Morphologically constrained signal subspace characterization of electrograms during ventricular fibrillation

We present one such targeted adaptive signal decomposition of intra-cardiac electrograms during VF, to automatically extract morphologies previously observed in relation to the vicinity of rotor phenomenon (i.e. an event of interest during VF). Using a Gabor dictionary and Matching Pursuits (MP), the proposed method successfully performed targeted signal decomposition in highlighting signal structures that could be used to determine the ventricular electrogram's vicinity to a rotor in the epicardium. A comparative analysis using the MP with a proposed choice function (P < 0.001) against the standard choice function (P < 0.5688) performed better in segregating the electrogram data (130 electrograms from 7 isolated human hearts) based on the signal structures.