Low noise level unmasks late potentials on signal-averaged electrocardiography

Abstract

INTRODUCTION:

Identification of late potentials requires the reduction of random noise by signal averaging. The importance of using a very low noise level (NL) as the end point of the averaging process in patients with ventricular tachycardia, the variation of results when a lower than standard NL was used and the modification of the sensitivity of the test when a very low NL was reached were evaluated.

METHODS AND RESULTS:

Signal-averaged electrocardiograms were recorded in 36 patients with ischemic heart disease and spontaneous or induced sustained or nonsustained ventricular tachycardia. Thirteen patients showed negative or indeterminate results on recordings with an NL of 0.3 μV. Eight patients (group 1) underwent a second recording with an NL of 0.1 μV. Eight normal volunteers constituted the control group (group 2). The total duration of the filtered QRS vector magnitude (QRSd), the root mean square voltage of the terminal 40 ms of the vector magnitude (RMS₄₀) and the low amplitude signal duration under 40 μV in the terminal portion of the vector magnitude (LAS) modifications were evaluated.

A significant difference (P<0.01) in these parameters was observed in group 1 (15.88%, 48.25% and 68.5%, respectively) when both recordings were compared. Tests were positive in all patients (100%) with NL reduction.

In group 2, tests were negative in all patients (100%) at both NLs (0.3 μV and 0.1 μV). QRSd was 1.18% longer, RMS₄₀ was 1.38% lower and LAS was 3.55% longer with NL reduction.

CONCLUSION:

Late potentials in patients with ischemic heart disease, ventricular tachycardia, and a negative or indeterminate signal-averaged electrocardiogram may be detected if the NL is reduced to 0.1 μV. Reduction of the NL increased the sensitivity of the test without modifying its specificity.

Late potentials (LPs) are low-amplitude, high-frequency cardiac potentials of less than 5 μV, located at the end of the QRS complex or in the early ST segment recorded in peri-infarct zones. These are areas of slow conduction and unidirectional block that represent the electrophysiological substrate for sustained, re-entrant ventricular tachycardia (VT) (1–4).

Recording of LP requires reduction of simultaneous biological and environmental electrical signals known as noise. The process of averaging QRS complexes reduces these noncardiac signals without modifying LPs. Thus, signal averaging exposes LPs by reducing the random noise. Although residual noise cannot be completely eliminated, it can be significantly reduced, and one can more precisely determine the QRS complex end point if the ratio of signal to baseline noise increases (5).

Until the beginning of the past decade, a fixed number of beats was used as the end point of the averaging process (150 to 200 beats) (6–8). Later, the American Heart Association guidelines (9) stated that noise must be the end point of the averaging process. They concluded that a noise level (NL) of lower than 1.0 μV or 0.7 μV using filters of 25 Hz or 40 Hz, respectively, is required to maximally detect body surface LP (9). However, a lower NL has been routinely used by many centres (10,11).

It was also proposed (12,13) that the use of low noise thresholds in signal-averaged electrocardiograms (SAECGs) could decrease the specificity of the test. Because we observed several patients with spontaneous or induced VT and ischemic heart disease showing negative or indeterminate results on their SAECG with a NL of 0.3 μV (14), we decided to test the hypothesis that further NL reduction would unmask LPs in these selected patients.

Accordingly, we compared SAECG results in the same patients using two different NLs: the NL commonly used in several published studies (0.3 μV) (10,11) and a further reduced NL (0.1 μV) that produces an almost flat, noise-free baseline recording.

Because the sensitivity and specificity of SAECGs reported in the literature greatly varied (5), we hypothesized that differences in NL reduction could partially explain these discrepancies.

METHODS

Study population

Thirty-six patients with sustained or nonsustained spontaneous or induced VT and ischemic heart disease were studied. Twenty-three patients had a positive and 13 had a negative or indeterminate SAECG. In eight of these patients, SAECGs were performed with NLs of 0.3 μV and 0.1 μV. These patients constituted group 1 – the study group – in which sensitivity was tested. Group 2 – the control group – consisted of eight healthy volunteers without heart disease or VT. They also underwent two recordings with NLs of 0.3 μV and 0.1 μV. In this group, specificity was tested. Subjects with bundle branch block or who were undergoing antiarrhythmic treatment were excluded because intraventricular conduction delay, whether intrinsic or induced by antiarrhythmic medications, may influence the sensitivity and specificity of the SAECG.

