ECG-based feature tracking in atrial tachyarrhythmias

A new method for extraction of general features in ECGs with atrial tachyarrhythmias is presented. The method is based on our recent method for atrial signal characterization which sequentially decomposes a time-frequency distribution into a set of parameters. In addition to rate and amplitude, the method tracks information on regularity, waveform, and structure of the atrial signal. The proposed method includes a feature tracker which continuously tracks the structure of the harmonic spectral pattern in order to determine which of a set of archetype waveforms that is best matched. The method also includes a trend detector, which detects significant long-term changes in the time series of frequency estimates. The results, illustrated by signals obtained during different interventions and a test signal, show that the algorithms can discriminate between different types of atrial rhythms.