Prediction of Paroxysmal Atrial Fibrillation through Spectral and Statistical Techniques

Abstract

Aims: The paper describes some methods based on statistical tools and time-frequency analysis for extracting features in frequency and temporal domain that may be used as predictors of Paroxysmal Atrial Fibrillation Place and Duration of Study: Department of Computer Engineering, College of Computers and IT, Taif University, Saudi Arabia (Jan 2014-Feb 2015) Methodology: The main focus of the work was on the investigation of useful tools emerging from statistics or quadratic transformations that can predict the possible onset of PAF in order to provide a prompt remedy. Our study was based on a database of two-channel ECG recordings which has been created for use in the Computers in Cardiology Challenge 2001. Real ECG records from 10 normal subjects and 10 subjects having PAF were studied. The specific features from the surface electrocardiogram immediately prior to the onset of Paroxysmal Atrial Fibrillation (PAF), showed that the occurrence of PAF can be predicted to some extend by analysing the spectral components, statistical and time-frequency parameters. The presence of PAF in p-records made the signal spectrum rich in energy. Alternatively, the variance and standard deviation of signals prior to the occurrence of PAF also gave valuable information which indicated the possible occurrence of PAF in the subsequent signal part. Results and Conclusion: The specific features from the surface electrocardiogram immediately prior to the onset of paroxysmal atrial fibrillation (PAF) proved that the occurrence of the latter can be predicted to some extent by analyzing the statistical and spectral parameters.