癲癇是一種常見的神經系統疾病，準確預測患者的下一次癲癇發作時間對於提高治療效果和生活品質非常重要。癲癇預測由於不同發作間期(Interictal)與發作前期(Preictal)信號的相似度頗高，導致分類效果不彰，本研究嘗試使用共同空間型樣法(Common Spatial Patterns, CSP)進行信號前處理，通過數據分析方法提升相似信號間的辨識度，並結合對數變異數(Log-variance)、功率頻譜(Power spectrum)做為特徵，結合L1正則化線性回歸(Least Absolute Shrinkage and Selection Operator, LASSO)進行模型訓練與特徵挑選，並比較CSP所提升的分類效果以及最佳特徵。除此之外，傳統頻帶做為經常使用的特徵擷取方法，其定義不明確且可繼續細分為諸多子頻帶區分生理活動，代表頻帶的詳細劃分能提供更多生物資訊。本研究將CSP視為信號源分析方法，透過分析EEG信號的動態特性並劃分數據驅動頻帶，希望以數據分析產生適用於癲癇預測的頻帶，並將其與傳統頻帶進行比較。本研究使用三組開放資料集並以AUC評估模型，最終得出CSP能在癲癇預測最多提升0.09的分類器，但仍未達顯著差異(paired t-test p=0.17)，在特徵選擇上，功率頻譜是比對數變異數更佳的特徵，平均AUC值相差0.12，且達到顯著差異(paired t-test p=0.04)。CSP數據驅動頻帶高於傳統頻帶，但未達顯著差異(p=0.29)。本研究所使用方法中，以CSP結合功率頻譜分類效果最佳，平均AUC達到0.7，並且標準差為0.13。AUC的提升在CSP空間濾波後比CSP數據驅動頻帶的提升更高。由此可見，劃分頻帶結合一般的功率頻譜效果有限，然而作為嘗試性的方法，仍有許多可提升與討論的空間。

Epilepsy is a common neurological disorder, and accurately predicting the next seizure onset time is crucial for epileptic patients who do not respond to medical or surgical treatment. Due to the high similarity between interictal and preictal signals, the classification performance for seizure prediction remains unsatisfactory. This study attempts to use Common Spatial Patterns (CSP) for signal preprocessing and employs data analysis methods to enhance the discriminability between similar signals. Log variance and power spectrum are used as features, combined with L1-regularized linear regression (LASSO) for model training and feature selection. Additionally, traditional frequency bands are often used as features, but their definitions are unclear. This study considers CSP as a source analysis method. It analyzes the dynamic characteristics of EEG signals to derive data-driven frequency bands, aiming to generate frequency bands suitable for seizure prediction and compare them with traditional bands. Three open-source datasets are used, and the models are evaluated using the area under the curve (AUC). The results of CSP feature selection, CSP combined with power spectrum outperforms log-variance, with a significant difference in average AUC of 0.12 (p-value < 0.05). The CSP data-driven frequency bands have higher average AUC than traditional bands, but the difference is not statistically significant (p-value = 0.29). The best classification performance is achieved by combining CSP with the power spectrum, with an average AUC of 0.7 and a standard deviation of 0.13. The AUC improvement after CSP spatial filtering is higher than that of the CSP data-driven frequency boundaries.

[1] B. Maimaiti et al., “An Overview of EEG-based Machine Learning Methods in Seizure Prediction and Opportunities for Neurologists in this Field,” Neuroscience, vol. 481, pp. 197–218, Jan. 2022, doi: 10.1016/j.neuroscience.2021.11.017.
[2] “On the predictability of epileptic seizures,” Clin. Neurophysiol., vol. 116, no. 3, pp. 569–587, Mar. 2005, doi: 10.1016/j.clinph.2004.08.025.
[3] B. Blankertz, R. Tomioka, S. Lemm, M. Kawanabe, and K. Muller, “Optimizing Spatial filters for Robust EEG Single-Trial Analysis,” IEEE Signal Process. Mag., vol. 25, no. 1, pp. 41–56, 2008, doi: 10.1109/MSP.2008.4408441.
[4] K. K. Ang, Z. Y. Chin, C. Wang, C. Guan, and H. Zhang, “Filter Bank Common Spatial Pattern Algorithm on BCI Competition IV Datasets 2a and 2b,” Front. Neurosci., vol. 6, Mar. 2012, doi: 10.3389/fnins.2012.00039.
[5] T. N. Alotaiby, S. A. Alshebeili, F. M. Alotaibi, and S. R. Alrshoud, “Epileptic Seizure Prediction Using CSP and LDA for Scalp EEG Signals,” Comput. Intell. Neurosci., vol. 2017, p. e1240323, Oct. 2017, doi: 10.1155/2017/1240323.
