Signal Averaging in Wolff-Parkinson-WhiteSyndrome: Evidence That FractionatedActivation Is Not Necessary for Body SurfaceHigh Frequency PotentialsWILLIAM M. SMITH,* HUMBERTO }. VIDAILLET, JR.,t SETH J. WORLEY,*JEFFREY K. POLLARD,§ LAWRENCE D. GERMAN,# DAVID W. MORTARA,**and RAYMOND E. IDEKER*From the Departments of *MGdicinG and Biomedical Engineering, The University of Alabama atBirmingham, Birmingham, Alabama, the tMarshfiold Clinic, Marshfield, Wisconsin,tThe Heart Group. Lancaster, Pennsylvania, the §University of Calgary, Calgary, Alberta, Canada,the New Mexico Heart Institute, Albuquerque, New Mexico and **Mortara Instruments,Incorporated, Milwaukee, Wisconsin

SMITH, W., ET AL.: Signal Averaging in Wolff-Parkinson-White Syndrome: Evidence that FractionatedActivation Is Not Necessary for Body Surface High Freqnency Potentials. It is commonly assumed thatthe presence of high frequency components in hody surface potentials implies that fractionated activationfronts, caused by heterogeneously viable tissue, are present in the heart. However, it is possible that non-fractionated activation fronts can also give rise to high freqaency surface potentials and that the relativeamount of high frequency power is related to the complexity of the activation sequence. In a test of thisidea, averaged body surface potentials were recorded during the entire QRS complex of nine Wolff-Parkin-son-White (WPW) patients in situations in which fractionated activation fronts should not have been pre-sent, but which represent increasing degrees of complexity of ventricular activation: (t) postoperative ec-topic pacing from subepicardial wires placed daring sargery, when a single coherent activation front waspresent throughout most of the QBS; (2) Preoperative preexcited rhythm, when a single coherent activa-tion front was present for one portion of the QRS (the delta wave); and (3) postoperative normal rhythm,when two or more activation fronts were present in the ventricles throughout most of the QRS. For com-parison, averaged body surface potentials were also analyzed during the last 40 ms of the QRS complexand the ST segment of 14 postinfarction patients with chronic ventricular tachycardia. In the patients withWPW syndrome, relatively high frequency content increased (attenuation -36.7 vs -27.2 vs -18.3 dB)and QRS width decreased (160.7 vs 125.9 vs 94.1 ms) significantly from paced to preoperative to postop-erative beats, Significant high frequency content was present in all cases, showing that coherent activa-tion fronts can give rise to high frequencies. Interestingly, the postoperative QRS of WPW patients con-tained a larger proportion of high frequency power than did the late potentials of the patients withventricular tachycardia. Thus, while the presence of late fractionated body surface potentials may be amarker for ventricular tachycardia, these potentials by themselves do not necessarily signify that the un-derlying cardiac activation giving rise to these signals is fractionated. (PACE 2000; 23:1330-1335)

high frequency electrocardiography, fractionated potentials

IntroductionIt has been demonstrated that the amplitude

of the high frequency components in body surfacepotentials is altered in survivors of myocardial in-

Supporled in part by research grants HL-33637 and HL-28429from The National Heart, Lung and Blood Institute, NationalInstitutes of Health, Bethesda. Maryland, and National ScienceFoundation Engineering Research Center Grant CDR-8622201.

Address for reprints: William M, Smith, Ph.D., The Universityof Alabama at Birmingham, B140C Volker Hall, 1670 Univer-sity Boulevard, Birmingham, AL 35294. Fax: (205) 975-4720.

