abnormal intra-qrs potentials associated with percutaneous transluminal coronary...

www.elsevier.com/locate/jelectrocard

Journal of Electrocard

Abnormal intra-QRS potentials associated with percutaneous transluminal

coronary angiography–induced transient myocardial ischemia

Paul Lander, PhDa, Pedro Gomis, PhDb,4, Stafford Warren, MDc, Gary Hartman, MDd,

Kathy Shuping, BScd, Ralph Lazzara, MDd, Galen Wagner, MDd

aFlow Metrix, Maynard, MA 01754, USAbCentre de Referencia en Bioenginyeria de Catalunya, Technical University of Catalonia, 08028 Barcelona, Spain

cCharleston Area Medical Center, Charleston, WV 25304, USAdDuke University Medical Center, Durham, NC 27705, USA

Abstract This article introduces a novel concept of abnormal intra-QRS potentials (AIQPs) associated with

0022-0736/$ – see fro

doi:10.1016/j.jelectroc

4 Corresponding

E-mail address: p

myocardial ischemia. AIQPs are microvolt-level potentials—subtle alterations in the QRS of the

high-resolution electrocardiogram (ECG)—isolated from the unfiltered signal-averaged ECG

(SAECG) by a method of mathematical modeling. The aims of the study were (1) to determine

the characteristics of potentials in the SAECG related to ischemically altered activation during

percutaneous transluminal coronary angiography (PTCA), (2) to determine their relationship with

standard 12-lead ECG variables, and (3) to investigate whether AIQPs have a specific

pathophysiologic basis in myocardial ischemia. Continuous high-resolution ECG data were

acquired from 12 patients before, during, and after PTCA. SAECGs were computed every

60 seconds using an enhanced method of signal averaging. AIQP, ST-segment deviation, and

changes in standard ECG QRS duration were measured in each 1-minute SAECG. AIQP

amplitudes increased significantly during balloon inflation, compared with the preinflation state.

AIQPs exhibited a greater prevalence (12 of 12 patients) than ST-segment deviation changes of

more than 100 lV (7 of 12 patients), or measurable changes in standard QRS duration (4 of

12 patients). In patients with significant changes in 12-lead ECG variables during balloon

inflation, AIQPs were strongly correlated with both ST-segment and QRS-duration changes. AIQP

timing was correlated with the artery occluded, suggesting a specific, ischemia-influenced origin

of the signal. AIQPs show promise as a time-localized, sensitive new ECG marker of ischemically

altered ventricular activation.

D 2006 Elsevier Inc. All rights reserved.

Keywords: Myocardial ischemia; Potentials; Angioplasty; Signal-averaged ECG; QRS

Introduction

Presently, no single noninvasive test is well suited to

continuous monitoring of myocardial ischemia, detecting

presence, extent, and severity. Emergency care of patients

with acute chest pain suggestive of myocardial infarction is

based on decisions regarding initiation and administration

of thrombolytic therapy. Standard 12-lead electrocardiogram

(ECG) remains the only readily available clinical method

to guide these decisions. The general objective of this

work was to compare the capabilities of standard and

high-resolution ECG (HRECG) to detect acute myocar-

dial ischemia.

nt matter D 2006 Elsevier Inc. All rights reserved.

ard.2006.02.006

author. Tel.: +34 934010763; fax: +34 93 4017045.

edro.gomis@upc.edu (P. Gomis).

Conventional large-scale notches and slurs in the QRS

complex of the standard ECG have long been associated

with altered conduction due to myocardial scarring.1-4 We

have defined abnormal intra-QRS potentials (AIQPs) as

low-amplitude notches and slurs (1-50 lV), or subtle

alterations, in the QRS complex of the HRECG. AIQPs

are quantified by using a computerized modeling technique

described previously.5,6 We have recently shown that AIQP

may be a better pathophysiologic signal representation of a

reentrant mechanism for ventricular tachycardia than ven-

tricular late potentials.7 AIQP and late potentials were

measured by using epicardial maps recorded during

experimental myocardial infarction.7 In one study, in the

presence of an arc of block, late potentials and AIQPs were

both present. Late potentials were distal, whereas AIQPs

were coincident with the localized site of reentry. This

iology 39 (2006) 282–289

P. Lander et al. / Journal of Electrocardiology 39 (2006) 282–289 283

points to AIQP as a marker for altered conduction in the

border zone of the myocardial infarction.

In both myocardial infarction and myocardial ischemia,

abrupt, irregular, and localized changes in ventricular

conduction will occur. We hypothesize that in both cases,

AIQPs will result from altered activation around an area

of functional block. Changes in magnitude and direction

of conduction velocity would be transient, resulting in

low-level alterations in QRS morphology in the body

surface HRECG.

Previous studies8-11 have documented changes in high-

frequency QRS energy during periods of complete acute

coronary occlusion in man, which were reversed when

blood flow to the ischemic myocardium was reestablished.

