Impact of filtering upon ventricular tachycardia identification by correlation waveform analysis.

Signal analysis of digitized waveforms has been postulated as a method for improving sensitivity and specificity of ventricular tachycardia (VT) detection in implantable antitachycardia devices. Such improvement may alleviate the problem of unwarranted delivery of therapy by adding precision to the identification of the pathological VT. Morphological analysis could also allow distinct therapies to be initialized for multiple VTs in the same patient. Correlation waveform analysis (CWA) has been demonstrated to be effective in separating benign rhythms from VT in wideband recordings (1-500 Hz) but the effect of filtering has not been previously examined. Bipolar (1 cm) intraventricular recordings (1-500 Hz) of sinus rhythm (SR) and 25 distinct VTs in 18 patients were analyzed by CWA using a signal-averaged SR template. Passages contained 65.9 +/- 19.8 VT depolarizations (range 45-108). Digital filtering was performed on all data passages with varying passbands. Results for passages with a bandwidth of 1-250 Hz were equivalent to wideband results, i.e., greater than or equal to 92% paired sets of SR and VT were separable at a 95% confidence level. A bandwidth of 1-100 Hz decreased discrimination to 84%. At a bandwidth of 1-80 Hz, 80% of cases were successfully separated, but at 10-80 Hz these results improved to 88%. Bandwidths of 20-80 and 30-80 Hz reduced reliability of CWA performance to 72% and 60%, respectively. Filtering at typical pacemaker/defibrillator passbands produced morphological analysis results equivalent to those yielded at wideband settings. Differences in the range between SR versus VT decreased in filtered recordings but overall detection of VT was not degraded.