Dynamic Response of Electrocardiographic Indices During Abrupt Heart

Rate Changes

Comparison between Young and Middle-aged Subjects

Marcos Javier Teperino

, Mar´ıa Paula Bonomini

, Pablo Daniel Cruces

1,2

and Pedro David Arini

1,2

Instituto de Ingenier´ıa Biom´edica, Facultad de Ingenier´ıa, Universidad de Buenos Aires, Buenos Aires, Argentina

Instituto Argentino de Matem´atica, ’Alberto P. Calder´on’ CONICET, Buenos Aires, Argentina

Keywords:

Ventricular Repolarization Dispersion, T-wave Morphology, Age.

Abstract:

Abnormal modiﬁcations in ventricular repolarization dispersion (VRD) have been shown to constitute a sub-

strate for malignant arrhythmias. In this work, we have induced abrupt heart rate (HR) changes to young

and middle-aged healthy subjects through a Tilt-test and have analyzed the evolution of several VRD indices.

Duration ones, based on electrocardiogram intervals; energy ones, developed through a Principal Components

Analysis (PCA) in T-wave; and the morphology ones, extracted feature from an absolute T-wave. In both

groups, results have shown signiﬁcant decreases in early repolarization duration. These changes are responsi-

ble for the alterations in the total repolarization duration, because T-wave peak-to-end has not shown statistical

signiﬁcance. Moreover, we have found signiﬁcant decreases in total, early and late repolarization energy, and

in the T-wave amplitude. In another sense, we have observed that the repolarization energy obtained by PCA

jointly with early T-wave slope and amplitude have been able to reﬂect VRD differences between young and

middle-aged subjects. Finally, this work provides the range of values for VRD in normal conditions during

abrupt HR changes. Outside this range, we could assume that it exists a cardiac risk.

1 INTRODUCTION

The electrical inhomogeneities in the cardiac repolar-

ization process along the ventricular wall are usually

called ventricular repolarization dispersion (VRD).

Several works have shown that abnormal alterations

of VRD are associated with a higher risk of develop-

ing ventricular arrhythmias (Surawicz, 1997).

Some authors have shown that alterations in VRD

are correlated with changes in T-wave width (Fuller

et al., 2000). It has also shown that T-wave widening

can result from a differential shortening or lengthen-

ing of the action potential duration in both apex-base

and transmural (Arini et al., 2008).

The T-wave peak-to-end interval has been sug-

gested as a marker of transmural VRD (Zareba et al.,

2000; Antzelevitch et al., 2007; Smetana et al.,

2011), and so, the T-wave peak as an indicator of

the full repolarization of epicardium. The transla-

tion of this concept to the standard electrocardiogram

(ECG) is not straightforward, making it difﬁcult to

interpret of the relationship between T-wave peak-to

end and transmural dispersion in a clinical population

(Smetana et al., 2011). Another study has shown that

during Valsalva maneuver the T-wave width shorten-

ing seems to result from a width reduction from the

onset to the T-wave peak rather than from the peak to

the T-wave end (Minchol´e et al., 2006).

Moreover, others have proposed repolarization in-

dices such as the QT interval (Pueyo et al., 2004)

or the T-wave peak-to-end interval (Minchol´e et al.,

2011) depend on heart rate (HR) and such dependence

has also been related to arrhythmic risk. However,

some researchers have objected the validity of these

two indices as markers of VRD (Malik et al., 2000;

Opthof et al., 2007) and have questioned their depen-

dence of HR (Andersen et al., 2008). It has also been

suggested that several morphological indices, such as

the slopes and the area of the T-wave, are independent

of HR (Merri et al., 1989).

On the other hand, several VRD descriptors based

on Principal Component Analysis (PCA) have been

used to differentiate normal and abnormal VRD pat-

terns (Zabel et al., 2002) and have been used to quan-

tify pathological characteristics of VRD at high HR

(Smetana et al., 2004).

Finally, some authors have found that most repo-

larization indices are independent of age (Merri et al.,

264

Javier Teperino M., Bonomini M., Cruces P. and Arini P..

Dynamic Response of Electrocardiographic Indices During Abrupt Heart Rate Changes - Comparison between Young and Middle-aged Subjects.

