Abstract:
In order to understand the mechanisms causing and maintaining atrial fibrillation (AFib) and atrial flutter (AFlut), the intraatrial excitation is measured with multi-electrode map- ping catheters. The unorganized excitation propagation during atrial fibrillation makes an interpretation of measured intracardiac electrogram data very difficult. Frequency do- main analysis methods are an approach to extract valuable diagnostic information.In this thesis, an approach for the statistical analysis of intracardiac mapping data was implemented and evaluated. With the partial directed coherence function (PDC), direc- tional couplings between electrodes are determined. It is a method to measure Granger causality in the frequency domain. Before computation of the PDC, a multivariate autore- gressive model (MVAR model) is fitted on the measured data by sparse modeling. The approach was implemented and evaluated on simulated and clinical data. Simulations of AFlut were realized with an implementation of the fast marching algorithm for which a graphical user interface was implemented. The intraatrial electrograms were forward cal- culated with an adapted implementation of the boundary element method. Furthermore, methods for the 3D visualization and evaluation of coupling information resulting from the statistical electrogram analysis were developed.The application of the statistical modeling approach on intracardiac electrogram data showed that recurrent propagation patterns could successfully be detected. The selected time window was found to have a strong influence on the result, especially during AFib, where the propagation patterns often stay only stable for several seconds.During the analysis of basket catheter data, the results were often limited by bad coverage of the atrium by the catheters. Due to spline deformation and the specific geometry of the catheter, large areas were frequently badly covered and no satisfactory results could be obtained for these areas. In the well covered areas, basket catheters were found to be very suitable for the detection of global patterns, whereas for the detection of localized patterns smaller catheter geometries were more appropriate.