Improved late potential analysis in frequency domain

This work presents a technique to improve the identification of late potentials (LP) in patients affected by greater arrhythmogenic right ventricular disease (GARVD). Several authors have documented the correlation between GARVD and LP by means of time domain analysis. Moreover, the high incidence of bundle branch block in patients affected by GARVD suggests LP analysis in the frequency domain be performed. The method of spectral mapping of the ECG with Fourier transform was adopted. This consists in dividing the ST segment into 25 subsegments and estimating their frequency components by means of the fast Fourier transform. Recently, it was documented that this technique suffers from poor reproducibility of results. Low reproducibility is the consequence of an improper localization of the analysed QRS segments. An algorithm to increase the QRS end point identification reproducibility is proposed. An optimal QRS filter was adopted as well as a technique based on the Hilbert transform. This technique allowed the reliability of the normality factor estimates to be improved. The computed normality factors on the XYZ leads and on the vector magnitude were used to classify patients and healthy subjects; 28 patients affected by greater arrhythmogenic right ventricular disease and 35 healthy subjects were analysed in the study. High sensitivity was obtained with respect to GARVD and clinical sustained ventricular tachycardia by means of a cluster analysis technique. By applying the technique proposed in this paper the identification of LP in GARVD was increased from 47% to 88%, when clinical sustained ventricular tachycardia was documented, whereas in patients affected by GARVD but not prone to sustained ventricular tachycardia LP identification increases from 18% to 64%.