In Fig. 7-1, the labeled useful component is marked with a blue circle, and the
prediction of useful component is marked with red rectangle. It presents that the
component which only has red rectangle or blue circle is the misclassification. In
order to make the clustering correctly, the component which only has both red
rectangle and blue circle is considered. However, it makes a problem between the
components in Fig. 7-1 (a) no. 7, and (b) no. 8 and 15. These components are useful
component (it is parietal) by manual selection, but the classifier selects the (a) no. 7
and (b) no. 8 as useful component without the (b) no.15. Because in the training data,
there are some components which have high weight in an electrode position on scalp
map, the similar component is in the Fig. 7-1 (b) no. 27 (this is an example in the
frontal region). Therefore, if we want to implement the component clustering, we
need to improve the automatic component selection more effective, or enhance the
dimension of the weight matrix to extract more detail features.
57
(a).
(b)
Figure 7-1. An illustration of how a scalp map is drawn.
58
7-1. The Enhancement of Feature
The enhancement of weight matrix is a good choice of future experiment. Not
only we can reduce the misclassification of insufficient data dimensions, but also the
new matrix has the possibility that the component clustering could remove the bad
component by itself. In our experiment, we used 28-channel recording. However,
there exist 32, 64, or 128 channels to record the EEG signal. It is a big trouble if we
change the recording channel, the classifier of automatic selection needs to be trained
again for new input data. Therefore, we could consider the scalp-map as our new
feature for training classifiers. The different channel recording would generate the
same scalp map topology, which is the matrix that we called color map. The color
map could be a large or small matrix, which is selected by the concentration of points
on the scalp map. We can train and evaluate classifiers by the color maps as input data.
The component clustering system could be performed by using the color maps, which
supplies the more detail from scalp-map for future experiment. These hypothesis need
to design a new system scheme to implement, that is another topics in future works.
59
7-2. Auto Clustering System
The component clustering system is a designing topic to extent the ability of
independent component analysis. In recent researches, the ICA can separate the EEG
signals into independent components and draw the scalp map by weight matrix. And
we can analyze specific independent component with corresponding scalp-map. Thus,
we could point out the EEG activities from different regions. Like the component
based on occipital lobe has the advantage on drowsiness analysis, the component
based on prefrontal cortex usually concentrates the eye-blinking affect. We could base
on these characteristic phenomena to understand the association between the brain
activities and behaviors. If we could achieve this system, for the engineers or
psychologists which has no neuroscience background is an important tool to introduce
them into brain analysis. Currently, mostly brain researches based on ICA have lots of
conclusion between component activities and behaviors. Therefore, we could
implement the component clustering system for these researches in brain computer
interface or real-time application. This work will changes the application of raw EEG
based analysis system in the world.
60
Reference
[1] M. Iwasaki, C. Kellinghaus, A. Alexopoulos, R. Burgess, A. Kumar, Y. Han, H. Luders, and R. Leigh, "Effects of eyelid closure, blinks, and eye movements on the electroencephalogram," Clinical Neurophysiology, vol. 116, pp. 878-885, 2005.
[2] I. Goncharova, D. McFarland, T. Vaughan, and J. Wolpaw, "EMG contamination of EEG: spectral and topographical characteristics," Clinical Neurophysiology, vol. 114, pp. 1580-1593, 2003.
[3] D. Brunner, R. Vasko, C. Detka, J. Monahan, C. Reynolds III, and D. Kupfer,
"Muscle artifacts in the sleep EEG: Automated detection and effect on all-night EEG power spectra," Journal of Sleep Research, vol. 5, pp. 155-164, 1996.
[4] D. Mantini, M. Perrucci, S. Cugini, A. Ferretti, G. Romani, and C. Del Gratta,
"Complete artifact removal for EEG recorded during continuous fMRI using independent component analysis," Neuroimage, vol. 34, pp. 598-607, 2007.
[5] P. Allen, G. Polizzi, K. Krakow, D. Fish, and L. Lemieux, "Identification of EEG Events in the MR Scanner: The Problem of Pulse Artifact and a Method for Its Subtraction," Neuroimage, vol. 8, pp. 229-239, 1998.
[6] P. Allen, O. Josephs, and R. Turner, "A Method for Removing Imaging Artifact from Continuous EEG Recorded during Functional MRI,"
Neuroimage, vol. 12, pp. 230-239, 2000.
[7] A. Erfanian and B. Mahmoudi, "Real-time ocular artifact suppression using recurrent neural network for electro-encephalogram based brain-computer
61
interface," Medical and Biological Engineering and Computing, vol. 43, pp.