Equipment and SAECG recording

The SAECGs were recorded with the Predictor system (Corazonix Corp, USA). The subjects’ skin was cleaned with gauze soaked in alcohol and, in some cases, shaved to improve electrode contact. The standard orthogonal lead arrangement with three pairs of electrodes, according to the American Heart Association guidelines (5), was followed.

The electrodes used in all recordings were silver/silver chloride positioned in the standard Frank orthogonal X, Y and Z lead system. Normal QRS complexes were averaged to the designated NL. The signal was digitized at a frequency of 2000 samples with 16-bit accuracy. A template was selected by the operator and QRS complexes that did not match the template with a 99% correlation coefficient were rejected automatically. Selected QRS complexes were acquired until the chosen noise end point was reached. The two tests were performed consecutively within an interval of 10 min. A bidirectional filter and a bandpass cut-off frequency of 25 Hz to 250 Hz used by the system were applied.

The equipment combines QRS complexes into a vector magnitude using the formula root mean square of X²+Y²+Z². The onset and offset of the QRS complexes were automatically calculated by the computer and manually determined to be the point on the tracing when the voltage exceeded the mean of the baseline plus three times the SD. Manual determination was preferred when there was no matching. Retrograde scanning to detect the end of the vector magnitude and anterograde scanning to detect its beginning were performed.

The three standard parameters of SAECGs were calculated for each study: the total duration of the filtered QRS vector magnitude (QRSd); the root mean square voltage of the terminal 40 ms of the vector magnitude (RMS₄₀) or V₄₀; and the low-amplitude signal duration under 40 μV in the terminal portion of the vector magnitude (LAS). Normal values for these parameters were as follows: QRSd, less than 110 ms to 112 ms; RMS₄₀, more than 25 μV; and LAS, less than 32 ms. An SAECG was considered to be positive if all three parameters showed abnormal values, negative if all values were normal and indeterminate if only one or two values were normal.

NL

Two NL end points were chosen for the study: 0.3 μV and 0.1 μV. The 0.3 μV NL was chosen as the baseline because it is the lowest NL used by most services. The 0.1 μV NL represents a very low NL, in which practically no noise can be observed on the recording. Values of each parameter were compared on recordings at both NLs. Variations of sensitivity and specificity were also determined.

Statistical analysis

Data analysis was performed using the Wilcoxon test for paired samples.

RESULTS

NL – beats averaged

Table 1 shows the results of SAECGs in group 1 recorded with NLs of 0.3 μV and 0.1 μV. The 0.3 μV NL was reached by averaging 265.75±121.16 beats; to reach the 0.1 μV NL, it was necessary to sample an average of 1614.37±884.82 beats.

TABLE 1.

Signal-averaged electrocardiography in patients with spontaneous or induced ventricular tachycardia (VT) in the electrophysiological study (EPS): Group 1

Patient	Age, years	Beats averaged, n		QRSd, ms		RMS40, μV		LAS, ms		Result		Diagnosis
		0.3 μV NL	0.1 μV NL	0.3 μV NL	0.1 μV NL	0.3 μV NL	0.1 μV NL	0.3 μV NL	0.1 μV NL	0.3 μV NL	0.1 μV NL
1	65	202	603	135	157	36.7	13.5	24	47	I	P	VT (EPS), MI, CAD, CHF
2	52	208	709	89	119	45	2.1	27	57	N	P	VT (EPS), MI
3	48	288	3057	130.5	142.5	31.78	24	25.5	34.5	I	P	VT, MI, UA, HT
4	67	518	1508	125	143.5	23.11	21.14	30.5	39	I	P	VT (EPS), MI, UA
5	68	165	1126	107	127.5	41.18	2.86	24.5	44.5	N	P	VT, MI (n=3)
6	70	259	2362	131	137	29.51	13.63	33	39	I	P	VT, MI, CHF
7	58	343	2365	125.5	140.5	27.7	22.9	24.5	54	I	P	VT (EPS), MI, syncope
8	63	143	1185	104	122	22.93	15.99	26	41	I	P	Recurrent VT (spont, EPS), MI

