RESEARCH ARTICLE Copyright 2015 American Scientific Publishers All rights reserved Printed in the United States of America Journal of Medical Imaging and Health Informatics Vol. 5, 1811 1815, 2015 Phase Average Waveform Analysis of Different Leads in Epileptic EEG Signals J. L. Zhang 1, N. Jiang 1, M. Y. Zhang 2, and Q. S. Feng 1 1 Department of Mathematics, School of Science, Tianjin University, Tianjin, 300072, China 2 Neurology Treatment Department, Hospital of People, Tianjin, 300130, China The aim of this paper is to analyze phase average waveforms for absence epileptic EEG signals and normal control EEG signals by using wavelet transformation. Conditional sampling method and phase averaging technique have also been used in this study. The results suggested that the cycle of phase average waveforms prior to absence epileptic seizure is almost the same with that of the normal control, but the 12th and 20th scale components were strengthened. During the seizure period, the frequency of phase average waveforms is 2.5 3.0 Hz and the cycle of phase average waveforms is 0.35 seconds. In addition, their phases were opposite. These results provide further understanding for the widespread brain disorders. Keywords: EEG, Conditional Sampling, Phase Average Waveform, Wavelet Transformation. 1. INTRODUCTION Copyright: American Scientific features. 10 Publishers However, Fourier transform is not suitable for analyzing Electroencephalogram (EEG) is a record of the electrical potentials non-stationary signals. Consequently, the development of generated by cerebral cortex nerve cells. The highly complex wavelet theory became inevitable. The root of wavelets can be EEG signals contain lots of valuable information for studying traced back to the thesis of Haar in 1909. 12 The transforma- brain function and neurological disorders. The first discover tion gives possibility to analyze both frequency and time domain of the human electroencephalogram signals was in 1929 by characteristic of a signal simultaneously. Some studies 13 15 have Hans Berger. 1 There are two different types of EEG measuring shown that the wavelet transformation could be really a powerful methods depending on where the electrodes are placed in the time-frequency analysis tool for EEG signals. head: scalp measurement way and intracranial measurement way. The purpose of this paper is to analyze phase average waveforms An EEG signal contains a wide range of frequency components, of different leads in absence epileptic EEG signals and which can classified as: rhythms (<4 Hz), rhythms (4 8 Hz), normal control EEG signals. We aim to detect and assess the rhythms (8 14 Hz), rhythms (14 30 Hz), rhythms abnormalities of absence epileptic based on phase average waveforms. (>30 Hz). 2 Absence epilepsy is one of the forms of nonconvulsive We first give a description of the EEG signals, mainly epilepsy. And absence seizures are a primarily generalized including the difference between the absence epileptic EEG sig- seizure type involving all cortical areas at once. The nals and the normal control EEG signals. Then comparison anal- basic characteristics of this type of epilepsy are the temporarily ysis has been made on their wavelet coefficients contour maps reduction in responsiveness and the accompanying occurrence of and phase average waveforms of different leads. These findings spike wave discharges in the electroencephalogram. 3 About 75% may benefit our understanding of the mechanisms causing epileptic of the patients can achieve effective seizure control by anticonvulsive disorders. medication or resective surgery. And the remaining 25% still have no sufficient treatment nowadays. 4 Using EEG to study the brain s electrical activities is one of the most in the treatment 2. MATERIALS AND METHODS of neurological diseases. 5 7 2.1. Subjects In order to extract the frequency domain features, discrete The absence epileptic EEG signals were tested in Tianjin Medical Fourier transformation was first applied to calculate the power University General Hospital. The subjects in control group are spectral density of EEG signals. 8 9 all healthy volunteers, who are normal growth and development Time and frequency analysis and cross-correlations calculation without any neurological disease. The clinical manifestations of are commonly used for classification of time domain the absence epileptic patient satisfy the CCTILAP. 16 And there is no use of anti-epileptic drugs or sedatives before EEG testing. Author to whom correspondence should be addressed. Related informed consent was also told. 17 J. Med. Imaging Health Inf. Vol. 5, No. 8, 2015 2156-7018/2015/5/1811/005 doi:10.1166/jmihi.2015.1650 1811
