Catheter ablation is generally used to treat atrial fibrillation (AF), but the identification of correct ablation targets remains challenging in persistent AF. The work aimed at investigating an approach based on directed networks, that can allow the automatic detection of arrhythmia mechanisms and be useful for guiding the ablation strategy.
Using a realistic 3D model, sustained rotor activation of the atria was simulated numerically and mapped using virtual electrodes placed on the endocardial surface. Unipolar electrograms (EGMs) were computed at each of the electrodes, and networks were obtained by processing electrograms. Rotors were tracked using novel directed network mapping and compared with the electrical propagation map, which served as reference. To be able to properly map complex arrhythmia such as AF, we divided EGMs into multiple short time frames, and generated a specific network for each of them. In order to find the center of the rotor, we looked for the cycles with three nodes that were repeating in consecutive networks.
We detected a total of 97 cycles across all 58 networks. The most frequent cycle occurred in 19 networks, two occurred in 10 networks, two in 9 networks, three in 8 networks, three in 7 networks, seven in 6 networks, and the remaining 79 in 5 or less networks. The longest sequence of consecutive occurrences was relatively low (less than 5 networks) in all but two cycles that occurred in 9 and 10 consecutive networks. The cycle that occurred the most times (19 total where 10 of these in consecutive time frames) successfully located a stable rotor.
In this simulations study, we demonstrated that directed networks can be useful to represent electrical activity in the atria during AF. We showed that directed networks can detect rotors in an in-silico heart model.