Signal-averaged electrocardiogram: Overview of technical aspects and clinical applications

INTRODUCTION — Over 300,000 individuals succumb to sudden cardiac death (SCD) per year in the United States alone, with the single biggest cause being ventricular tachycardia (VT) or ventricular fibrillation (VF). The approach to the prevention of SCD depends upon the identification of those patients who are most likely to have a VT or VF and the effectiveness of the available preventive measures. (See "Pathophysiology and etiology of sudden cardiac arrest" and "Incidence of and risk stratification for sudden cardiac death after myocardial infarction".)

The signal-averaged electrocardiogram (SAECG) is a noninvasive signal-processing technique to detect subtle abnormalities in the surface ECG, not visible to the naked eye, that are related to the pathophysiology underlying reentrant arrhythmias such as VT. The SAECG has most often been used to identify low-amplitude signals at the terminus of the QRS complex, referred to as "ventricular late potentials." These late potentials represent delayed ventricular activation, which may reflect the presence of myocardial scar tissue and identify patients at increased risk for reentrant ventricular tachyarrhythmias. The SAECG has been studied in an effort to identify individuals at risk for SCD, particularly in the context of coronary artery disease, acute myocardial infarction (MI), and left ventricular dysfunction. However, the SAECG can also be useful in evaluating the risk for atrial arrhythmias, with a prolonged SAECG P wave, equivalent to "atrial late potentials," which may potentially identify patients at risk for atrial fibrillation.

In patients with the substrates for VT, slow conduction through myocardium disrupted by inflammation, edema, fibrosis, or scar tissue results in electrical potentials that extend beyond the activation time of normal surrounding myocardium but that are too small for detection on the surface ECG. The SAECG uses computerized averaging of ECG complexes during sinus rhythm to facilitate the detection of these small microvolt-level signals, which occur later than rapid ventricular activation and are termed late potentials. (See 'Techniques' below.)

The SAECG is particularly useful in understanding the arrhythmic substrate and stratifying risk for ventricular tachyarrhythmias in patients with cardiomyopathies of various etiologies, most commonly in association with arrhythmogenic (right) ventricular cardiomyopathy [1], but also in the context of coronary heart disease, healed MI, or left ventricular dysfunction. The clinical uses and technical aspects of the SAECG and the characteristics of late potentials will be reviewed here. Specific conditions in which the SAECG is more regularly used are discussed separately. (See "Arrhythmogenic right ventricular cardiomyopathy: Diagnostic evaluation and diagnosis", section on 'Signal-averaged ECG' and "Sustained monomorphic ventricular tachycardia: Clinical manifestations, diagnosis, and evaluation", section on 'Signal-averaged electrocardiogram' and "Brugada syndrome: Clinical presentation, diagnosis, and evaluation", section on 'Signal-averaged ECG (SAECG)'.)

DEFINITIONS AND EPIDEMIOLOGY

Definition — Late potentials are low-amplitude, high-frequency signals that are thought to reflect slow and fragmented myocardial conduction (figure 1). Late potentials are defined by three criteria (waveform 1):

●Total filtered QRS duration

●Voltage in the terminal portion of the QRS (usually the last 40 milliseconds)

●Duration of the terminal QRS that is below a particular amplitude (usually 40 microvolts), called the late potential duration

This definition of late potentials is empiric, yet has become validated over time by clinical observations. Notably, values for each parameter depend upon the type and frequency response of the filters used in the signal-averaging system. The most commonly used algorithms employ a 40- or 25-Hz high-pass filter. Using the 40-Hz filter, a Task Force Committee of the European Society of Cardiology, the American Heart Association, and the American College of Cardiology suggested the following definition of late potentials:

●Filtered QRS duration >114 milliseconds

●Terminal (last 40 milliseconds) QRS root mean square voltage <20 microvolts

●Low-amplitude (<40 microvolts) late potentials with duration >38 milliseconds

Despite this attempt at standardization, authors have used variations of these criteria, and definitions for spectral (frequency-domain) analysis of late potentials have not been formalized.

