Analysis of 12-lead classification models for ECG classification

In this work we studied the improvement achieved by including information from the 12 ECG leads, in a previously developed classification model. This model includes features from the RR interval series and morphology descriptors calculated from the wavelet transform. The experiments were carried out in the INCART database, available in Physionet, and the generalization was corroborated in a private database. In both databases the AAMI recommendations for class labeling and results presentation were followed. Different approaches to integrate the additional information available in the 12-leads were studied. The best performing approach obtained for normal beats, sensitivity (S) 98%, positive predictive value (P⁺) 94%; for supraventricular beats, S 88%, P⁺ 91%; and for ventricular beats S 91%, P⁺ 92%. The generalization capability was confirmed in a private database with comparable results. The performance of the reference two-lead classifier was improved by taking into account additional information from the 12-leads.