引用本文
郑润泽, 冯袁盟, 胡丁寒, 蒋铁甲, 高峰, 曹九稳. 儿童癫痫综合征智能分析:综述与展望. 模式识别与人工智能, 2024,37(2): 144-161
ZHENG Runze, FENG Yuanmeng, HU Dinghan, JIANG Tiejia, GAO Feng, CAO Jiuwen. Intelligent Analysis of Childhood Epileptic Syndrome: Overview and Prospect. PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE, 2024,37(2): 144-161.  
Doi: 10.16451/j.cnki.issn1003-6059.202402004
Permissions
Copyright©2024, 《模式识别与人工智能》编辑部
《模式识别与人工智能》编辑部
儿童癫痫综合征智能分析:综述与展望
郑润泽1,2, 冯袁盟1,2, 胡丁寒1,2, 蒋铁甲3,4, 高峰3,4, 曹九稳1,2
1.杭州电子科技大学 自动化学院 杭州 310018
2.杭州电子科技大学 浙江省机器学习与智慧健康国际合作基地 杭州 310018
3.浙江大学 医学院 附属儿童医院 神经内科 杭州 310052
4.浙江大学 医学院 附属儿童医院 国家儿童健康与疾病临床医学研究中心 杭州 310003
通讯作者:

曹九稳,博士,教授,主要研究方向为深度学习、神经网络、医学信号处理.E-mail:jwcao@hdu.edu.cn.

作者简介:

郑润泽,博士研究生,主要研究方向为智能信号分析与处理.E-mail:runzewuyu@hdu.edu.cn.

冯袁盟,博士研究生,主要研究方向为智能信号分析与处理.E-mail:jsnt_fym@126.com.

胡丁寒,博士,讲师,主要研究方向为机器学习、脑电信号分析与处理.E-mail:hdh@hdu.edu.cn.

蒋铁甲,硕士,副主任医师,主要研究方向为儿童神经系统疾病的电生理信号量化分析.E-mail:jiangyouze@zju.edu.cn.

高 峰,硕士,教授,主要研究方向为癫痫的电生理特点和病因学.E-mail:epilepsy@zju.edu.cn.

收稿日期: 2023-11-01 修回日期: 2024-01-18
资助项目: 国家重点研发计划项目(No.2021YFE0100100,2021YFE0205400)、国家自然科学基金联合项目(No.U1909209)、浙江省自然科学基金重点项目(No.LZ24F030010)、浙江省教育厅一般科研项目(No.Y202249784)资助
摘要

儿童癫痫综合征智能分析是指通过统计分析、机器学习等数据驱动方法,挖掘临床有效生物标志物,构建相应的专家系统,以解决临床和预后管理问题的研究.文中首先简述儿童癫痫综合征的定义、发作类型和分类等临床基础知识.然后,回顾基于脑电信号的儿童癫痫综合征智能分析框架和各组成部分典型方法存在的优缺点,包括数据收集及预处理、特征提取、决策器系统和专家系统.其中,将专家系统分为特定波形检测系统、诊断分类系统、发作检测系统、发作预测系统和量化评估系统,并进行全面概括与理论解释.最后,结合儿童癫痫综合征智能分析领域现有研究的局限性和挑战,展望未来研究方向,以推动儿童癫痫综合征智能分析系统的研究进展,减轻该病带来的负面影响.

关键词: 儿童癫痫综合征; 生物标志物; 脑电信号; 智能分析; 专家系统
中图分类号:TN911.7;TP183;R742.1
Intelligent Analysis of Childhood Epileptic Syndrome: Overview and Prospect
ZHENG Runze1,2, FENG Yuanmeng1,2, HU Dinghan1,2, JIANG Tiejia3,4, GAO Feng3,4, CAO Jiuwen1,2
1. School of Automation, Hangzhou Dianzi University, Hangzhou 310018
2. Machine Learning and I-health International Cooperation Base of Zhejiang Province, Hangzhou Dianzi University, Hangzhou 310018
3. Neurology Department, Children's Hospital, School of Medi-cine, Zhejiang University, Hangzhou 310052
4. National Clinical Research Center for Child Health, Children's Hospital, School of Medicine, Zhejiang University, Hangzhou 310003
Corresponding author:
CAO Jiuwen, Ph.D., professor. His research interests include deep learning, neural networks and medical signal processing.

About Author:
ZHENG Runze, Ph.D. candidate. His research interests include intelligent signal ana-lysis and processing.
FENG Yuanmeng, Ph.D. candidate. His research interests include intelligent signal analysis and processing.
HU Dinghan, Ph.D., lecturer. Her research interests include machine learning and EEG signal analysis and processing.
JIANG Tiejia, Master, deputy director. His research interests include quantitative analysis of electrophysiological signals in children's neurological diseases.
GAO Feng, Master, professor. His research interests include electrophysiological characteristics and etiology of epilepsy.

Fund:National Key Research and Development Program of China(No.2021YFE0100100,2021YFE0205400), Joint Funds of National Natural Science Foundation of China(No.U1909209), Key Program of Natural Science Foundation of Zhejiang Province(No.LZ24F030010), Scientific Research Fund of Zhejiang Provincial Education Department(No.Y202249784)
Abstract

The intelligent analysis of childhood epileptic syndrome refers to the research which aims at addressing clinical and prognostic management issues by data-driven methods such as statistical analysis and machine learning to explore clinically effective biomarkers and construct corresponding expert systems. Firstly, the definition, seizure types and classification of childhood epileptic syndrome are briefly introduced. Then, the advantages and disadvantages of the framework and typical methods of the intelligent analysis of childhood epileptic syndrome based on scalp electroencephalogram are reviewed, including data collection and preprocessing, feature extraction, decision-making systems, and expert systems. Specifically, the expert systems are divided into specific waveform detection systems, diagnostic classification systems, seizure detection systems, seizure prediction systems and quantitative assessment systems with a comprehensive summary and theoretical explanation. Finally, with the consideration of the limitations and challenges of the existing research in the field of intelligent analysis of childhood epileptic syndrome, future research directions are proposed to advance the study of intelligent analysis systems for childhood epileptic syndrome and alleviate the negative impact of the disease.

Key words: Key Words Childhood Epileptic Syndrome; Biomarker; Electroencephalogram; Intelligent Analysis; Expert System

儿童癫痫综合征(Childhood Epileptic Syndrome)是一种具有独特临床和脑电表征、特定病因(结构、遗传、代谢、免疫和感染)的癫痫疾病[1, 2].相比成人, 儿童癫痫综合征具有年龄依赖性、诱因复杂和预后管理困难等问题, 可引发严重的合并症、癫痫猝死、意外伤害等不良后果[3].

婴幼儿期是大脑发育的关键时期, 频繁的大脑异常放电可导致发育的不可逆性损伤, 如注意力缺陷、认知下降等, 并且伴有严重的精神、行为障碍, 给患儿家庭和社会带来巨大的负担[4].

截止目前, 对于儿童癫痫综合征病理的理解存在较多的局限性, 无法精准定义和划分[5].部分类型的儿童癫痫综合征依旧无法完全治愈[6].因此, 如何通过有效的工具研究儿童癫痫综合征病理, 明确病因, 精准定义对患者的早期诊断、治疗、评估和预后管理都具有至关重要的作用和价值.

