Independent component analysis applied to electrogram classification during atrial fibrillation

Cardiac arrhythmia analysis is one important biomedical application of pattern recognition. We present a pattern recognition technique applied to the analysis of electrograms during atrial fibrillation. Atrial fibrillation (AF) is a common arrhythmia which has a high rate of incidence among the elderly. Besides being poorly tolerated, it greatly increases the risk of embolic stroke. We propose an algorithm based on independent component analysis for classifying multichannel electrograms from an ovine model of AF into one of four classes-normal sinus rhythm, atrial flutter, paroxysmal AF and chronic AF. The success rates achieved indicate great potential of the method in automated electrogram analysis and classification.