Received August 3, 1999; revised January 4, 2000; acceptedJanuary 10, 2000.

farction who develop ventricular tachycardia^ andthat late ventricular potentials are markers for ar-rhythmias in other nonischemic heart diseases.^^For this application, electrocardiograms (ECGs)are generally signal averaged, filtered, and highlyamplified to allow detection of low level, high fre-quency potentials in the time domain,'''^ the fre-quency domain,^ or a combination of the two.**"̂ "Techniques have also been developed for estimat-ing late potentials on a beat-to-beat basis.^ '̂̂ ^ Sev-eral groups have developed quantitative criteriafor identifying patients at risk for ventriculartachycardia. Cain et al.̂ '̂ used parameters derivedfrom the power spectrum of a portion of the QRS

1330 September 2000 PACE, Vol. 23

SIGNAL AVERAGING IN WPW SYNDROME

complex or the late QRS Gomplex and T wave asclassifiers. Simson et al.^"'̂ ^ studied the fiUeredECG waveform during late QRS and used the am-plitude of the high frequency waveform compo-nents to discriminate between patients with andwithout ventricular arrhythmias. Even thoughthere is a disagreement as to whether late poten-tials and ventricular dysfunction are independentpredictors of arrhythmias/"-^^ it is fairly well ac-cepted that the frequency characteristics of theECG waveform in some way are related to the abil-ity of the ventricle to support slow conductionand reentrant rhythms.

The correlation between the high frequencycharacteristics of the EGG and the incidence ofventricular ectopy in patients with chronic my-ocardial infarction is presumed to be due to thefragmentation of activation wavefronts as theypass through heterogeneous viable and infarctedmyocardium with resulting reentrant circuits.^"-^^However, it is possible that high frequencies canresult from other kinds of complexities in the acti-vation sequence.^" In particular, such complexi-ties could arise from multiple activation fronts co-existing and interacting in the myocardium orfrom discontinuities introduced into the ECGwaveform by the creation and disappearance ofwavefronts. Simple, uniform activation sequenceswould then contain relatively less high frequencypower, and the high frequency content of thewaveform should be related to the amount of timeduring the QRS complex that multiple, complexwavefronts are present. In a test of this hypothesis,relatively high frequency content was measured inthe surface EGGs in patients before and aftersurgery for Wolff-Parkinson-White [WPW) syn-drome during normal sinus rhythm and ventricu-lar pacing. The simplest activation sequenceshould occur during pacing with a single wave-front proceeding almost uniformly across the ven-tricles.^^ The activation sequence of spontaneousrhythms should be more complex; in patients withpreexcitation, tbe preoperative QRS complexshould initially consist of a single wavefront, ex-cited by the accessory pathway, later joined bywavefronts from the normal conduction system.^^The normal postoperative activation sequenceshould contain the most complexity with morethan one wavefront present throughout most ofthe QRS.^^ For comparison, the high frequencyGomponent of ECGs of patients with chronic ven-tricular tachycardia was also measured. The re-sults of these analyses are presented here.

MethodsNine patients with WPW syndrome, who

were to undergo surgery for division of an acces-sory conduction pathway, were included in the

study with 14 patients with chronic ventriculartachycardia secondary to ischemic heart disease.All WPW patients had accessory pathways thatconducted in the antegrade direction. Each patientgave informed consent for the collection of signal-averaged EGGs. Data collection was performedusing an EGG manufactured by ART, Inc [Austin,TX, USA). This system has been described indetail elsewhere. '̂'• '̂̂ EGGs were measured in allWPW patients during normal rhythm pre- andpostoperatively and in seven of them during epi-cardial ventricular pacing postoperatively. Ven-tricular tachycardia patients were studied innormal sinus rhythm. In each patient study, sig-nal-averaged EGGs from X, Y, and Z leads wereacquired and proGessed. For the WPW patients,the power spectrum for each averaged QRS com-plex was estimated using the fast Fourier trans-form (FFT). For patients with ventricular tachy-cardia, the last 40 ms of the QRS and a portionof the ST segment were analyzed similarly. Inaccordance with one of the techniques of Gainet al.,^ the 40-Hz attenuation of the power, in de-cibels, of each power speGtrum was measuredand used as an index of the relative high fre-quency content of the corresponding signal. Inaddition, the vector magnitude of each EGG wascomputed as V J P +~Y^ + Z" where X', Y', andZ' are the X, Y, and Z leads filtered from 25 to 250Hz with a hidirectional filter, as described hy Sim-son et al. '̂*'̂ '' From this vector magnitude, theroot-mean-squared amplitude of the signal duringthe last 40 ms of the QRS complex [LPRMSA) andthe duration of the late high frequency potentials< 40 (xV (LPD40) were recorded.