Epidemiology — This abnormal conduction, which may occur in regions of prior infarction, nonischemic scar, or fibrosis, provides a substrate for reentrant ventricular tachycardia (VT) (figure 2). Late potentials are present in 6 percent of asymptomatic normal subjects, but their incidence increases progressively when examining patients with a recent myocardial infarction (MI) without VT, patients with a remote MI who have not had VT, and patients with a remote MI who have had sustained monomorphic VT. Accordingly, the prevalence of late potentials in patients with documented VT and coronary artery disease ranges from 70 to 92 percent, and the predictive value of late potentials for VT increases with the number of other clinical risk factors.

MECHANISM OF VENTRICULAR ARRHYTHMIAS DUE TO LATE POTENTIALS — Late potentials are primarily used to reflect the presence of substrates for ventricular tachycardia (VT) rather than a dynamic predisposition to form a reentrant circuit. This is because reentrant arrhythmias require a critical balance between conduction delay and heterogeneous recovery of excitability. (See "Reentry and the development of cardiac arrhythmias".)

Therefore, while late potentials reflect slow conduction related to arrhythmogenic substrates, episodes of VT initiation in such patients are uncommon in the absence of additional triggers such as ventricular premature beats, electrolyte abnormalities, increased circulating catecholamines, or ischemia. (See "Sustained monomorphic ventricular tachycardia: Clinical manifestations, diagnosis, and evaluation", section on 'Additional diagnostic evaluation'.)

Despite their link with structural substrates, there is also evidence to suggest that late potentials may indicate dynamic predisposition to ventricular arrhythmias:

●Diurnal variations in late potentials correspond to those reported in both the incidence of myocardial infarction (MI) and in the relative resistance of acute MI to thrombolysis.

●Despite variability in the characteristics of late potentials, they consistently correlate with the subsequent induction of VT.

There has been ongoing debate about whether variability in late potentials reflects a dynamic susceptibility to VT or influences such as autonomic tone without direct arrhythmogenic impact. The synergy between slow conduction and additional clinical triggers in initiating VT is supported by the improvement in predictive accuracy for VT when combining the SAECG with assessments of left ventricular ejection fraction, autonomic tone, the presence of complex ventricular ectopy on ambulatory ECG recordings, or ischemia on exercise testing. An evolving area of study is in correlating late potentials on the ECG with intracardiac ventricular late potentials that are increasingly used as targets for the ablation of VT [2].

TECHNIQUES — The electrical signals of interest that reflect delayed conduction through ventricular or atrial scar are minute (in the microvolt range) compared with the size of the QRS complex or P wave seen on a surface ECG. Although simple amplification of the ECG may reveal these signals, it also amplifies ambient noise, which typically masks the useful low-amplitude signals.

The SAECG is derived by computing the arithmetic mean of multiple ECG complexes. This process requires consistent signals between complexes while diminishing the more variable noise components, thus increasing the signal-to-noise ratio of cardiac potentials to enable detection of smaller (ie, microvolt-level) signals than would otherwise not be discernible on visual inspection of the surface ECG. Signals from the His bundle as well as subtle abnormalities of atrial or ventricular complexes, anomalies not visualized on a surface ECG, are detectable using the SAECG. The ECG can be averaged over time ("temporal averaging") or spatially, and this process has been further enhanced by digital filtering and spectral analysis.

Acquiring the data — The ideal number and configuration of ECG leads used to record the SAECG is unclear, but the most commonly used lead set uses three bipolar leads in a standard X, Y, and Z (orthogonal) arrangement (figure 3). In most cases, the averaging of 200 to 400 QRS complexes with the same morphology, over approximately three to seven minutes, is sufficient to record an adequate SAECG.

Signal filtering further augments signal detection of the SAECG and minimizes random noise that is not synchronized to the QRS complex. As an example, filters facilitate easy recognition of QRS onset and offset. They may also help with the detection of low-amplitude, high-frequency signals, such as late potentials or atrial activity, during the otherwise low-frequency or isoelectric ST segment.