CAD Coronary artery disease; CHF Congestive heart failure; HT Hypertension; I Indeterminate (one or two negative parameters); LAS Low-amplitude signal duration under 40 μV in the terminal portion of the vector magnitude; MI Myocardial infarction; N Negative (three negative parameters); NL Noise level; P Positive (three positive parameters); QRSd Total duration of the filtered QRS vector magnitude; RMS₄₀ Root mean square voltage of the terminal 40 ms of the vector magnitude; spont Spontaneous; UA Unstable angina

Table 2 shows the results of SAECGs in group 2. The 0.3 μV NL was reached by averaging 1464±827.37 beats; to reach 0.1 μV, 5180.63±2775.64 beats were averaged.

TABLE 2.

Signal-averaged electrocardiography in control subjects: Group 2

Patient	Age, years	Beats averaged, n		QRSd, ms		RMS40, μV		LAS, ms		Result
		0.3 μV NL	0.1 μV NL	0.3 μV NL	0.1 μV NL	0.3 μV NL	0.1 μV NL	0.3 μV NL	0.1 μV NL	0.3 μV NL	0.1 μV NL
1	42	2700	9146	103	105.5	45.86	42.26	14.5	15	N	N
2	41	535	6047	102	101.5	47.88	48.61	29.5	30	N	N
3	8	1495	7626	101.5	102	325.01	308.83	13.5	14.5	N	N
4	71	1556	7522	87	87.5	134.63	134.63	15	15	N	N
5	19	646	2942	95.5	95.5	190.97	185.27	16.5	16.5	N	N
6	71	1807	1629	106.5	109	56.84	55	16.5	16.5	N	N
7	37	2382	3790	95	93.5	91.69	98.3	10	9.5	N	N
8	57	591	2643	96.5	102	35.9	35.63	24	29	N	N

LAS Low-amplitude signal duration under 40 μV in the terminal portion of the vector magnitude; N Negative (three negative parameters); NL Noise level; QRSd Total duration of the filtered QRS vector magnitude; RMS₄₀ Root mean square voltage of the terminal 40 ms of the vector magnitude

As mentioned in previous reports (10,15,16), recording time depended on the heart rate and NL at the beginning of the study, which was affected by the environmental noise, electrode contact and noise from the power source. As shown in Tables 1 and 2, the average number of beats to reach 0.1 μV greatly varied for these reasons in our patient population.

Detection of LP

In group 1 (n=8), six patients (75%) showed indeterminate SAECG results in the 0.3 μV NL study. Four patients had one abnormal variable, and two patients had two abnormal variables. Two patients had negative SAECG results.

When NL was reduced to 0.1 μV, all eight patients (100%) showed positive results and the three evaluated variables became positive in each patient. Figure 1 shows the SAECG for patient 5 in group 1 using NLs of 0.3 μV and 0.1 μV. Low-amplitude LPs were clearly seen when the NL was reduced to 0.1 μV.

Figure 1).

Modification of the signal-averaged electrocardiogram (SAECG) results in patient 5 of group 1. A Vector magnitude obtained with a noise level (NL) of 0.3 μV. NL: 0.3 μV; beats averaged: 165; total duration of the filtered QRS vector magnitude (QRSd): 107 ms; root mean square voltage of the terminal 40 ms of the vector magnitude (RMS₄₀): 41.18 μV; low-amplitude signal duration under 40 μV in the terminal portion of the vector magnitude (LAS): 24.5 ms. Result: negative. QRSd, RMS₄₀ and LAS were normal. B Vector magnitude obtained with an NL of 0.1 μV. NL: 0.1 μV; beats averaged: 1126; QRSd: 127.5 ms; RMS₄₀: 2.86 μV; LAS: 44.5 ms. Result: positive. QRSd, RMS₄₀ and LAS were abnormal. NL reduction unmasked low-amplitude late potential in the final part of the vector magnitude

In group 2 (n=8), all patients (100%) showed negative results on SAECG recordings using NLs of 0.3 μV and 0.1 μV (Table 2). No terminal late activity suggestive of LP was recorded when the NL was reduced.

Quantitative variations of parameters

In group 1, values of the three standard SAECG variables (ie, QRSd, RMS₄₀ and LAS) (Tables 1 and 2) varied between the 0.3 μV and 0.1 μV NL studies performed in the same patient.