脑内神经元组织活动伴随着明显的电生理反应, 当癫痫发作时, 神经元出现短暂或持续的异常同步, 电生理信号呈现明显变化[7].脑电信号(Electroencephalogram, EEG)不仅能长时间准确记录与这些与癫痫密切相关的特定波形, 包括慢波、棘波或复合波等, 并且价格低廉、实用性强[8, 9].

早在1985年国际抗癫痫联盟(International Lea-gue Against Epilepsy, ILAE)首次提出癫痫病和癫痫综合征分类之前, 不同的儿童癫痫综合征就已明确定义独特的EEG临床特征[10].因此EEG广泛应用于儿童癫痫综合征临床, 医生依此对儿童癫痫综合征进行诊断和评估.

EEG存在较大局限性, 容易受到临床经验、主观情感等因素的干扰[11, 12].由于儿童癫痫综合征机理尚不完全清楚, 对于特异性强的癫痫综合征无法精准诊断, 致使目前还没有有效的工具帮助临床医生客观标准的评估和诊断[13, 14].

随着人工智能技术的不断突破[15, 16, 17], 机器学习、深度学习等数据驱动的智能分析方法为解决儿童癫痫综合征面临的挑战提供新的思路[18, 19].机器学习通过任务相关的特征词典, 不断优化模型, 直至准确度阈值, 从而挖掘数据之间的关联性[20].机器学习虽能较好地克服统计分析的弊端, 但性能受到特征词典中有效信息的约束[21, 22].相反, 深度学习采用自主、多层和多样化的深度架构, 自动提取高层次和潜在的复杂特征, 但模型训练时需要大量的数据, 可解释性较差[23, 24].

综上所述, 各种方法均存在如下明显的优缺点.

1)EEG生物标志物的研究可解释性较强, 但受限于临床经验, 精度过度依赖于统计结果p值.

2)机器学习可挖掘EEG隐藏模式和关联性, 但精度和泛化性容易受数据质量、类别不均衡、个体差异等影响.

3)深度学习可直接解码、学习和理解原始EEG信号的深度特征, 但模型训练过拟合、泛化性等问题仍有待解决.

因此, 如何综合使用智能分析方法对儿童癫痫综合征EEG 进行准确分类和识别是构建智能诊断、预后、评估等专家系统研究的核心问题[25].

开发以特征工程和统计分析为辅助、机器学习和深度学习为决策器的智能分析系统[26]是解决该问题的有效研究方式之一.

儿童癫痫综合征智能分析系统[27, 28]一般由EEG采集及预处理、特征提取、决策器和专家系统组成.儿童癫痫综合征智能分析系统从预处理后的EEG中提取特征或生物标志物.决策器学习相关特征变化, 为专家系统提供决策.专家系统依靠决策器结果做出预警或为癫痫控制器提供反馈信号.癫痫干预闭环系统研究迄今只存在少数成功的临床研究, 几乎都是基于成人颅内脑电, 并未应用于儿童癫痫综合征[29, 30]分析.

本文首先简述儿童癫痫综合征的基础理论, 详细描述儿童癫痫综合征的定义、发作类型和分类等临床知识, 为构建该领域的专家系统提供理论基础.然后, 系统阐述儿童癫痫综合征智能分析框架和各组成部分, 概括预处理、特征提取、决策器设计及专家系统的经典方法优缺点.重点对儿童癫痫综合征专家系统进行详细的调研和总结, 归纳为5类:特定波形检测系统、诊断分类系统、发作检测系统、发作预测系统和量化评估系统.最后, 结合临床实际问题、现有技术和该领域研究成果, 总结儿童癫痫综合征研究领域面临的挑战, 并提出前瞻性的解决方案.

1 儿童癫痫综合征理论基础
1.1 儿童癫痫综合征定义

癫痫发作是指大脑神经元细胞或神经环路出现异常过度、频繁且同步放电所造成的短暂的、一过性的临床表征, 发作起始和结束脑电图均呈现明显变化[31].

癫痫发作大致可根据发作诱因和发作类型两种方式进行分类.2017年ILAE提出的癫痫发作分类框架在临床上认可度较高[32], 将发作类型分为如下3类.

1)全面性发作.起源于双侧大脑皮质及皮质下结构构成的致痫网络中的某一点, 并快速波及整个网络.

2)局灶性发作.恒定地起源于一侧大脑半球内、呈局限性或更广泛分布的致痫网络, 并具有放电的优势传导途径, 可继发累及对侧半球.

3)未知癫痫发作[33].无法确认癫痫发作的开始时间.

癫痫是指具有持久、反复癫痫发作的慢性脑部疾病病状, 不是单一的疾病实体, 病因复杂、临床表征各异.按照癫痫的病因、临床变现和EEG变化进行分类后的疾病实体称为癫痫综合征[34].因此, 癫痫综合征诊断需要结合癫痫发作类型、病因学、共患病、发病年龄等因素, 整体流程如图1所示.

在2022年癫痫综合征指南中儿童癫痫综合征定义为一组具有特征性临床和脑电表现, 通常具有特定的病因(结构、遗传、代谢、免疫和感染性) 的癫痫疾病 [35, 36], 并强调EEG是诊断婴幼儿癫痫发作的“ 金标准” , 即可仅靠EEG 表现诊断儿童癫痫综合征[37].

图1 儿童癫痫综合征诊断框架Fig.1 Framework of childhood epileptic syndrome

1.2 儿童癫痫综合征分类

癫痫是儿童时期(0~18岁)常见的一种病因复杂、反复发作、阵发性、暂时性脑功能紊乱所致神经系统综合征[38].新生儿期直至青春期, 神经系统结构和功能都处于快速发育塑形过程, 因此, 不同年龄段的儿童癫痫综合征病因、发病机制、临床特征表现、预后等差异巨大, 诊断困难[39].癫痫综合征分类是临床和学术研究交流的基础工具, 同时反映当前对于病理的理解和临床认识, 对于癫痫的诊断、治疗和预后都具有重要作用.

儿童癫痫综合征具有较强的年龄依赖性, 在固定的年龄段表现为特定的癫痫综合征[40].临床研究者试图从发作类型、年龄、病因对儿童癫痫综合征进行详细分类, 具体如图2所示.然而癫痫病理尚不清晰, 存在较强的特异性, 导致稀有的癫痫综合征无法进行归类.

图2 儿童癫痫综合征的分类Fig.2 Classification of childhood epileptic syndrome

本文主要回顾典型儿童癫痫综合征相关的智能分析研究, 包括婴儿癫痫性痉挛综合征(Infantile Epileptic Spasms Syndrome, IESS)[41]、伴中央颞区棘波的儿童良性癫痫(Benign Childhood Epilepsy with Centro-Temporal Spikes, BECT)[42]、儿童失神(Child-hood Absence Epilepsy, CAE)[43]、睡眠癫痫持续状态(Encephalopathy Related to Status Epilepticus du-ring Slow Sleep, ESES)[44]等.

2 脑电信号数据收集及预处理

图3详细描述基于EEG儿童癫痫综合征智能分析系统的组成, 图中虚线表示存在闭环癫痫发作控制的可能性.