DOI: 10.5220/0005253102640269

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2015), pages 264-269

ISBN: 978-989-758-069-7

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

1989). Nevertheless, others have shown that aging

modulates the normal VRD (Huang et al., 2010).

In the present work, we have analyzed the re-

sponse of several ECG indices to HR changes. So,

we have determined the range of normal values out-

side which we could assume that it exists a cardiac

risk. Finally, we have studied how aging might inﬂu-

ence in the dynamic response of electrocardiographic

indices during abrupt HR changes.

2 METHODS

2.1 Database

A total of 17 subjects without previous cardiovascu-

lar diseases were retrospectively studied. Two groups

were selected: Young Subjects (8 subjects, with a

mean of 26.5±7.5 years old) and Middle-aged Sub-

jects (9 subjects, with a mean of 65.8±11.2 years

old). Subjects were recruited in the medical insti-

tution Cl´ınica San Camilo of the Buenos Aires, Ar-

gentina and in all cases informed consent was signed.

Each subject recorded has undergone a head-up tilt

test trial according to the following protocol: 5 min

in the supine position, 5 min tilted head-up to an an-

gle of 70 degrees, and 18 sec during table movement.

This method generates an abrupt acceleration of the

heart rate. The ECG leads I, III, V1-V6 were recorded

during the whole test using ECG View Eccosur equip-

ment with a sampling rate of 1000 Hz.

2.2 Preprocessing

The ECG signals have been preprocessed as fol-

lows: 1) QRS complexes have been detected and

normal beats have been selected according to the

method in (Mendieta, 2012), 2) A Butterworth high-

pass ﬁlter (0.5Hz, bidirectional) has been applied

for baseline wander rejection and in order to reduce

high frecuency noise, a Butterworth low-pass ﬁlter

(100Hz, bidirectional) has been used, and 3) T-waves

and QRS-complexes have been delineated using the

wavelet-transform based method in (Mendieta, 2012).

2.3 Repolarization Indices

Several indices have been selected to describe the fol-

lowing characteristics of VRD on the surface ECG: 1)

Duration, 2) Energy and 3) Morphology.

For all indices we have applied a multilead cri-

terion to determine wave boundaries, where QRS

END

and T

END

are respectively the lastest reliable QRS-

complex and T-wave ends at any lead (I, III, V1-

V6). Also, the T-wave peak (T

PEAK

) and R-wave peak

PEAK

) as median values of all leads have been com-

puted with an outlier protection rule.

For each i

beat, we havecomputed the aforemen-

tioned characteristics of VRD.

1) Indices of Duration: The total repolarization

duration index (T

), quantifying the total ventric-

ular repolarization time, has been calculated as;

= T

END

− QRS

END

(1)

The early repolarization duration index (E

)

and the T-wave peak-to-end interval index (T

which several authors have linked to the full repolar-

ization of epicardium and transmural repolarization

respectively (Antzelevitch et al., 2007), have been

computed as;

= T

PEAK

− QRS

END

(2)

and,

= T

END

− T

PEAK

(3)

2) Indices of Energy: We have obtained T-wave

energy based on the PCA of the ECG leads. For the

calculus of PCA indices in each i

beat, it has been

considered three windows, one for total repolariza-

tion duration (time between QRS

END

and T

END

) and

one for each portion (early and late) of the same T-

wave (partitioned in T

PEAK

). PCA has been applied in

the set of the independent leads, from which 8 eigen-

values have been obtained. We have denoted them

by λ

i, j

(j = 1, . . . , 8), where they are sorted so that

i,1

≥ λ

i,2

≥ ..... ≥ λ

i,8

≥ 0. Then, we have computed

the Total energy as the sum of the eight eigenvalues.

∑

l=1

i,l

(4)

This way, we have obtained E

for the full T-

wave; E

for the ﬁrst half of the T-wave (or Early T-

wave) and E

for the Late T-wave (or T-wave peak-

to-end).

