Combining atrial activation maps for the big picture of simulated atrial fibrillation mechanisms

Abstract

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – EU funding. Main funding source(s): This work was supported by PersonalizeAF project. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 860974. Background Pulmonary vein isolation (PVI) alone is not sufficient for all persistent atrial fibrillation (AF) patients. Ablation of localized AF drivers might be a promising adjunctive ablation strategy but detecting such drivers is limited by the small field of view of current mapping catheters. Purpose We propose a technique to detect localized AF drivers that combines sequential mapping of repetitive activation patterns into a larger composite map. The performance of this technique was evaluated in several simulated scenarios of known mechanisms during AF. Methods Representative repetitive conduction patterns that may occur in AF were simulated in a detailed volumetric 3D model of the atria: a large peripheral wave, stable and meandering re-entries, and single and multiple ectopic foci. Endocardial electrograms were sequentially generated from overlapping regions in the left posterior wall (4x4 grid, 3mm spacing, 0.5-1.5 seconds), and repetitive activation patterns were detected. Overlapping areas between sequential maps were used to align the repetitive patterns. Local patterns were then combined into larger composite maps (Fig. 1A). Reconstruction performance was quantified by the average local activation time (LAT) difference between original and reconstructed activation patterns. The relation between electrode overlaps and the quality of reconstructions were assessed at different degrees of spatial overlap (approximately 10% to 25% of the grid area, corresponding to 1.6 ± 0.6 and 4.1 ± 1.2 overlapping electrodes, respectively). Results All simulated patterns could be reconstructed from sequential recordings (Fig. 1B). LAT errors in recordings with 25% overlap were low in all cases: 5±2 msec for the peripheral wave, 5±2 and 3±1 msec for focal and multifocal activities and 6±2 and 11±2 & 17±3 msec for stable, meandering and highly meandering re-entries respectively (Fig. 2). Composite maps of meandering and highly meandering re-entries presented significantly higher reconstruction errors than other more stable patterns (p<0.01). Reducing spatial overlap to 10% did not significantly increase the errors, except for the highly meandering re-entrant activity (p<0.05). Conclusion We developed a novel methodology to combine high-density sequential recordings with limited coverage into larger composite activation maps. Using this method, we reconstructed highly accurate activation maps for multiple simulated AF mechanism scenarios. Results were promising even with a small overlap between sequential recording sites and relatively unstable patterns, providing a proof of principle to apply this approach in clinical recordings of human AF.