图3 儿童癫痫综合征智能分析框图Fig.3 Intelligent analysis framework of childhood epileptic syndrome

反映大脑和身体活动的生理信号在预处理后可提取特征或生物标志物.使用决策系统学习特征表征, 完成相应专家系统任务.

2.1 数据收集

癫痫发作的潜在机制是大脑皮层中神经元细胞出现过度或同步活动[45], 导致血流和电流出现异常变化.该变化可通过电场、磁场和光学进行测量, 如图4所示.

图4 儿童癫痫综合征大脑活动常用采集技术Fig.4 Commonly used acquisition techniques for brain activity in childhood epileptic syndrome

临床检测癫痫的设备大致可分为侵入式设备和非侵入式设备.侵入式设备需要在颅内或皮层内植入电极, 用于测量大脑内部的电活动.常用的技术有脑皮层电图(Intraoperative EEG Recording, ECoG)[46]、颅内脑电图(Intracranial EEG, iEEG)[47], 可记录非常高的空间和时间分辨率, 提供有关大脑活动的更精确信息.但其需要复杂、昂贵的设备, 同时ECoG采集还需要手术配合, 并不适用于儿童癫痫综合征的检测.非侵入式设备是在人体头皮放置多个电极进行检测[48], 主要包括脑磁图(Magnetoencephalo-graphy, MEG)、功能核磁共振(Functional Magnetic Resonance Imaging, fMRI)[49]、功能性近红外光谱(Functional Near-Infrared Spectroscopy, fNIRS)、EEG等.由于儿童癫痫综合征具有独特的临床脑电表征, 并且EEG容易使用、便于携带、成本低廉、时间分辨率高, 是诊断癫痫发作、确定发作和癫痫类型重要的辅助手段之一[50].因此, 基于EEG构建儿童癫痫综合征智能辅助分析系统具有较强的实用价值.

癫痫发作始于大脑内部, 传输至头皮表面受到较多阻碍, 如中间组织、头骨、头皮等[51].婴幼儿大脑发育并不完整, 头皮较薄, 因此EEG中记录的棘波、尖波和复合波等与痫样放电具有密切的关联性.为了获取更多信息, 可通过增加空间分辨率和采样率的方法提高EEG质量.空间分辨率是指头皮上电极的空间分布位置, 电极数量通常为21~64, 其中10~20个的国际标准导联系统被临床广泛使用[52].时间分辨率表示电极每秒采集信号的数量, 采样频率通常为128 Hz~1 000 Hz.随着快速震荡和高频震荡研究兴起, 开始应用2 000 Hz采样频率的EEG系统[53].

基于EEG的儿童癫痫综合征智能分析系统的研究多数采用医院非公开的匿名数据库.为了促进该领域的发展, 部分研究者同时致力于制定标准统一的公开数据集.

表1简要概述常用的癫痫数据集, 包括浙江大学医学院附属儿童医院(CHZU)神经内科与本文团队合作收集整理并创建的目前样本量最大的儿童癫痫及癫痫综合征脑电数据库.

表1 常用癫痫数据集 Table 1 Commonly used epilepsy dataset
2.2 数据预处理

脑电采集设备通过连接电极的差分放大器以捕捉皮层神经元的突触后电位的变化.过高的电压增益导致捕捉到大脑自发性活动之外, 也记录较多的生理伪迹与非生理伪迹[60], 如图5所示.

图5 儿童癫痫综合征患者的脑电信号Fig.5 EEG signals of children with epilepsy syndromes

生理伪迹通常来源受试者本身, 如眼动伪迹、心电伪迹、汗水、肌电伪迹等.非生理伪迹来源于多种因素, 如头皮与电极的接触情况、设备性能、工频噪声、环境因素等.生理伪迹的形态较典型, 容易辨别, 而非生理伪迹的形态众多、较难区分[61].这些伪迹严重干扰EEG分析.

通常有3种去除伪迹的策略[62, 63].

1)自动去除.可通过带通滤波器、切换参考导联等方法抑制EEG中的伪迹[64].例如:利用滤波器去除肌电伪迹; 参考导联切换双极导联, 避免眨眼伪迹干扰[65, 66].

2)手动去除.主要依靠视觉检测, 剔除存在伪迹的EEG段.为了获取连续、较干净的数据, 研究者常采用睡眠期的脑电数据进行相关分析[67].Nariai等[53]为了避免伪迹干扰高频振荡识别, 通常采用睡眠期脑电数据进行分析.

3)不去除伪迹.有时伪迹的存在对于分析起到正向指导的作用.Zheng等[68]在婴儿痉挛症的预测研究中并未去除伪迹, 其认为IESS发作时EEG 伴随的伪迹对于预测起到辅助分析作用.其次, 通过深度学习可直接从原始数据中提取有用特征, 因此在分析前可不进行伪迹去除[69].

儿童癫痫综合征发作时具有显著的肢体表征, 这些动作会使EEG产生相应的生理伪迹, 对于全面发作的分析具有辅助作用[70], 如强直-阵挛发作、阵挛发作、癫痫性痉挛、肌阵挛发作等.因此, 并不是所

有的生理伪迹都是干扰, 分析时需要根据研究目的和拟开展的后续分析采用合适的伪迹处理策略.单纯通过去除伪迹提高EEG信号的信噪比存在较大局限, 如何有效检测与区分生理伪迹和非生理伪迹, 并加以区分特定EEG自发性波形对儿童癫痫综合征临床研究者更具有实用性.

3 特征提取

如何将儿童癫痫综合征患者的EEG信号进行量化, 提取有效关键的信息是决定发作检测、预测、预后评估等专家系统性能的核心问题之一.按照时间、空间和频率3个维度可将特征划分成4类:时域、频域、时频域和空域, 此外也常使用深度模型特征.因此, 本文将EEG量化特征进行如图6所示的分类.

图6 EEG量化特征分类Fig.6 EEG quantitative feature categorization

1)时域特征.棘波尖波、棘慢波、复合波等EEG波形都是痫样放电典型波形[71].根据这些波形出现的脑区和频率、持续时间、演变过程, 临床医生做出相应病理诊断.时域特征可有效直观地量化EEG 中波形变化的情况[72], 因此被广泛用于儿童癫痫综合征量化分析.

根据时域特征的度量方式, 可分为有无量纲两类[73].有量纲特征计算快捷、便于理解, 如极值、峰峰值、平均值、方差、波形的半波持续时长、波形振幅、斜率等[74, 75], 但要求EEG数据保持一致性, 如脑电图机的灵敏度、导联方式、输入输出阻抗等均会影响有量纲特征的分布.无量纲特征表达抽象, 更注重描述特征统计分布状态, 如偏态度量特征分布对称程度, 若特征均值大于中位数为右偏态, 反之为左偏态[76, 77].常见的无量纲特征有峭度、偏度、波形因子、峰值因子、脉冲因子、裕度因子等.EEG信号的非平稳性、个体差异和外界干扰等都容易影响时域特征.

2)频域特征.EEG信号记录多种频段大脑的自发性电活动, 每个频段节律活动都具有特定的头皮分布和生物学意义[78], 因此频域特征分析是经典且常用的.现已证实, 儿童癫痫综合征在不同任务态的节律频率能量存在差异性变化[79].EEG 在时域中表征为信号幅值随时间的变化, 而在频域表征为信号功率随频率的变化分布.频谱估计可有效将EEG转换成频域信号, 并计算EEG功率、幅值等与频率的关联性.