[6] M. X. Cohen, “A data-driven method to identify frequency boundaries in multichannel electrophysiology data,” J. Neurosci. Methods, vol. 347, p. 108949, Jan. 2021, doi: 10.1016/j.jneumeth.2020.108949.
[7] B. Litt and K. Lehnertz, “Seizure prediction and the preseizure period,” Curr. Opin. Neurol., vol. 15, no. 2, p. 173, Apr. 2002.
[8] L. Kuhlmann, K. Lehnertz, M. P. Richardson, B. Schelter, and H. P. Zaveri, “Seizure prediction — ready for a new era,” Nat. Rev. Neurol., vol. 14, no. 10, pp. 618–630, Oct. 2018, doi: 10.1038/s41582-018-0055-2.
[9] M. O. Baud, T. Proix, V. R. Rao, and K. Schindler, “Chance and risk in epilepsy,” Curr. Opin. Neurol., vol. 33, no. 2, p. 163, Apr. 2020, doi: 10.1097/WCO.0000000000000798.
[10] M. M. Ahsan, S. A. Luna, and Z. Siddique, “Machine-Learning-Based Disease Diagnosis: A Comprehensive Review,” Healthcare, vol. 10, no. 3, p. 541, Mar. 2022, doi: 10.3390/healthcare10030541.
[11] F. Mormann, R. G. Andrzejak, C. E. Elger, and K. Lehnertz, “Seizure prediction: the long and winding road,” Brain J. Neurol., vol. 130, no. Pt 2, pp. 314–333, Feb. 2007, doi: 10.1093/brain/awl241.
[12] L. Hu and Z. Zhang, Eds., EEG Signal Processing and Feature Extraction. Singapore: Springer Singapore, 2019. doi: 10.1007/978-981-13-9113-2.
[13] R. Kohavi, “A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection,” vol. 14, Mar. 2001.
[14] P.-N. Yu, S. A. Naiini, C. N. Heck, C. Y. Liu, D. Song, and T. W. Berger, “A sparse Laguerre-Volterra autoregressive model for seizure prediction in temporal lobe epilepsy,” in 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Aug. 2016, pp. 1664–1667. doi: 10.1109/EMBC.2016.7591034.
[15] P.-N. Yu, C. Y. Liu, C. N. Heck, T. W. Berger, and D. Song, “A sparse multiscale nonlinear autoregressive model for seizure prediction,” J. Neural Eng., vol. 18, no. 2, p. 026012, Feb. 2021, doi: 10.1088/1741-2552/abdd43.
[16] 蔡明倫 and M.-L. Tsai, “定向轉移函數基於卷積神經網路與 L1 正則化線性迴歸之癲癇預測,” thesis, 2023. Accessed: Jul. 09, 2024. [Online]. Available: https://nckur.lib.ncku.edu.tw/handle/987654321/302894
[17] B. H. Brinkmann et al., “Crowdsourcing reproducible seizure forecasting in human and canine epilepsy,” Brain, vol. 139, no. 6, pp. 1713–1722, Jun. 2016, doi: 10.1093/brain/aww045.
[18] D. H. Wolpert, “The Lack of A Priori Distinctions Between Learning Algorithms,” Neural Comput., vol. 8, no. 7, pp. 1341–1390, Oct. 1996, doi: 10.1162/neco.1996.8.7.1341.
[19] K. Zhou, Z. Liu, Y. Qiao, T. Xiang, and C. C. Loy, “Domain Generalization: A Survey,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 45, no. 4, pp. 4396–4415, Apr. 2023, doi: 10.1109/TPAMI.2022.3195549.
[20] “Trends in biomedical signal feature extraction,” Biomed. Signal Process. Control, vol. 43, pp. 41–63, May 2018, doi: 10.1016/j.bspc.2018.02.008.
[21] B. Russell and J. Han, “Jean Morlet and the Continuous Wavelet Transform,” vol. 28, 2016.
[22] Z. J. Koles, M. S. Lazar, and S. Z. Zhou, “Spatial patterns underlying population differences in the background EEG,” Brain Topogr., vol. 2, no. 4, pp. 275–284, Jun. 1990, doi: 10.1007/BF01129656.
[23] Z. J. Koles, “The quantitative extraction and topographic mapping of the abnormal components in the clinical EEG,” Electroencephalogr. Clin. Neurophysiol., vol. 79, no. 6, pp. 440–447, Dec. 1991, doi: 10.1016/0013-4694(91)90163-X.
[24] Z. J. Koles, J. C. Lind, and P. Flor-Henry, “Spatial patterns in the background EEG underlying mental disease in man,” Electroencephalogr. Clin. Neurophysiol., vol. 91, no. 5, pp. 319–328, Nov. 1994, doi: 10.1016/0013-4694(94)90119-8.
