Time-Averaged Wavefront Analysis Demonstrates Preferential Pathways of Atrial Fibrillation, Predicting Pulmonary Vein Isolation Acute Response

Figures

• 

  Figure 1. Schematic showing the steps in model construction, simulation, and post-processing. (A) Imaging data were segmented, and simulation meshes were created incorporating conductivity and ionic changes in areas of fibrotic tissue identified using LGE-MRI intensity values. Atrial fibrillation (AF) was simulated and analysed at a basket arrangement of points modelled based on clinical electrode locations. Pulmonary vein isolation (PVI) was applied after 5 s. (B) Models were generated for different combinations of anatomy and fibrosis distribution to build a virtual patient cohort of 100 models. (C) Simultaneous biatrial contact mapping was performed with two 64 electrode Constellation catheters.

• 

  Figure 2. Methodology used for calculating the average activation flow vector field from unipolar electrograms. (A) The normalised filtered derivative signal for each electrode on the basket catheter was displayed in a regular 9 × 8 grid. These signals were interpolated to a higher resolution grid, and optical flow vectors were calculated. The vectors were averaged over time for isolines of constant normalised filtered derivative signal (indicated by white and black dots; wavefront and wave back). Finally, streamlines were calculated from average optical flow vectors to visualise activation pathways. The normalised derivative, interpolate and isoline plots are snapshots representing the wavefronts at a given time in the recording, while the average activation flow and streamlines are calculated as an average across the 10 s recording. (B) Example PV activation flow calculation for a case with low PV activation flow. Vectors within the blue boxes were assessed, with red vectors tagged as PV antra to LA activation flow. (C) Example PV activation flow calculation for a case with high PV activation flow.

• 

  Figure 3. Validation using simulated data: flutter and focal activity. (A) Simulated atrial flutter with activation streamlines from the IVC to SVC along the septal wall and from the SVC to IVC along the lateral wall, to complete the re-entry circuit. Basket location, isopotential plots and the activation streamlines are shown. The wavefront propagation path direction is indicated by the green arrows. The orientation of the RA in each sub-figure was chosen to offer easier visualisation of the electrode locations and the wavefront propagation direction. (B) A simulated repetitive focal source manifests as a divergent activation streamline field. TV, tricuspid valve; SVC, superior vena cava; IVC, inferior vena cava; RAA, right atrial appendage; MV, mitral valve; LSPV, left superior pulmonary vein; LIPV, left inferior pulmonary vein; RIPV, right inferior pulmonary vein; RSPV, right superior pulmonary vein; LA, left atrium; RA, right atrium.

• 

  Figure 4. Algorithm validation using clinical atrial tachycardia data. (A) Clinical atrial tachycardia phase maps, exhibiting re-entry on the LA posterior wall and passive RA activation are shown on the left of the figure. Normalised filtered derivative signals interpolated to a regular grid are shown on the top right for a single time point to show LA re-entry. Streamlines constructed from the average optical flow activation field for this clinical atrial tachycardia case demonstrate a re-entrant activation pattern on the LA posterior wall, and regular RA activation starting at the RA septal wall as it is activated from the LA. (B) Four LA basket electrograms are shown at the grid locations indicated by boxes with E1 to E4 on the streamline map. Locations E1-E4 are sequentially activated, demonstrating the presence of reentry. TV, tricuspid valve; SVC, superior vena cava; IVC, inferior vena cava; RAA, right atrial appendage; MV, mitral valve; LSPV, left superior pulmonary vein; LIPV, left inferior pulmonary vein; RIPV, right inferior pulmonary vein; RSPV, right superior pulmonary vein; LA, left atrium; RA, right atrium.

