Available online at www.sciencedirect.

com

ScienceDirect
Procedia Materials Science 10 (2015) 307 – 313

2nd International Conference on Nanomaterials and Technologies (CNT 2014)

Detection of T-Wave Alternans in ECGs by Wavelet Analysis

Vasudha Nannaparajua,* and Sundararajan Narasimmanb
a
Department of Mathematics, Vasavi College of Engineering,Ibrahimbagh,Hyderabad,Telengana-500029,India
b
Department of Earth Science, Sultan Qaboos University, Al Khod, Muscat,Post Box 36, Postal Code123, Sultanate of Oman

Abstract

T wave alternans (TWA) is an acute and serious problem associated with the heart and mostly they are reflected in
Electrocardiogram (ECG). It is identified as an alteration of the amplitude and morphology of the T wave that occurs in every
other beat indicating electrical instability in acute ischemia, where it may precede ventricular tachyarrhythmia. The analysis of
TWA is introduced recently as a new diagnostic tool for identification of patients with an increased risk of ventricular
tachyarrhythmia or sudden cardiac death. ECGs recorded by sophisticated equipment with specific software can only diagnose
TWA. The hard and software are exorbitantly costlier and therefore not available everywhere. In the present study, the TWA is
brought out very well from the amplitude of continuous Wavelet transform (CWT) which is not only cheap and simple but also
highly reliable and may be considered as an alternate tool for such investigations. This study is also substantiated by the
exclusively proposed new mother wavelet namely ECG wavelet. The results are presented highlighting the salient features.
© 2015
© 2015 The
TheAuthors.
Authors.Published
Publishedby by
Elsevier Ltd.Ltd.
Elsevier This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the International Conference on Nanomaterials and Technologies (CNT 2014).
Peer-review under responsibility of the International Conference on Nanomaterials and Technologies (CNT 2014)
Keywords: T wave alternans; Electrocardiogram; continuous Wavelet transform; discrete Wavelet transform

1. Introduction

Electrical alternans of the T wave describe an alternating beat-to-beat amplitude change in the T wave. It was
reported as a marker for ventricular arrhythmias Rosenbaum et al. (1994). Considerable interest has recently been
shown in the detection of microvolt TWA as a noninvasive marker of the risk of ventricular tachyarrhythmia in
patients with chronic heart disease.

* Corresponding author. Mobile: +919441779840

E-mail address: nvasudha_raj@yahoo.co.in

2211-8128 © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the International Conference on Nanomaterials and Technologies (CNT 2014)
doi:10.1016/j.mspro.2015.06.055
308 Vasudha Nannaparaju and Sundararajan Narasimman / Procedia Materials Science 10 (2015) 307 – 313

Macroscopic alternans of the T wave is a rare appearance and it is not only linked to malignant ventricular
arrhythmias more frequently in the long QT syndrome Khalameizer et al. (2005), but also after administration of
some drugs Tan and Wilde (1998). TWA can be found even in Normal ECGs and is generally known that the
variation of magnitude and/or shape of T wave in the microvolt domain involve the same arrhythmic risk as do
macroscopic alternans. The exact mechanism of alternans is yet to be known. However, it can be determined by
spatial and/or temporal differences in repolarization of different regions of myocardium Chinushi et al. (1998)
generating unidirectional block and re-entrant arrhythmias or Ventricular Fibrillation (VF) Pastore et al. (1999).

2. Electrocardiogram (ECG)

To interpret an ECG it is necessary to understand the components making up the ECG signal. Each wave and
interval appears on ECG is a result of particular electrical function of the heart. Each of three major waves of
electric signals appears on ECG represents a different part of the heartbeat. The first wave that appears on ECG is
called P wave which records the electrical activity of the heart's two upper chambers (atria).The second and largest
wave, the QRS, records the electrical activity of the two lower chambers (ventricles) of the heart. The third wave is
the T wave that records the returning to the resting state of heart.
The time interval (duration), or distances between two points in the ECG is as significant as identifying these
waves themselves. In other words, the onset and offset of P, QRS and T waves are of great significance in
deciphering the health of a heart Sundararajan et al. (2011). They provide information about the progression of
impulses through the cardiac cycle. The basic component known as cycle of ECG rhythm is given hereunder in
Figure 1.

