A comparison of methods for the classification of atrial fibrillation from intra-atrial electrograms

Reliable classification of atrial fibrillation from intra-atrial signals is a central task in the development of implantable atrial defibrillators. Methods that satisfactorily discriminate normal sinus rhythms, atrial tachycardia and atrial fibrillation include atrial period (AP) amplitude probability density function (APDF), correlation waveform analysis (CWA) and power spectrum density (PSD). Nevertheless their sensitivity in discriminating among and classifying AF types has been seldom addressed. We compared parameters obtained by the following methods for the classification of the AF rhythm, according to Wells´ types: AP and its coefficient of variation (CV), number of points in the baseline (NO) and Shannon entropy (ENTR) from APDF, correlation coefficient from CWA and indexes obtained from PSD. To evaluate and compare the proposed methods ANOVA and Student´s test were used. Cluster analysis was also used to define the best set of parameters to be used and to test the overall performances. We used data from intra-atrial recordings from chronic AF patients and from subjects with electrically induced AF. The best discriminating parameters were AP, CV, NO and CWA. Cluster analysis using this set satisfactorily discriminated AF1 and AF3 electrograms. Type II AF is still difficult to identify