Large-scale QRS changes in the standard ECG have also

been documented during the transient acute transmural

ischemia produced by angioplasty balloon occlusion.12 The

present study introduces a new concept—that low-level

abnormal intra-QRS potentials not seen in the standard ECG

may be a direct, sensitive marker of ischemically altered

ventricular activation. Percutaneous transluminal coronary

angioplasty (PTCA) provides an excellent model in which

to investigate the electrophysiologic changes of transmural

ischemia resulting from the spontaneous thrombotic occlu-

sion responsible for acute myocardial infarction. The PTCA

laboratory selected for the present study is unique in that it

uses a single, prolonged balloon inflation of up to 5 minutes

when clinically feasible. AIQP in the SAECG, and changes

in ST-segment deviation and QRS duration from the

standard 12-lead ECG, were measured every minute before,

during, and after PTCA. This permitted comparison of serial

changes during each minute of balloon inflation in both

standard and high-resolution ECG signals. Signal-averaged

ECGs with enhanced signal-to-noise ratio were obtained

every 60 seconds when the time-frequency plane Wiener

filter was used.13-15

The present study had 3 specific aims: (1) to determine

the prevalence and natural progression of AIQPs (abnor-

mal intra-QRS potentials not seen in the standard ECG)

during PTCA, (2) to determine whether correlations exist

among AIQP and ST-segment deviation and QRS prolon-

gation in the standard 12-lead ECG, and (3) to investigate

whether AIQPs are specific to acute PTCA-induced

myocardial ischemia by relating their timing to the

particular artery occluded.

Methods

Data acquisition

Continuous high-resolution ECGs were recorded in 12

patients undergoing elective PTCA in the catheterization

laboratory at the Charleston Area Medical Center, WVa.

Recording commenced at least 50 seconds before balloon

inflation. Balloon inflation periods ranged from 3 to

5 minutes, with the exact times of inflation and deflation

noted. HRECG data were recorded continuously by use of

orthogonal XYZ leads with the Predictor system (Corazonix

Corp., Oklahoma City, Okla), at a sampling rate of 1000 Hz.

The resolution of the analog to digital converter was 16 bits,

resulting in a 0.3-lV quantization step. Particular attention

was paid to obtaining a high-fidelity recording by careful

skin preparation and electrode application, and by minimal

use of fluoroscopy and other devices that are significant

sources of electromagnetic interference.

Patients who met the following criteria were enrolled in

the study: no history of coronary bypass surgery; no

previous myocardial infarction; absence of ST-segment

deviation, ventricular hypertrophy, bundle branch block,

and extrasystoles on the baseline standard 12-lead ECG.

Perfusion balloons or atherectomy devices were not used.

Of the 12 patients studied, 4 had occlusion of the left

anterior descending coronary artery (LAD), 4 of the left

circumflex coronary artery (LCX), and 4 of the right

coronary artery (RCA). One subject had ECG evidence of

a prior myocardial infarction.

In addition to the HRECG, a standard, 12-lead ECG

(Hewlett-Packard 4765A, Palo Alto, CA) was recorded at

1-minute intervals during PTCA. ST-segment deviations

(either depression or elevation) were measured 60 milli-

seconds after the J point and defined by the maximum value

present in any lead. During and after the period of balloon

inflation, ST-segment deviations were compared with a

reference ECG recorded just before balloon inflation.

Off-line processing of the continuous HRECG record

The continuous HRECG from each study was stored

digitally and subsequently transferred to a recordable

compact disc for later processing. An annotation process

was performed for each subject’s HRECG file before signal

averaging and analysis. Custom-designed computer algo-

rithms were used, which provided a visual overview of the

automatic annotation process. Initially, a normal sinus

rhythm QRS complex was selected from within the

preinflation segment of the HRECG record. This beat

served as a template for both classification (sinus or

nonsinus beat) and alignment for enhanced signal averag-

ing. The QRS onset and offset were marked visually on the

template beat. All other beats in the HRECG record were

then compared with the template. Nonsinus, fusion, or

excessively noisy beats were detected by separately

comparing the integrals (ie, areas) of each half of the QRS

with those of the template. Sinus rhythm beats that passed

this test were then aligned by using a correlation method

described previously.16 The correlation region was the first

half of the QRS, with a threshold value for the correlation

coefficient of .99 set for beat acceptance. Noise in each beat

was measured by the ensemble signal variance method

described previously.16 The result of the annotation process

was a list of all beats, with their alignment fiducial point,

classification, and noise measurement.

The time-frequency plane Wiener–filtered SAECG

Signal averaging of several hundred beats is convention-

ally required to detect high-resolution ECG signals, such as

ventricular late potentials.17,18 In this study, the hypothe-

sized dynamic nature of abnormal intra-QRS potentials

during PTCA necessitated forming the signal-averaged ECG

P. Lander et al. / Journal of Electrocardiology 39 (2006) 282–289284

(SAECG) with a significantly reduced number of beats, on

the order of 50 to 100 beats. The time-frequency plane

Wiener (TFPW)–filtered SAECG was used to accelerate

noise reduction. This is a new method of signal averaging,

which uses processing in the time-frequency plane to

enhance the signal-to-noise ratio of the average. The method

has previously been used successfully to detect late

potentials, with a reduced number of beats and averaged

noise levels less than 0.3 lV root-mean-square (rms).19 The

TFPW filtering method is summarized in the following. (A

full description of the method is given in References 13-15).

