Towards standardization of non-invasive atrial fibrillation substrate complexity quantification: Effect of choice of ECG-leads and complexity measure on prediction of pharmacological cardioversion

A large variety of parameters have been proposed for quantification of atrial fibrillation substrate complexity computed on a standard 12-lead ECG. A direct comparison of the predictive performance of these complexity parameters using a single dataset has not been performed. We retrospectively studied 221 patients undergoing pharmacological cardioversion (CV) with flecainide to compare the predictive performance of dominant atrial frequency (DF), spectral organization index (OI), spectral entropy (SE), sample entropy (SampEn) and fibrillation wave amplitude (FWA). Prediction of successful CV using a single parameter computed on a single lead was poor for all parameters with a maximum AUC of 0.67 (01 on lead II). Performance improved by incorporating multiple leads, with maximum A UC of o. 75 (SampEn on I, II, a VR and V6), and by combining parameters and leads with a maximum AUC of 0.83, using DF (I), 01 (II), SE (I), SampEn (I, II and V
₆
) and FWA (aVF and V
₁
). The approach taken in this study to compare the performance of diferent non-invasive AF complexity parameters can be seen as a first step towards a standardization of AF substrate complexity quantification.