Improved identification of late potentials by adjustment of the number of analyzed segments of the spectral temporal mapping of the signal-averaged electrocardiogram

Abstract

Late potentials (LP), microvolt oscillations occurring at the end and after the QRS complex, identify patients at risk of ventricular tachycardia and sudden death. 1,2 They can be detected by the time domain analysis of the surface signal-averaged electrocardiogram (SAECG). 1 LP can also be identified by the frequency domain analysis of the SAECG because they are characterized by higher frequencies than the ST segment. 3 This technique offers some advantages over time domain analysis: patients with bundle branch block need not be excluded, a noise-dependent algorithm for detection of LP offset is not applied, and high-pass filters, distorting the signal, are not used. Spectral temporal mapping (STM) of the SAECG is a new frequency domain technique that uses fast Fourier transform analysis. 4 Routinely, 25 electrocardiographic segments are analyzed, first starting 20 ms before the standard QRS offset. The frequency content in the most distal 5 segments, which are located well within the ST segment (where LP are not likely to be) is used as a reference spectrum, and compared with the spectra of earlier segments (where LP are often present) in order to calculate the result of the STM. It is crucial for correct analysis to avoid the presence of high frequency components in the last 5 reference segments. However, when the total QRS duration exceeds the standard QRS duration by >20 ms owing to the presence of a long LP, the reference segments include LP (Figure 1A). Therefore, a new formula is needed to calculate the number of segments to be analyzed in such patients.