脑电信号自适应加权模型在癫痫发作检测中的应用

Adaptive Weighted Modeling of EEG Signals in Seizure Detection

摘要

癫痫发作的自动检测在癫痫的预防和治疗策略中扮演着至关重要的角色,基于深度学习的脑电图(EEG)分析技术在该领域取得了显著进展．然而,现有方法在处理多通道 EEG 数据时,未能充分考虑不同脑电通道在癫痫发作检测中的独特影响,导致模型缺乏可解释性．为此,提出两种深度学习模型,识别在检测过程中发挥关键作用的脑电通道,旨在提高癫痫发作检测的可解释性．第1种模型是结合压缩激励模块的全卷积网络检测模型(FCNse),首先,利用其特有的卷积结构,模型能够在时间轴上进行有效的特征提取和空间信息的增强；其次,引入上采样机制和跳级结构处理 EEG 信号中的局部时空特征；最后,结合压缩激励模块以关注更重要的特征信息进行癫痫检测．第 2 种模型是基于 Transformer 的网络检测模型(Transet),首先,将不同的脑电通道视为一句话当中的不同单词输入到 Transformer 网络；其次,利用其特有的自注意力机制,模型能够更有效地学习 EEG 特征,并专注于与癫痫发作密切相关的信息；最后,探索自注意力机制在模型自适应分配通道权重过程中的作用．在 CHB-MIT 数据集上的实验结果表明,FCNse 模型实现了 0.88 的A_UC值和89.2%的准确率,Transet模型达到了0.90的A_UC值和87.4%的准确率．这两种不同类型的深度学习模型在通道权重图结果上表现出高度的一致性,进一步验证了其准确性和可靠性,为癫痫发作的自动检测提供了新的解决方案,有助于推动癫痫诊断和治疗策略的发展．

Abstract

Automated seizure detection plays a crucial role in the prevention and treatment of epilepsy. Deep learning-based electroencephalography(EEG)analysis techniques have made significant progress in this field. However,extant methods have failed to fully consider the unique influences of different EEG channels in seizure detection when processing multichannel EEG data,resulting in models that lack interpretability. To this end,two innovative deep learning models are proposed to identify which EEG channels play key roles in the detection process,with the aim of improving the interpretability of seizure detection. The first model is the fully convolutional network(FCNse) detection model,which is combined with a compressed excitation module. The model’s unique convolutional structure enables effective feature extraction and spatial information enhancement on the time axis. Next,an upsampling mechanism and a skip-level structure are introduced to address the local spatio-temporal characteristics present within the EEG signal. Finally,a compression excitation module is combined to focus on more important feature information for epilepsy detection. The second model is the Transformer-based network detection model(Transet). In this model,different EEG channels are considered as different words in a sentence. Then,these words are input into the Transformer network. Next,the model employs a distinctive self-attention mechanism,enabling it to learn EEG features with greater efficiency and focus on information closely related to seizures. Finally,the role of the self-attention mechanism is explored in the process of the model’s adaptive allocation of channel weights. The experimental results on the CHB-MIT dataset show that the FCNse model achieves an A_UC value of 0.88 and an accuracy rate of 89.2%,and the Transet model achieves an A_UC value of 0.90 and an accuracy rate of 87.4%. The findings indicate a high degree of consistency between the two distinct types of deep learning models in terms of the results obtained from the channel weight map,thereby validating their accuracy and reliability. This research provides a novel solution for the automatic detection of epileptic seizures and contributes to the advancement of diagnostic and therapeutic strategies for epilepsy.

关键词

癫痫发作检测 / 脑电图 / 深度学习

Key words

seizure detection / electroencephalography(EEG) / deep learning

参考文献

[1] Fiest K M, Sauro K M, Wiebe S, et al. Prevalence and incidence of epilepsy：A systematic review and meta—analysis of international studies[J]. Neurology, 2017, 88(3):296-303.

[2] Freestone D R, Karoly P J, Cook M J. A forward—looking review of seizure prediction[J]. Current Opinion in Neurology, 2017, 30(2):167-173.

[3] 于淼, 刘秀云, 岳伟. 临床脑电图技术发展历程与前沿展望[J]. 中国现代神经疾病杂志, 2024, 24(12):1063-1070.

[4] Yu Miao, Liu Xiuyun, Yue Wei. Development of clinical electroencephalogram technology and frontier outlook[J]. Chinese Journal of Modern Neurological Diseases, 2024, 24(12):1063-1070(in Chinese).

[5] Usman S M, Khalid S, Bashir Z. Epileptic seizure prediction using scalp electroencephalogram signals[J]. Biocybernetics and Biomedical Engineering, 2021, 41(1):211-220.

[6] Dash D P, Kolekar M, Chakraborty C, et al. Review of machine and deep learning techniques in epileptic seizure detection using physiological signals and sentiment analysis[J]. ACM Transactions on Asian and Low—Resource Language Information Processing, 2024, 23(1):1-29.

[7] Babb T L, Mariani E, Crandall P H. An electronic circuit for detection of EEG seizures recorded with implanted electrodes[J]. Electroencephalography and Clinical Neurophysiology, 1974, 37(3):305-308.

