1900 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO. 10, OCTOBER 2007
in the literature: In [2], 69%; in [4], 54.79%; in [5], 95.1%; in [10],
62.33%; in [12], 91%; in [13], 67.2%; in [14], 84%; in [18], 86.89%
were 13.11%, 4.92%, 4.92%, 6.56%, 8.20%, and 13.11% for groups 1
to 6, respectively. In assessing the subjects belonging to the 7th group,
the system gave some incorrect classifications, patients from postoper-
atory were identified as belonging to the chronic laryngitis because of
the extreme similarity of voice quality. Due to the high average ,
causal by the small number of subjects, it is interesting that future
works use bigger databases. Other suggestions are: assessing different
cost functions in BBA and/or the use of radial base function ANN. All
in all, it is important to remark that this system is able to classify the
probability of some subject having a specific disease in physicians’ of-
fice, using just a simple set of hardware/software. This approach can
be very useful to help diagnosis.
REFERENCES
[1] R. J. Baken and R. F. Orlikoff, Clinical Measurement of Speech and
Voice, 2nd ed. San Diego, CA: Singular Thomsom Learning, 2000.
[2] D. G. Childers and K. S. Bae, “Detection of laryngeal function using
speech and electroglottographic data,” IEEE Trans. Biomed. Eng., vol.
39, no. 1, pp. 19–25, Jan. 1992.
[3] B. Fritzell, “Inverse filtering,” J. Voice, vol. 6, no. 2, 1992.
[4] M. O. Rosa, M. J. C. Pereira, and M. Grellet, “Adaptive estimation
of residue signal for voice pathology diagnosis,” IEEE Trans. Biomed.
Eng., vol. 47, no. 1, pp. 96–104, Jan. 2000.
[5] S. Hadjitodorov, B. Boyanov, and B. Teston, “Laryngeal pathology
detection by means of class-specific neural maps,” IEEE Trans. Inf.
Technol. Biomed., vol. 4, no. 1, pp. 68–73, Mar. 2000.
[6] C. E. Martinez and H. L. Rufiner, “Acoustic analysis of speech for
detection of laryngeal pathologies,” in Proc. IEEE-EMBS Int. Conf. Inf.
Tech. Appl. Biomed., 2000, pp. 2369–2372.
[7] R. T. Ritchings, M. Mcguillion, and C. J. Moore, “Pathological voice
quality assessment using artificial neural networks,” presented at
the 2nd Int. Workshop Models and Analysis of Vocal Emissions for
Biomedical Application, Firenze, Italy, 2001.
[8] M. Frohlich, D. Michaelis, and S. H. Werner, “Acoustic ’breathiness
measures’ in the description of pathologic voices,” in Proc. 1998 IEEE
Int. Conf. Acoustics Speech and Signal Processing., vol. 2, pp. 937–940.
[9] M. O. Rosa, M. Greller, and A. Carvalho, “Signal processing and sta-
tistical procedures to identify laryngeal pathologies,” in 6th Int. Conf.
IEEE Circuits and Systems Society Electronics, Circuits and Systems.,
1999.
[10] M. O. Rosa, C. J. Pereira, and A. Carvalho, “Evaluation of neural classi-
fiers using statistic methods for identification of laryngeal pathologies,”
in Proc. 5th Brazilian Symp. Neural Networks, Brazil, Dec. 9–10, 1998,
pp. 220–225.
[11] G. V. D. Wouwer, P. Scheunders, and D. V. Dyck, “Wavelet-Filvq
Classifier for Speech Analysis,” in Proc. 13th Int. Conf. Pattern Recog-
nition (ICPR’96), Antwerp, Belgium, 1997, vol. 4, p. 214.
[12] K. Umapathy et al., “Discrimination of pathological voices using a
time-frequency approach,” IEEE Trans. Biomed. Eng., vol. 52, no. 3,
pp. 421–421, Mar. 2005.
[13] A. Parraga, M. A. Zaro, and A. Schuck Jr., “Quantitative assessment
of the use of continuous wavelet transform in the analysis of the funda-
mental frequency disturbance of the synthetic voice,” Med. Eng. Phys.
PII: S, vol. 2, no. 00050-4, pp. 1350–4533.
[14] A. Schuck, Jr., and A. Parraga, “On the use of the wavelet packet trans-
form as a feature extractor for pathological voice assessement,” pre-
sented at the IFMBE Proceedings EMBEC’02 2nd Eur. Medical and
Biological Engineering Conf., Vienna, Austria, 2002.