频谱估计是分析EEG各类振荡和节律模式的基础且重要的工具, 常用方法有周期图、Welch 法、自回归模型和多窗口法.基于频谱分析, 使用不同方法如小波包分解[80]、经验模态分解[81]等, 提取差异性较大的频率段能量或节律信号的特征用于任务态分析.此外, 可将EEG 频谱作为随机过程, 进行二次特征提取[82].因此, 频谱估计中如窗口大小、类型、重叠率等参数选取对分析结果存在一定影响.

3)时频域特征.时域特征和频域特征均基于EEG信号近似平稳的假设, 然而EEG往往是高度非稳态的, 即EEG的时域和频域特性均会随时间变化.因此, 时域和频域特征并无法准确刻画EEG信号的时变性.时频分析利用时频功率分布估计计算非平稳的EEG信号在特定时间和特定频率的功率量值, 通用的时频分析方法包括:短时傅里叶变换(Short-Time Fourier Transform, STFT)[83]、连续小波变换(Continuous Wavelet Transform, CWT)[84].STFT和CWT均基于滑动窗口的思想, 计算每个窗口的周期图, 并堆叠形成时频谱.CWT因其特定时间分辨率和空间分辨率的基函数获得比STFT更精细的频谱图, 但计算复杂度更高[85].近年来, 新的时频分析技术得到快速发展, 如梅尔频率倒谱系数 (Mel Frequency Cepstral Coefficient, MFCC)[86, 87]和线性预测倒谱系数(Linear Predictive Cepstral Coefficient, LPCC)[88], 在描述非稳态信号方面较有效, 也被引入EEG研究中.时频图可反映任务态相关的细节, 常被直接或拼接成3D形式, 用于深度模型训练学习和测试应用.

4)空域特征.多导联EEG信号反映大脑复杂的神经活动, 具有较强的非线性动力学特性和脑区关联性.空域特征如非线性动力学特征[89]、导联关联性[90]、脑网络[91]、脑拓扑[92], 可衡量EEG随时间变化的空间特性.非线性动力学特征是描述和解释EEG信号在不同任务态下随时间变化的空间特性, 如复杂度[93]和熵值[94]用于衡量EEG信号空间变化复杂度和概率密度, 广泛应用于儿童癫痫综合征的发作检测[95].同时脑区关联性指标和脑网络特征用于评估神经活动中不同脑区之间的信息交互.基于关联性指标计算的原理, 可将关联性指标划分为4类:相干性指标[96]、相位同步指标[97]、广义同步性指标[98]和因果相关性指标[99].癫痫发作时的共同源问题是选择关联性指标时必需考虑的, 建议在预处理时增加空间滤波器, 避免来自头皮深层生理电干扰.结合图理论, 可将上述计算关联性指标转化成图并提取图论特征, 即脑网络特征.导联关联性和脑网络有效解释不同任务态下大脑不同分区之间的信息交流, 可辅助解释背后的机制, 用于儿童癫痫综合征的分析.此外, 将不同导联的特征映射到脑地形图上进行辅助分析也是常用手段.

5)深度模型特征.时域、频域、时频域和空域特征能有效表征和量化大部分不同任务态下EEG的变化趋势, 但容易受到数据质量、生理差异等多种因素的干扰.深度学习可直接从原始EEG信号中学习深层次表征信息[100], 对于儿童癫痫综合征专家系统决策研究具有巨大潜力.可基于深度迁移学习的思想, 将训练好的深度学习模型, 如长短期记忆网络(Long Short-Term Memory, LSTM)、卷积神经网络(Convolutional Neural Network, CNN)等[101], 在提取EEG特征后, 将迁移特征输入分类器模型进行任务态决策[102].也可通过自编码器的编码解码重构过程, 最优化损失, 挖掘深层次特征[103].这两种方式在构建决策模型时均可取得不错效果, 但深度模型特征缺乏可解释性, 在儿童癫痫综合征分析领域存在较大争议.综上所述, 需要根据构建专家系统的任务, 依赖特征优化方法, 灵活选择合适特征或特征组合, 而单纯依靠增加特征维度提升专家系统决策模型性能会带来“ 维度灾难” .

4 决策系统设计

如何从高维度EEG特征中区分和描述与儿童癫痫相关的大脑活动状态, 是构建相关专家系统的难点.传统方法依据统计分析寻找合适的生物标志物.机器学习、深度学习和模式识别技术的不断突破为解决专家系统的决策问题提供新思路[104, 105, 106].机器学习和深度学习分类器旨在通过训练获得相关先验知识以区分未知大脑状态.根据训练分类器时是否需要标签, 可将方法分为有监督学习和无监督学习.本节概述用于儿童癫痫专家系统的典型决策方法, 即统计分析、无监督学习和有监督学习.

统计学方法可对EEG特征进行描述和分析, 挖掘与儿童癫痫大脑活动相关的规律性、定性或定量的结论, 从而辅助专家系统决策.统计量是较常见的统计分析指标, 如平均值、方差、偏度和峰值等, 用于描述单一状态EEG特征的整体分布趋势.统计推断方法对比分析不同状态下EEG特征分布的差异性和关联性.特征整体分布趋势影响统计推断方法的选择.根据已知特征和未知特征或少知特征分布趋势, 可将推断方法分为参数检测和非参数检测[107, 108].在儿童癫痫综合征分析领域, 不同任务态EEG特征均衡且已知分布的情况下, 针对不同儿童癫痫综合征EEG差异性分析、儿童癫痫综合征认知任务分析等研究, 常使用单样本t检测、配对t检测、单因素方差分析、多因素方法分析、Pearman相关分析等参数检测方法[109, 110, 111].相反, 对于基于EEG的儿童癫痫综合征预测、睡眠分析等数据不均衡、特征分布未知的研究[112], 选择非参数检验, 如符号检验、置换检验、Friedman检验及Spearman相关性分析等[113, 114, 115].

统计分析方法选择并不是一成不变的, 其严重依赖已知和未知的数据分布.当样本量足够大时, 参数检验对于非正态分布依旧有效.因此正确选择统计检验方法对于分析不同任务态的EEG数据至关重要, 直接影响相关决策系统的性能.统计检验结果过度依赖p值概率, 导致在分析时存在较多的假阳性事件, 这是构建基于统计方法的儿童癫痫综合征专家系统所面临的一大挑战.