The most commonly used filters are bidirectional filters. In this setup, the high-pass component removes low-frequency activity, such as that due to baseline drift of the ECG signal and low-frequency components of the ST segment and T waves. The low-pass components remove high-frequency noise, such as pectoral muscle components.

Temporal signal averaging — The most commonly applied SAECG method, temporal signal averaging, averages a number of QRS complexes over time. Beats are detected by computer algorithms, based on voltage thresholds or other criteria, then aligned by a constant feature of the signal, the "fiducial" (figure 3 and waveform 1). Once beats have been aligned, their arithmetic mean is taken. This process diminishes random noise that is not synchronized to the QRS complex. The disadvantage of this process is that oscillating, transient, or other signals that are not present in each beat are diluted by temporal averaging. An approach to retain this information is spatial averaging. (See 'Spatial signal averaging' below.)

Since averaging is performed until a "noise floor" is reached, the lower the initial baseline noise level, the fewer the number of beats required to generate the SAECG, and the better the resultant signal. However, in order to obtain a reliable averaged signal, only QRS complexes of similar beat-to-beat morphology should be analyzed. Thus, premature ventricular beats, aberrantly conducted beats, or grossly "noisy" beats are excluded from the analysis. An automated algorithm typically generates an "acceptable" template from the first few QRS complexes, then compares each successive beat for "closeness of fit" with the template before incorporating the beat into the average.

Temporal signal averaging has some inherent limitations:

●This technique can only enhance signals that occur in a fixed relationship to the fiducial point. If there is significant variability in the morphology of late potentials from beat to beat, they may actually be considered to be "noise" and diminished by averaging.

●Ventricular late potentials can only be detected from the body surface when the fragmented activity outlasts the normal ventricular activation. Thus, in patients with bundle branch block, late potentials may be obscured by the delayed activation of normal ventricular myocardium.

Spatial signal averaging — ECG spatial averaging involves the summation and averaging of electrical potentials simultaneously recorded from multiple pairs of closely spaced electrodes. This process allows a real-time, beat-by-beat analysis of individual QRS complexes but requires electrical shielding of the patient and equipment. The advantage of this technique over temporal signal averaging is that it allows assessment of irregular rhythms with changing conduction or signals that may oscillate or otherwise vary between beats. The disadvantage is that by averaging across spatial locations, it dilutes information specific to any one location.

Although often considered distinct from the SAECG, spatial averaging is the basis of Laplacian filtering embodied in electrodes used to record beat-to-beat dynamics of T-wave alternans. The principle of spatial averaging is also the basis for high-resolution body surface potential mapping that may detect the risk for sudden cardiac arrest or other arrhythmias. Finally, spatial averaging also allows dynamic changes to be evaluated after drug therapy, after ectopic beats, or after the onset of ischemia. (See "T wave (repolarization) alternans: Overview of technical aspects and clinical applications".)

Other techniques

●Spectral analysis – Spectral analysis considers the QRS complex (or P wave) to be composed of multiple, simple waveforms, typically sinusoids. Spectral analysis thus decomposes the QRS complex or P wave into these constituent signals, represented by component frequencies and corresponding phase and amplitudes. Spectral decomposition is typically performed using Fourier analysis, although methods such as wavelet transform have also been successfully applied.

●Spectrotemporal analysis – Spectrotemporal analysis is the combination of time- and frequency-domain techniques. This procedure presents the Fourier transform of multiple segments of the ECG signal, shifted in time as a three-dimensional plot of the magnitude of individual signal frequencies against time. Thus, this technique may overcome the disadvantages of temporal signal averaging by allowing identification of fractionated activity occurring within the QRS complex (eg, in patients with an intraventricular conduction delay or a bundle branch block).

CLINICAL SETTINGS — SAECG abnormalities are associated with an increased risk of ventricular arrhythmias and cardiac and sudden death mortality in the following clinical settings:

●Ischemic heart disease (table 1), including post-myocardial infarction (MI) and ischemic cardiomyopathy. (See 'Ischemic heart disease' below.)