Figure 1 Cycle of ECG

3. Wavelet Transform

Over the last two decades, the Wavelet transform (WT) has proven to be a valuable tool in many applications of
non-stationary signals including the biomedical signals in general and ECG in particular. The WT provides a time
frequency representation of the signal, and thus suitable for the inspection of characteristic waves of the ECG signal
at different scales with different resolutions. WT has already been used successfully in various areas of electro
 Vasudha Nannaparaju and Sundararajan Narasimman / Procedia Materials Science 10 (2015) 307 – 313 309

cardiology namely to identify the markers from a multi-scale approach Sahambi et al. (1997), detection of ischemia
in the QRS complex Gramatikov and Thakor (1993) and to characterize the time frequency signature of HRV
Thurner et al. (1998). The time series pertaining to inter-beat interval (distance between two consecutive R waves)
of the human heart exhibits scaling behavior, multiresolution wavelet analysis etc. Mallat (1989), Meyer (1990) and
Daubechies (1992) provides an ideal means of decomposing ECG signal into its components at different scales, and
at the same time has the salutary effect of eliminating non stationarities Arneodo et al. (1988), Abry and Flandrin
(1996) and Teich et al. (1996). Wavelet based methods for compression of ECG signals has been proved to perform
well Lu et al. (2000) and Miaou et al. (2002).

4. Materials and Methodology

Wavelets and all their variants result in two dimensional continuous and discretized outputs of two parameters
namely the scale and time shift. While the time shift corresponds to time axis of the signal to be analyzed, scale
corresponds but is not equal to frequency. However, in wavelet analysis of a given signal, we have to take into
account the type of the Wavelet transform, type of the mother wavelet and the desired scale.
An infinite number of options are available in wavelet analysis involving the above aspects which are closely
related and must be considered at the same time. However, the present study confines to a few limited options. The
first step usually is the type of the Wavelet transform to be considered leading to decide whether to use continuous
Wavelet transform (CWT) or discrete Wavelet transform (DWT). The CWT is much more accurate measure of ECG
pattern Aldroubi and Unser (1996) and Rajendra Acharya et al. (2002). The type of mother wavelet (here after called
wavelet) should be set according to the signal being analyzed and/or the event being detected in the signal. Based on
a previous extended survey on wavelets suitable to ECG analysis Andreao (2004) the wavelet namely, complex
Gaussian wavelet given as

d 2 iZ0t t 2
\ t
dt2

e e  Z0 t 5 (1)

is found to be most useful in such a study. Real and imaginary parts of complex Gaussian wavelet are given in
Figure 2.

Figure 2 Real and imaginary parts of Gaussian wavelet

310 Vasudha Nannaparaju and Sundararajan Narasimman / Procedia Materials Science 10 (2015) 307 – 313

On the other hand, the proposed wavelet namely the ECG wavelet Vasudha (2008) exclusively for the present
study is investigated. The proposed wavelet is somewhat similar to Morlet wavelet and is defined as

2 1 t
4

\t S 4
1t e
4 iZ0t
e 2
Z0 t 5 (2)
3
The real and imaginary parts of ECG wavelet are given in Figure 3.

Figure 3 Real and imaginary parts of ECG wavelet.

The proposed ECG wavelet was established based on trial and error basis and equally applicable in the
study of ECG signal even in noisy environment. This wavelet qualify to be standard wavelets as it satisfy the
conditions of zero mean ( f   ) and compact support. The requirement of zero mean is called the admissibility
³\ t dt 0
f

condition of the wavelet. The selection of scale is based on the number of details in what frequency range is needed
for a particular application.

