Detection of late potentials in patients with and without myocardial infarction. (IV) Passband analysis

The presence of ventricular late potentials in patients is regarded as a prognostic indicator of a sudden and possibly fatal arrhythmic event. Ventricular late potentials can be identified in both the Time and Frequency domain methods. The present study investigated the frequency range over which late potentials are suggested to exist within the frequency spectra, using the frequency domain method. Fast Fourier Transformation of the signal-averaged surface ECG was performed in subjects with and without a prior myocardial infarction, looking specifically at 14 different passbands. Areas and Area Ratios for the X, Y and Z leads were calculated separately and then combined (averaged X, Y and Z). Discriminant analysis was performed on the Area and Area Ratio data to identify which passband provided optimum separation (`Predictiveness') of those subjects with and without ventricular late potentials (as defined by the time domain method). Following discriminant analysis the results indicated that the Area data consistently provided better `true predictions' than the Area Ratio data for the discriminant analysis for all leads whether individual or combined (87.1, 91.8, 89.6 and 88.1% vs 68.1, 77.0, 49.6 and 62.9% for Area and Area Ratio for the X, Y, Z and combined leads respectively). For the Area data the best passband was 85-105 Hz for the X lead (90.3%) compared to 65-95 Hz/25-120 Hz for the X lead (77.0%) of the Area Ratio data. In conclusion: a) normalization of data provided less discrimination, b) a greater frequency content was evident in subjects with late potentials at higher frequency passbands, c) individual leads and not combined leads could provide greater discrimination between subjects.