ECG-based waveform characterization of atrial fibrillation

Abstract:
In atrial fibrillation (AF), the body surface signal pattern varies considerably from one patient to another as well as over time. We have developed a new method for ECG-based characterization of AF which explores the morphology of the f-waves. Following QRST cancellation, the method divides the atrial signal into short blocks and performs a model-based analysis of each block. The blocks are then clustered into different waveform patterns. The method was applied to a database of 36 patients (10-s recordings) with organized AF. The results show that the different waveforms in AF are well represented by the proposed model. Each recording was, in average, represented by 3 clusters of which the main cluster covered 31% of the blocks. In lead V1, the typical phase relationship between the fundamental and first harmonic was -82 degrees, indicating sawtooth-like waves with a steeper downslope than upslope.