Abstract The 2006 Physionet/Computers in Cardiology

Challenge is motivated by the expected requirement of regulatory agencies for skilled readers to conduct QT duration studies of all drugs in clinical trials. However, clinically valid automatic methods would reduce the cost and time need to conduct such QT studies, which are crucial in determining potential proarrhythmic side effects, such as torsade de pointes, of non-antiarrhythmic drugs. The challenge dataset is the PTB Diagnostic ECG Database, which consists of 549 records from 294 subjects. This work proposes an integrative approach for the identification of the PQ junction and T-end of the QT interval. This novel approach uses the curvature of the Qwave and the T-wave in order to identify an accurate isoelectric curve, between two consecutive peaks, that intersects the extremities of the features. The score for the challenge is 35.19 ms for this method compared to top score of 17.33 ms. 1