ResultsFigure 1 shows typical time-domain results

from one of the WPW patients in the study. Theleft columns show the signal-averaged X leadsduring preoperative, postoperative, and pacedrhythms, and the right columns show the corre-sponding vector magnitudes of the EGGs. Powerspectra of the signal-averaged X lead from the EGGof one WPW patient as estimated by the FFT areshown in Figure 2. The spectra were derived fromthe QRS complex during postoperative ventricu-lar pacing [dashed), preoperative normal sinusrhythm [dotted), and postoperative normal sinusrhythm [solid). The data interval used for the FFTwas determined automatically for preoperativeand postoperative spontaneous beats, but waschosen by inspection for the paced beats, since thesignal averager was triggered by the pacing arti-fact. We estimated the relative high frequencycontent of each waveform by measuring the atten-uation of the corresponding spectrum at 40 Hz,called A40, as indicated.

PACE, Vol. 23 September 2000 1331

SMITH, ET AL.

Figure 1. Signal-averaged electrocardiograms from apatient with Wolff-Parkinson-White syndrome. (Panel A)Signal-averaged X lead during preoperative normalrhythm. The preexcitation is evident before the main QRScomplex. (Panel B) Vector magnitude of the electro-cardiogram (see text) from the same patient duringpreoperative normal rhythm. (Panels C and D) Sameplots for postoperative normal rhythm. (Panels E and F)Same plots for postoperative ventricularly paced rhythm.

\\\

\\\

> \'. \

POST-OPPRE-OPPACING

\

\

40 60FREQUENCY. Hi

Figure 2. Normalized power spectra of the signal-averaged X lead of the electrocardiogram for one patientwith Wolff-Parkinson-White syndrome. The spectra forthe electrocardiogram during the QRS complex wereestimated by the fast Fourier transform for preoperative(dotted) and postoperative (solid) normal rhythms andfor postoperative ventricular pacing (dashed). The 40-Hz intercepts, A40. are largest for postoperative normalrhythm and smallest for paced rhythm.

POST-OP PRE-OP PACING

Figure 3. The 40-Hz intercept for nine patients withWolff-Parkinson-White syndrome during the conditionsstudied. The decreases in high frequency power frompostoperative normal rhythm to preoperative normalrhythm to postoperative pacing are statisticallysignificant. The studies duringpacing were unsuccessfulin two patients for a technical reason.

Figure 3 shows A40 for all WPW patients forthe three measurement conditions. The studiesduring pacing were unsuccessful in two patientsfor technical reasons. A higher (less negative)value corresponds to a relatively more high fre-quency content in the signal. There is a generaltrend toward decreasing high frequency powerfrom postoperative sinus rhythm to preoperativesinus rhythm to postoperative pacing. These dif-ferences are significant by paired Ntest as shownin Table I. There was a similar significant in-crease in the width of the QRS complex in thethree categories (Table I). A surprising finding isthat there was significantly more relative high fre-quency content in the QRS complex of the WPWpatients during postoperative sinus rhythm thanin the late QRS-ST segment of the patients withventricular tachycardia (P < 0.001, Student's t-test). Similar tests comparing ventricular pacedrhythm and normal preoperative rhythm to ven-tricular tachycardia were not statistically signifi-cant. Figures 4 and 5 are scatterplots of LPRMSAand LPD40, respectively, against 40-Hz attenua-tion for the three conditions under which signalaveraging was performed for the WPW patientsand for the ventricular tachycardia patients.These indicate that there is considerable overlapbetween the values for WPW patients and thosewith ventricular tachycardia. Consistent withprevious reports, the ventricular tachycardiawaveforms tend to have lower amplitude high fre-quency signals late in the QRS when comparedwith those from patients without ventriculartachycardia. As shown in Figure 4, the 25-(iV

1332 September 2000 PACE, Vol. 23

SIGNAL AVERAGINC IN WPW SYNDROME

MeanSDP value

MeanSDP value

Postoperative

-18.34.4

94.214.0

Table 1.

