Exploring alternans characteristics of electrocardiogram for prediction of Paroxysmal Atrial Fibrillation

Abstract

The goal of our thesis was to investigate if P-wave and PQ segment alternans can be used to detect and predict Paroxysmal Atrial Fibrillation (PAF) from an Electrocardiogram (ECG). The work involved implementing an algorithm for computing alternans information for a region of ECG, applying the algorithm to derive P-wave and PQ segment alternans information, and analyzing data generated for normal and PAF ECGs for evidence of any distinguishable predictive characteristics. Based on analysis of total 80 patient records (35 normal and 45 PAF) with validation on 30 records randomly selected from those (achieving c-indexes between 0.66 and 0.70), we concluded that alternans can potentially be a useful predictor for PAF.