选择合适的EEG特征对于机器学习至关重要, 但跨越多域的高维度EEG特征集合存在较高噪音和信息冗余.无监督学习可在无标注的情况下挖掘数据隐藏模式和关联性, 获取相关任务的先验知识, 从而实现特征降维、聚类和分类等功能.主成分分析(Principal Component Analysis, PCA)[116]、独立成分分析(Independent Components Analysis, ICA)[117]、t分布随机邻域嵌入(t-Distributed Stochastic Neighbor Embedding, t-SNE)[118]、自编码器[119]均为EEG 分析中常用特征降维方法.ICA可将含噪音EEG信号分离成线性且互相统计独立的信号源组合, 广泛用于EEG伪迹去除.PCA和t-SNE通过线性运算和非线性运算将高维EEG特征映射至低维空间, 减少时间、频率、通道等维度信息, 筛选与任务态相关的重要信息.与上述方法不同, 自编码器旨在学习数据重建的映射关系, 通过梯度下降优化编码器和解码器的压缩和重构数据, 构建有效的潜在特征空间, 实现数据降维.无监督学习的聚类算法是基于特征集合构建的高维度空间中数据的距离、密度等相似性进行不同聚类, 经典方法有K-means[120]、DBSCAN(Density-Based Spatial Clustering of Applications with Noise)[121]、分层聚类、高斯混合模型[122]等.

由于儿童癫痫综合征临床定义并不精准, 存在EEG表征重叠现象, 因此无监督聚类算法的结果可用于挖掘病理相关的特征, 数据驱动定义亚型或争议较大的疾病实体的划分.综上所述, 无监督学习对于未知数据的适应性和鲁棒性更强, 在数据降维、特征优化、特征关联性挖掘等方面起到较突出的作用[123], 但其敏感度较低, 导致分类准确率与不同任务态EEG特征空间表征差异性密切相关.

相比无监督学习, 有监督学习算法利用标签与EEG特征关系, 实现不可分类高维数据到可分类低维数据的映射.根据类标签与学习特征之间的映射关系, 将有监督学习分为判别模型和生成模型.判别学习通过设置参数构建决策边界, 实现不同任务态的划分, 如线性判别分析(Linear Discriminant Analy- sis, LDA)[124]、支持向量机(Support Vector Machine, SVM)[125]、随机森林(Random Forests, RF)[126]等.生成模型通过生成给定类别的统计模型进行分类, 如朴素贝叶斯(Naive Bayes, NB)[127]、K近邻(k-Nea- rest Neighbor, KNN)[128]等.生成模型对于特征数据分布具有一定假设, 错误较高, 在儿童癫痫综合征专家系统中较多使用的为判别模型.但儿童癫痫综合征数据集存在类别不均衡、样本较少、个体差异性较大等问题, 导致传统的机器学习模型存在不能实现较精准的判别、模型泛化性有限等问题.

不断发展的深度学习技术逐步克服机器学习方法在儿童癫痫综合征专家系统决策方面的局限性.深度学习以人工神经网络模块为构架, 可直接从原始EEG信号中解码、学习和理解深度表征, 实现更精准、复杂的任务.本文在此主要介绍该领域中常用的深度学习基础模型:卷积神经网络(CNN)和循环神经网络(Recursive Neural Network, RNN).

CNN通过多层卷积层、池化层不断学习、提炼EEG局部特征, 如时间维度、空间维度等, 获得优秀的辨识性能.CNN广泛用于儿童癫痫综合征EEG原始信号解码研究, 并且根据其利用EEG维度信息的不同, 衍化为1D-CNN、2D-CNN和3D-CNN.Jana等[129]设计捕捉EEG时间维度特征的1D-CNN, 用于癫痫病灶定位.Avcu 等[130]提出SeizNet, 通过傅立叶变换将EEG信号转换为时频图, 利用卷积层学习时间维度和频域维度信息, 实现癫痫发作的自动检测.Feng 等[131]提出AR3D(3D Residual-Attention-Module-Based Deep Network), 探索脑电波的空间和时间频率特性以及通道的作用, 实现不同类别儿童癫痫综合征的识别.

相比CNN, RNN通过内部存储状态学习时间维度特征之间的关联性以实现相关任务, 其中LSTM、门控循环单元(Gated Recurrent Unit, GRU)使用较频繁[132, 133].决策系统的核心是针对不同任务态设计有效的EEG分类模型.

因此, 如何设计合理结构, 提升模型泛化性, 避免模型训练过拟合和梯度消失问题, 以及选用恰当的模型评价指标等是构建相关专家决策系统极具挑战性的研究方向.基于不同任务态所面临的实际问题, 迁移学习[134]、多视图学习[135]、集成学习[136]、主动学习[137]等方法为相关专家决策系统提供新的思路.

5 基于脑电信号的儿童癫痫综合征专家系统

针对儿童癫痫综合征的临床问题, 研究者们提出一系列基于EEG的专家系统, 本节总结近年来的相关研究, 划分为5类任务系统:特定波形检测系统、诊断分类系统、发作检测系统、发作预测系统、预后评估系统.本节简单阐述各类系统的作用以及智能模型在系统中的应用, 具体如表2所示.

表2 儿童癫痫综合征专家系统总结 Table 2 Summary of expert systems for childhood epileptic syndrome
5.1 特定波形检测系统

如何从儿童癫痫综合征患者的EEG信号中识别特定的病理波[156], 如棘波、慢波、棘复合波、高频震荡波等, 是癫痫综合征类别判定、预后效果评估的重要依据和基础.然而癫痫患儿的EEG生理伪迹与病理波十分相似, 会干扰临床医生EEG判读.

因此, 研究者提出痫样放电的伪迹检测技术和特定病理波检测技术.Cao等[138]设计无监督高斯混合模型, 有效检测儿童癫痫综合征患者EEG中的眨眼伪迹.在此基础上, Wang等[139, 140]通过多维特征优化, 进一步提升眨眼伪迹检测的精度.Wę sierski等[141]提出伪迹检测协议, 可实现对于眼动、肌肉和电极伪迹的自动标记.棘波是癫痫发作时EEG主要波形, 以BECT患儿中央和颞区局部放电最经典. Wang等[142]通过不同导联下的EEG特征结合机器学习模型设计多通道的BECT棘波检测算法, 辅助临床分析.Monsoor 等[143]采用深度学习检测器, 有效识别耐药性癫痫儿童的发作间期病理性高频震荡波.

智能模型虽在儿童癫痫EEG伪迹和痫样放电波形识别上有一定潜力, 但依旧存在如下局限性:1)这些算法均在理想数据集上训练而成, 但常规EEG和动态EEG数据上波形变化的差异还是巨大的[157, 158].2)不同癫痫综合征患者的波形特点存在较大的差异.因此, 如何将现有的高精度波形检测系统实时应用于不同类别的癫痫综合征临床分析中依旧存在较多挑战.

5.2 诊断分类系统

对儿童癫痫综合征准确的临床分类是神经科医生进行针对性治疗的基础[159], 对患者的预后管理等起着至关重要的作用.不同癫痫综合征的发作类型、持续时间、临床表征表现各异, 要求临床医生具有丰富的EEG阅读经验, 这给医疗资源匮乏的基层医院诊断带来困扰.基于数据驱动的智能诊断分类模型则较好地解决此类问题.Cao等[144]提取发作间期EEG数据的空间和时频特征, 提出TSA3-D(Two-Stream 3-D Attention Module-Based Deep Network), 实现BECT、CAE、IESS等5种癫痫综合征与对照组的区分.与上述研究不同, Misi u~nas等[145]以发作间期棘波为特征训练神经网络, 实现良性局灶性儿童癫痫与结构性局灶性癫痫的分类.

在儿童癫痫综合征诊断分类的相关研究中, EEG明显差异变化的常见病症类别分类已取得较高准确率[160].但是, 由于数据库的限制, 对于稀有和变异性癫痫综合征的研究依然较少.同时癫痫综合征亚型之间存在较明显的相似性, 模型分类准确率并不理想.

