簡易檢索 / 詳目顯示

研究生: 陳政淳
Chen, Zheng-Chun
論文名稱: 定向轉移函數基於域對抗式神經網路應用於跨受試者癲癇偵測
Cross-subject Seizure Detection Using Directed Transfer Function Through Domain Adversarial Neural Network
指導教授: 游本寧
Yu, Pen-Ning
學位類別: 碩士
Master
系所名稱: 工學院 - 機械工程學系
Department of Mechanical Engineering
論文出版年: 2025
畢業學年度: 113
語文別: 中文
論文頁數: 91
中文關鍵詞: 癲癇偵測領域適應CNNDANN定向轉移函數
外文關鍵詞: seizure detection, deep learning, directed transfer function, domain adaptation
相關次數: 點閱:21下載:0
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報
  • 癲癇為慢性神經系統疾病之一,有效的癲癇偵測方法可以幫助受試者改善其生活品質。通過發展自動化癲癇偵測方法可以減少醫師人工判讀腦波訊號的時間。癲癇偵測方法可藉由特徵擷取找出發作期(Ictal)以及發作間期(Interictal)之間腦波訊號差異,並以之進行癲癇偵測。本研究使用定向轉移函數(Directed Transfer Function, DTF)擷取腦波訊號於頻域上頻道之間的傳遞方向性以及其強弱關係,通過特徵於兩腦波狀態上的差異進行分類。
    由於受試者個體因素影響,癲癇偵測方法通常針對不同受試者個體建立模型,此方法在標記資料上的需求量高,因此當癲癇偵測模型欲應用於新受試者資料上,需要專業人員標記,而通過建立跨受試者間模型建立可以降低於新受試者標記資料的需求,進而減少標記所需的時間成本。隨著卷積神經網路(Convolutional Neural Networks, CNN)於圖像分類應用快速發展,本研究使用DTF堆疊成的二維圖像作為特徵,依照有無使用跨受試者無標記資料兩種情況建立兩種跨受試者模型,分別為未使用無標記資料的CNN與使用無標記資料的域對抗式神經網路(Domain Adversarial Neural Network, DANN),並比較兩種模型之分類能效。
    本研究使用Kaggle開源資料集中四隻狗腦波訊號建立跨受試者模型並以AUC評估模型的分類效能。最終在測試資料集上,未使用無標記資料下得出平均AUC 0.84、使用無標記資料下得出平均AUC 0.87,顯示使用DTF特徵配合CNN模型於四隻狗資料建立之跨受試者模型上即可達到有效分類;雖然在有無使用受試者標記資料改善並無達到顯著差異(paired t-test, p = 0.31),在特定受試者上分類能效仍有小幅改善,因此推薦以DTF作為特徵配合CNN建立跨受試者模型,並嘗試使用以DANN進一步改善模型分類能效。

    Epilepsy is a chronic neurological disorder, and effective automated seizure detection can improve patients' quality of life by reducing reliance on manual EEG interpretation. This study employs the directed transfer function (DTF) to extract frequency-domain features that capture directional connectivity between EEG channels for distinguishing ictal and interictal states. To reduce the need for labeled data in new subjects, we develop cross-subject models using DTF-based images with two approaches: convolutional neural network(CNN) without data of target subject and domain adversarial neural network (DANN) leveraging both unla-beled and a small amount of labeled data of target subject.
    Using EEG data from four dogs in a public Kaggle dataset, we evaluate model performances via AUC. The CNN and DANN models achieve average AUCs of 0.84 and 0.87, respectively. While the improvement is not statistically significant (p = 0.31), DANN offers better performance in certain subjects. These results suggest DTF features with CNN are effective for cross-subject seizure de-tection, and DANN may further enhance classification performances of seizure detection.