3) Indices of Morphology: Afterwards, we have

computed an absolute T-wave (T

ABS

), through the sum

of the eight T-waves in each i

beat, let

ABS

(k) =

∑

j=I,III,V1-V6

(k)| k = QRS

END

, ... , T

END

(5)

where x

(k) is the ECG signal at lead j, then a

polynomial ﬁtting has been applied for each i

ABS

obtaning

ABS

. Then the following indices were cal-

culated: T-wave Amplitude index (T

), calculated

as the amplitude of the

ABS

wave peak; Slope of

DynamicResponseofElectrocardiographicIndicesDuringAbruptHeartRateChanges-ComparisonbetweenYoungand

Middle-agedSubjects

265

0 100 200 300 400

175

185

195

205

215

[ms]

t [s]

[s]

∆E

[ms]

Figure 1: Example of characterization of E

index. The

dotted line shows the index value and bold line represents

the ﬁtting curve.

Early T-wave index (S

), maximum slope obtained

through a ﬁve point centered derivative in whole ﬁrst

half of

ABS

wave and Slope of Late T-wave index

), absolute value of the maximum slope obtained

through the ﬁve point centered derivative in whole

second half of

ABS

wave.

2.4 Series of Repolarization Indices

For each index, the values beat-to-beat have been con-

catenated and so obtained the series of values over

time: RR, T

, E

, T

, E

, T

, S

and S

, during the Tilt-test Maneuver. A numerical

interpolation has been applied using the R

PEAK

values

as time reference for all beats in order to resample to

1 Hz. It has also been applied a median ﬁlter with a

windows size of 20 seconds. Series have been charac-

terized through a numerical ﬁtting with a linear com-

bination of two exponentials, as shown in Eq. (6),

where a

, ... , a

are the ﬁtting parameters.

(n)

= a

+ a

(6)

The optimization is based on the minimization of

the sum squared error (SSE) of each series, as illus-

trated in Eq. (7) where I

(n)

represents the index under

study.

∂e

∂a

∂

∂a

∑

n=1

(n)

−

(n)

)

= 0 (7)

We have obtained the optimum starting point of

the Tilt-test Maneuver (so-called t

), by minimizing

the global error function for each adjustment an so,

each series have been computed. An example of this

ﬁtting is shown in Fig. 1.

160

180

200

220

240

260

50 100 150 200 250 300 350

-40

-60

-20

[ms]

t [s]

∆E

[ms]

Figure 2: Evolution of E

for several subjects of the

database. Example of normalization for statistical analysis

(SA#1).

2.5 Characterization of Repolarization

Series and Statistical Analysis

In order to analyze just the dynamics of each series

following abrupt HR changes, we have performed a

normalization procedure by subtracting, for all series

samples, the value at t

of each series, see Eq. (8).

For convenience, we have named I = I(t) where I rep-

resents the evaluated index.

∆I = I− I(t

) (8)

We have characterized each series of repolariza-

tion indices through two parameters, such as

1) Index variation (∆), which characterize the in-

dex value difference between t

(0%) and 100% of

change. Also, we have determined the indices with

statistically signiﬁcant changes respect to zero value

) applying a two-sided Wilcoxon signed rank test.

This procedure has been so-called Statistical Analy-

sis #1, SA#1 (see Fig. 2).

2) Response time (t

), the time required to change

from 10% to 90% of index value, as can be observed

in Fig. 1.

BIOSIGNALS2015-InternationalConferenceonBio-inspiredSystemsandSignalProcessing

266

Moreover, another statistical analysis has been

performed for the comparison between two popu-

lations, Young and Middle-aged Subjects. A non-

parametric two-sided Mann-Whitney U test has been

used and when p < 0.05, differences have been con-

sidered statistically signiﬁcant. This procedure has

been so-called Statistical Analysis #2 , SA#2.

3 RESULTS

The statistical results obtained for each series are

shown in Fig. 3.a for young subjects and in Fig.

3.b for middle-aged subjects. Except ∆T

in both

population and ∆S

in middle-aged subjects, all in-

dices have shown signiﬁcant differences in their val-

ues in response to physiological changes induced by

abrupt HR changes. Moreover, not all indices reach

the steady state at the same time. Table 1 shows t

val-

ues in Mean±SEM for all proposed VRD indices in

both population. Finally, in Table 2 we have pointed

out the comparison between the two groups.