5.3 发作检测系统

儿童癫痫综合征发作检测是该领域研究的热点之一, 基于成人的癫痫发作检测技术较成熟.与成人癫痫不同, 儿童癫痫综合征发作类型更复杂, 自动检测技术旨在通过有效表征, 将发作期和非发作期EEG进行区分.Jiang等[146]通过深度学习自编码器学习和通道相关的脑电图特征融合以检测儿童局灶性癫痫的早期癫痫发作.Shoji等[147]以原始EEG信号为特征, 通过多维EEG网络, 对青少年和儿童的失神发作异常进行检测.深度学习技术的成熟同样也为解决儿童癫痫综合征个体性差异及复杂任务带来新的思路.Cao等[148]提出DSMN-ESS(Dual-Stream Multitask Network for Joint Childhood Epilepsy Syndrome Classification and Seizure Detection), 可完成癫痫类型分类和相关类型发作的检测, 具有较高的实用性.Cui等[149, 150]设计领域适应模型, 解决儿童局灶性癫痫发作检测中的个体差异问题.同时, 虽然发作检测公开数据集[161]解决深度学习训练所需的数据问题, 但发作期数据远少于发作间期数据, 因此生成式网络也被用于发作检测.

综上所述, 现有研究大部分基于理想的数据集, 存在较多局限性:1)实际临床中一种癫痫综合征往往伴随不同的发作类型[162, 163], 未知且易受到环境的干扰, 如何有效准确识别不同的发作类型并定位依旧是有待解决的问题.2)为了满足临床需求, 应使用实时信号进行检测和分类, 但现有的数据集大多包含选定的脑电信号片段, 并且已开发的算法之间缺乏标准化.

5.4 发作预测系统

癫痫是否可预测是一个极具争议的问题[164], 直至2013年, Cook等[165]基于iEEG开发的闭环发作检测系统, 证实局灶性癫痫是可以预测的.根据预测的时间尺度, 可将癫痫预测分为长期预测和短期预测[166].长期预测针对发作次数并不频繁的癫痫, 主要是根据与病理密切相关的波形预测癫痫下次发作的可能, 预测时间为下次发作前几周到几个小时不等.临床研究者将EEG中的高频振荡作为癫痫发作预测的生物标志物[151].相反, 短期预测针对具有频繁发作的癫痫, 用于在发作前的十几分钟对患者进行迷走神经刺激、惊厥药物注射或磁刺激等, 从而阻止癫痫发作.现有多数癫痫发作预测算法均基于CHB-MIT数据进行设计的, 而数据集使用的患者为成人局灶性发作数据[152], 理论上也可用于儿童局灶性癫痫.与此不同的是, 全局性发作癫痫发作前并无EEG明显的变化, 因而预测更困难.Zheng 等[167]通过Resnet18学习不同状态的脑关联性特征, 对IESS进行发作预测研究, 结果具有进一步提升的空间.

综上所述, 儿童癫痫综合征发作存在差异性变化, 如起始区域、起因等, 现有预测模型[153]均基于已收集的数据学习先验知识, 评价指标也是依据此计算出的, 对于实时癫痫发作是否能检测到而言仍是研究难点[168].

5.5 预后评估系统

儿童癫痫综合征存在独特的癫痫状态, 其中以BECT和睡眠癫痫持续状态患者的睡眠期癫痫状态(ESES)研究最为集中[169, 170].ESES特征是睡眠期强烈的痫样放电活动, 严重影响患儿的认知发育[171].如何检测并定量描述该状态有助于患者的诊断和预后.传统量化方法通过临床专家的经验标注患者睡眠期放电尖波并进行放电指数(Spike-Wave Index, SWI)统计[172], 标注存在主观化、过程繁琐、比较耗时等问题[173, 174].SWI[175]为非快速眼动睡眠中尖波和尖慢波异常的百分比.研究者采用形态学、模板匹配和模板聚类的方法实现自动SWI的统计, 但存在一定的局限性.Zhao等[154]使用形态学方法, 解决异常慢波和尖波开始位置检测精度较低的问题.Zhou等[155]使用遗传优化算法结合尖波特征, 进一步降低评估误差.

目前的大多数研究仅能识别尖峰波, 忽视在ESES演变过程中经常出现的重叠尖峰波和慢波, 依旧需要提升精度.上述的SWI 指数仅针对特殊癫痫综合征有效, 不同儿童癫痫综合征仍存在很多需要量化评估的状态, 如儿童癫痫综合征的预后用药效果、癫痫术后效果等.同时需要针对不同的量化任务构建统一的评价指标, 体现专家系统的性能.

6 结束语

基于EEG儿童癫痫综合征智能分析系统旨在通过数据驱动模型分析和解决儿童癫痫综合征临床面临的问题, 减轻患者及家属的负担, 辅助医生诊断.本文首先概括性介绍儿童癫痫综合征的定义, 再系统性阐述智能分析系统的流程及各组成涉及的具体方法, 最后详细调研和总结现有儿童癫痫综合征专家系统.新技术的不断发展, 如大数据、云平台、人工智能、高精度传感器等, 为儿童癫痫综合征的综合治理提供新的解决思路.结合上述现象、临床问题和现有的技术, 本文从如下几方面展望该领域的未来研究方向.

1)儿童作为特殊群体, 对于抗癫痫药物使用要求极为严格, 但目前相关研究依旧缺乏对于精确治疗的精准评估, 无法量化药物和手术对于癫痫的控制效果.因此, 需要寻找合适的生物标志物和构建精准量化模型, 避免儿童用药的不依从性, 提升癫痫治疗效果.

2)儿童时期是大脑发育的特殊时期, 神经元异常同步放电严重阻碍大脑正常发育, 导致患儿认知、睡眠、运动等障碍和精神共患病.相关研究主要集中在不同患者人群的差异性, 缺乏相关评估.因此, 有效量化癫痫共患病影响对儿童癫痫综合管理具有重要的意义.

3)儿童癫痫综合征个体性差异较大, 存在亚型和变异性, 同时, 不同的癫痫综合征病理生理特性存在部分相似性, 干扰临床诊断.因此, 仍需要进一步提升癫痫分类精准度, 实现个体差异化、异质性癫痫综合征的划分.

4)虽然研究已证明不同时期大脑差异性变化, 但癫痫综合征病理尚不清晰, 需进一步理解从正常大脑活动向癫痫不同阶段过渡的机制, 更好地定义并利用机制、模型、数据、设备和算法之间的协同作用, 更全面地研究病理, 提升不同儿童癫痫综合征专家系统的性能.

5)重症癫痫综合征的发作会带来灾难性的结果, 不同与常见的癫痫综合征, 重症癫痫综合征缺乏标准化、精准化的定义及划分, 存在争议.传统癫痫综合征研究方法大部分对其并不适用.但可解释的智能模型可挖掘其有效生物标志物, 为个性化、有效的治疗方案设计提供新思路.

