FREQUENCY DOMAIN BASED AUTOMATIC EKG ARTIFACT REMOVAL FROM EEG DATA features FOR BRAIN such as entropy COMPUTER and kurtosis for INTERFACING artifact rejection. V. Viknesh B.E.,(M.E) - Lord Jeganath College of Engineering & Technology, Anna Univ. CHAPTER -1 Mr.Franklin George Jobin, M.E(Ph.D) Asst. Professor Lord Jeganath College of Engineering INTRODUCTION & Technology, Anna University. ABSTRACT An automated method for detecting and eliminating electrocardiograph (ECG) artifacts from electroencephalography (EEG) without an additional synchronous a variety of reasons it is often desirable to automatically detect and remove these artifacts. Especially, for accurate source localization of epileptic spikes in an EEG recording from a patient with epilepsy, it is of great importance to remove any concurrent artifact. Due to similarities in morphology between the EKG artifacts and epileptic spikes, any automated artifact removal algorithm must have an extremely low false-positive rate in addition to a high detection rate. In this paper, an automated algorithm for removal of EKG artifact is proposed that satisfies such criteria. The proposed method, which uses combines independent component analysis and continuous wavelet transformation, uses both temporal and spatial characteristics of EKG related potentials to identify and remove the artifacts. The method outperforms algorithms that use general statistical The need for ambulatory electroencephalographic monitoring has increased in both clinical practice and research, in areas such as sleep/wake state or epilepsy monitoring. However, longterm recordings are vulnerable to various artifacts. In particular, cardiac activity may have pronounced effects on the electroencephalogram (EEG) because of its relatively high electrical energy, especially upon the noncephalic reference recordings of EEG. Algorithms have been proposed to eliminate electrocardiogram (ECG) artifacts from the EEG. Nakamura and Shibasaki proposed an ECG artifact elimination algorithm, which we call the ensemble average subtraction (EAS) method, whereby ECG-contaminated EEG series are synchronously segmented with respect to the timing of consecutive ECG R-peaks. By subtracting the ensemble average across EEG segments from the contaminated EEG, the algorithm eliminates ECG artifacts. EAS is based on the strict assumptions of homogeneity across segments and Gaussian property of the EEG. Using a different concept, the 105
independent component analysis (ICA) method was also applied to eliminate ECG artifacts using multichannel signals. Previously, we adopted adaptive noise canceling theory to eliminate such ECG artifacts using a reference ECG channel. It should be noted that these algorithms use consecutive R-waves in a separate ECG channel as a reference, and therefore, cannot be applied when an ECG channel is not available. Objective: A fundamental problem in neural network research, as well as in many other disciplines, is finding a suitable representation of multivariate data, i.e. random vectors. For reasons of computational and conceptual simplicity, the representation is often sought as a linear transformation of the original data. In other words, each component of the representation is a linear combination of the original variables. To propose, which uses combines independent component analysis ICA and continuous wavelet transformation CWT, uses both temporal and spatial characteristics of EKG related potentials to identify and remove the artifacts. CHAPTER 2 PROPOSED METHOD 2.1. EEG Recording All EEG files used in this study were recorded (from adult patients and for clinical purposes) at the Epilepsy Center of the University Hospitals Case Medical Center. An IRB approval was obtained to retrospectively use the data for optimizing and validating our algorithm. Four sets of files were chosen for analysis. The first set, comprising of five files (about 30 h of data) from five different patients, was used for computing the reference scalp distribution of EKG artifact. To be selected as a member of this set, the EEG file had to have prominent EKG artifact in most of the recording. The second set, containing 18 files (11 EKG artifact-free, 7 contaminated, about 15 h of data), was used to find the optimum threshold for similarity (to the template) test. The second group consisted of awake, asleep, normal, and abnormal EEG files with a variety of abnormalities