First, the ensemble (or signal) average is given by

xP tð Þ ¼XRi¼1

xi tð Þ=R ¼ s tð Þ þ nP tð Þ=ffiffiffiR

pð1Þ

where xi(t) are the R beats of the ensemble, s(t) is the

deterministic, cardiac signal component, and n(t) is the

assumed additive and Gaussianly distributed noise of a

typical beat. The quadratic time-frequency representation

(TFRQ, or spectrotemporal map) of the signal average is

given by

TFRQ xP tð Þ½ � ¼ XPt; fð Þ ¼O

ZxP sð Þw s � tð Þ e�j2pf s dsO

2 ð2Þ

where w(t) is the analysis window used to compute the

spectrotemporal map.20 Ideally, the quantity TFRQ[x(t)] can

be expressed mathematically as

E TFRQ xP tð Þ½ �� �

¼ S t; fð Þ þ NP

t; fð Þ=R ð3Þ

where S(t,f) and N(t,f)/R represent the cardiac signal and

averaged noise components in the time-frequency plane.

Similarly, the average of the TFRs computed for each beat

in the ensemble can be ideally expressed as

EXRi¼1

TFRQ xi½ �=R#¼ TFRQ xi tð Þ½ �P

¼ S t; fð Þ þ NPt; fð Þ

"

ð4Þ

In practice, the 2 quantities in Eqs. (3) and (4) are

smoothed with a 2-dimensional filter15 to reduce interfer-

ence and variance terms before being combined into a

weighting function, given by

w t; fð Þ ¼ S t; fð ÞS t; fð Þ þ N

Pt; fð Þ=R

ð5Þ

This is the TFPW filter, which is then applied to the

linear TFR (TFRL) of the ensemble average, x(t,f), by a

multiplicative operation. The resulting TFPW-filtered

SAECG, s(t), is recovered by an inverse transformation

from the time-frequency plane to the time domain, that is15

ss tð Þ ¼ TFRL�1

xP t; fð Þw t; fð Þ� �

ð6Þ

Subensemble averaging

The ensemble of all beats in the continuous HRECG

record before, during and after PTCA, was split into

subensembles by using the list of annotated beats described

here. A series of sequential, TFPW-filtered, subensemble

averages was computed as follows:

(1) A preinflation subensemble average, composed of

up to 128 beats available immediately before

balloon inflation.

(2) to (6) Between 3 and 5 subensemble averages,

each of 60 seconds duration, computed sequentially

during the balloon inflation period. All available

beats within each 1-minute epoch that passed the

alignment, classification, and noise criteria, were

included.

(7) to (10) Up to 4 sequential, postdeflation sub-

ensemble averages, each of 60 seconds duration.

Each subensemble average was TFPW filtered and

then analyzed.

Analysis of subensemble SAECGs

Abnormal intra-QRS potentials were measured in each

subensemble average with the computerized modeling

technique developed in References 5 and 6. This approach

is summarized as follows. First, QRS onset, offset, and

duration (QRSD) were measured automatically and then

manually overread, as has been described previously.16 A

discrete cosine transform (DCT) of the QRS waveform was

computed and then mathematically represented by using an

ARX model.21 This is given by

y nð Þ ¼ �Xnai¼1

aiy n� 1ð Þ þXnbj¼0

bju n� jð Þ þ e nð Þ

¼ yy nð Þ þ e nð Þð7Þ

where y(n) is the DCT of the QRS complex, y(n) is the

modeled signal DCT, u(n) is an impulse, and e(n) is the

residual, following standard terminology for an ARX

model.21 The model, h = [a1a2. . .ana b0b1. . .bnb], is found

by a nonlinear iterative search for the minimum mean

squared error between y(n) and y(n), the original and

modeled signals, respectively. A model order of [na = 7,

nb = 8]; that is, 7 autoregressive and 8 moving average

terms, was used for all subjects and all leads. This model

order was selected on the basis of previous experience with

SAECG waveforms.5,6,22,23

The residual signal, e(n) = y(n) � y(n), is the AIQP

waveform, which was quantified by calculating its rms

value within the duration of the high-resolution QRS,

that is

AIQPrms ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiXT

OAIQP tð ÞO2=T

sð8Þ

The timing of the abnormal intra-QRS potential was also

computed as the bcenter of massQ of AIQP(t), defined by

AIQPt0¼

Pt0QRSOn OAIQP tð ÞO

2

P QRSOffQRSOn OAIQP tð ÞO

2c0:5 ð9Þ

Fig. 1. Progression of AIQP amplitudes before, during, and after the PTCA

procedure for each of the 12 patients.