    摘要i 致謝ix 目錄x 表目錄xiii 圖目錄xiv 符號表xvii 第一章 緒論1 1.1 癲癇(Epilepsy)1 1.2 腦電圖(Electroencephalogram, EEG)1 1.3 癲癇偵測文獻回顧3 1.3.1 特徵擷取3 1.3.2 深度學習於癲癇偵測5 1.4 不平衡資料集處理方式6 1.5 遷移學習(Transfer Learning)7 1.5.1 域適應(Domain Adaptation)8 1.6 研究動機與目的10 第二章 研究方法11 2.1 實驗分析及流程11 2.2 資料集使用12 2.3 特徵工程(定向轉移函數)14 2.3.1 多變量自迴歸模型(Multivariate Autoregressive Model, MVAR)14 2.3.2 MVAR模型階數選取15 2.3.3 定向轉移函數(Directed transfer function, DTF)18 2.3.4 定向轉移函數特徵形式21 2.4 訓練模型23 2.4.1 卷積神經網路(Convolutional Neural Networks, CNN)23 2.4.1.1 二維卷積神經網路(Two-dimensional Convolution Neural Network, 2D-CNN) 29 2.4.2 非監督式域適應(Unsupervised Domain Adaptation)31 2.4.3 域對抗式神經網路(Domain Adversarial Neural Networks, DANN)32 2.5 訓練流程35 2.5.1 訓練步驟以及資料集使用方式35 2.5.2 不平衡資料集處理方式37 2.5.3 Z-Score標準化39 2.5.4 CNN與DANN訓練架構以及流程40 2.6 t-SNE(t-Distributed Stochastic Neighbor Embedding)42 2.7 評估指標 AUC(Area Under the Curve)43 第三章 實驗結果44 3.1 MVAR模型擬合44 3.1.1 受試者EEG訊號對應MVAR模型階數44 3.1.2 MVAR擬合誤差結果45 3.2 DTF結果對照48 3.3 2D-CNN與2D-DANN於驗證集結果52 3.4 2D-DANN 特徵分布 t-SNE結果56 3.5 測試資料集結果57 第四章 討論58 4.1 DTF特徵選擇資訊建議58 4.2 患者模型建立與特徵影響探討60 4.2.1 患者模型建立60 4.2.2 MVAR模型擬合探討61 第五章 結論與未來展望63 5.1 結論63 5.2 未來展望63 參考文獻65 附錄70 A. 患者資料集內容70 B. MVAR於患者資料階數選擇71

    [1] E. Beghi, “The Epidemiology of Epilepsy,” Neuroepidemiology, vol. 54, no. 2, pp. 185–191, 2020.
    [2] X. Zhang, L. Yao, M. Dong, Z. Liu, Y. Zhang, and Y. Li, “Adversarial Represen-tation Learning for Robust Patient-Independent Epileptic Seizure Detection,” IEEE J. Biomed. Health Inform., vol. 24, no. 10, pp. 2852–2859, Oct. 2020.
    [3] M. K. Siddiqui, R. Morales-Menendez, X. Huang, and N. Hussain, “A review of epileptic seizure detection using machine learning classifiers,” Brain Inf., vol. 7, no. 1, p. 5, Dec. 2020.
    [4] S. N. Baldassano et al., “Crowdsourcing seizure detection: algorithm development and validation on human implanted device recordings,” Brain, vol. 140, no. 6, pp. 1680–1691, Jun. 2017.
    [5] “Epilepsy.” Accessed: Sep. 04, 2024. [Online].
    [6] A. Ulate-Campos, F. Coughlin, M. Gaínza-Lein, I. S. Fernández, P. L. Pearl, and T. Loddenkemper, “Automated seizure detection systems and their effectiveness for each type of seizure,” Seizure, vol. 40, pp. 88–101, Aug. 2016.
    [7] R. S. Fisher et al., “Operational classification of seizure types by the International League Against Epilepsy: Position Paper of the ILAE Commission for Classifica-tion and Terminology,” Epilepsia, vol. 58, no. 4, pp. 522–530, Apr. 2017.