RESEARCH ARTICLE J. Med. Imaging Health Inf. 5, 1811 1815, 2015 In this test, the scalp electrodes were set in accordance with the international standard 10 20 lead system with regarding the ears electrodes as the reference electrodes. 18 We use an EEG-2130 digit electroencephalogram to record 20 minutes long EEG under the subjects of quiet, waking, eye closure state. The sampling rate in this experiment was 200 Hz. We used an A/D card to assign the 16-bit EEG signal to a block of data every 20 seconds. And there was 3 minutes excessive ventilation between each interval. 2.2. Methods Phase average waveforms analysis relative to rhythm in EEG signals were accomplished by using continuous wavelet transformation. The continuous wavelet transformation of a specific EEG signal f t is: + W f a b = f t a b t dt a b t = 1 ( ) t b a a where a, b are real numbers and a must be a positive real number. In the above definition, a b t denote a sequence of wavelet functions, and denotes the complex conjugation of the mother wavelet In this paper, we use a Gaussian wavelet function as the continuous mother wavelet function. The Gaussian wavelet function is given by the following expression: 2 t = te t2 /2 Obviously, it is easy to see that wavelet transformation represents the similarity degree between a specific wavelet basis function and a complex signal. Specific methods are as follows. We considered 30 scales in all. We set 0.00075 second as the minimal scale and the scale magnification rate is 1.5. The computing method was performed by measuring EEG series in space and time domain at a specific scale. The detecting criteria t = 1 if and only if W f a b i >0 and W f a b i reaches the local maximum, otherwise, t = 0. On account of the wavelet coefficients, we can extract the phase average waveforms from the time series through phase average and conditional sampling technique. 19 20 In detail, in order to get the conditional phase averaged waveforms, we should overlap and average the segments of multi-scale rhythmic signals according to phase alignments method. 3. MAIN RESULTS 3.1. EEG Signals of Normal Control and the Absence Epileptic The EEG of the normal control mainly contains rhythm whose frequency ranges in 8 14 Hz, as shown in Figure 1(a). No obvious abnormal was observed. Particularly, the rhythm possesses occipital advantage. Figure 1(b) is the EEG signals of a epileptic before absence seizure, which is similar to the EEG signal of the normal control. Figure 1(c) was the beginning of the absence (a) normal control 20 seconds Fig. 1. EEG for normal control and absence epileptic. 1812
J. Med. Imaging Health Inf. 5, 1811 1815, 2015 RESEARCH ARTICLE seizure, and the seizure begun at about 10 seconds. Figure 1(d) is the end of the absence epilepsy seizure, and the seizure ended at about 15 seconds. The main abnormal waves include spike waves, sharp waves, high attitude slow waves and spike slow complex waves. 3.2. Wavelet Coefficients Contour Maps of the EEG Signals The wavelet coefficients contour maps are given in Figure 2(a), which corresponding to the normal control s visual EEG signals. And Figures 2(b) (d) are the wavelet coefficients contour maps corresponding to Figures 1(b) (d), respectively. The horizontal axis represents time variable, and the vertical axis represents scale variable. The color in the map stands for the value of wavelet coefficients, which are classified into 8 grades marked with different colors. It is easy to observe clearly that rhythm distribute mainly in 9th scale, as shown in Figure 2(a). Figure 2(a) also presents that the wavelet coefficients of small scale highly correlated to the wavelet coefficients of large scale. Compared