P. Lander et al. / Journal of Electrocardiology 39 (2006) 282–289 285

Statistical analysis

Comparison of means between SAECG parameters

recorded in a series of subensemble averages (epochs) was

analyzed by repeated measures analysis of variance with

1-minute epoch measurements as within-subject factors.

Post hoc comparisons of each epoch with each of the others

were performed with the Bonferroni adjustment for multiple

comparisons. Correlation was computed by use of the

Pearson correlation coefficient. The probability that no

correlation was present (q = 0) was also calculated. All

analyses were performed with the SPSS for Windows

release 11.0 (SPSS Inc., Chicago, Ill).

Results

Changes in AIQP amplitudes during PTCA

Fig. 1 shows the progression of measured AIQP

amplitudes (mean of the 3 XYZ leads) in each of the

SAECG subensembles before, during, and after the PTCA

procedure for each of the 12 subjects. If an inflation lasted

only 4 or 3 minutes (3 cases), no data are plotted in the latter

Fig. 2. Illustration of HRECG waveforms showing (A) the preinflation SAECG,

waveforms recorded in the preinflation, inflation, and postdeflation periods; and

1 or 2 minutes, respectively, of the inflation period. That is,

all postdeflation data commence at minute/subensemble

number 7. AIQP amplitudes rise sharply during the first

minute of inflation (AIQPrms = 51.8 F 27.3 lV rms)

compared with the preinflation recording (AIQPrms =

9.3F3.2 lV rms). The increase typically is maintained

throughout the inflation period, and AIQP values slightly

decrease after balloon deflation. An analysis of variance

showed that changes in mean AIQP amplitudes between

some epochs were significant (F = 6.27, P b .0001).

Bonferroni corrected pairwise comparisons showed signif-

icant differences between preinflation epoch and any minute

of inflation values. Pairwise comparison also found signif-

icant increase from the second to the third minute of

inflation (P = .02). A marked decrease is observed during

the second minute after balloon deflation with respect to the

fourth minute of inflation but without statistical significance

(P = .07). No other differences between epochs were

statistically significant.

Fig. 2A illustrates, for 1 subject, the unfiltered, prein-

flation SAECG (0.05-300 Hz), the modeled preinflation

SAECG (dotted trace), and the sequence of AIQP wave-

forms recorded in the preinflation, five 1-minute inflations,

and first postdeflation epochs. The AIQP analysis interval

(ie, the SAECG QRS duration) is denoted by the vertical

dashed lines. Fig. 2B shows the progression of AIQP

amplitudes relative to the preinflation state, increasing

sharply in the first minute of inflation and decreasing after

balloon deflation.

Relationship between AIQP timing and artery occluded

Mean AIQP timing in the 3 XYZ leads is shown in

Fig. 3 for each artery. Timing expresses AIQPt0as a

percentage of the QRS duration. The vertical limit bars

show the minimum and maximum values of AIQPt0in the

4 patients for each of the LAD, LCX, and RCA arteries.

No statistical analysis was performed because of the small

number of subjects.

the modeled preinflation SAECG (dotted trace), and the sequence of AIQP

(B) the progression of AIQP amplitudes in the same SAECGs.

Fig. 3. Mean AIQP timing for each artery, expressed as a percentage of

QRS duration.

Table 1

Correlations and level of significance between AIQP and the standard ECG

variables, ST-segment deviation, and sQRSD measured during the

preinflation and inflation periods

ST segment sQRSD

AIQP 0.51 ( P = .0006)4 0.37 ( P = .015)4

ST segment 0.37 ( P = .017)4

4 P b .05.

P. Lander et al. / Journal of Electrocardiology 39 (2006) 282–289286

Relationship between changes in AIQP and ST-segment

deviations and QRS duration measured from the standard

ECG

Seven (58.3%) of the 12 subjects studied exhibited a

clinically significant change in ST-segment level (greater

than 100 lV in any lead), measured in the standard 12-lead

ECG, during the course of PTCA (range, 0 to 4 mm, ie, 0 to

0.4 mV). Considering the transient ST-segment change

(=0.05 mV) criterion used in acute coronary syndrome,24

8 (66.7%) of 12 patients presented ST-segment changes.

Using a more sensitive criterion (= 25 lV), we found that

10 (83.3%) of 12 patients showed ST-segment changes.

Four subjects (33%) exhibited a measurable change in QRS

duration, as seen in the standard 12-lead ECG (range, 0-

80 milliseconds), during the PTCA procedure. In contrast,

all subjects (100%) showed a significant increase in AIQP

amplitudes between the preinflation SAECG and the first

subensemble average recorded during inflation. AIQP

amplitudes increased by 5.99 F 3.75 times (range, 2.28-

14.8 times) between these 2 recordings. AIQP amplitudes in

the preinflation recording ranged from 3.83 to 14.8 lV rms.