[25] A. Daffertshofer, C. J. C. Lamoth, O. G. Meijer, and P. J. Beek, “PCA in studying coordination and variability: a tutorial,” Clin. Biomech., vol. 19, no. 4, pp. 415–428, May 2004, doi: 10.1016/j.clinbiomech.2004.01.005.
[26] 莊瑋智, “應用共同空間型樣法改善EEG 控制矯型手於中風病患之復健,” 國立成功大學, Jun. 2013.
[27] M. Tangermann et al., “Review of the BCI Competition IV,” Front. Neurosci., vol. 6, Jul. 2012, doi: 10.3389/fnins.2012.00055.
[28] G. Zheng et al., “Seizure prediction model based on method of common spatial patterns and support vector machine,” in 2012 IEEE International Conference on Information Science and Technology, Mar. 2012, pp. 29–34. doi: 10.1109/ICIST.2012.6221603.
[29] Y. Zhang, Y. Guo, P. Yang, W. Chen, and B. Lo, “Epilepsy Seizure Prediction on EEG Using Common Spatial Pattern and Convolutional Neural Network,” IEEE J. Biomed. Health Inform., vol. 24, no. 2, pp. 465–474, Feb. 2020, doi: 10.1109/JBHI.2019.2933046.
[30] M. Amiri, H. Aghaeinia, and H. R. Amindavar, “Automatic epileptic seizure detection in EEG signals using sparse common spatial pattern and adaptive short-time Fourier transform-based synchrosqueezing transform,” Biomed. Signal Process. Control, vol. 79, p. 104022, Jan. 2023, doi: 10.1016/j.bspc.2022.104022.
[31] “The Science of Brainwaves - the Language of the Brain,” NeuroHealth Associates. Accessed: Jul. 15, 2024. [Online]. Available: https://nhahealth.com/brainwaves-the-language/
[32] H. C.s, A. A, C. C. Nair, S. M. J. Nair, and F. B. U, “A Review on Brainwave Therapy,” World J. Pharm. Sci., pp. 59–66, Oct. 2020.
[33] “Electrophysiological CNS-processes related to associative learning in humans,” Behav. Brain Res., vol. 296, pp. 211–232, Jan. 2016, doi: 10.1016/j.bbr.2015.09.011.
[34] B. Ghojogh, F. Karray, and M. Crowley, “Eigenvalue and Generalized Eigenvalue Problems: Tutorial,” May 20, 2023, arXiv: arXiv:1903.11240. doi: 10.48550/arXiv.1903.11240.
[35] L. Kuhlmann et al., “Epilepsyecosystem.org: crowd-sourcing reproducible seizure prediction with long-term human intracranial EEG,” Brain, vol. 141, no. 9, pp. 2619–2630, Sep. 2018, doi: 10.1093/brain/awy210.
[36] A. H. Shoeb, “Application of machine learning to epileptic seizure onset detection and treatment,” Thesis, Massachusetts Institute of Technology, 2009. Accessed: Jun. 19, 2024. [Online]. Available: https://dspace.mit.edu/handle/1721.1/54669
[37] M. R. Bower and P. S. Buckmaster, “Changes in Granule Cell Firing Rates Precede Locally Recorded Spontaneous Seizures by Minutes in an Animal Model of Temporal Lobe Epilepsy,” J. Neurophysiol., vol. 99, no. 5, pp. 2431–2442, May 2008, doi: 10.1152/jn.01369.2007.
[38] I. Jemal, N. Mezghani, L. Abou-Abbas, and A. Mitiche, “An Interpretable Deep Learning Classifier for Epileptic Seizure Prediction Using EEG Data,” IEEE Access, vol. 10, pp. 60141–60150, 2022, doi: 10.1109/ACCESS.2022.3176367.
[39] S. Haufe et al., “On the interpretation of weight vectors of linear models in multivariate neuroimaging,” NeuroImage, vol. 87, pp. 96–110, Feb. 2014, doi: 10.1016/j.neuroimage.2013.10.067.
[40] F. Lotte and C. Guan, “Regularizing common spatial patterns to improve BCI designs: unified theory and new algorithms,” IEEE Trans. Biomed. Eng., vol. 58, no. 2, pp. 355–362, Feb. 2011, doi: 10.1109/TBME.2010.2082539.
[41] K. K. Ang, Z. Y. Chin, H. Zhang, and C. Guan, “Filter Bank Common Spatial Pattern (FBCSP) in Brain-Computer Interface,” in 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence), Jun. 2008, pp. 2390–2397. doi: 10.1109/IJCNN.2008.4634130.
[42] A. Grossmann and J. Morlet, “Decomposition of Hardy Functions into Square Integrable Wavelets of Constant Shape,” SIAM J. Math. Anal., vol. 15, no. 4, pp. 723–736, Jul. 1984, doi: 10.1137/0515056.