Comparison of 40-Hz Attenuation

40-Hz Attenuation (dB)

Preoperative

-27.28.4

0.002* 0.015"

QRS Width (ms)

125.921.4

0.001* 0.012**

Values

Pacing

-36.75.6

160.710.0

Ventricular Tachycardia

-32.910.1<o.ooit

121.716.2<0.001t

* Postoperative versus preoperative.*' Preoperative versus pacing.t Postoperative versus ventricular tachycardia.

threshold proposed hy Simson '̂* separates ven-tricular tachycardia patients reasonably well.Table II contains comparisons between the ven-tricnlar tachycardia patients and the WPW pa-tients for the two time-domain parameters,LPRMSA and LPD40. The late potential ampli-tude. LPRMSA. was significantly different be-tween the ventricular tachycardia group and theWPW group during nonpaced rhythms, and theduration of the late potentials, LPD40, was signif-icantly different between the ventricular tachy-cardia gronp and the WPW group during pacingand postoperative spontaneous rhythm.

o PACING• PRE-OPn POST-OP• VT

•leo -40 -20 040 Hz ATTENUATION, dB

Figure 4. Late potential amplitude versus 4O'Hzattenuation. The ordinate is the root-mean-squaredamplitude of the X, Y. and Z leads of theelectrocardiogram during the last 40 ms of the QRScomplex. As anticipated, the value of this variableclusters below 25 fiV for patients with ventriculartachycardia.

DiscussionThe data presented here demonstrate that the

relative high frequency content of ECG waveformsis not necessarily caused by fragmented wave-fronts passing through patchily infarcted tissue. Inpatients with WPW syndrome, the power spec-trum of the QRS complex changes significantlyduring different electrophysiological conditions,even though the myocardium involved should berelatively homogeneous and healthy. There areseveral possible explanations for these changes.

First, increased high frequency componentscan be enhanced as an artifact of the analysismethod. The FFT is based on the assumption thatthe signal being transformed is one cycle of arepetitive signal. This assumption implies conti-

£ 505

o PACING•PRE-OP• POST-OP• VT

o 0-40 -20

40 Hz ATTENUATION. dB

Figure 5. Late potential duration versus 40-Hzattenuation. The ordinate is the duration of the late highfrequency potentials below 40 /iV.

PACE, Vol. 23 September 2000 1333

SMITH, ET AL.

MeanSDP value

MeanSDP value

Table II.

Comparison of Time-Domain

Postoperative Preoperative

123.870.70.005

29.931.90.003

LPRMSA (M.V)

93.452.40.003

LPD40 (ms)

22.46.20.13

Parameters

Pacing

47.235.4

0.2

24.019.30.02

Ventricular Tachycardia

26.424.1

50.629.4

LPRMSA - the root-mean-squared amplitude of the signal during the last 40 ms of the QRS complex; LPD40 =duration of the late high frequency potentials below 40 \LV.The probability values are for the comparisons of the parameters in patients with ventricular tachycardia to thosein patients during the specified rhythm.

nuity hetween the last data point and the first datapoint of the samples analyzed. If there is a largedifference between the values of the two samples,a discontinuity is introduced in the waveformwith resultant artifactual increases in the high fre-quency components. Using windows with thedata is a partial solution to this dilemma.^ In thisstudy, the entire QRS complex of the WPW pa-tients was included, meaning that both ends of thedata to he analyzed were near the same value.Thus, the high frequency components should nothave been artifactual.

Second, the high frequency content of theQRS complex can he increased by the sudden ap-pearance or disappearance of activation wave-fronts in the myocardium. The collision and anni-hilation of two wavefronts or the suddenbreakthrough to the epicardium or endocardiumof a single wavefront can quickly change the totalpotential signal sensed by a remote electrode, giv-ing rise to a discontinuity in the signal. Related tothis phenomenon, the low frequency componentsof multiple wavefronts can cancel leaving morerelative high frequency power so that the lowestfrequencies should come from the simplest, mostuniform activation sequences. These mechanismsare relevant to the QRS complexes in this studybecause of the difference in complexity of the ac-tivation sequences for the paced, preoperative,and postoperative cardiac cycles.