6)随着基因学、神经化学、流行病学等技术的不断进步, 促使儿童癫痫综合征的定义和划分体系不断变化, 同时带来较大的分歧.因此, 如何通过智能分析模型挖掘不同组学之间的关联性, 以灵活、多维度的框架定义儿童癫痫综合征, 对癫痫药物研发、临床具有指导意义.

7)儿童癫痫综合征临床管理中依旧存在较多问题, 其中长期有效规范的治疗和管理对于患儿预后至关重要.如何通过智能模型结合大数据、云平台技术, 解决长程管理困难, 如随访、预防接种、青春期管理等, 进一步降低患者家庭的经济和精神负担, 是一个值得研究的方向.

本文责任编委 吕宝粮

Recommended by Associate Editor LÜ Baoliang

参考文献
[1] LEE S, YEOM J S. Revised Classification and Definition of Epilepsy Syndromes with Onset in Childhood by the International League Against Epilepsy Task Force. Epilia: Epilepsy and Community, 2023, 5(1): 28-39. [本文引用:1]
[2] ZUBERI S M, WIRRELL E, YOZAWITZ E, et al. ILAE Classification and Definition of Epilepsy Syndromes with Onset in Neonates and Infants: Position Statement by the ILAE Task Force on Nosology and Definitions. Epilepsia, 2022, 63(6): 1349-1397. [本文引用:1]
[3] HARRIS L, ANGUS-LEPPAN H. Epilepsy Update: Diagnosis, Cla-ssification and Management. Medicine, 2023, 51(8): 545-551. [本文引用:1]
[4] 张瑞, 李亚红, 李荣, 等. 癫痫患儿疾病家庭负担的现状及影响因素研究. 神经疾病与精神卫生, 2023, 23(7): 483-489.
(ZHANG R, LI Y H, LI R, et al. Current Status and Influencing Factors of Family Burden of Epilepsy Children. Journal of Neuroscience and Mental Health, 2023, 23(7): 483-489. ) [本文引用:1]
[5] The Lancet Neurology. Epilepsy Syndromes: An Essential Piece of the Puzzle. The Lancet Neurology, 2022, 21(7): 577. [本文引用:1]
[6] WIRRELL E C, RINEY K, SPECCHIO N, et al. How Have the Recent Updated Epilepsy Classifications Impacted on Diagnosis and Treatment? Expert Review of Neurotherapeutics, 2023, 23(11): 969-980. [本文引用:1]
[7] KOUTROUMANIDIS M, ARZIMANOGLOU A, CARABALLO R, et al. The Role of EEG in the Diagnosis and Classification of the Epilepsy Syndromes: A Tool for Clinical Practice by the ILAE Neurophysiology Task Force(Part 1). Epileptic Disorders, 2017, 19(3): 233-298. [本文引用:1]
[8] NOACHTAR S. RMI J. The Role of EEG in Epilepsy: A Critical Review. Epilepsy and Behavior, 2009, 15(1): 22-33. [本文引用:1]
[9] SMITH S J M. EEG in the Diagnosis, Classification, and Management of Patients with Epilepsy. Journal of Neurology, Neurosurgery and Psychiatry, 2005, 76(S2): ii2-ii7. [本文引用:1]
[10] FISHER R S, VAN EMDE BOAS W, BLUME W, et al. Epileptic Seizures and Epilepsy: Definitions Proposed by the International League Against Epilepsy(ILAE) and the International Bureau for Epilepsy(IBE). Epilepsia, 2005, 46(4): 470-472. [本文引用:1]
[11] PRESSLER R M, LAGAE L. Why We Urgently Need Improved Seizure and Epilepsy Therapies for Children and Neonates. Neuropharmacology, 2020, 170. DOI: DOI:10.1016/j.neuropharm.2019.107854. [本文引用:1]
[12] TRINKA E, KWAN P, LEE B, et al. Epilepsy in Asia: Disease Burden, Management Barriers, and Challenges. Epilepsia, 2019, 60(S1): 7-21. [本文引用:1]
[13] MILÀ B R, SINDHU K R, MYTINGER J R, et al. EEG Biomar-kers for the Diagnosis and Treatment of Infantile Spasms. Frontiers in Neurology, 2022, 13. DOI: DOI:10.3389/fneur.2022.960454. [本文引用:1]
[14] LHATOO S D, BERNASCONI N, BLUMCKE I, et al. Big Data in Epilepsy: Clinical and Research Considerations. Report from the Epilepsy Big Data Task Force of the International League Against Epilepsy. Epilepsia, 2020, 61(9): 1869-1883. [本文引用:1]
[15] 李莹, 欧阳楷. 自动检测儿童脑电中癫痫波的方法研究. 中国生物医学工程学报, 2005, 24(5): 541-545.
(LI Y, OUYANG K. Study of Automatic Detection of Epileptiform Waves in Children EEG. Chinese Journal of Biomedical Enginee-ring, 2005, 24(5): 541-545. ) [本文引用:1]
[16] SUERI C, GASPARINI S, BALESTRINI S, et al. Diagnostic Biomarkers of Epilepsy. Current Pharmaceutical Biotechnology, 2018, 19(6): 440-450. [本文引用:1]
[17] JACOBS J, LEVAN P, CHANDER R, et al. Interictal High-Frequency Oscillations(80-500 Hz) Are An Indicator of Seizure Onset Areas Independent of Spikes in the Human Epileptic Brain. Epilepsia, 2008, 49(11): 1893-1907. [本文引用:1]
[18] GAO Z K, DANG W D, WANG X M, et al. Complex Networks and Deep Learning for EEG Signal Analysis. Cognitive Neurodynamics, 2021, 15(3): 369-388. [本文引用:1]
[19] AHMED M I B, ALOTAIBI S, DASH S, et al. A Review on Machine Learning Approaches in Identification of Pediatric Epilepsy. SN Computer Science, 2022, 3(6). DOI: DOI:10.1007/s42979-022-01358-9. [本文引用:1]
[20] RASHEED K, QAYYUM A, QADIR J, et al. Machine Learning for Predicting Epileptic Seizures Using EEG Signals: A Review. IEEE Reviews in Biomedical Engineering, 2021, 14: 139-155. [本文引用:1]
[21] FAROOQ M S, ZULFIQAR A, RIAZ S. Epileptic Seizure Detection Using Machine Learning: Taxonomy, Opportunities, and Cha-llenges. Diagnostics, 2023, 13(6). DOI: DOI:10.3390/diagnostics13061058. [本文引用:1]
[22] HOSSEINI M P, HOSSEINI A, AHI K. A Review on Machine Lear-ning for EEG Signal Processing in Bioengineering. IEEE Reviews in Biomedical Engineering, 2020, 14: 204-218. [本文引用:1]
[23] NHU D, JANMOHAMED M, ANTONIC-BAKER A, et al. Deep Learning for Automated Epileptiform Discharge Detection from Scalp EEG: A Systematic Review. Journal of Neural Engineering, 2022, 19(5). DOI: DOI:10.1088/1741-2552/ac9644. [本文引用:1]
[24] SHOEIBI A, GHASSEMI N, KHODATARS M, et al. Applications of Epileptic Seizures Detection in Neuroimaging Modalities Using Deep Learning Techniques: Methods, Challenges, and Future Works[C/OL]. [2023-10-12]. https://arxiv.org/abs/2105.14278v1. [本文引用:1]
[25] AN S, KANG C, LEE H W. Artificial Intelligence and Computational Approaches for Epilepsy. Journal of Epilepsy Research, 2020, 10(1): 8-17. [本文引用:1]