Abnormal and normal variables with an NL of 0.3 μV had significantly different values when the NL was reduced to 0.1 μV. An average percentage of variation for each variable was calculated as follows: QRSd was 15.88% longer; RMS₄₀ was 48.25% lower; and LAS was 68.5% longer (Table 1). These changes were caused by the detection of low-amplitude signals at the end of the QRS complexes. LPs were present in both tracings, but visualization and detection were improved by noise reduction to the 0.1 μV level (Figure 1).

In group 2, changes of variables were as follows: QRSd was 1.18% longer; RMS₄₀ was 1.38% lower; and LAS was 3.55% longer (Table 2). The results of the SAECGs were also negative when the NL was reduced to 0.1 μV. Therefore, no decline in the specificity of the test was demonstrated in the group of healthy subjects.

As reported in Tables 1 and 2, quantitative variations of each variable in group 2 were significantly smaller than in group 1. Results had statistical significance at P<0.01 using the Wilcoxon test for paired samples.

In summary, all patients with spontaneous or induced VT in the electrophysiological study (group 1) had LP on their SAECGs when the NL was reduced, while none of SAECGs of the healthy subjects (group 2) showed abnormal parameters and, therefore, abnormal SAECG results.

DISCUSSION

By the end of the 1980s, SAECGs emerged as helpful diagnostic tools for the diagnosis, risk stratification and treatment of patients with VT. However, initial enthusiastic expectations diminished over time, due to significant variability in the sensitivity of the test to detect patients predisposed to developing VT. During the past few years, its predictive value for VT and ventricular fibrillation has been re-evaluated either alone or in conjunction with other parameters for several cardiac diseases (17–21).

Surprisingly, the technical aspects of the test were investigated as the cause of the variability by only a few authors. Observations made by Steinberg and Bigger (10), our group and others clearly demonstrated that the NL should be considered an important technical aspect influencing the results of the test. In 1989, Steinberg and Bigger (10) determined that the degree of noise reduction can significantly alter the results of SAECGs. They stated that the 0.3 μV criterion specifically improves detection of LP. Thus, a 0.3 μV NL was accepted as the end point of the averaging process for many centres. We reported a patient with VT and a negative SAECG who had a positive SAECG when a lower than standard NL was used (14). Conversely, Engel et al (22) concluded that noise does not influence the value of SAECG variables in a systematic manner. Moreover, Christiansen et al (23) suggested that SAECG becomes LP positive at low NLs in healthy subjects.

We observed that values of the variables greatly varied in some patients when we used a very low NL in our recordings. Testing the hypothesis than a 0.3 μV NL does not detect LP in all patients with VT, we undertook the present study to determine whether a lower than standard NL can affect the SAECG results when they are negative or indeterminate in patients with ischemic heart disease and VT.

We examined the prevalence of LP using the current acceptable NL and a lower NL in two groups of patients. As shown, an important difference was observed in the results of studies performed with 0.3 μV and 0.1 μV NLs in patients with VT and ischemic heart disease (group 1). While SAECG results were negative or indeterminate with a 0.3 μV NL, new SAECGs recorded with a lower NL (0.1 μV) all demonstrated LPs. Prolongation of the QRSd in these patients accounted for a longer terminal low-amplitude segment and an increase in LAS. A decrease in RMS₄₀ revealed abnormal values in all eight patients. In the control group, minimal, nonsignificant changes in the three parameters when using a 0.1 μV NL did not modify the previous negative results obtained by using an NL of 0.3 μV.

These findings proved our hypothesis that the current acceptable NL for SAECGs may not be low enough to detect the presence of low-amplitude signals in some patients with VT and ischemic heart disease; a lower NL should be used to unmask LP. A very low NL did not ‘create’ LP and, consequently, false-positive results were not demonstrated in healthy subjects (group 2). Therefore, the specificity of the test did not change in our group of patients.

Our results provided evidence that LP detection improved in a subgroup of patients with VT by lowering the NL. Thus, the predictive value of SAECG to induce VT improves without modification of its specificity. This modification of the recording technique will probably decrease the range of variation of the sensitivity reported in the literature. Although the recording time is prolonged, this modification of the recording technique is justified and should be applied to a subgroup of patients with VT in whom accurate SAECG results have important clinical implications (24).