[8] 王晶晶, 郝思瑶, 胡珊珊. 基于机器学习的癫痫诊疗研究综述[J]. 山东师范大学学报(自然科学版), 2023, 38(4):304-325.

[9] Wang Jingjing, Hao Siyao, Hu Shanshan. A review of research on epilepsy diagnosis and treatment based on machine learning[J]. Journal of Shandong Normal University(Natural Science Edition), 2023, 38(4):304-325(in Chinese).

[10] Maimaiti B, Meng H, Lü Y, et al. An overview of EEG—based machine learning methods in seizure prediction and opportunities for neurologists in this field[J]. Neuroscience, 2022, 481:197-218.

[11] Abdulwahhab A H, Abdulaal A H, Al—Ghrairi A H T, et al. Detection of epileptic seizure using EEG signals analysis based on deep learning techniques[J]. Chaos,Solitons & Fractals, 2024, 181:114700.

[12] Georgis—Yap Z, Popovic M R, Khan S S. Supervised and unsupervised deep learning approaches for EEG seizure prediction[J]. Journal of Healthcare Informatics Research, 2024, 8(2):286-312.

[13] 张汉明, 马金刚, 张宁宁, 等. 深度学习在癫痫检测中的应用进展[J]. 计算机工程与应用, 2023, 59(10):35-47.

[14] Zhang Hanming, Ma Jingang, Zhang Ningning, et al. Advances in deep learning for epilepsy detection[J]. Computer Engineering and Applications, 2023, 59(10):35-47(in Chinese).

[15] Ahmedt—Aristizabal D, Armin M A, Hayder Z, et al. Deep learning approaches for seizure video analysis：A review[J]. Epilepsy & Behavior, 2024, 154:109735.

[16] Hassan M M, Haque R, Shariful Islam S M, et al. NeuroWave—Net：Enhancing epileptic seizure detection from EEG brain signals via advanced convolutional and long short—term memory networks[J]. AIMS Bioengineering, 2024, 11(1):85-109.

[17] Liu Y S, Jiang Y H, Liu J, et al. Efficient EEG feature learning model combining random convolutional kernel with wavelet scattering for seizure detection[J]. International Journal of Neural Systems, 2024, 34(11):2450060.

[18] Pan Y Y, Dong F Y, Yao W, et al. Empirical mode decomposition for deep learning—based epileptic seizure detection in few—shot scenario[J]. IEEE Access, 2024, 12:86583-86595.

[19] Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach,USA, 2017:6000-6010.

[20] Bhattacharya A, Baweja T, Karri S P K. Epileptic seizure prediction using deep transformer model[J]. International Journal of Neural Systems, 2022, 32(2):2150058.

[21] Peng R M, Du Z B, Zhao C M, et al. Multi—branch mutual—distillation transformer for EEG—based seizure subtype classification[J]. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2024, 32:831-839.

[22] Zhang J W, Swinnen L, Chatzichristos C, et al. Learning robust representations of tonic—clonic seizures with cyclic transformer[J]. IEEE Journal of Biomedical and Health Informatics, 2024, 28(6):3721-3731.

[23] Jebin B M, Rejula M A, Eberlein G. Neonatal seizure detection using GLCM feature extraction & AlexNet classification[J]. Multimedia Tools and Applications, 2024, 83(35):83139-83155.

[24] Khurshid D, Wahid F, Ali S, et al. A deep neural network—based approach for seizure activity recognition of epilepsy sufferers[J]. Frontiers in Medicine, 2024, 11:1405848.

[25] Yochum M, Kachenoura A, Aud’hui M, et al. Deep learning—based early detection of absence seizures in children[J]. Biomedical Signal Processing and Control, 2025, 100:106845.

[26] Lawhern V J, Solon A J, Waytowich N R, et al. EEGNet：A compact convolutional neural network for EEG—based brain—computer interfaces[J]. Journal of Neural Engineering, 2018, 15(5):056013.

[27] Wei Z C, Zou J Z, Zhang J, et al. Automatic epileptic EEG detection using convolutional neural network with improvements in time—domain[J]. Biomedical Signal Processing and Control, 2019, 53:101551.

[28] Craley J, Johnson E, Jouny C, et al. Automated inter—patient seizure detection using multichannel convolutional and recurrent neural networks[J]. Biomedical Signal Processing and Control, 2021, 64:102360.

[29] Pedoeem J, Abittan S, Bar Yosef G, et al. TABS：Transformer based seizure detection[C]// 2020 IEEE Signal Processing in Medicine and Biology Symposium(SPMB). Philadelphia,USA, 2022:133-160.

[30] Zhao Y N, Chu D Y, He J T, et al. Interactive local and global feature coupling for EEG—based epileptic seizure detection[J]. Biomedical Signal Processing and Control, 2023, 81:104441.

[31] Zhao W, Wang W F, Patnaik L M, et al. Residual and bidirectional LSTM for epileptic seizure detection[J]. Frontiers in Computational Neuroscience, 2024, 18:1415967.

基金资助

超声医学工程国家重点实验室基金资助项目(2022KFKT004)

国家自然科学基金资助项目(62071323)

天津市科技计划重点项目(22JCZDJC00220)

天津市自然科学基金资助项目(23JCZDJC0020)

河南省重点研发与推广专项科技攻关资助项目(232102210030)