●Nonischemic cardiomyopathies. (See 'Dilated cardiomyopathy' below.)

●Arrhythmogenic right ventricular cardiomyopathy (ARVC) [3].

●Brugada syndrome. (See "Brugada syndrome: Clinical presentation, diagnosis, and evaluation".)

●Noninvasive diagnosis of transplant rejection in cardiac allograft recipients. (See "Heart transplantation in adults: Diagnosis of allograft rejection".)

●Identification of individuals with paroxysmal atrial fibrillation (AF) prone to frequent recurrences based upon abnormalities in the P-wave SAECG. (See 'P-wave SAECG and atrial fibrillation' below.)

●Although abnormalities in the SAECG have been identified in divergent populations with a number of other cardiac diseases, there is little or no clinical role for SAECG in their management.

●There are emerging reports of the use of SAECG in other conditions. Recent studies have shown that a positive SAECG in patients with thalassemia is indicative of the extent of iron overload [4].

Patients with an abnormal SAECG appear to be at prognostically greater risk of ventricular tachycardia (VT) or sudden cardiac arrest (SCA), but the SAECG is only one of many prognosticating factors for VT and has a limited role for guiding therapy in modern cardiology practice. We concur with the 2008 American Heart Association/American College of Cardiology/Heart Rhythm Society (AHA/ACC/HRS) scientific statement on noninvasive risk stratification [5] and the 2006 ACC/AHA/European Society of Cardiology guidelines for management of patients with ventricular arrhythmias [6], which concluded that the SAECG may be useful to identify patients at low risk for sudden cardiac death (SCD), but its routine use to identify patients at high risk for SCD is not yet adequately supported.

Ischemic heart disease

Patients with prior VT — Most patients with sustained VT or ventricular fibrillation (VF), with or without coronary heart disease, will be treated with an implantable cardioverter-defibrillator (ICD) for secondary prevention of SCD. In such patients, further risk stratification testing including the SAECG is not likely to impact management. Because of this, we do not routinely perform SAECG in all patients with ischemic heart disease and documented VT or VF.

Abnormalities on the SAECG are common among patients with ischemic heart disease and prior MI, with a notable difference between patients with and without VT. SAECG abnormalities are present in up to 93 percent of those with a history of VT, compared with only 18 to 33 percent without prior VT. This observation has inspired efforts to use the SAECG to predict future arrhythmic risk.

Abnormalities in both time- and frequency-domain analyses are common in patients with a history of VT. Such abnormalities may occur at the terminal QRS, within the ST segment, or throughout the entire cardiac cycle. By contrast to time-domain analysis, the predictive accuracy of frequency-domain abnormalities is not affected by bundle branch block. This implies that the physiologic substrate critical to the initiation or maintenance of reentrant arrhythmias is independent of the deranged sequence or total duration of ventricular activation. SAECG abnormalities are less common in patients with a history of VF, demonstrable in only 21 to 65 percent of patients, which may reflect pathophysiologic differences compared with VT.

Patients without prior VT — Due to the reduced rates of arrhythmic events in patients treated with contemporary reperfusion and medical therapies, the utility of the SAECG is limited in low-risk, unselected post-MI populations. As such, we do not routinely perform SAECG for risk stratification in all patients with ischemic heart disease without documented VT or VF. However, there may be a role for the SAECG in selected higher-risk, post-MI cohorts defined in conjunction with additional risk stratification tests, or for patients at high risk of complications from an ICD, in whom the absence of late potentials would impart a better prognosis and potentially shift the risk/benefit ratio. (See "Incidence of and risk stratification for sudden cardiac death after myocardial infarction".)

Following an acute MI, the absence of late potentials on SAECG has a very high negative predictive accuracy (95 to 99 percent), but the presence of late potentials is associated with a low positive predictive accuracy (14 to 29 percent). Although the absence of late potentials on the SAECG indicates an excellent arrhythmia-free prognosis, their positive predictive accuracy is too low to guide therapy. Another limitation is that most of the data predate the use of contemporary therapies for acute MI that have been shown to improve outcome and reduce the rates of SCD (eg, rapid reperfusion, beta blockers, angiotensin converting enzyme inhibitors, statins). (See "Incidence of and risk stratification for sudden cardiac death after myocardial infarction".)