including epileptic periodic patterns that could be mistaken for EKG artifact. The third set, which was used for validation of the algorithm, had 29 EEG files (about 55 h of data) from 29 randomly selected adult patients. Eighteen files (approximately 49 h of EEG data) in the third group were contaminated with EKG artifact. Finally, the fourth set of EEG files (5 files, about 3.4 h of data) were almost entirely contaminated with periodic epileptic patterns. All EEG files were recorded using a Nihon Kohden system (Neurofax, Nihon Kohden, Inc., Japan) at a sampling rate of 200 Hz from the standard 10 20 system of electrodes including A1 and A2. The files had a 106
simultaneously recorded EKG channel that was used for testing purposes. The EEG and EKG data were filtered with 1 Hz high-pass and 70-Hz low-pass filters. Filtering was performed using the MATLAB built-in function filtfilt to eliminate phase shift and used Butterworth filter designs. The EEG data were analyzed in the original (referential) montage, i.e., no change was done to the montage. 2.2 Algorithm Fig. 1 shows a flow chart of the proposed algorithm. ICA is first applied to the EEG data. ICs whose scalp distributions have large correlation coefficients with a precomputed pattern of EKG artifact are considered for further analysis. A CWT is then performed on these ICs and the square of the continuous wavelet coefficients are checked for having multiple, quasiperiodic peaks. The ICs that pass the second step, if any, are then sorted based on the correlation between their scalp distribution and the reference pattern and the one with the highest correlation is reported as the EKG component. This component is then rejected and the artifact-free EEG signals are reconstructed. Although it is possible to have more than one EKG-related IC, our Fig. 2.1. Flowchart of the proposed algorithm is shown. results and those reported in the literature suggest that, at least in the case of lowdensity EEG recordings, only one IC is the main source of contamination. The algorithm, however, can be modified to reject more than one component. These steps are described in detail in following paragraphs. The algorithm was implemented in MATLAB 7.1 (The MathWork, Inc., USA). 107
CHAPTER -3 RESULT & DISCUSSION 3.1 Input EEG & EKG Waves for Mixing: 108
3.2 Normalization: 109
3.3 EKG Separation from Mixed EEG Signal: 110
3.4 Detected EKG peaks from the EEG: REFERENCES 1. J. S. Barlow and J. Dubinsky, EKG-artifact minimization in referential EEG recordings by computer subtraction, Electroencephalogr. Clin. Neurophysiol., vol. 48, no. 4, pp. 470 472, 1980. 2. Y. Ishiyama, M. Ebe, I. Homma, and Z. Abe, Elimination of EKG artifacts from EEGs recorded with balanced non-cephalic reference electrode method, Electroencephalogr. Clin. Neurophysiol., vol. 53, no. 6, pp. 662 665, 1982. 3. C. Fortgens and M. P. De Bruin, Removal of eye movement and 111
ECG artifacts from the noncephalic reference EEG, Electroencephalogr. Clin. Neurophysiol., vol. 56, no. 1, pp. 90 96, 1983. 4. H. J. Park, D. U. Jeong, andk. S. Park, Automated detection and elimination of periodic ECG artifacts in EEG using the energy interval histogram method, IEEE Trans. Biomed. Eng., vol. 49, no. 12, pp. 1526 1533, Dec.2002. 5. H. N. Suresh and C. Puttamadappa, Removal of EMG and ECG artifacts from EEG based on real time recurrent learning algorithm, Int. J. Phys.Sci., vol. 3, no. 5, pp. 120 125, 2008. 6. J. A. Jiang, C. F. Chao,M. J. Chiu, R. G. Lee, C. L. Tseng, and R. Lin, An automatic analysis method for detecting and eliminating ECG artifacts in EEG, Comput. Biol. Med., vol. 37, no. 11, pp. 1660 1671, 2007. 7. G. Correa, E. Laciar, H. D. Pati no, and M. E. Valentinuzzi, Artifact removal from EEG signals using adaptive filters in cascade, in 16 th Argentine Bioeng. Congr. 5th Conf. Clin. Eng., J. Phys.: Conf. Series.,vol. 90, Bristol, U.K., 2007, pp. 012081 012090. 8. Hyv arinen and E. Oja, Independent component analysis: Algorithms and applications, Neural Netw., vol. 13, no. 4, pp. 411 430, 2000. 9. Hyv arinen, Fast and robust fixed point algorithms for independent component analysis, IEEE Trans. Neural Netw., vol. 10, no. 3, pp. 626 634, May 1999. 10. R. Vig ario, J. Sarela, V. Jousmiki, Matti H am al ainen, and Erkki Oja, Independent component approach to the analysis of EEG and MEG recordings, IEEE Trans. Biomed. Eng., vol. 47, no. 5, pp. 589 593, May 2000. 112