In the SAECG recorded during the first minute of inflation,

AIQP amplitudes ranged from 17.2 to 116.5 lV rms. There

is no overlap between these 2 ranges. Note that all

measurements throughout the PTCA procedure were made

relative to the model of the preinflation SAECG.

Fig. 4 shows the mean progressions of AIQP, ST

segment, and standard QRS duration (sQRSD) changes,

Fig. 4. Progression of AIQP, ST-segment, and sQRSD changes, averaged

over the 12 subjects.

averaged over the 12 subjects. The traces show similarities

in form. Table 1 gives the significant correlations present

between any 2 of the 3 variables during the first 6 minutes

(preinflation plus balloon inflation epochs). The post-

deflation AIQP measurements were not correlated with

the standard ECG variables of ST deviation and QRS

duration change.

Discussion and conclusions

Significance and pathophysiologic basis of AIQP during

acute transmural myocardial ischemia

The results of the present study suggest that AIQP may

be a time-localized representation of ischemically altered

activation. The evidence for this is the strong correlation

present between AIQP amplitudes and ST-segment devia-

tion and between AIQP timing and the coronary artery

occluded. We hypothesize that AIQPs result from disrupted

activation around an area of functional block, and not from

altered properties of the myocardial action potential. This

seems likely because AIQPs appear to be more sensitive to

the presence of ischemia than ST-segment deviation, yet

ischemic injury currents occur before changes to the

upstroke of the action potential during ischemia. AIQPs

are also present when there is no prolongation of the QRS,

further suggesting that the phenomenon is based on changed

activation patterns contained within the normal period of

ventricular activation. The significance of ST depression

and elevation with acute transmural ischemia is a subject of

ongoing research.25,26 Some ECG leads may show elevation

while others exhibit depression. In the present study, the

intent was to maximize the sensitivity of ST measures by

considering the greatest change in any lead, whether

depression or elevation. With this comparison, AIQPs

emerge as the most sensitive ECG index of ischemia.

As measured by AIQP timing, ischemically altered

activation of the myocardium occurred earliest in the time

course of ventricular activation for LAD occlusions, later for

circumflex, and latest for RCA occlusions. The LAD

supplies 60% of the left ventricle, including the portions

likely to be activated earliest. In contrast, the RCA, as

evidenced by ECG changes during right bundle branch

block, supplies the ventricular myocardium activated latest.

The circumflex artery with its less distinct anatomical path

could be expected to be intermediate and to supply areas

overlapping those supplied by the LAD and RCA. These

projections are supported by the data of Fig. 3. AIQP

timing, that is, the point of disruption in ventricular

activation, occurs first with occlusion of the LAD, next in

the circumflex artery, and latest with occlusion of the RCA.

Fig. 5. Illustration of the differences in AIQP caused by large-scale QRS

morphologic changes and low-amplitude notches and slurs. A, Shows a

patient where large-scale QRS changes during PTCA created AIQPs (cf

Fig. 1). From top to bottom, 4-second ECG rhythm strips are shown for the

preinflation epoch, the five 1-minute inflation epochs, and the postdeflation

epoch. The corresponding subensemble-averaged AIQP waveforms are

shown to the right. B, Shows a patient with no large-scale QRS

morphologic changes. The ECG rhythm strips are invariant and AIQP

waveforms result from low-level QRS changes. Note: AIQP are not shown

to scale with the ECG rhythm strips, or between the 2 patients. The duration

of AIQP shown is the high-resolution QRS duration.

P. Lander et al. / Journal of Electrocardiology 39 (2006) 282–289 287

The variability in AIQP timing with occlusion of the

circumflex artery is also greater (see error bars) than with

the LAD and RCA, as anticipated. These data support the

assertion that the timing of AIQP is consistent with AIQP

representing localized, ischemically altered activation of the

ventricular myocardium. Possibly owing to the small

number of subjects (n = 12), no relationship between AIQP

timing and location of the occlusion site within a specific

artery (proximal, mid, or distal) could be discerned. In

addition, AIQP timing did not appear to be lead specific for

particular arteries in this study. Future work will consider

standard precordial and limb leads measurements.

AIQP amplitudes showed a statistically significant

increase between the preinflation and first minute of balloon

inflation states in all 12 patients studied. The tendency in the

second to fourth minutes of inflation was a further, gradual

increase in AIQP amplitude. Comparison between paired

epochs showed a significant AIQP amplitude augmentation

between the second and third minute of inflation. However,

during the fifth minute, AIQP amplitudes declined in some

cases. One possible explanation may be that the myocardi-

um in and around the region of functional block had become

stunned after 4 minutes of occlusion. Postdeflation, AIQP

amplitudes typically decreased upon reperfusion and then

increased slightly or remained constant. A marked AIQP

reduction was observed during the second minute after

balloon deflation compared with the fourth minute of

occlusion. Further studies are needed to determine if and

when AIQP amplitudes return to baseline (preinflation)

values, and whether there may be any significance, relating

to reperfusion, in the postdeflation AIQP time course.