The positive predictive value of the SAECG is modest but improves considerably when combined with other risk factors including reduced left ventricular ejection fraction (LVEF), attenuated heart rate variability, high-grade ventricular ectopy, or inducible VT on electrophysiologic study (table 1).

●In one series of 102 post-MI patients, the combination of an abnormal T-wave alternans study and an abnormal SAECG had a positive predictive value for ventricular arrhythmias of 50 percent (figure 4).

●In the MUSTT trial, which enrolled 1925 patients with chronic coronary disease, left ventricular dysfunction, and asymptomatic nonsustained VT, an abnormal SAECG (filtered QRS duration >114 milliseconds) was associated with a significantly increased risk of arrhythmic death or cardiac arrest, cardiac death, and total mortality. The combination of an abnormal SAECG and LVEF <30 percent identified patients at highest risk for arrhythmic or cardiac death (figure 5).

The optimal time at which to obtain the SAECG post-MI is uncertain, although it should be at least two to three weeks following the index MI. Late potentials are present within three hours of an acute MI in up to 52 percent of patients, and the incidence increases during the electrically unstable initial 24 to 48 hours post-MI. Although their prevalence decreases after this initial phase, late potentials recorded as early as 24 to 72 hours post-MI may portend increased risk for VT and VF. At 6 to 14 days post-MI, late potentials are detectable in up to 93 percent of patients who eventually develop VT or VF, particularly in the first six months. The detectability of late potentials falls over time, and when measured five years post-MI, late potentials are no longer seen in up to 50 percent of patients in whom they were detectable early after the infarction. It is uncertain whether the arrhythmic risk of these individuals diminishes compared with those in whom late potentials persist.

Among patients treated with thrombolytic therapy or primary percutaneous coronary intervention (PCI), the prevalence of SAECG abnormalities is lower (5 to 24 percent) than in patients with MI who are not reperfused (18 to 43 percent).

●In a substudy of a randomized trial of thrombolytic therapy in which 310 patients had an SAECG prior to discharge, the incidence of late potentials was 37 percent lower in patients assigned to thrombolytic therapy compared with placebo.

●In a cohort of 1800 survivors of acute MI, among whom 99 percent underwent primary reperfusion therapy (91 percent primary PCI), only 90 of 968 patients who had an SAECG (9.3 percent) were identified as having late potentials (ie, abnormal SAECG). Over a mean follow-up of 34 months, SAECG abnormalities did not correlate with the primary combined end point of cardiac death and arrhythmic events.

Although primary reperfusion does not affect the physiologic basis of the SAECG, it is not known if the lower prevalence is due to reductions in infarct size or restoration of patency in the infarct-related artery.

Nonischemic heart disease

Dilated cardiomyopathy — There are conflicting data on the predictive value of the SAECG in patients with dilated nonischemic cardiomyopathy, with the SAECG being predictive of total mortality, cardiac death, and/or arrhythmic events in some studies but not others. Given the inconsistent data, we do not routinely use the SAECG for risk stratification of patients with nonischemic cardiomyopathy.

As examples of the mixed data on SAECG in patients with dilated nonischemic cardiomyopathy:

●In a prospective study of 114 patients with nonischemic dilated cardiomyopathy, in which 20 of the 86 patients without bundle branch block had an abnormal SAECG, one-year survival free of VT was markedly better in patients with a normal SAECG (95 percent) compared with those with late potentials on the SAECG (39 percent). Similarly, a study of 131 patients with dilated cardiomyopathy followed for 54 months found that those with late potentials had an increased risk of all-cause cardiac death (relative risk [RR] 3.3, 95% CI 1.5-7.5) and arrhythmic events (RR 7.2, 95% CI 2.6-19.4). Recent studies continue to suggest that a positive SAECG predicts appropriate ICD shocks in such patients [7].

