Wavelet bidomain sample entropy analysis to predict spontaneous termination of atrial fibrillation

The ability to predict if an atrial fibrillation (AF) episode terminates spontaneously or not through non-invasive techniques is a challenging problem of great clinical interest. This fact could avoid useless therapeutic interventions and minimize the risks for the patient. The present work introduces a robust AF prediction methodology carried out by estimating, through sample entropy (SampEn), the atrial activity (AA) organization increase prior to AF termination from the surface electrocardiogram (ECG). This regularity variation appears as a consequence of the decrease in the number of reentries wandering throughout the atrial tissue. AA was obtained from surface ECG recordings by applying a QRST cancellation technique. Next, a robust and reliable classification process for terminating and non-terminating AF episodes was developed, making use of two different wavelet decomposition strategies. Finally, the AA organization both in time and wavelet domains (bidomain) was estimated via SampEn. The methodology was validated using a training set consisting of 20 AF recordings with known termination properties and a test set of 30 recordings. All the training signals and 93.33% of the test set were correctly classified into terminating and sustained AF, obtaining 93.75% sensitivity and 92.86% specificity. It can be concluded that spontaneous AF termination can be reliably and noninvasively predicted by applying wavelet bidomain sample entropy.