Spatial Characterization and Estimation of Intracardiac Propagation Patterns During Atrial Fibrillation


Abstract: This doctoral thesis is in the field of biomedical signal processing with focus on methods for the analysis of atrial fibrillation (AF). Paper I of the present thesis addresses the challenge of extracting spatial properties of AF from body surface signals. Different parameters are extracted to estimate the preferred direction of atrial activation and the complexity of the atrial activation pattern. In addition, the relation of the spatial properties to AF organization, which is quantified by AF frequency, is evaluated. While no significant correlation between the preferred direction of atrial activation and AF frequency could be observed, the complexity of the atrial activation pattern was found to increase with AF frequency. The remaining three papers deal with the analysis of the propagation of the electrical activity in the atria during AF based on intracardiac signals. In Paper II, a time-domain method to quantify propagation patterns along a linear catheter based on the detected atrial activation times is developed. Taking aspects on intra-atrial signal organization into account, the detected activation times are combined into wavefronts, and parameters related to the consistency of the wavefronts over time and the activation order along the catheter are extracted. Furthermore, the potential relationship of the extracted parameters to established measures from body surface signals is investigated. While the degree of wavefront consistency was not reflected by the applied body surface measures, AF frequency could distinguish between recordings with different degrees of intra-atrial signal organization. This supports the role of AF frequency as an organization measure of AF. In Paper III, a novel method to analyze intracardiac propagation patterns based on causality analysis in the frequency domain is introduced. In particular, the approach is based on the partial directed coherence (PDC), which evaluates directional coupling between multiple signals in the frequency domain. The potential of the method is illustrated with simulation scenarios based on a detailed ionic model of the human atrial cell as well as with real data recordings, selected to present typical propagation mechanisms and recording situations in atrial tachyarrhythmias. For simulated data, the PDC is correctly reflecting the direction of coupling and thus the propagation between all recording sites. For real data, clear propagation patterns are identified which agree with previous clinical observations. Thus, the results illustrate the ability of the novel approach to identify propagation patterns from intracardiac signals during AF which can provide important information about the underlying AF mechanisms, potentially improving the planning and outcome of ablation. However, spurious couplings over long distances can be observed when analyzing real data comprised by a large number of simultaneously recorded signals, which gives room for further improvement of the method. The derivation of the PDC is entirely based on the fit of a multivariate autoregressive (MVAR) model, commonly estimated by the least-squares (LS) method. In Paper IV, the adaptive group least absolute selection and shrinkage operator (LASSO) is introduced in order to avoid overfitting of the MVAR model and to incorporate prior information such as sparsity of the solution. The sparsity can be motivated by the observation that direct couplings over longer distances are likely to be zero during AF; an information which has been further incorporated by proposing distance-adaptive group LASSO. In simulations, adaptive and distance-adaptive group LASSO are found to be superior to LS estimation in terms of both detection and estimation accuracy. In addition, the results of both simulations and real data analysis indicate that further improvements can be achieved when the distance between the recording sites is known or can be estimated. This further promotes the PDC as a method for analysis of AF propagation patterns, which may contribute to a better understanding of AF mechanisms as well as improved AF treatment.