    [8] J. A. French, “Refractory Epilepsy: Clinical Overview,” Epilepsia, vol. 48, no. s1, pp. 3–7, Mar. 2007.
    [9] S. Ramgopal et al., “Seizure detection, seizure prediction, and closed-loop warn-ing systems in epilepsy,” Epilepsy & Behavior, vol. 37, pp. 291–307, Aug. 2014.
    [10] S. Shorvon, R. Guerrini, M. Cook, and S. Lhatoo, Oxford Textbook of Epilepsy and Epileptic Seizures. OUP Oxford, 2012.
    [11] D. P. Subha, P. K. Joseph, R. Acharya U, and C. M. Lim, “EEG signal analysis: a survey,” J Med Syst, vol. 34, no. 2, pp. 195–212, Apr. 2010.
    [12] B. Pohlmann-Eden and M. Newton, “First seizure: EEG and neuroimaging fol-lowing an epileptic seizure,” Epilepsia, vol. 49, no. s1, pp. 19–25, 2008.
    [13] K. J. Blinowska and J. Zygierewicz, “Practical Biomedical Signal Analysis Using MATLAB,” 2012.
    [14] S. R. Mousavi, M. Niknazar, and B. V. Vahdat, “Epileptic Seizure Detection using AR Model on EEG Signals,” in 2008 Cairo International Biomedical Engineering Conference, Feb. 2008, pp. 1–4.
    [15] A. Alkan, E. Koklukaya, and A. Subasi, “Automatic seizure detection in EEG us-ing logistic regression and artificial neural network,” Journal of Neuroscience Methods, vol. 148, no. 2, pp. 167–176, Oct. 2005.
    [16] A. Bozdogan, M. Ustu, R. Ileri, and F. Latifoglu, “Detection of Epilepsy Using Wavelet Coherence and Convolutional Neural Networks,” in 2021 Medical Technologies Congress (TIPTEKNO), Antalya, Turkey: IEEE, Nov. 2021, pp. 1–4.
    [17] A. T. Tzallas, M. G. Tsipouras, and D. I. Fotiadis, “Epileptic Seizure Detection in EEGs Using Time–Frequency Analysis,” IEEE Transactions on Information Technology in Biomedicine, vol. 13, no. 5, pp. 703–710, Sep. 2009.
    [18] A. Supratak, L. Li, and Y. Guo, “Feature extraction with stacked autoencoders for epileptic seizure detection,” in 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Aug. 2014, pp. 4184–4187.
    [19] X. Hu, S. Yuan, F. Xu, Y. Leng, K. Yuan, and Q. Yuan, “Scalp EEG classification using deep Bi-LSTM network for seizure detection,” Computers in Biology and Medicine, vol. 124, p. 103919, Sep. 2020.
    [20] I. Assali, A. Ghazi Blaiech, A. Ben Abdallah, K. Ben Khalifa, M. Carrère, and M. Hédi Bedoui, “CNN-based classification of epileptic states for seizure prediction using combined temporal and spectral features,” Biomedical Signal Processing and Control, vol. 82, p. 104519, Apr. 2023.
    [21] A. Chattopadhyay and M. Maitra, “MRI-based brain tumour image detection us-ing CNN based deep learning method,” Neuroscience Informatics, vol. 2, no. 4, p. 100060, Dec. 2022.
    [22] A. Farooq, S. Anwar, M. Awais, and S. Rehman, “A deep CNN based multi-class classification of Alzheimer’s disease using MRI,” in 2017 IEEE International Conference on Imaging Systems and Techniques (IST), Oct. 2017, pp. 1–6.
    [23] Y. Pan, X. Zhou, F. Dong, J. Wu, Y. Xu, and S. Zheng, “Epileptic Seizure Detec-tion with Hybrid Time-Frequency EEG Input: A Deep Learning Approach,” Computational and Mathematical Methods in Medicine, vol. 2022, no. 1, p. 8724536, 2022.
