Screening and Prediction of Paroxysmal Atrial Fibrillation using Adaptive Neuro-fuzzy Classifier

Abstract: Paroxysmal atrial fibrillation (PAF) is intermittent atrial fibrillation which is not present all the time. Atrial fibrillation (AF) is one of the prominent causes of stroke and its risk increases with age. This paper explains a method to screen the patients with PAF from the normal sinus rhythm patients using Heart Rate Variability (HRV) parameters and also the prediction of onset of PAF. The time domain, frequency domain and nonlinear parameters are estimated. Feature selection is done with the use of Linguistic Hedges (LH) of fuzzy set. The selected features are fed to an Adaptive Neuro-fuzzy Classifier (ANFC) for screening as well as prediction. The accuracy obtained for screening and prediction were 94% and 93.75% respectively. The sensitivity in both cases found to be 100%. Keywords: Paroxysmal Atrial Fibrillation, Heart rate variability, Time domain, Frequency domain, Nonlinear, Linguistic Hedges, Adaptive Neuro-fuzzy Classifier.