●Conversely, in a study of 343 patients with idiopathic dilated cardiomyopathy followed for 52 months, late potential on the SAECG did not predict arrhythmia risk or transplantation-free survival. Some studies have found that SAECG abnormalities are predictive of progressive heart failure in patients with dilated cardiomyopathy.

Detection of cardiac allograft rejection — Cardiac allograft rejection is characterized by myocardial inflammation followed by necrosis. Early studies suggested that the SAECG may be sensitive to disruptions in myocardial conduction caused by rejection, in addition to reductions in ECG voltage, conduction delay, and other ECG abnormalities that may occur despite immunosuppression. However, the SAECG is rarely used for this purpose in clinical practice.

Arrhythmogenic right ventricular cardiomyopathy — The SAECG has long been considered a diagnostic test for ARVC as one of the task force minor electrical criteria [8], although its availability is limited in many medical centers. A recent study demonstrated that a positive SAECG identified structural abnormalities on imaging, separating patients with severe from mild disease from control subjects, but was less effective at identifying patients with ARVC without structural phenotypes [9]. A detailed discussion of the role of SAECG in the diagnosis of ARVC is presented separately. (See "Arrhythmogenic right ventricular cardiomyopathy: Diagnostic evaluation and diagnosis", section on 'Signal-averaged ECG'.)

Brugada syndrome — SAECG is not performed in all patients with Brugada-pattern ECG findings, but it can be helpful when there is a high suspicion of Brugada syndrome but the diagnosis remains uncertain after other testing. Additionally, the SAECG may be helpful in identifying a subset of patients at higher risk for VT and SCD. This is discussed in detail separately. (See "Brugada syndrome: Clinical presentation, diagnosis, and evaluation", section on 'Initial steps to diagnose Brugada syndrome'.)

Late potentials and catheter ablation for ventricular tachycardia — We do not routinely perform SAECG in all patients with VT going for catheter ablation, but in select patients it can provide prognostic information. Some data support the prognostic utility of intracardiac late potentials in patients undergoing percutaneous catheter ablation, particularly in the SMASH-VT study, in which catheter ablation at regions of intraventricular late potentials reduced the incidence of ICD therapy on long-term follow-up [10]. In a study of 18 patients with ARVC, elimination of late ventricular potentials by radiofrequency ablation predicted freedom from ventricular arrhythmias after a mean follow-up of five years [11]. In two other studies, abnormal SAECG indices of ventricular depolarization were associated with cardiomyopathy in patients with right ventricular outflow tract-related ventricular arrhythmias, while epsilon waves in late depolarization and negative T waves were associated with the extent of ventricular scar on electroanatomic mapping [12,13]. The clinical role of ventricular late potentials in these contexts is being actively investigated, with some more modern studies using novel high resolution mapping also increasingly identifying late potentials as targets for ablation [14,15] (See "Invasive diagnostic cardiac electrophysiology studies".)

P-wave SAECG and atrial fibrillation — Reentrant arrhythmias in the atrium, analogous to those in the ventricle, also likely require heterogeneities in slow conduction and repolarization. This provides a theoretical basis for prediction of the propensity for AF by SAECG by examining slow conduction in the atrium, analogous to the detection of ventricular late potentials. However, further data are needed to clarify the role of P-wave SAECG before its widespread use can be recommended. (See "Reentry and the development of cardiac arrhythmias" and "The electrocardiogram in atrial fibrillation".)

Patients with paroxysmal AF may be identified by P-wave prolongation on the SAECG during sinus rhythm, although the threshold P-wave duration that best discriminates such patients is controversial. A vector P-wave duration of 155 milliseconds predicted AF with a sensitivity, specificity, and positive predictive accuracy of 80, 93, and 92 percent, respectively. Unlike the ventricle, the utility of atrial late potentials in the terminal 10 to 20 milliseconds of the P wave remains controversial. It is also unclear whether low root mean square voltage at the P-wave terminus indicates AF risk independent of structural heart disease or left atrial dilatation. A biatrial mapping study showed that the site of AF initiation typically shows slow conduction at rapid rates, but not at baseline, just prior to AF onset [16].