with the wavelet coefficients contour map of normal control group, the rhythm begun to reduce before absence seizure and eventually vanished during the seizure period in the wavelet coefficients contour map of absence epileptic. In particular, the wave frequency band appear a slow decrease in frequency but increase in amplitude. We can see that the 3 Hz rhythm in 12th scale was the most intense, which shown the spike and slow complex waves. On the other hand, Figures 2(b) (d) shows that the wavelet coefficients between 12th scale and 20th scale correlated to each other. Consequently, during the seizure period 3 Hz rhythmic spike and slow complex waves were the characteristics of EEG signals. 3.3. Phase Average Waveforms Analysis of Different Leads Figure 3 shows phase average waveforms of FP1 (F3, C5, P7, O1 respectively) at scale 9, which were rhythmic synchronous and obtained by regarding O1 lead as a benchmark. Figure 3(a) shows an rhythm frequency and the cycle of phase average waveforms in normal control group was about 0.12 seconds. In Figure 3(b), before absence seizure the frequency of FP1 lead phase average waveform was 8 9 Hz and the cycle of FP1 lead was almost the same with that of the normal adults. But we (a) normal control Fig. 2. Wavelet coefficient contour maps of the normal control and absence epileptic EEG signals. 1813
RESEARCH ARTICLE J. Med. Imaging Health Inf. 5, 1811 1815, 2015 (a) normal control Fig. 3. Phase average waveforms of different leads in normal control and absence epileptic EEG signals. can observe that in Figures 3(c) and (d) the frequency changed to 2.5 3.0 Hz and the cycle changed to 0.35 seconds, that is, a slow decrease in frequency but increase in amplitude. As a consequence, the main characteristics of absence seizure EEG signals were rhythmic activities. The maximum amplitude of O1 lead is 250 (Fig. 3(a)), however, it can reach 650 during the absence epileptic seizure (Fig. 3(d)). Furthermore, the phase of the rhythmic activities was almost completely opposite. 4. DISCUSSION AND SUMMARY EEG is a complex neural electrophysiological signals, which reflects the complex electro-physiological activities in central nervous system. The conventional visual inspection of EEG recordings includes the examination of the following features: wave form regularity, frequency or wavelength, voltage or amplitude, and reactivity to eye opening and so on. Due to the limitations of human recognition, it is always hard to describe the frequency, amplitude, phase index, etc. Fourier transform can provide quantitative power value of different frequency bands, but it lost the time domain information. It is just the wavelet transformation that can cover the shortage. And researches shown 21 22 that in EEG signals phase correlation changes in brain areas. This paper analyzes phase average waveforms in EEG signals of different leads by using wavelet transformation. In the EEG signals of normal adults, waves distribute mainly in 9th scale. During the seizure period, we can see that the 12th and 20th scale components strengthened, meanwhile, the rhythm vanished. On the other hand, the cycle of phase average waveforms were different from that of the normal control group. We can observe that the rhythm show a slow decrease in frequency but increase in amplitude when absence epileptic onset. Particularly, the frequency of the phase average waveform is 2.5 3.0 Hz and the cycle of phase average waveform in the course of absence seizure is 0.35 seconds. In addition, the opposite phase implies abnormal discharge of the epileptic patient. Above all, rhythmic activities were the main characteristics of absence seizure in EEG signals. These results can benefit our understanding for the widespread brain disorders. Acknowledgments: This work has been supported by the National Natural Science Foundation of China 1814
