Conduction patterns in the coronary sinus during atrial fibrillation and their modification by pulmonary vein isolation

Abstract

mathematical AV node model for which the parameters can be estimated from short-term analysis of the surface electrocardiogram. Methods: Atrial impulses are assumed to arrive to the AV node according to a Poisson process with a mean arrival rate. Each atrial impulse arriving at the AV node will result in a ventricular contraction, unless the atrial impulse is blocked. The atrial impulses are blocked at the AV node according to a timedependent probability, modeling AV nodal refractory period and concealed conduction. An atrial impulse arriving close in time to the previous ventricular contraction is more likely to be blocked; all atrial impulses arriving before the end of the AV nodal refractory period are blocked. Two different refractory periods (tau1 and tau2), corresponding to dual AV nodal paths, as well as the probability of an atrial impulse choosing either of these, are used in the model. The mean arrival rate is estimated by the AF frequency, obtained from the atrial activity of the surface electrocardiogram, and the shortest refractory period (tau1) is estimated from the lower envelope of the Poincare plot of the RR series. The other parameters characterizing the AV model are obtained from the series by means of maximum likelihood estimation. The output of the AV model is the probability density function (PDF) of the time for which the ventricular activations occur. The average absolute error between the normalized RR histogram and the estimated PDF is computed for bins of 20-millisecond size, spaced between 0 and 2 seconds. In addition, the absolute error between the mean and the variance of RR series and of the PDF is assessed. The model was tested on 115 recordings from patient with AF and congestive heart failure (NYHA II-III) enrolled in the MUerte Subita en Insuficiencia Cardiaca study. Results: The median error is 0.0021, and the 25th/75th percentile error is 0.0017/0.0026. We judged the maximum acceptable error to be 0.0025. Using this threshold, 81 (70%) of 115 recordings were well fitted by the model. The average error between the mean and the variance is 0.01 ± 0.02 and 0.01 ± 0.01, respectively. Conclusion: These preliminary results are encouraging because certain AV nodal properties can be noninvasively characterized by a set of statistical parameters with electrophysiological interpretation.