INSERM and Hopital Cardiologique, Lyon, France.

We developed a new technique to identify the post infarction patients (MI pts) prone to ventricular tachycardia (VT) by decomposing the unfiltered X.Y.Z signal-averaged high-resolution ECGs following a series of dyadic orthogonal Meyer wavelets with central frequencies at 28, 47, 97 and 195 Hz .

The time-scale transforms were computed from the onset of ORS (ONQRS) to OnORS+256ms on a set of 40 MI pts comprising 20 MI pts with documented VT (MI+VT) and 20 Ml pts without VT (MI-VT). The 40 MI pts were randomly split into a leaning and a test set. The reproducibility was tested with an additional set of 1 0 MI pts and 2 healthy subjects (HS) recorded twice at 10min interval. MI+VT and MI-VR were dichotomized using stepwise linear discriminant analysis. The 3 most relevant wavelets are located, at ONORS+ 128ms and in the ORS complex at ONQRS+ 48ms and at OnRS+64ms. A fourth wavelet located at OnORS+224ms was introduced for separating HS tram Ml. The selected wavelets are in the highest frequency scale (195 Hz), which corresponds to a 135-25OHz bandpass filter. Results are listed thereafter. A test performed on 1 0 HS confirmed that none was categorized as MI+TV.

In conclusion, we designed a new method able to localize micropotentials within QRS, which significantly improves the detection of patients prone to VT.