A methodology for predicting paroxysmal atrial fibrillation based on ECG arrhythmia feature analysis

This article addresses the Computers in Cardiology Challenge 2001 for predicting the onset of paroxysmal atrial fibrillation (PAF) from the surface electrocardiogram (ECG). To predict PAF, we developed an algorithm based upon the number and timing of atrial premature complexes (APCs) in the ECG. The algorithm detects classical isolated APCs, then predicts PAF based on a measurement of APC rate that favors the most recent APCs. The challenge database consists of 100 pairs of 30-minute ECG segments that may or may not directly precede a PAF episode. We used the learning set of the challenge database to optimize our algorithm. On the test set, it achieved scores of 40 out of 50 for PAF screening (event 1) and 44 out of 50 for PAF prediction (event 2)