Session S94.2

An Algorithm for Robust and Efficient Location of T-Wave Ends in Electrocardiogram

Q Zhang, A Illanes Manriquez, C Medigue,
Y Papelier, M Sorine

Université de Rennes
Rennes, France

Computer-aided analysis of electrocardiogram (ECG) is widely used in cardiac disease diagnosis. It is acknowledged that T-wave end location is the most difficult one among ECG wave form detection and location problems. The purpose of this paper is to propose a new algorithm for T-wave end location, which is conceptually original, robust to noise and disturbance, and computationally efficient.
Many algorithms for ECG wave form detection and location have been reported in the literature. Typically, these algorithms are based on numerical differentiation, on pattern recognition with pre-defined wave form templates, or on mathematical models of the considered wave forms.
The algorithm proposed in this paper mainly consists of the computation of an indicator related to the area covered by the T-wave curve and delimited in a special manner. As a remarkable originality of the proposed algorithm, it is formally proved that, based on simple assumptions, essentially on the concavity of the T-wave form, the maximum of the computed indicator inside each cardiac cycle coincides with the T-wave end. Moreover, the algorithm has the following advantages: (a) it is robust to measurement noise, since the computation of the algorithm mainly consists of an integration operation; (b) it is robust to wave form morphological variations and to baseline wander, since the consistency of the algorithm is essentially based on the assumption of T-wave form concavity; (c) it is computationally very simple: the main computation can be implemented as a simple finite impulse response (FIR) filter.
When the proposed algorithm is evaluated with the PhysioNet QT database, the mean and the standard deviation of the T-wave end location errors, over the whole available 3542 annotated T-wave ends in 105 records, are respectively 0.31 millisecond and 17.43 milliseconds. This result slightly outperforms the other algorithms evaluated with the same data base, according to the most recent available publications up to our knowledge.