Computers in cardiology/physionet challenge 2004: AF classification based on clinical features

The Computers in Cardiology / Physionet Challenge 2004 deals with the classification of ECG signals from AF patients into three categories: types N, S and T corresponding to AF episodes terminating never; soon and immediately, respectively. In our study, diflerent features were used, extracted by the experienced clinician among the authors (LK) on the supplied training set. Algorithms were developped to quantify these features from provided ECG data. A Support Vector Muchine was used to classic these features. In this papel: we present our method, results and conclusion about this clinically-oriented approach.