Classification of atrial tachyarrhythmias in electrocardiograms using time frequency analysis

In this work a novel automated method, combining time-frequency analysis and expert's knowledge, for the classification of atrial tachyarrhythmias is presented. It is based on the analysis of small ECG segments and their classification into three categories of cardiac rhythm: (a) atrial fibrillation, (b) atrial flutter and (c) normal sinus rhythms. Time-frequency analysis is used to calculate the power spectrum density for each segment. Several spectral characteristics are extracted from the power spectrum density, representing the energy distribution on the time-frequency plane. These characteristics are used as input in an artificial neural network, which classifies each ECG segment into one of the three categories. The method is validated using the MIT-BIH arrhythmia database and the obtained average sensitivity and specificity are 93.4% and 96.5%, respectively.