• 

  Figure 5. Simulated data PV activation flow metric is significantly different for PVI responder and non-responder cases. Isopotential snapshots (left), streamlines (middle), and optical activation flow (right) plots for examples of (A) PVI responder and (B) PVI non-responder. PV activation flow was quantified as % of vectors pointing into the LA body in a 90° range within the PV regions (blue boxes). PV activation flow vectors are shown in red. (A) For this simulation, re-entry around the right PV drives AF pre-ablation and PVI terminates AF (an example of a responder), with a PV activation flow metric of 22.2%. (B) For this simulation, multiple drivers exist in the LA body pre-ablation, and PVI did not terminate AF (an example of a non-responder); the PV activation flow metric is 4.4%. (C) Simulated PV activation flow metric is significantly different pre-ablation for PVI responder and non-responder cases (median 21.1 vs. 14.1%, p = 0.018, Wilcoxon signed rank).

• 

  Figure 6. Clinical average vector fields demonstrate activation flow from the PV antra into the atrial body for PVI responder cases pre-ablation. Paths are shown for the left and right atria as streamlines with the magnitude dependent on the magnitude of the underlying average vector field, signifying how often a direction is repeated. The direction of the centre of each streamline is also indicated. Cases are arranged in order of decreasing PV activation flow metric: (A) 37.5%, (B) 28.8%, (C) 26.9%, (D) 16.4%, (E) 12.5%, (F) 10.6%, and (G) 5.8%.

• 

  Figure 7. Clinical average vector fields did not demonstrate activation flow from the PV antra into the atrial body for PVI non-responder cases pre-ablation. Visualisation details are as in Figure 6. Cases are arranged in order of increasing PV activation flow metric: (A) 0, (B) 1.9, (C) 2.9, (D) 4.5, (E) 6.7, and (F) 27.9%. For case (D), multiple splines of the LA basket were not in contact and so these were excluded from the analysis.

• 

  Figure 8. The clinical PV activation flow metric is significantly different for PVI responder and non-responder cases. (A) Clinical PV activation flow (%) pre-ablation was significantly higher for PVI responder than PVI non-responder cases (PVI responder: 16.3% vs. non-responder: 3.71%, p = 0.035). (B) LA cycle length was not significantly different between groups (Wilcoxon signed-rank). (C) RA cycle length was not significantly different between groups.

• 

  Figure 9. Effects of recording duration on streamlines for example AF episode. Streamlines are shown for 10 s recordings taken from intervals spread through the entire recording. The top row shows the left atrium, and the bottom row shows the right atrium. The streamline plots in (A) show the original interval; panels (C,D) are visually similar to (A), while the LA for (B) demonstrates larger differences. The PV activation flow metrics from the right PV are as follows: (A) 37.5%, (B) 96%, (C) 31.7%, (D) 38.5%. There were 9 out of 10 recordings that had no activation flow from the left PV to the LA body, while (B) has a PV activation flow metric from the left PV of 27.3%. The start times for these recordings are 0, 6, 125, and 187 s.

• 

  Figure 10. Effects of basket size on preferential pathways. Reducing the basket size changed the significance of the PV flow metric: larger basket: median 21.1 vs. 14.1%, p = 0.018, Wilcoxon signed-rank median; smaller basket: 24 vs. 17.7%, not significant.

• 

  Figure 11. Effects of basket contact on preferential pathways. Effects of poor contact on preferential pathway analysis. PV activation flow metrics are shown for cases with 10–50% of electrodes removed. (A) All electrodes: median 21.1 vs. 14.1%, p = 0.018, Wilcoxon signed rankmedian. (B) 90% of electrodes included: 17.9 vs. 11.9%, p = 0.07. (C) 75% of electrodes included: 16.7 vs. 10.3%, p = 0.09. (D) 50% of electrodes included: 6.4 vs. 10.7%, p = 0.49.

• 

  Figure 12. Simulating PV triggers vs. PV rotational drivers. (A) Isopotential plots, (B) average optical flow activation vector fields and (C) activation streamlines are shown for a case sustained by PV triggers (left) and for a PV rotational driver (right).