AIQPs generally result from small-signal changes in QRS

shape. However, 1 subject exhibited a large-scale change in

AIQP amplitudes between the second and third minutes of

balloon inflation. In this case, AIQPs were the result of gross

changes in QRS morphology, as illustrated in Fig. 5A. The

traces from top to bottom show representative 4-second ECG

rhythm strips, and the corresponding AIQP waveforms, for

the preinflation, inflation, and postdeflation 1-minute

epochs. AIQPs emerge because large-scale changes in

QRS morphology occur with significant ST elevation. This

was an unusual phenomenon. Fig. 5B shows a more typical

case. The ECG rhythm strips did not show significant

variation in either QRS morphology or ST-segment deviation

throughout the PTCA procedure. The corresponding AIQP

waveforms reflect low-amplitude changes in QRS morphol-

ogy. (The AIQP waveforms are not shown to scale, either

between patients or with the ECG rhythm strips. Their

duration is that of the high resolution QRS.)

Relationship of the present work to previous studies

Abboud and colleagues8,27-29 have shown that the high-

resolution ECG complex exhibits reduced high-frequency

energy in the bandwidth 150 to 250 Hz with myocardial

ischemia. These changes are not apparent in the standard

ECG. The phenomenon has been demonstrated in both

animal27 and clinical8,9 settings. A computer simulation

suggested that low-level QRS morphology changes could be

attributable to slowing of conduction velocity in a region of

ischemia.29 Aversano et al10 and Berkalp et al11 have

demonstrated that similar changes in high-frequency

content of the QRS after myocardial infarction could be a

marker of successful reperfusion after thrombolytic therapy.

These studies lend weight to the hypothesis of this article,

that ischemically altered activation can often induce

electrical changes visible as microvolt-level potentials in

the HRECG.

However, the techniques used in these prior studies are

oriented toward the frequency domain, considering only the

150- to 250-Hz band. The fixed-frequency spectral window

filter used produces phase distortion in the QRS.30 By their

nature, there is no time resolution available with frequency

domain methods.20 Neither frequency domain methods nor

conventional filtering will allow an assessment of low-level,

time-localized activity in the QRS. The modeling process

used with AIQP filters the HRECG with an arbitrary

frequency characteristic to extract an abnormal signal

wholly in the time domain. In contrast to reduced high-

frequency energy, subtle, time-localized alterations in the

HRECG QRS complex (ie, AIQP) may be a direct marker of

ischemically altered activation.

A quantitative comparison of ST segment and tomo-

graphic scintigraphy changes during PTCA has recently

been reported.31 This study concluded that ST-segment

deviations, when present, yielded only a crude estimate of

the extent of myocardial ischemia. In our study, AIQPs had

a prevalence of 100%, as opposed to 60% for ST-segment

deviation (considering ST deviation greater than 100 lV)and up to 83.3% with a more sensitive criterion (= 25 lV).If AIQPs are verified as a time-localized marker of

ischemically altered activation, they may yield more

accurate electrocardiographic estimates of ischemia. Further

P. Lander et al. / Journal of Electrocardiology 39 (2006) 282–289288

work is planned to characterize AIQP during and up to 24

hours after PTCA. Scintigraphic estimates of ischemia in

both the occluded and reperfused states will be compared

with abnormal HRECG signals in a large cohort of patients.

Advantages and limitations of the experimental design

Handling large amounts of continuous HRECG data

presented significant challenges, including quality control of

signal averaging (final noise levels and correct beat

classification) and developing reliable methods of automat-

ed analysis. The TFPW-filtered SAECG13-15 greatly en-

hanced noise reduction, enabling the computation of high

fidelity SAECGs every 60 seconds. This technique enabled

measurement of serial changes in AIQP amplitude and their

comparison with ST-segment deviation and standard QRS

duration. The modeling of the HRECG QRS complex to

extract microvolt-level AIQPs is a computational method

described in detail in References 5 and 6. The principal step

in the process is choosing the model order, which will affect

the computed AIQP waveform. The model order used (na =

7, nb = 8) was selected on the basis of prior experi-

ence.5,6,22,23 We expect the major findings of this study to

be substantially independent of the choice of model order.

Further studies are planned to determine the most appropri-

ate methods for AIQP calculation.

Balloon occlusions of exact timing and duration during

PTCA provide a model of acute transmural ischemia in a

controlled setting. The clinical environment imposed some

limitations. The most important of these was a limited

period of HRECG acquisition after balloon deflation. This

was typically 3 to 4 minutes. Interpretation of the HRECG

in the short term (up to 1 hour) after balloon deflation

would be problematic because of the expected electrophys-

iologic effects of reperfusion and coronary dilation. The

effects of reperfusion and the potentially altered substrate

for AIQP, due to changed myocardial blood flow, are

unpredictable. Therefore, the significance of AIQP changes

after balloon deflation is difficult to determine. Some

patients had a subsequent balloon inflation shortly after

the first PTCA procedure. However, reperfusion effects and

altered myocardial blood flow preclude using subsequent

inflations to test the reproducibility of the ECG changes

observed after the first balloon inflation. Further experi-

ments are planned to address the long-term (up to 24 hours)

behavior of AIQP.