Data suggest that the development of AF can be predicted using analyses of the P wave from the SAECG in certain patient populations; for example, patients undergoing cardiac surgery, patients with a recent MI, and patients with hypertrophic cardiomyopathy (figure 6). A substudy analysis of MADIT II found that abnormalities of P-wave shape in orthogonal X, Y, and Z axis ECGs predicted AF onset and SCA in 802 patients with ischemic cardiomyopathy and LVEF ≤30 percent [17]. However, the traditional SAECG indices of P-wave duration and terminal amplitude did not predict AF onset or SCA. These results suggest a wider significance of the P wave than is generally recognized, perhaps as an indicator of atrial enlargement from ventricular dysfunction, and warrant further study. In a related study, P-wave morphology changed over time in patients from the MADIT II study who later developed AF, but remained unchanged in those who did not [18]. Finally, a recent study in two Framingham cohorts suggested that a positive P-wave SAECG is associated with several aspects of AF risk, including heritability [19]. Whether this reflects incipient atrial cardiomyopathy or other physiological features remains to be determined.

SUMMARY AND RECOMMENDATIONS

●Introduction – Late potentials reflect the presence of substrates for ventricular tachycardia (VT). Although late potentials reflect slow conduction related to arrhythmogenic substrates, episodes of VT initiation in such patients are uncommon in the absence of additional triggers such as ventricular premature beats, electrolyte abnormalities, increased circulating catecholamines, or ischemia. (See 'Introduction' above and 'Definitions and epidemiology' above.)

●Definitions – The following time analyses, derived from temporal signal averaging, are the most frequently used criteria to define an abnormal signal-averaged electrocardiogram (SAECG) (waveform 1) (see 'Definition' above):

•Filtered QRS duration >114 milliseconds

•Terminal (last 40 milliseconds) QRS root mean square voltage <20 microvolts

•Low-amplitude (<40 microvolts) late potentials with duration >38 milliseconds

●Mechanism of ventricular arrythmia due to late potentials – The SAECG can detect subtle abnormalities in the surface ECG that are not visible to the naked eye. One example of such an abnormality is the "ventricular late potential," a low-amplitude signal near the end of the QRS complex that can be used to stratify risk for ventricular tachyarrhythmias in patients with cardiomyopathies of various etiologies. (See 'Mechanism of ventricular arrhythmias due to late potentials' above and "Reentry and the development of cardiac arrhythmias".)

●Techniques – The SAECG can be acquired by one of three methods (temporal signal averaging, spatial signal averaging, or spectral analysis), all with inherent benefits and limitations (see 'Techniques' above):

•Temporal signal averaging, the most common method for obtaining the SAECG, averages a number of QRS complexes over time.

•Spatial signal averaging analyzes electrical potentials simultaneously recorded from multiple pairs of closely spaced electrodes.

•Spectral analysis considers the QRS complex (or P wave) to be composed of multiple simple waveforms, typically sinusoids. Spectral analysis thus decomposes the QRS complex (or P wave) into these constituent signals for analysis.

●Clinical settings – SAECG abnormalities are associated with an increased risk of ventricular arrhythmias and cardiac and sudden death mortality in various clinical settings including post-myocardial infarction (figure 1 and figure 4 and table 1) , ischemic cardiomyopathy, nonischemic cardiomyopathy, hypertrophic cardiomyopathy, arrhythmogenic right ventricular cardiomyopathy, and the Brugada syndrome. (See 'Clinical settings' above.)

●Late potentials and catheter ablation for ventricular tachycardia – Ventricular late potentials on the intracardiac electrogram may represent critical sustaining mechanisms for VT and may be targeted for catheter ablation with successful arrhythmia elimination. (See 'Late potentials and catheter ablation for ventricular tachycardia' above.)

●P-wave SAECG and atrial fibrillation – The P-wave SAECG may indicate slow conduction in the atrium and in early work has been used to predict the propensity for atrial fibrillation. (figure 6)