J. Med. Imaging Health Inf. 5, 1811 1815, 2015 RESEARCH ARTICLE (Grant Nos. 11332006, 11272233 and 11411130150), National key basic research and development program (plan 973) (No. 2012CB720101), Tianjin University Research and Innovation Foundation. References and Notes 1. E. Niedermeyer, Historical aspects, Electroencephalography, Basic Principles, Clinical Applications and Related Fields, edited by E. Niedermeyer and F. Silva, Williams and Wilkins, Philadephia, PA, USA (1999), pp. 1 14. 2. P. Kellaway, An orderly approach to visual analysis: Characteristics of the normal EEG of adults and children, Current Practice of Clinical Electroencephalography, edited by D. Daly and T. Pedley, Raven Press, Philadephia, PA, USA (1990), pp. 139 99. 3. A. M. L. Coenen and E. L. J. M. Van Luijtelaar, Genetic animal models for absence epilepsy: A review of the WAG/Rij strain of rats. Behav. Genet. 33, 635 (2003). 4. Y. Zheng, G. Wang, K. Li, G. Bao, and J. Wang, Epileptic seizure prediction using phase synchronization based on bivariate empirical mode decomposition. Clinical Neurophysiology 125, 1104 (2014). 5. P. Matthew, T. Manolis, and L. Norbert, Vision-based motion detection, analysis and recognition of epileptic seizures A systematic review. Comput. Methods Programs Biomed. 108, 1133 (2012). 6. M. Ihle, H. Feldwisch-Drentrup, C. A. Teixeira, A. Witon, B. Schelter, J. Timmer, and A. Schulze-Bonhage, EPILEPSIAE A European epilepsy database. Comput. Methods Programs Biomed. 106, 127 (2012). 7. L. Guo, D. Rivero, and A. Pazos, Epileptic seizure detection using multiwavelet transform based approximate entropy and artificial neural networks. J. Neurosci. Methods 1, 156 (2010). 8. Z. Iscan, Z. Dokur, and T. Demiralp, Classification of electroencephalogram signals with combined time and frequency features. Expert Syst. Appl. 38, 10499 (2011). 9. T. Nguyen-Ky, P. Wen, and Y. Li, Theoretical basis for identification of different anesthetic states based on routinely recorded EEG during operation. Comput. Biol. Med. 39, 40 (2009). 10. F. Stephen and M. William, Dynamic, location based channel selection for power consumption reduction in EEG analysis. Comput. Methods Programs Biomed. 108, 1206 (2012). 11. E. Avci, D. Hanbay, and A. Varol, An expert discrete wavelet adaptive network based fuzzy inference system for digital modulation recognition. Expert Syst. Appl. 33, 582 (2007). 12. I. Daubechies, Ten Lectures on Wavelets, Society for Industrial and Applied Mathematics, Philadelphia, PA (1992). 13. D. Bosnyakova, A. Gabova, A. Zharikova, V. Gnezditski, G. Kuznetsova, and G. van Luijtelaar, Some peculiarities of time-frequency dynamics of spikewave discharges in humans and rats. Clin. Neurophysiol. 8, 1736 (2007). 14. E. Sitnikova, A. E. Hramov, A. A. Koronovsky, and G. van Luijtelaar, Sleep spindles and spike-wave discharges in EEG: Their generic features, similarities and distinctions disclosed with Fourier transform and continuous wavelet analysis. J. Neurosci. Methods 2, 304 (2009). 15. O. Faust, U. Rajendra Acharya, H. Adeli, and A. Adeli, Wavelet-based EEG processing for computer-aided seizure detection and epilepsy diagnosis. Seizure 26, 56 (2015). 16. The Commission on Classification and Terminology of the International League against Epilepsy, Proposal for a revised classification of the epilepsies and epileptic syndromes. Epilepsia 4, 389 (1989). 17. State Council of the People s Republic of China, Administrative Regulations on Medical Institution (1994). 18. X. Liu, Clinical Electroencephalography, People s Medical Pubishing House, Chaoyang, Beijing, China (2006). 19. N. Jiang and J. Zhang, Detecting multi-scale coherent eddy structure and intermittency in turbulent boundary layer by wavelet analysis. Chinese Phys. Lett. 8, 1968 (2005). 20. N. Jiang, W. Liu, J. H. Liu, and Y. Tian, Phase-averaged waveforms of Reynolds stress n wall turbulence during the burst events of coherent structures. Sci. Chin. G-Phys. Mech. Astron. 7, 857 (2008). 21. O. A. Rosso, W. R. Hyslop, and R. Gerlach, Quantitative EEG analysis of the maturational changes associated with childhood absence epilepsy. Physica A 1, 184 (2005). 22. F. Amor, D. Rudrauf, V. Navarro, K. N diaye, L. Garnero, J. Martinerie, and M. L. Van Quyen, Imaging brain synchrony at high spatio-temporal resolution: Application to MEG signals during absence seizures. Signal Processing Delivered by Publishing Technology to: Nanyang 11, 2101 (2005). Technological University Received: 25 March 2015. Revised/Accepted: 2 July 2015. 1815