    [24] S. Wong et al., “EEG datasets for seizure detection and prediction— A review,” Epilepsia Open, vol. 8, no. 2, pp. 252–267, Jun. 2023.
    [25] J. M. Johnson and T. M. Khoshgoftaar, “Survey on deep learning with class im-balance,” J Big Data, vol. 6, no. 1, p. 27, Dec. 2019.
    [26] M. R. Longadge, S. S. Dongre, and D. L. Malik, “Class Imbalance Problem in Data Mining: Review,” vol. 2, no. 1, 2013.
    [27] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE: Synthetic Minority Over-sampling Technique,” jair, vol. 16, pp. 321–357, Jun. 2002.
    [28] M. Kukar and I. Kononenko, “Cost-Sensitive Learning with Neural Networks,” Machine Learning and Data Mining.
    [29] H. Kaur, H. S. Pannu, and A. K. Malhi, “A Systematic Review on Imbalanced Data Challenges in Machine Learning: Applications and Solutions,” ACM Com-put. Surv., vol. 52, no. 4, pp. 1–36, Jul. 2020.
    [30] S. J. Pan and Q. Yang, “A Survey on Transfer Learning,” IEEE Transactions on Knowledge and Data Engineering, vol. 22, no. 10, pp. 1345–1359, Oct. 2010.
    [31] J. Quiñonero-Candela, Ed., Dataset shift in machine learning. in Neural infor-mation processing series. Cambridge, Mass.: MIT Press, 2009.
    [32] V. M. Patel, R. Gopalan, R. Li, and R. Chellappa, “Visual Domain Adaptation: A survey of recent advances,” IEEE Signal Processing Magazine, vol. 32, no. 3, pp. 53–69, May 2015.
    [33] B. Zhang et al., “Cross-Subject Seizure Detection in EEGs Using Deep Transfer Learning,” Computational and Mathematical Methods in Medicine, vol. 2020, pp. 1–8, May 2020.
    [34] M. Nasseri et al., “Non-invasive wearable seizure detection using long–short-term memory networks with transfer learning,” J. Neural Eng., vol. 18, no. 5, p. 056017, Oct. 2021.
    [35] F. Sun, J. Li, H. Liu, and Z. Chu, Eds., Cognitive Computation and Systems: First International Conference, ICCCS 2022, Beijing, China, December 17–18, 2022, Revised Selected Papers, vol. 1732. in Communications in Computer and Infor-mation Science, vol. 1732. Singapore: Springer Nature Singapore, 2023.
    [36] Y. Ganin and V. Lempitsky, “Unsupervised Domain Adaptation by Backpropaga-tion,” Feb. 27, 2015, arXiv: arXiv:1409.7495. Accessed: Sep. 03, 2024.
    [37] S. J. Pan, I. W. Tsang, J. T. Kwok, and Q. Yang, “Domain Adaptation via Transfer Component Analysis,” IEEE Transactions on Neural Networks, vol. 22, no. 2, pp. 199–210, Feb. 2011.
    [38] E. Tzeng, J. Hoffman, N. Zhang, K. Saenko, and T. Darrell, “Deep Domain Confusion: Maximizing for Domain Invariance,” Dec. 10, 2014, arXiv: arXiv:1412.3474. Accessed: Sep. 03, 2024.
    [39] Y. Li, N. Wang, J. Shi, X. Hou, and J. Liu, “Adaptive Batch Normalization for practical domain adaptation,” Pattern Recognition, vol. 80, pp. 109–117, Aug. 2018.
    [40] A. Sicilia, X. Zhao, and S. J. Hwang, “Domain adversarial neural networks for domain generalization: when it works and how to improve,” Mach Learn, vol. 112, no. 7, pp. 2685–2721, Jul. 2023.