Support for a local ischemic origin of AIQP comes

primarily from the association between AIQP timing and the

artery occluded. This was a cornerstone of the experimental

design. The number of subjects in the study (n = 12) did not

enable the consistency of AIQP measurements to be

thoroughly evaluated. An expanded study is currently in

progress that will address the issue of consistency by

concentrating on subjects with occlusions of, for example,

the proximal and mid LAD.

Because all subjects had AIQP and acute transmural

ischemia was present by definition during arterial occlusion,

the specificity of AIQP could not be tested. In previous

work, studying a canine model of infarction, we presented

evidence that AIQP are the pathophysiologic signal repre-

sentation of a reentrant substrate for ventricular tachycar-

dia.7 In studies of 2 clinical populations, we have

demonstrated that the amplitude of AIQP is progressively

greater in subjects with myocardial scarring and with a

future arrhythmic event post-MI, compared with subjects

with no prior myocardial infarction.5,22,23 This evidence

suggests that the characteristics of AIQP will depend on the

extent of myocardial ischemia.

Clinical relevance and perspective for future studies

Changes in AIQP that are observed only in the HRECG

appear to be more sensitive for detection of acute transmural

ischemia than QRS or ST-segment changes in the standard

ECG. Therefore, further investigation of their potential

clinical usefulness is indicated, and expanded studies of

HRECG signals during PTCA are in progress. QRS changes

apparent on the standard ECG are even less sensitive than

ST-segment deviations.32 They are incapable of providing

the clinician with the information that appears to be

available in the high-resolution ECG.

From a clinical perspective, the most interesting result of

the present study is the presence of AIQP in patients without

any ST-segment deviation during the period of prolonged

balloon inflation. The strong correlation between AIQP

amplitudes and ST-segment deviations suggests that these

methods might have future potential for a noninvasive

clinical indication of the extent or the severity of acute

ischemia. Future studies with an independent measurement

of ischemia, such as tomographic scintigraphy,33 are needed

to investigate whether the HRECG could provide a more

accurate electrocardiographic measurement of ischemically

altered electrophysiologic changes.

Acknowledgments

This work was supported in part by CICYT grant

TEC2004-02274 from the Spanish government.

References

1. Langner PH. The value of high fidelity electrocardiography using the

cathode ray oscillograph and an expanded time scale. Am Heart J

1952;5:429.

2. Langner PH, Geselowitz DB, Mansure FT. High-frequency compo-

nents in the electrocardiograms of normal subjects and of patients with

coronary heart disease. Am Heart J 1961;62:749.

3. Flowers NC, Horan LG, Thomas R, et al. The anatomic basis for high

frequency components in the electrocardiogram. Circulation

1969;39:531.

4. Flowers NC, Horan LG, Tolleson WG, et al. Localization of the site of

myocardial scarring in man by high-frequency components. Circulation

1969;40:927.

5. Gomis P, Jones DL, Caminal P, Berbari EJ, Lander P. Analysis of

abnormal signals with the QRS complex of the high resolution

electrocardiogram. IEEE Trans Biomed Eng 1997;44:681.

6. Gomis P, Lander P, Caminal P. Parametric linear and non-linear

modeling techniques for estimating abnormal intra-QRS potentials in

the high resolution ECG. In: Cerrolaza M, Jugo D, Brebbia C, editors.

Simulation Modeling in Bioengineering. Southampton (UK)7 Compu-

tational Mechanics Publications; 1996. p. 271.

7. Lander P, Gomis P, Caminal P, Berbari EJ, Ballouz G, Patterson E, et al.

Pathophysiological insights into abnormal intra-QRS signals in the

high resolution ECG. Comput Cardiol 1995;22:273.

P. Lander et al. / Journal of Electrocardiology 39 (2006) 282–289 289

8. Abboud S, Cohen RJ, Selwyn A, Ganz P, Sadeh D, Friedman PL.

Detection of transient myocardial ischemia by computer analysis of

standard and signal-averaged high frequency electrocardiograms in

patients undergoing percutaneous transluminal coronary angioplasty.

Circulation 1987;76:585.

9. Pettersson J, Pahlm O, Carro E, Edenbrandt L, Ringborn M, Sfrnmo L,

et al. Changes in high-frequency QRS components are more sensitive

than ST-segment deviation for detecting acute coronary artery

occlusion. J Am Coll Cardiol 2000;36:1827.

10. Aversano T, Rudikoff B, Washington A, Traill S, Combs V, Raqueno J.

High frequency QRS electrocardiography in the detection of reperfu-

sion following thrombolytic therapy. Clin Cardiol 1994;17:175.