4.1 Denoising of ECG using DWT

ECGs are invariably contaminated by drift and interference caused by several bioelectric phenomena, or by
various types of noise, like intrinsic noise from the recorder and noise from electrode skin contact etc. Unfortunately
there is no universal method to reduce noise, since the probability distributions of noise are different. However,
wavelet analysis ensures minimization of such noise background associated with ECGs and the process is known as
wavelet ‘denoising’ Li et al. (1995), Hardle et al. (1997) and Jansen (2001). In this study an eighth lead (V2) of
ECG was digitized at the rate of 100 Hz which is adequate enough to accommodate the useful information in an
ECG. It was denoised using the DWT with the biorthogonal wavelet ‘bior1.1’ before subjecting for further analysis.
The wavelet ‘bior1.1’ is given in Figure 4.
 Vasudha Nannaparaju and Sundararajan Narasimman / Procedia Materials Science 10 (2015) 307 – 313 311

Figure 4 Decomposition and Reconstruction filters of ‘bior1.1’

In DWT, a time scale representation of a digital signal is obtained using digital filtering techniques. Further,
filters of different cut-off frequencies are used to analyse the signal at different scales. The signal is passed through a
series of high-pass filters to analyse high frequencies, and then it is passed through a series of low-pass filters to
analyse low frequencies. The resolution of the signal, which is a measure of the amount of detail information in the
signal, is changed by the filtering operations, and the scale is changed by up-sampling and down-sampling (sub-
sampling) operations. While up-sampling of the signal corresponds to increasing the sampling rate of a signal by
adding new samples to the signal and sub-sampling of a signal corresponds to reducing the sampling rate, or
removing some of the samples of the signal. A denoised three cycles of an eighth lead from ECG of a normal
person with an obscured T wave located at 40, 100 and 160 milli seconds is given in Figure 5.

Figure 5 Denoised eighth lead of normal ECG

4.2 Detection of TWA by CWT

On computing the CWT of the denoised lead results amplitude and phase for further analysis. The CWT of a
signal is computed by changing the scale of the mother wavelet, shifting it along the time and convolving it with the
signal. The amplitude of CWT coefficients was obtained using the complex Gaussian wavelet and shown in the
form of 2D contoured images in Figure 6.
312 Vasudha Nannaparaju and Sundararajan Narasimman / Procedia Materials Science 10 (2015) 307 – 313

Figure 6 Amplitude of the CWT with the Gaussian wavelet

Figure 7 shows the computed amplitude of CWT in the form of 2D contour image using the ECG wavelet.

Figure 7 Amplitude of the CWT with the ECG wavelet

5. Results and Discussions

The study of TWA of normal ECG although results three prominent QRS complexes, the presence of T wave is
obscure in time domain (Figure (5)). However, the amplitude of CWT with both the Gaussian and the ECG wavelets
has brought out very explicitly the T phases. It is to be noticed that the amplitude level of T wave vary marginally
in time domain but prominent variation is observed in Figures (6 and 7) based on CWT analysis. This suggests that
the subject although is normal ECG, evidences of TWA is clear and is suggestive of either an existing problem with
heart or it may also be due to the side effect of drugs. It may be ascertained that the proposed wavelet viz. the ECG
wavelet is found to be reliable in the detection of TWA and may be a potent tool to study wavelet analysis of ECG.

References

Abry, P., and Flandrin, P., 1996. Wavelets in Medicine and Biology, pp. 413-437
Aldroubi, A., and Unser, M., 1996. Wavelets in Medicine and Biology. New York: CRC Press.
Andreao, R .V., 2004. Segmentation de battements ECG par approche markovienne: application à la détection d’ischémies, INT-UFC, Evry,
coopération CAPES-COFECUB.
Arneodo, G., Grasseau, and Holschneider, M., 1988. Phys. Rev. Lett. 61, 2281-2284.Time Frequency and Wavelets in Biomedical Signal
Processing, IEEE Press, Piscataway, NJ.
 Vasudha Nannaparaju and Sundararajan Narasimman / Procedia Materials Science 10 (2015) 307 – 313 313

