Cancellation of ventricular activity in endocavitary recordings during atrial fibrillation by Particle Swarm Optimization

The cancellation of ventricular activity (VA) from atrial electrogram (AEG) is commonly performed by template matching and subtraction (TMS): a running template, built by adaptive averaging of AEG segments in correspondence of QRS, is subtracted from AEG to uncover atrial activity (AA). In our approach, before subtraction, templates are modulated by a set of coefficients which are estimated by maximizing, via Multiple Particle Swarm Optimization (MPSO), a fitness function based on: 1) the energy of the estimated and measured AA; 2) the first derivative of the estimated and measured AA; 3) the similarity between the template and its modulated version. To validate the method, three datasets of 500 synthetic AEG were built. Each signal included background AA, localized AA and VA. We observed that TMS+MPSO provided better performances then TMS alone when the ratio of VA/AA amplitude is large (VA/AA ≥ 3), while the performances get closer when the ratio decreases.