    [41] G. Dharani. Y, N. G. Nair, P. Satpathy, and J. Christopher, “Covariate Shift: A Review and Analysis on Classifiers,” in 2019 Global Conference for Advance-ment in Technology (GCAT), Oct. 2019, pp. 1–6.
    [42] A. Bhattacharyya and R. B. Pachori, “A Multivariate Approach for Pa-tient-Specific EEG Seizure Detection Using Empirical Wavelet Transform,” IEEE Trans. Biomed. Eng., vol. 64, no. 9, pp. 2003–2015, Sep. 2017.
    [43] Hao Qu and J. Gotman, “A patient-specific algorithm for the detection of seizure onset in long-term EEG monitoring: possible use as a warning device,” IEEE Trans. Biomed. Eng., vol. 44, no. 2, pp. 115–122, Feb. 1997.
    [44] X. Zhang, L. Yao, M. Dong, Z. Liu, Y. Zhang, and Y. Li, “Adversarial Represen-tation Learning for Robust Patient-Independent Epileptic Seizure Detection,” IEEE Journal of Biomedical and Health Informatics, vol. 24, no. 10, pp. 2852–2859, Oct. 2020.
    [45] K.-O. Cho and H.-J. Jang, “Comparison of different input modalities and network structures for deep learning-based seizure detection,” Sci Rep, vol. 10, no. 1, p. 122, Jan. 2020.
    [46] G. Wang et al., “Seizure Prediction Using Directed Transfer Function and Con-volution Neural Network on Intracranial EEG,” IEEE Trans Neural Syst Rehabil Eng, vol. 28, no. 12, pp. 2711–2720, Dec. 2020.
    [47] 蔡明倫, “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.”.碩士論文, 國立成功大學機械工程學系, 臺灣臺南市, 2023.
    [48] F. Wendling, A. Hernandez, J.-J. Bellanger, P. Chauvel, and F. Bartolomei, “In-terictal to Ictal Transition in Human Temporal Lobe Epilepsy: Insights From a Computational Model of Intracerebral EEG,” Journal of Clinical Neurophysiology, vol. 22, no. 5, 2005.
    [49] C. Nwankpa, W. Ijomah, A. Gachagan, and S. Marshall, “Activation Functions: Comparison of trends in Practice and Research for Deep Learning,” Nov. 08, 2018, arXiv: arXiv:1811.03378. Accessed: Oct. 08, 2024.
    [50] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” in Advances in Neural Information Pro-cessing Systems, Curran Associates, Inc., 2012. Accessed: Oct. 09, 2024.
    [51] Y. Ganin et al., “Domain-Adversarial Training of Neural Networks,” May 26, 2016, arXiv: arXiv:1505.07818.
    [52] “One dimensional convolutional neural networks for seizure onset detection using long-term scalp and intracranial EEG,” Neurocomputing, vol. 459, pp. 212–222, Oct. 2021.
    [53] G. Wang et al., “Seizure Prediction Using Directed Transfer Function and Con-volution Neural Network on Intracranial EEG,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 28, no. 12, pp. 2711–2720, Dec. 2020.
    [54] X. Wang, et al, " One dimensional convolutional neural networks for seizure onset detection using long-term scalp and intracranial EEG ", Neurocomputing, vol 459, pp. 212-222, Oct. 2021.
    [55] S. Kiranyaz, O. Avci, O. Abdeljaber, T. Ince, M. Gabbouj, and D. J. Inman, “1D convolutional neural networks and applications: A survey,” Mechanical Systems and Signal Processing, vol. 151, p. 107398, Apr. 2021.
    [56] R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra,“Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Lo-calization,” in 2017 IEEE International Conference on Computer Vision (ICCV), Oct. 2017, pp. 618–626.

    無法下載圖示 校內:2027-08-31公開
    校外:2027-08-31公開
    電子論文尚未授權公開,紙本請查館藏目錄
    QR CODE