Enhancement of accuracy and reproducibility of parametric modeling for estimating abnormal intra-QRS potentials in signal-averaged electrocardiograms.

This work analyzes and attempts to enhance the accuracy and reproducibility of parametric modeling in the discrete cosine transform (DCT) domain for the estimation of abnormal intra-QRS potentials (AIQP) in signal-averaged electrocardiograms. One hundred sets of white noise with a flat frequency response were introduced to simulate the unpredictable, broadband AIQP when quantitatively analyzing estimation error. Further, a high-frequency AIQP parameter was defined to minimize estimation error caused by the overlap between normal QRS and AIQP in low-frequency DCT coefficients. Seventy-two patients from Taiwan were recruited for the study, comprising 30 patients with ventricular tachycardia (VT) and 42 without VT. Analytical results showed that VT patients had a significant decrease in the estimated AIQP. The global diagnostic performance (area under the receiver operating characteristic curve) of AIQP rose from 73.0% to 84.2% in lead Y, and from 58.3% to 79.1% in lead Z, when the high-frequency range fell from 100% to 80%. The combination of AIQP and ventricular late potentials further enhanced performance to 92.9% (specificity=90.5%, sensitivity=90%). Therefore, the significantly reduced AIQP in VT patients, possibly also including dominant unpredictable potentials within the normal QRS complex, may be new promising evidence of ventricular arrhythmias.