Table 1: Response time t

expressed in seconds

(Mean±SEM) for both groups Young and Middle-aged

Subjects.

Indice Young Middle-aged

RR 171.3± 27.0 90.7± 27.3

163.3± 21.9 100.0± 11.4

150.6± 26.3 107.0± 11.3

177.6± 30.2 104.2± 30.8

135.3± 26.9 149.3± 23.7

133.1± 25.5 145.2± 23.0

129.3± 24.5 147.8± 26.2

122.9± 26.7 134.1± 27.6

114.8± 27.0 163.8± 29.8

150.6± 33.8 158.6± 29.7

Table 2: Comparison (SA#2) between Young and Middle-

aged Subjects (Mean±SEM).

Young Middle-aged p value

∆E

35.59±8.30 7.40±2.38 <0.05

∆E

23.68±5.92 5.46±1.95 <0.05

∆E

7.19±2.09 1.74±0.47 <0.05

∆T

1.03±0.28 0.43±0.14 <0.05

∆S

0.018±0.005 0.006±0.004 <0.05

4 DISCUSSION

In this work we have proposed a multilead crite-

rion to analyze VRD alterations caused by abrupt HR

changes using duration, energy and morphology in-

dices. To do so, we have studied each of them sepa-

rately and then, we have hypothesized about their re-

lationships and implications.

Duration variables. We have observed, in both

young and middle-aged subjects, a statistical signif-

icant decrease of ∆T

and ∆E

in response to RR

interval decrease, while ∆T

has not changed (see

Fig. 3.a and Fig. 3.b). These results are consistent

with a previous work (Cruces et al., 2014). In the lat-

ter, we have shown alterations in both T

(T-wave

width) and T

(T-wave onset-to-peak duration) dur-

ing HR changes in another database (Autonomic Ner-

vous System Database, University of Zaragoza). Our

results are comparable to those in (Merri et al., 1989),

where it has shown that QT interval and early du-

ration (time interval between QRS-complex-end and

T-wave-peak) are the only intervals with signiﬁcant

changes in VRD under HR changes . Conversely, An-

dersen et al. have calculated HR dependence for ECG

duration markers showing independence in both T

and T

(Andersen et al., 2008). In another sense,

as it can be seen in Table 2, no signiﬁcant changes

in the duration indices have been observed in young

with respect to middle-aged subjects according to the

procedure SA#2 described in Subsection 2.5.

Energy Variables. We have performed the energy of

the T-wave by applying PCA method with the infor-

mation of all ECG leads. In response to HR increases

we have found statistically signiﬁcant decreases of

∆E

, ∆E

and ∆E

in both young and middle-aged

subjects (applying SA#1), as we have shown in Fig.

3.a and Fig. 3.b.

Andersen et al. have computed and analyzed sim-

ilar indices, such as early and late T-wave areas. They

have shown dependence of HR in each one (Andersen

et al., 2008), then we have consistent results with our

energy indices obtained by PCA procedure.

Also, as we have observed in Table 2, the energy

of total (∆E

), early (∆E

) and late (∆E

) repolar-

ization have been able to reﬂects VRD differences be-

tween young subjects with respect middle-aged sub-

jects (Subsection 2.5 SA#2).

Morphology Variables. Regarding morphologic

analysis in young subjects, we have found statistically

signiﬁcant decreases in ∆T

, ∆S

and ∆S

(see Fig.

3.a). Moreover, ∆T

and ∆S

were signiﬁcantly in

middle-aged subjects and not signiﬁcance of ∆S

in-

dex (see Fig. 3.b). Results obtained in young sub-

jects were consistent with another work (Andersen

et al., 2008), who have shown signiﬁcant changes in

both up and down slopes T-wave. Andersen et al.

have concluded that those variations were just about

T-wave amplitude decrease, because they have not

DynamicResponseofElectrocardiographicIndicesDuringAbruptHeartRateChanges-ComparisonbetweenYoungand

Middle-agedSubjects

267

mv111

(a) (b)

∆RR

∆E

∆T

∆S

∆T

∆E

[%]

