycles, 95% confidence interval (54-94) vs. 22 wkAF: 41 cycles (21-62), p = 0.03], and smaller recurrent regions within repetitive patterns [3 wkAF 1.7 cm² (1.0-2.3) vs. 22 wkAF 0.5 cm² (0.0-1.2), p = 0.02]. Both breakthrough patterns and re-entry were identified as repetitive conduction patterns.

Conclusion: Recurrence plots provide a novel way to delineate high-density contact mapping of AF. Dominant repetitive conduction patterns were identified in a goat model of sustained AF. Application of the developed methodology using the new generation of multi-electrode catheters could identify additional targets for catheter ablation of AF.

Keywords: atrial fibrillation; mapping; mechanisms; recurrence plots; repetitive conduction patterns.

FIGURE 1

Recurrence plot construction. Block diagram and example of recurrence plot reconstruction for a recording of a paced rhythm. Unipolar electrograms were annotated, converted to activation-phase signals and used to create activation phase snapshots. Next, the distances δ_i,j between all pairs of snapshots (i,j) were used to construct a distance matrix. The distance threshold δ_max was computed based on the maximum recurrence rate per AF cycle RR_max (default RR_max = 1) and applied to the distance matrix to construct a recurrence plot. Consecutive recurrences were eroded, removing false positives from the recurrence plot. HD, high density; LA, left atrium; PCL, pacing cycle length.

FIGURE 2

Repetitive pattern detection. Intervals containing repetitive patterns were detected by traversing the main diagonal of the recurrence plot and computing the duration of the square block surrounding the diagonal with the maximum recurrence rate at each time point. Intervals with peak duration and a recurrence rate above a minimum recurrence rate per AF cycle RR_min (0.9) were selected. In the four detected intervals the average activation time map was constructed, representing the original four pacing directions. For each pattern a heat map was constructed that indicated the average recurrent activation-phase distance ${\mathrm{\delta }}_{\text{k}}^{\text{p}}$ for a pattern (p) for each electrode (k) separately. From this heat map the pattern size was computed by determining the area of the electrodes with ${\mathrm{\delta }}_{\text{k}}^{\text{p}}<{\mathrm{\delta }}_{\text{max}}$. In this example of a paced rhythm the repetitive patterns were very consistent: almost all electrodes contributed to the repetitive pattern with an average activation-phase distance below the distance threshold. HD, high density; LA, left atrium; PCL, pacing cycle length.

FIGURE 3

Example of repetitive pattern detection in simultaneous left (LA) and right (RA) atrial recordings. For both locations the recurrence plot and interval detection (red blocks) are depicted. Those intervals are shown that belong to the three clustered patterns with longest duration (in cycles). Cross-recurrence between intervals belonging to the same cluster is indicated in blue in the recurrence plot. The average activation pattern for each pattern is illustrated, together with the heat map of the average activation-phase distance per electrode.

FIGURE 4

Recurrence detection results in 3 and 22 wkAF goats. (A) Correlation between AF cycle length (AFCL) and the adaptive distance threshold δ_max, and maximum pattern duration per recording in AF cycles. (B) Variety of number of detected intervals containing a repetitive interval and clustered patterns compared to the total recording coverage. Only clustered patterns with a combined duration exceeding 10 AF cycles were included. (C) Average size of the most recurrent pattern region within each recording, computed using the adaptive distance threshold δ_max (left) and a fixed maximum time difference Δt (10 ms) applied to the average activation time difference between pattern recurrences (right).

FIGURE 5

Examples of AF mechanisms detected using recurrence analysis. Two conduction patterns associated with candidate AF mechanisms are depicted: an intermittent, unstable local re-entry within the mapping area (left) and a repetitive focal/breakthrough wave (right). Intervals containing recurrent patterns are indicated in red blocks, while cross-recurrences between clustered intervals are indicated in blue. The average activation pattern for each pattern is illustrated, together with the heat map of the average activation-phase distance per electrode and the average conduction direction during the pattern duration.