Chinushi, M., Restivo, M., Caref, E. B., and El Sherif, N., 1998. Electrophysiological basis of arrhythmogenicity of QT/T alternans in the long
QT syndrome: tri-dimensional analysis of the kinetics of cardiac repolarization. Circ Res; 83: 614 -28.
Daubechies, I., 1992. Ten lectures on wavelets: Society for Industrial and Applied Mathematics.
Gramatikov, B., and Thakor, N., 1993. Wavelet analysis of coronary artery occlusion related changes in ECG. in Proc. 15th Intern. Annual Conf.
IEEE/EMBS, San Diego. p.731.
Hardle, W., Kerkyacharian, G., Picard, D., and Tsybakov, A., 1997. Wavelets, approximation and statistical applications, Seminaire Paris-Berlin,
Paris, 1-284.
Jansen, M., 2001. Noise reduction by wavelet thresholding. New York, Springer Verlag.
Khalameizer, V., Pancheva, N., Reizin, L., and Ovsyshcher, I. E., 2005. Benign course and malignant clinical presentations of congenital long
QT syndrome. Europace; 7: 50-3.
Li, C., Zheng, C., and Ta,i C., 1995. Detection of ECG characteristic points using Wavelet transforms, IEEE Trans. on Biomedical Engineering,
42(1), 21-28.
Lu, Z., Kim, D. Y., and Pearlman, W. A., 2000. Wavelet compression of ECG signals by Set Partitioning in hierarchical trees algorithm. IEEE
Trans, Biomedical Medicine Engineering, vol. 47, No. 7.
Mallat, S. G., 1989. A theory for multiresolution signal decomposition: the wavelet representation, IEEE Transactions on Pattern Analysis and
Machine Intelligence, 11, pp. 674–693.
Meye, Y., 1990. Ondelettes et op´erateurs.
Miaou, S. G., Yen, H. L.,and Lin, C. L., 2002. Wavelet- based ECG Compression Using Dynamic Vector Quantization with Tree Code vectors in
Single Codebook, IEEE. Trans. Biomedical Eng., vol. 49, NO. 7.
Pastore, J. M., Girouard, S. D., Laurita, K. R., Akar, F. G., and Rosenbaum, D. S., 1999. Mechanism linking T wave alternans to the genesis of
cardiac fibrillation. Circulation; 99: 1385-94.
Rajendra Acharya, Hegde, B. M., Subbanna, Bhat, P., Ashok Rao, Niranjan, U., Cholayyam, 2002. Wavelet Analysis of Heart Rate Variability:
New method of studying the heart’s function, Kuwait Medical Journal, 34(3): 195-200.
Rosenbaum, D. S., Jackson, L, E., Smith, J, M., Garan, H., Ruskin, J. N., and Cohen, R. J., 1994. Electrical alternans and vulnerability to
ventricular arrhythmias. N Engl J Med; 330: 235–241
Sahambi, J., Tandon, S., and Bhatt, R., 1997. Using Wavelet Transform for ECG Characterization An On-line Digital Signal Processing System,
IEEE Eng. in Med. And Bio. Mag., pp. 77-83.
Sundararajan,N., N.,Vasudha, and S., Farook, 2011. A novel approach for Wave Segments in ECG by wavelet analysis. IEEEXplore
Tan, H. L., and Wild,e A. A., 1998. T wave alternans after sotalol: evidence for increased sensitivity to sotalol after conversion from atrial
fibrillation to sinus rhythm. Heart; 80: 303-6.
Teich, M. C., Heneghan, C., Lowen, S. B., and Turcot, R. G., 1996. Wavelets in Medicine and Biology pp. 383-412.
Thurner, S., Feurstein, M., and Teich, M., 1998. Multiresolution Wavelet Analysis of heartbeat intervals dicriminates healthy patients from those
with cardiac pathology, Phys. Rev. Letter, vol. 80, no. 7, pp. 1544-1547.
Vasudha, N., 2008, Wavelet transform and its applications to signal enhancement studies. Ph.D thesis submitted to Osmania University, India.