Finally, some of the differences in high fre-quency content in the three conditions analyzedin this study could be attributable to the differ-ences in QRS duration. The total QRS lengths asmeasured by the signal averager were significantlydifferent with paced greater than preoperative andpreoperative greater than postoperative (P < 0.001

in both cases). A shorter QRS complex should gen-erate relatively more high frequency than a QRScomplex that is similar in morphology but ofgreater duration. The width of the QRS complexand the complexity of the activation sequencecould be related if increased cancellation of com-peting wavefronts causes a shorter QRS duration.

The power spectra of the QRS complexes an-alyzed in this study are influenced by all of theabove considerations, but primarily by the lasttwo. During ventricular pacing, the ventriclesshould be excited by a single, fairly uniform wave-front from the pacing site with minimal fusion andcollisions. There should be few sudden changes inthe FCG during activation so that there should beless contrihution to the spectrum at high frequen-cies. At the same time, the QRS duration duringpacing is the longest for auy of the three condi-tions.

During preoperative sinus rhythm, at leastpart of the QRS complex is generated by a wave-front from the accessory pathway existing alone inthe myocardium (i.e.. the delta wave). The singleactivation front generates relatively more low fre-quency signal and allows a more gradual onset ofthe QRS, smoothing the discontinuities from nor-mal sinus epicardial breakthrough. The longerQRS duration due to preexcitation should makethe high frequencies relatively less prominent.

The postoperative QRS complexes containthe most high frequency power of the three condi-tions studied and have more relative high fre-quency than the late QRS complex and ST seg-ment of the ECGs of patients with ventriculartachycardia when assessed in the frequency do-main [Table I). This is probably due to the com-plexity of the activation front with multiple break-

1334 September 2000 PACE, Vol. 23

SIGNAL AVERAGING IN WPW SYNDROME

through sites and to the short duration of the QRScomplex. The fact that the time-domain measure-ments in patients with ventricular tachycardiawere different from the postoperative measure-ments [Tahle II) indicate that the terminal QRScomplexes in the patients with WPW syndromewere relatively normal. Even though low ampli-tude high frequency signals late in the QRS com-plex and relatively large high frequency compo-nents of the power spectra have been shown to be

markers for patients at risk for ventricular tachy-cardia, this study demonstrates that there areother physiological states capable of generatingwaveforms with significant amounts of high fre-quency power. Since activation during some ofthese other physiological states is not ahnormallyfragmented, the presence of high frequency com-ponents in the body surface potentials does notnecessarily signify that the underlying cardiac ac-tivation giving rise to these signals is fractionated.

References1, Breitbardt G. Becker R, Seipel L. et al. Non-invasive detection of

late potentials in man-A new marker for ventricular tachycardia,Eur Heart I 1981; 2:1-11,

2, Fauchier )-P, Fauchier L, Babuly D, et ai. Time-domain signai-av-eraged electrocardiogram in noni.schemic ventricular tachycardia,PAGE 1996: 19:231-244,

3, lanoiisek ]. Paul T, Bartakova H. Role of late potentials in identify-ing paticnis at risk for ventricular tachycardia after surgical cor-rection of congenital heart disease. Am I Cardiol 1995: 75:146-150,

4, Stelling [A, Danford DA, Kugier |D. ul al. Late potentials and in-ducible ventricular taubycardia in surgically repaired congenitalbeart disease. Circulation 1990: 82:1690-1696,

5, Berbari EJ, .Scberlag B|, Hopo RR, et al. Recording from the bodysurface of arrhytbmogenic ventricular activity during tbe S-T seg-ment. Am J Cardiol 1978: 41:697-702,

6, Lander P, Berbari EJ, Rajagopalan GV, e{ al, Critical analysis of tbesignal-averagod electrocardiogram: Improved identificationof latepotentials, Circulalion 1993; 87:195-117.