[msec]

[msec][msec][msec][msec]

[mV/sec]

Figure 3: Box and Whisker diagram for all VRD indices. (a) Young subjects; (b) Middle-aged subjects. Statistical signiﬁcance

(SA#1) has been represented as ’x’ for p<0.05 and ’#’ for p<0.005. Non signiﬁcant changes have been indicated as NS.

found modiﬁcations in T-wave onset-to-peak. How-

ever, our results suggest that both T-wave amplitude

and early repolarization duration were responsible for

those variations. Also, we have found signiﬁcant

differences between young subjects and middle-aged

subjects in T

and S

indices (as we have explained

in SA#2). Conversely, no signiﬁcant changes in S

has been observed with SA#2.

Response Time. Table 1 shows that RR index reach

the steady state after duration, energy and morphol-

ogy VRD indices for young subjects. Conversely,

for middle-aged subjects the time required to change

from 10 % to 90 % (t

) in the VRD indices are greater

than the t

measured in the RR index. These results

allows us to hypothesize that there are different times

of adaptation under abrupt HR changes, possibly re-

lated to aging effects. Also, as it can be seen, dura-

tion indices have shown faster adaptation to high HR

in middle-aged subjects than in young subjects. How-

ever, energy and morphology indices are faster in the

latter population.

5 CONCLUSION

Given that E

and T

indices have shown statis-

tical signiﬁcance and T

has not, we have also con-

cluded that HR increases induce a shift in the T-wave

peak position towards the QRS-complex. We have

concluded that under abrupt changes in HR, the main

alterations of VRD correspond to the variations in the

the duration of action potentials which do not affect

differentially epicardium from endocardium tissues.

Moreover, T

, S

, E

and E

have

shown statistically signiﬁcant differences under HR

increases. We can hypothesize that these effects

have a physiological limit because several repolariza-

tion indices reach the steady state before RR interval

(Young Subjects) and others indices have presented

aging effects due to they reach the steady state after

RR interval (Middle-aged Subjects).

Some authors have shown that several repolar-

ization indices are independent of age (Merri et al.,

1989), nevertheless others have shown that aging

modulates the normal VRD (Huang et al., 2010). In

our study we have concluded that the energy of to-

tal (E

), early (E

) and late (E

) repolarization have

BIOSIGNALS2015-InternationalConferenceonBio-inspiredSystemsandSignalProcessing

268

been able to reﬂects VRD differences between young

and middle-aged healthy subjects.

Finally, this study constitutes a basis for setting

normal conditions of the ventricular repolarization

process in young and middle-aged subjects.

ACKNOWLEDGEMENTS

This work was supported by Consejo Nacional de In-

vestigaciones Cient´ıﬁcas y T´ecnicas CONICET, un-

der project PIP #538 and PICT 2008 #2108 Agencia

Nacional de Promoci´on Cient´ıﬁca y Tecnol´ogica, Ar-

gentina.

REFERENCES

Andersen, M., Xue, J., Graff, C., Kanters, J., Toft, E., and

Struijk, J. (2008). New descriptors of T-wave mor-

phology are independent of heart rate. J Electrocar-

diol, 41:557–561.

Antzelevitch, C., Viskin, S., Shimizu, W., Yan, G., Kowey,

P., Zhang, L., Sicouri, S., Di Diego, J., and Burash-

nikov, A. (2007). Does Tpeak-Tend provide an index

of transmural dispersion of repolarization? Hearth

Rhythm, 4(8):1114–1119.

Arini, P. D., Bertr´an, G. C., Valverde, E. R., and Laguna, P.

(2008). T-wave width as an index for quantiﬁcation of

ventricular repolarization dispersion: Evaluation in an

isolated rabbit heart model. Biomed Signal Proc and

Control, 3:67–77.

Cruces, P., Bonomini, M., Teperino, M., Minchol´e, A., La-

guna, P., and Arini, P. (2014). Normal ventricular

repolarization dispersion range with abrupt heart rate

changes. In Computing in Cardiology. In press.

