Measuring shape variations of ECG waves through Time-Frequency Representations
Résumé
The Distribution Function Method in two dimensions (DFM-2D) is briefly presented as an extension of the previous Distribution Function Method in one dimension (DFM-1D). This novel method is applied to the estimation of shape difference between ECG waves through time-frequency representations (spectrograms). Shape variations due to drug action were followed along one day: for T-waves with Quinidine, for P-waves and QRS complexes with Cibenzoloine. Clustering in two and three classes according to the shape distance was made too. The both methods, DFM-2D and DFM-1D, were compared using a "pure shape distance", i.e. independently of scale changes, and a distance including width variations of the waves. The results concerning shape variations are coherent for the both methods, but in any case they are enhanced by using DFM-2D. In conclusion this novel method will be interesting for early detection and fine measurements of electrophysiological effects of cardiac drugs.
Origine : Fichiers éditeurs autorisés sur une archive ouverte