296-305, 2005.
[8] P. He, M. Kahle, G. Wilson, and C. Russell, "Removal of Ocular Artifacts from EEG: A Comparison of Adaptive Filtering Method and Regression Method Using Simulated Data," in Engineering in Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual International Conference of the, 2005, pp. 1110-1113.
[9] P. He, G. Wilson, and C. Russell, "Removal of ocular artifacts from electro-encephalogram by adaptive filtering," Medical and Biological Engineering and Computing, vol. 42, pp. 407-412, 2004.
[10] P. He, G. Wilson, C. Russell, and M. Gerschutz, "Removal of ocular artifacts from the EEG: a comparison between time-domain regression method and adaptive filtering method using simulated data," Medical and Biological Engineering and Computing, vol. 45, pp. 495-503, 2007.
[11] A. Soong and Z. Koles, "Principal-component localization of the sources of the background EEG," IEEE Transactions on Biomedical Engineering, vol. 42, pp. 59-67, 1995.
[12] Z. Koles, J. Lind, and A. Soong, "Spatio-temporal decomposition of the EEG:
a general approach to the isolation and localization of sources,"
Electroencephalography and Clinical Neurophysiology, vol. 95, pp. 219-230, 1995.
[13] T.-P. Jung, S. Makeig, C. Humphries, T. Lee, M. McKeown, V. Iragui, and T.
Sejnowski, "Removing electroencephalographic artifacts by blind source separation," Psychophysiology, vol. 37, pp. 163-178, 2000.
62
[14] T.-P. Jung, C. Humphries, T. Lee, S. Makeig, M. McKeown, V. Iragui, and T.
Sejnowski, "Removing electroencephalographic artifacts: comparison between ICA and PCA," in Neural Networks for Signal Processing VIII, 1998.
Proceedings of the 1998 IEEE Signal Processing Society Workshop, 1998, pp.
63-72.
[15] S. Makeig, A. Bell, T. Jung, and T. Sejnowski, "Independent Component Analysis of Electroencephalographic Data," Advances in Neural Information Processing Systems, pp. 145-151, 1996.
[16] A. Flexer, H. Bauer, J. Pripfl, and G. Dorffner, "Using ICA for removal of ocular artifacts in EEG recorded from blind subjects," Neural Networks, vol.
18, pp. 998-1005, 2005.
[17] C. Joyce, I. Gorodnitsky, and M. Kutas, "Automatic removal of eye movement and blink artifacts from EEG data using blind component separation,"
Psychophysiology, vol. 41, pp. 313-325, 2004.
[18] T.-P. Jung, C. Humphries, T. Lee, S. Makeig, M. McKeown, V. Iragui, and T.
Sejnowski, "Extended ICA Removes Artifacts from Electroencephalographic Recordings," Advances in Neural Information Processing Systems, pp.
894-900, 1998.
[19] P. LeVan, E. Urrestarazu, and J. Gotman, "A system for automatic artifact removal in ictal scalp EEG based on independent component analysis and Bayesian classification," Clinical Neurophysiology, vol. 117, pp. 912-927, 2006.
[20] Y. Li, Z. Ma, and W. Lu, "Automatic removal of the eye blink artifact from EEG using an ICA-based template matching approach," Physiological Measurement, vol. 27, p. 425, 2006.
63
[21] S. Romero, M. A. Mananas, S. Clos, S. Gimenez, and M. J. Barbanoj,
"Reduction of EEG artifacts by ICA in different sleep stages," in Engineering in Medicine and Biology Society, 2003. Proceedings of the 25th Annual International Conference of the IEEE, 2003, pp. 2675-2678 Vol.3.
[22] R. Vigario, "Extraction of ocular artefacts from EEG using independent component analysis," Electroencephalography and Clinical Neurophysiology, vol. 103, pp. 395-404, 1997.
[23] A. Bell and T. Sejnowski, "An information-maximization approach to blind separation and blind deconvolution," Neural computation, vol. 7, pp.
1129-1159, 1995.
[24] T. Lee, M. Girolami, and T. Sejnowski, "Independent component analysis using an extended infomax algorithm for mixed subgaussian and supergaussian sources," Neural computation, vol. 11, pp. 417-441, 1999.
[25] F. De Martino, F. Gentile, F. Esposito, M. Balsi, F. Di Salle, R. Goebel, and E.
Formisano, "Classification of fMRI independent components using IC-fingerprints and support vector machine classifiers," Neuroimage, vol. 34, pp. 177-194, 2007.
[26] T. Kochiyama, T. Morita, T. Okada, Y. Yonekura, M. Matsumura, and N.