[15] A. Schuck, Jr, L. V. Guimarães, and O. Wisbeck, “Dysphonic voice
classification using wavelets packets transform and artificial neural net-
works,” in Proc. 25th Annu. Int. Conf. IEEE-EMBS, Cancún, Mexico,
2003, pp. 2958–2961.
[16] S. Haykin, Neural Network: A Comprehensive Foundation, 2nd ed.
Upper Saddle River, NJ: Prentice-Hall, 1998.
[17] M. V. Wickerhauser, Adapted Wavelet Packet Analysis From Theory to
Software. Wellesley, MA: AK Peters, Ltd., 1994.
[18] C. D. P. Crovato and A. Schuck, Jr, “Classificação de sinais de voz
utilizando a transformada wavelet packet e redes neurais artificiais,”
in Proc. III Congresso Latinoamericano De Engenharia Biomedica,
João Pessoa, Paraíba, Brazil, 2004, vol. 5, pp. 1027–1030, ISBN:
85-98739-01-04.
Use of Sample Entropy Approach to Study Heart Rate
Variability in Obstructive Sleep Apnea Syndrome
Haitham M. Al-Angari* and Alan V. Sahakian
Abstract—Sample entropy, a nonlinear signal processing approach, was
used as a measure of signal complexity to evaluate the cyclic behavior
of heart rate variability (HRV) in obstructive sleep apnea syndrome
(OSAS). In a group of 10 normal and 25 OSA subjects, the sample entropy
measure showed that normal subjects have significantly more complex
HRV pattern than the OSA subjects ( ). When compared with
spectral analysis in a minute-by-minute classification, sample entropy
had an accuracy of 70.3% (69.5% sensitivity, 70.8% specificity) while
the spectral analysis had an accuracy of 70.4% (71.3% sensitivity, 69.9%
specificity). The combination of the two methods improved the accuracy to
72.9% (72.2% sensitivity, 73.3% specificity). The sample entropy approach
does not show major improvement over the existing methods. In fact, its
accuracy in detecting sleep apnea is relatively low in the well classified
data of the physionet. Its main achievement however, is the simplicity
of computation. Sample entropy and other nonlinear methods might be
useful tools to detect apnea episodes during sleep.
Index Terms—Approximate entropy, heart rate variability, nonlinear
signal processing, obstructive sleep apnea, power spectral density, sample
entropy.
I. INTRODUCTION
Heart rate variability (HRV) varies from wakefulness to sleep due
to normal changes in the autonomic system activities. Sympathetic
tone drops from wakefulness over nonrapid eye movement (NREM)
sleep stages, while it shows an increase in REM sleep [1]. Parasympa-
thetic activity increases from wakefulness over NREM sleep [2]. Spec-
tral analysis of HRV is used to evaluate the activity of the autonomic
nervous system. Low-frequency (LF) components (0.04–0.15 Hz) eval-
uate the sympathovagal balance while high-frequency (HF) compo-
nents (0.15–0.4 Hz) estimate the parasympathetic tone related to res-
piratory rhythm [3]. In sleep disorders, impairment of the autonomic
nervous system is observed. Studies of muscle sympathetic nerve ac-
tivity (MSNA) have shown an increase in sympathetic tone in patients
with obstructive sleep apnea syndrome (OSAS) during sleep and wake-
fulness [4], [5]. These findings were supported by results from spec-
tral analysis. The HF power was significantly diminished and LF/HF
ratio was enhanced in awake OSA patients, which indicates a drop in
the parasympathetic tone associated with an increase in the sympa-
thetic tone [6]. At the start of the apnea episode however, RR inter-
vals lengthen which indicates an increase in the vagal activation [7].
There is also a noticeable increase in MSNA, peaking immediately
prior to apnea cessation. Arousal at the termination of an apnea ini-
tiates a burst of sympathetic activity (associated with an increase in
blood pressure and heart rate). This is observed as cyclical variation
(progressive bradycardia followed by abrupt tachycardia) of the heart
rate [8].
Nonlinear analysis of time series provides a parameter set that quan-
tifies the characteristics of the system attractor even when the model
Manuscript received January 2, 2006; revised October 29, 2006. Asterisk in-
dicates corresponding author.
*H. M. Al-Angari was with Northwestern University, Evanston, IL 60208
USA. He is now with the Electrical and Computer Engineering Department,
King AbdulAziz University, P O Box 80204, Jeddah, 21589 Saudi Arabia
(e-mail: hangari@kau.edu.sa).
A. V. Sahakian is with the Electrical Engineering and Computer Science De-
partment, Evanston, IL 60208 USA (e-mail: sahakian@ ece.northwestern.edu).
Digital Object Identifier 10.1109/TBME.2006.889772
0018-9294/$25.00 © 2007 IEEE