Evaluation of Fourier transform filter for high-resolution ECG

The high-resolution electrocardiography has become an important clinical tool for analyzing the high-frequency content of electrocardiograms (ECGs). Recent emphasis has been on the detection of ventricular late potential activity due to its ability to predict ventricular tachycardia (VT) in myocardial infarction (MI) patients. To accentuate the high-frequency components, the signal-averaged ECG data are filtered using high-pass filters. Two types of filters used in commercial systems, bidirectional Butterworth and Fourier transform filters, are compared using a common signal-averaged ECG data base. Signal-averaged ECG data acquired at two clinical sites (Mayo Clinic and Bowman Gray School of Medicine) using the MAC15 HIRES system were filtered using a 40-Hz fast Fourier transform (FFT) filter with a 6 dB/octave rolloff on an IBM-compatible personal computer. The same average data were filtered using a 40-Hz bidirectional Butterworth filter with similar rolloff. Using a common algorithm, outputs of both filters were used to compute vector magnitude and to obtain the measurements to quantify high-frequency, low-amplitude (HFLA) signals. The measurements include total QRS duration, duration of HFLA signals, root mean square voltage, and mean voltage in the terminal 40 msec. Applying the published criteria for late potentials, sensitivity and specificity were computed for both filters on data obtained from 59 patients of old MI (group 1, 29 with clinical or inducible VT and group 2, 30 with no history of VT and noninducible). The results were very similar, and both filters were found to be functionally equivalent.