Sadato, "Removing the effects of task-related motion using independent-component analysis," Neuroimage, vol. 25, pp. 802-814, 2005.
[27] M. McKeown, S. Makeig, G. Brown, T. Jung, S. Kindermann, A. Bell, and T.
Sejnowski, "Analysis of fMRI data by blind separation into independent spatial components," Human Brain Mapping, vol. 6, pp. 160-188, 1998.
64
[28] S. Fitzgibbon, D. Powers, K. Pope, and C. Clark, "Removal of EEG Noise and Artifact Using Blind Source Separation," Journal of Clinical Neurophysiology, vol. 24, p. 232, 2007.
[29] F. La Foresta, N. Mammone, and F. Morabito, "PCA–ICA for automatic identification of critical events in continuous coma-EEG monitoring,"
Biomedical Signal Processing and Control, 2009.
[30] F. La Foresta, F. C. Morabito, B. Azzerboni, and M. Ipsale, "PCA and ICA for the extraction of EEG components in cerebral death assessment," in Neural Networks, 2005. IJCNN '05. Proceedings. 2005 IEEE International Joint Conference on, 2005, pp. 2532-2537 vol. 4.
[31] C. Bugli and P. Lambert, "Comparison between principal component analysis and independent component analysis in electroencephalograms modelling,"
Biometrical Journal, vol. 49, pp. 312–327, 2007.
[32] T.-P. Jung, S. Makeig, M. Westerfield, J. Townsend, E. Courchesne, and T.
Sejnowski, "Analysis and visualization of single-trial event-related potentials,"
Human Brain Mapping, vol. 14, pp. 166-185, 2001.
[33] H.-S. Huang, N. R. Pal, L.-W. Ko, and C.-T. Lin, "Automatic Identification of Useful Independent Components with a View to Removing Artifacts from EEG Signal," in Neural Networks, 2009., 2009 IJCNN International Joint Conference on, Atlanta, USA, 2009, p. 5.
[34] C.-T. Lin, Y.-C. Chen, R.-C. Wu, S.-F. Liang, and T.-Y. Huang, "Assessment of driver's driving performance and alertness using EEG-based fuzzy neural networks," in Circuits and Systems, 2005. ISCAS 2005. IEEE International Symposium on, 2005, pp. 152-155.
65
[35] C.-T. Lin, L.-W. Ko, I.-F. Chung, T.-Y. Huang, Y.-C. Chen, T.-P. Jung, and S.-F. Liang, "Adaptive EEG-Based Alertness Estimation System by Using ICA-Based Fuzzy Neural Networks," IEEE Transactions on Circuits and Systems Part 1 Regular Papers, vol. 53, p. 2469, 2006.
[36] C.-T. Lin, R.-C. Wu, S.-F. Liang, W.-H. Chao, Y.-J. Chen, and T.-P. Jung,
"EEG-based drowsiness estimation for safety driving using independent component analysis," Circuits and Systems I: Regular Papers, IEEE Transactions on, vol. 52, pp. 2726-2738, 2005.
[37] A. Delorme and S. Makeig, "EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis,"
Journal of Neuroscience Methods, vol. 134, pp. 9-21, 2004.
[38] A. Hyvarinen, J. Karhunen, and E. Oja, Independent Component Analysis.
New York: John Wiley & Sons, 2001.
[39] T. Lee, Independent component analysis: theory and applications: Kluwer Academic Publishers Hingham, MA, USA, 1998.
[40] V. Vapnik, The nature of statistical learning theory: springer, 2000.
[41] S. Kumar, Neural Networks: a classroom approach: Tata McGraw-Hill, 2004.
[42] S. Haykin, Neural networks: a comprehensive foundation: Prentice Hall, 2008.
66
Appendix
The confusion matrix from evaluation results:
(a). MLPTAN
T0.9 C1 C2 T1.0 C1 C2 C1: Good Component Class
C1 35 1 C1 x x C2: Bad Component Class
T0.9 C1 C2 T1.0 C1 C2 C1: Good Component Class
C1 32 1 C1 x x C2: Bad Component Class
C2 51 196 C2 x x TX.X: Threshold X.X
0.81
(228/280) X
67
T0.9 C1 C2 T1.0 C1 C2 C1: Good Component Class
C1 30 1 C1 x x C2: Bad Component Class
T0.9 C1 C2 T1.0 C1 C2 C1: Good Component Class
C1 74 20 C1 x x C2: Bad Component Class
C2 9 177 C2 x x TX.X: Threshold X.X
0.90
(251/280) X