7, Gain ME, Ambos D, Witkowski FX, et al, Fast-Fourier transformanalysis of signal-averaged electrocardiograms for identification ofpatients prone to .sustained ventricular tacbycardia. Girculation1984;69:711-720,

B. MestD O, Rix H. Caminal P. el al. Ventricular late potentials char-acterization in time-frequency domain by moans of a wavelettransform, IEEE Trans Biomed Eng 1994; 41:625-634,

9, Sierra G. Fetscb T, Reinhardt L, et al, Multiresolution decomposi-tion of the signal-averaged EGG using the Mallat approach for pre-diction of arrhythmic events after myocardial infarction, | Electro-cardiol 1996; 29:223-234,

10. Steinbigler P, Haberl R, [ilge G, et al. Singlo-heat analysis of ven-tricular late potentials in the surface electrocardiogram using tbespectrotemporal pattern recognition algoritbm in patients withcoronary artery disease. Heart 1998: 19:435-^46,

11. Al-Nashasb HAM, Kelly SW, Taylor D)E, Noninvasive beat-to-beatdetection of ventricular late potentials, Med Biol Eng Gomput1989; 27:64-68,

12. Zimmerman M. Adamec R. Simonin P. et al. Beat-to-beat detectionof ventricular late potentials witb bigb-resolution electrocardiog-rapby. Am Heart ) 1991; 121:576-585.

13, Cain ME, AnibosD,Markham),etal, Quantification of differencesin frequency content of signal-averaged electrocardiograms in pa-tients with compared to tbose without sustained ventricular tachy-cardia. Am I Gardiol 1985: 55:1500-1505,

14, Simson MB, Use of signals in tbe terminal QRS complex to iden-tify patients witb ventricular tacbycardia after myocardial infarc-tion, Girculation 1981; 64;23S-242,

15, Simson MB, Euler D, Micbelson EL, et al. Detection of deiayedventricular activation on the body surface in dogs. Am | Physiol1981: 241:H363-H369,

16, Breitbardt G, Borggreft; M, Karbenn L), et al. Prevalence of late po-tentials ill patients witb and without ventricular tacbycardia; Cor-relation witb angiographic finding.s. Am J Gardiol 1982;49:1932-1937,

17, Polldk SI, Kertes PI, Uredlaii CE, et al. Influence of left ventricularfunction oil signal averaged late potentials in patients witb coro-nary artery disease and witbout ventricular tacbycardia. Am HeartI 1985;110:747-752,

18, Gardner PI, Ursell PG, Fenoglio J] )r, el al, Electropbysiologic andanatomic basis for fractionated electrograms recorded from bealedmyocardial inlarcts, Girculation 1985; 72:596-611.

19, El-Sberif N, Electropbysiologic basic of ventricular late potentials.Prog Gardiovasc Dis 1993; 35:417-427,

20, Ideker RE, Mirvis DM, Smitb WM, Late, fractionated potentials.Am I Gardiol 1985; 55:1614-1621,

21, Scber AM. Young AG. Spread of excitation during premature ven-tricular systoles, Girc Res 1955: 3:535-542,

22, Gallagbcr (I, Kasell JH, Sealy WG. et al, Epicardial mapping in tbeWolff-Parkinson-Wbite syndrome. Circulation 1978: 57:854-866.

23, DurrerD, van Dam RT, Freud GE, et al. Total excitation of the iso-lated human beart, Circulation 1970; 41:899-912,

24, Worley SJ. Mark DB, Smith WM. et al, Comparison of time do-main and frequency domain parameters from the signal averagedEGG: A multivariable analysis, J Am Goll Gardioi 1988; 11:1041-1051.

25, Mortara DW. Signal averaging in electrocardiograpby. In ILWillems, IH van Bemmei, G Zywielz (eds.): Gomputer ECG Analy-sis: Towards Standardization, Amsterdam, NY, Elsevier SciencePublishers B,V, (North-Holland], 198fi. pp, 123-127,

PAGE, Vol. 23 September 2000 1335