Effect Of Interpolation On Electroanatomical Mapping

Cardiac navigation systems (CNS) are often used in electrophysiological studies to create spatial-electrical maps supporting the arrhythmia mechanism identification. Sequentially recorded electrograms yield the bioelectrical information from features such as voltage and activation times in terms of their spatial location, which are subsequently interpolated for building the electroanatomical map (EAM) of the cardiac chamber Our goal was to evaluate quantitatively the effect of interpolation in the EAM accuracy when reconstructed from a set of samples. Triangulated irregular networks (TIN), thin plate spline (TPS), and support vector machines (SVM) were assessed by using: (a) two detailed simulated time activation maps during flutter and sinus rhythm in both atria; (b) a set of real CNS maps, given by 13 activation time and 19 voltage maps, with 6 right atria (RA), 6 left atria (LA), 4 right ventricles (RV), and 16 left ventricles (LV). Interpolation methods were benchmarked using root mean squared error (RMSE), efficiency (EF), and Willnott distance (WD). On the one hand, EF and WD were similar for yielding a clearer cut-off point than RMSE for the number of required samples, which was about 100. Better EAM accuracy was obtained using TPS, followed by SVM and TIN, except for flutter in the RA, where early-meets-late was smoothed by SVM. On the other hand, EAM accuracy (in terms of the average WD) was slightly outperformed by RA than LA (0.57 vs 0.52), whereas RV and LV were similar (0.71 vs 0.71). In reference to the methods, similar average WD was given by the interpolation methods (TIN 0.64 +/- 0.14; TPS 0.66 +/- 0.15; SVM 0.65 +/- 0.18). The EAM accuracy is dependent on the map nature and on the cardiac chamber.