11. Berkalp B, Oral D, Caglar N, Omurlu K, Pamir G, Alpman A, et al.

Effects of percutaneous transluminal coronary angioplasty on late

potentials and high frequency mid-QRS potentials. Cardiology

1994;85:216.

12. Wagner GS, Selvester RH, Wagner NB, Krucoff MW. QRS changes

during acute ischemia induced by balloon occlusion of LAD artery.

J Electrocardiol 1988;21(Suppl):S18.

13. Lander P, Berbari EJ. Performance assessment of optimal filtering of

the high resolution electrocardiogram. Comput Cardiol 1994;21:693.

14. Lander P, Berbari EJ. Time-frequency plane Wiener filtering of the

high resolution ECG: Background and time-frequency representations.

IEEE Trans Biomed Eng 1997;44:247.

15. Lander P, Berbari EJ. Time-frequency plane Wiener filtering of the

high resolution ECG: Development and application. IEEE Trans

Biomed Eng 1997;44:256.

16. Lander P, Berbari EJ. Principles and signal processing techniques of

the high-resolution electrocardiogram. Prog Cardiovasc Dis 1992;

35:169.

17. Steinberg JS, Bigger Jr JT. Importance of the endpoint of noise

reduction in analysis of the signal-averaged electrocardiogram. Am J

Cardiol 1989;63:556.

18. Lander P, Berbari EJ, Rajagopalan CV, Vatterott P, Lazzara R. Critical

analysis of the signal-averaged electrocardiogram. Improved identifi-

cation of late potentials. Circulation 1993;87:105.

19. Lander P, Berbari EJ, Lazzara R. Optimal filtering and quality control

of the signal-averaged ECG: high-fidelity 1-minute recordings.

Circulation 1995;91:1495.

20. Lander P, Albert DE, Berbari EJ. Spectrotemporal analysis of

ventricular late potentials. J Electrocardiol 1990;23:95.

21. Ljung L. System identification: theory for the user, chapter 4 and 10.

Englewood Cliff (NJ)7 Prentice-Hall; 1987.

22. Lander P, Berbari EJ, Lazzara R, Gomis P, Steinberg JS. Analysis of

intra-QRS late potentials improves the predictive value of the signal-

averaged ECG. Pace 1995;18:796.

23. Lander P, Gomis P, Goyal RM, Berbari EJ, Caminal P, Lazzara R, et al.

Analysis of abnormal intra-QRS potentials: Improved predictive value

for arrhythmic events of the signal-averaged electrocardiogram.

Circulation 1997;95:1386.

24. Braunwald E, Antman EM, Beasley JW, et al. ACC/AHA guidelines for

the management of patients with unstable angina and non-ST-segment

elevation myocardial infarction. J Am Coll Cardiol 2000;36:970.

25. Atar S, Birnbaum Y. Ischemia-induced ST-segment elevation: classi-

fication, prognosis, and therapy. J Electrocardiol 2005;38(Suppl):1.

26. Peterson ED, Hathaway WR, Zabel KM, Pieper KS, Granger CB,

Wagner GS, et al. Prognostic significance of precordial ST segment

depression during inferior myocardial infarction in the thrombolytic

era: results in 16,521 patients. J Am Coll Cardiol 1996;28:305.

27. Mori-Avi V, Shargorodsky B, Abboud S, Laniado S, Akselrod S.

Effects of coronary occlusion on high frequency content of the

epicardial electrogram and body surface electrocardiogram. Circulation

1987;76:237.

28. Abboud S. High-frequency electrocardiogram analysis of the entire

QRS in the diagnosis and assessment of coronary artery disease. Prog

Cardiovasc Dis 1993;35:311.

29. Abboud S, Berenfeld O, Sadeh D. Simulation of high-resolution QRS

complex using a ventricular model with a fractal conduction system.

Effects of ischemia on high-frequency QRS potentials. Circ Res

1991;68:1751.

30. Lander P, Berbari EJ. Use of highpass filtering to detect late potentials

in the signal averaged ECG. J Electrocardiol 1989;22(Suppl):7.

31. Steg PG, Faraggi M, Himbert D, Juliard JM, Cohen-Solal A, Lebtahi

R, et al. Comparison using dynamic vector cardiography and MIBI

SPECT of ST-segment changes and myocardial MIBI uptake during

percutaneous transluminal coronary angioplasty of the left anterior

descending coronary artery. Am J Cardiol 1995;75:998.

32. Mager A, Sclarovsky S, Wurtzel M, Menkes H, Strasberg B, Rechavia

E. Ischemia and reperfusion during intermittent coronary occlusion in

man. Studies of electrocardiographic changes and CPK release. Chest

1991;99:386.

33. Gibbons RJ, Verani MS, Behrenbeck T, Pellikka PA, O’Connor MK,

Mahmarian JJ, et al. Feasibility of tomographic technetium-99m-

hexakis-2-methylpropyl-isonitrile imaging for the assessment of

myocardial area at risk and the effect of acute treatment in myocardial

infarction. Circulation 1989;80:1277.