Fuller, M. S., Sndor, G., Punske, B., Taccardi, B., MacLeod,

R. S., Ershler, P. R., Green, L. S., and Lux, R. L.

(2000). Estimates of repolarization and its dispersion

from electrocardiographic measurements: direct epi-

cardial assesment in the canine heart. J Electrocardiol,

33:171–180.

Huang, J.-H., Lin, Y.-Q., Pan, N.-H., and Chen, Y.-J.

(2010). Aging modulates dispersion of ventricular re-

polarization in the very old of the geriatric population.

Heart Vessels, 25:500–508.

Malik, M., Acar, B., Gang, Y., Yap, Y. G., Hnatkova, K., and

Camm, A. J. (2000). QT dispersion does not represent

electrocardiographic interlead hetereogeneity of ven-

tricular repolarization. J Cardiovasc Electrophysiol,

11:835–843.

Mendieta, J. G. (2012). Algoritmo para el delineado

de se˜nales ECG en un modelo animal empleando

t´ecnicas avanzadas de procesamiento de se˜nales.

Master Thesis. , Facultad de Ingenier´ıa de la Univer-

sidad de Buenos Aires.

Merri, M., Benhorin, J., Alberti, M., Locati, E., and Moss,

A. J. (1989). Electrocardiographic quantitation of ven-

tricular repolarization. Circulation, 80:1301–1308.

Minchol´e, A., Mart´ınez, J., Arini, P., Risk, M., and Laguna,

P. (2006). T wave width alterations during valsalva

maneuver in diabetic patients. In Computers in Car-

diology, volume 33, pages 709–712.

Minchol´e, A., Pueyo, E., Rodriguez, J. F., Zacur, E.,

Doblar´e, M., and Laguna, P. (2011). Quantiﬁcation

of restitution dispersion from the dynamic changes of

the T wave peak to end, measured at the surface ECG.

IEEE Trans Biomed Eng, 58(5):1172–1182.

Opthof, T., Coronel, R., Wilms-Schopman, F. J. G., Plot-

nikov, A. N., Shlapakova, I. N., Danilo, P., Rosen,

M. R., and Janse, M. J. (2007). Dispersion of re-

polarization in canine ventricle and the electrocardio-

graphic T wave: Tp-e interval does not reﬂect trans-

mural dispersion. Heart Rhythm, 4:341–348.

Pueyo, E., Smetana, P., Caminal, P., Bayes de Luna, A.,

Malik, M., and Laguna, P. (2004). Characterization

of QT interval adaptation to RR interval changes and

its use as a risk-stratiﬁer of arrythmic mortality in

amiodarone-treated survivors of acute myocardial in-

farction. IEEE Trans Biomed Eng, 51(9):1511–1520.

Smetana, P., Batchvarov, V., Hnatkova, K., Camm, A. J.,

and Malik, M. (2004). Ventricular gradient and

nondipolar repolarization components increase at

higher heart rate. Am. J Physiol. Heart Circ. Physiol.,

286:H131–H136.

Smetana, P., Schmidt, A., Zabel, M., Hnatkova, K., Franz,

M., Huber, K., and Malik, M. (2011). Assessment of

repolarization heterogeneity for prediction of mortal-

ity in cardiovascular disease: peak to the end of the

T wave interval and nondipolar repolarization compo-

nents. J Electrocardiol, 44:301–308.

Surawicz, B. (1997). Ventricular ﬁbrillation and disper-

sion of repolarization. J. Cardiovasc. Electrophysiol.,

8:1009–1012.

Zabel, M., Malik, M., Hnatkova, K., Papademetriou, M. D.,

Pittaras, A., Fletcher, R. D., and Franz, M. R. (2002).

Analysis of T-wave morphology from the 12-lead

electrocardiogram for prediction of Long-Term prog-

nosis in male US veterans. Circulation, 105:1066–

1070.

Zareba, W., Couderc, J., and Moss, A. (2000). Automatic

detection of spatial and temporal heterogeneity of re-

polarization In: Dispersion of ventricular repolariza-

tion pp.85-107. Futura Publishing Company, Inc.

DynamicResponseofElectrocardiographicIndicesDuringAbruptHeartRateChanges-ComparisonbetweenYoungand

Middle-agedSubjects

269