[26] GOMBOLAY G Y, GOPALAN N, BERNASCONI A, et al. Review of Machine Learning and Artificial Intelligence(ML/AI) for the Pediatric Neurologist. Pediatric Neurology, 2023, 141: 42-51. [本文引用:1]
[27] ACHARYA U R, SREE S V, SWAPNA G, et al. Automated EEG Analysis of Epilepsy: A Review. Knowledge-Based Systems, 2013, 45: 147-165. [本文引用:1]
[28] 彭睿旻, 江军, 匡光涛, 等. 基于EEG的癫痫自动检测: 综述与展望. 自动化学报, 2022, 48(2): 335-350.
(PENG R M, JIANG J, KUANG G T, et al. EEG-Based Automatic Epilepsy Detection: Review and Outlook. Acta Automatica Sinica, 2022, 48(2): 335-350. ) [本文引用:1]
[29] KUHLMANN L, KAROLY P, FREESTONE D R, et al. Epilep-syecosystem. org: Crowd-Sourcing Reproducible Seizure Prediction with Long-Term Human Intracranial EEG. Brain, 2018, 141(9): 2619-2630. [本文引用:1]
[30] MEISEL C, LODDENKEMPER T. Seizure Prediction and Intervention. Neuropharmacology, 2020, 172. DOI: DOI:10.1016/j.neuropharm.2019.107898. [本文引用:1]
[31] GLAUSER T, BEN-MENACHEM E, BOURGEOIS B, et al. Updated ILAE Evidence Review of Antiepileptic Drug Efficacy and Effectiveness as Initial Monotherapy for Epileptic Seizures and Syndromes. Epilepsia, 2013, 54(3): 551-563. [本文引用:1]
[32] FISHER R S, CROSS J H, FRENCH J A, et al. Operational Cla-ssification of seizure Types by the International League Against Epilepsy: Position Paper of the ILAE Commission for Classification and Terminology. Epilepsia, 2017, 58(4): 522-530. [本文引用:1]
[33] ANDRADE-MACHADO R, CUARTAS V B, MUHAMMAD I K. Recognition of Interictal and Ictal Discharges on EEG. Focal vs Generalized Epilepsy. Epilepsy and Behavior, 2021, 117. DOI: DOI:10.1016/j.yebeh.2021.107830. [本文引用:1]
[34] PRESSLER R M, CILIO M R, MIZRAHI E M, et al. The ILAE Classification of Seizures and the Epilepsies: Modification for Seizures in the Neonate. Position Paper by the ILAE Task Force on Neonatal Seizures. Epilepsia, 2021, 62(3): 615-628. [本文引用:1]
[35] SPECCHIO N, WIRRELL E C, SCHEFFER I E, et al. International League Against Epilepsy Classification and Definition of Epilepsy Syndromes with Onset in Childhood: Position Paper by the ILAE Task Force on Nosology and Definitions. Epilepsia, 2022, 63(6): 1398-1442. [本文引用:1]
[36] RINEY K, BOGACZ A, SOMERVILLE E, et al. International Lea-gue Against Epilepsy Classification and Definition of Epilepsy Syndromes with Onset at a Variable Age: Position Statement by the ILAE Task Force on Nosology and Definitions. Epilepsia, 2022, 63(6): 1443-1474. [本文引用:1]
[37] TATUM W O, MANI J, JIN K, et al. Minimum Stand ards for Inpatient Long-Term Video-EEG Monitoring: A Clinical Practice Gui-deline of the International League Against Epilepsy and International Federation of Clinical Neurophysiology. Clinical Neurophysiology, 2022, 134: 111-128. [本文引用:1]
[38] FISHER R S, ACEVEDO C, ARZIMANOGLOU A, et al. ILAE Official Report: A Practical Clinical Definition of Epilepsy. Epilepsia, 2014, 55(4): 475-482. [本文引用:1]
[39] KATYAYAN A, DIAZ-MEDINA G. Epilepsy: Epileptic Syndromes and Treatment. Neurologic Clinics, 2021, 39(3): 779-795. [本文引用:1]
[40] GUERRINI R. Epilepsy in Children. The Lancet, 2006, 367(9509): 499-524. [本文引用:1]
[41] KNIGHT E M P, WYLLIE E. West Syndrome and the New Classification of Epilepsy. The Lancet Neurology, 2022, 21(8): 689. [本文引用:1]
[42] FEJERMAN N, CARABALLO R H, DALLA BERNARDINA B. Benign Childhood Epilepsy with Centrotemporal Spikes // IUPANC M L, ed. Benign Focal Epilepsies in Infancy, Childhood and Adolescence. Paris, France: John Libbey Eurotext, 2007: 77-113. [本文引用:1]
[43] GROSSO S, GALIMBERTI D, VEZZOSI P, et al. Childhood Absence Epilepsy: Evolution and Prognostic Factors. Epilepsia, 2005, 46(11): 1796-1801. [本文引用:1]
[44] RUBBOLI G, GARDELLA E, CANTALUPO G, et al. Encephalopathy Related to Status Epilepticus During Slow Sleep(ESES). Pathophysiological Insights and Nosological Considerations. Epilepsy and Behavior, 2023, 140. DOI: DOI:10.1016/j.yebeh.2023.109105. [本文引用:1]
[45] NAKASATO N, LEVESQUE M F, BARTH D S, et al. Comparisons of MEG, EEG, and ECoG Source Localization in Neocortical Partial Epilepsy in Humans. Electroencephalography and Clinical Neurophysiology, 1994, 91(3): 171-178. [本文引用:1]
[46] KOVAC S, VAKHARIA V N, SCOTT C, et al. Invasive Epilepsy Surgery Evaluation. Seizure, 2017, 44: 125-136. [本文引用:1]
[47] BARKLEY G L, BAUMGARTNER C. MEG and EEG in Epile-psy. Journal of Clinical Neurophysiology, 2003, 20(3): 163-178. [本文引用:1]
[48] DETRE J A. fMRI: Applications in Epilepsy. Epilepsia, 2004, 45(S4): 26-31. [本文引用:1]
[49] PILLAI J, SPERLING M R. Interictal EEG and the Diagnosis of Epilepsy. Epilepsia, 2006, 47(S1): 14-22. [本文引用:1]
[50] FLINK R, PEDERSEN B, GUEKHT A B, et al. Guidelines for the Use of EEG Methodology in the Diagnosis of Epilepsy: International League Against Epilepsy: Commission Report Commission on European Affairs: Subcommission on European Guidelines. Acta Neurologica Scand inavica, 2002, 106(1): 1-7. [本文引用:1]
[51] YOSHINAGA H, HATTORI J, OHTA H, et al. Utility of the Scalp-Recorded Ictal EEG in Childhood Epilepsy. Epilepsia, 2001, 42(6): 772-777. [本文引用:1]
[52] KLEM G H, IÜDERS H O. Ten-Twenty Electrode System of the International Federation. Electroencephalogr and Clinical Neurophysiology, 1999, 52: 3-6. [本文引用:1]
[53] NARIAI H, HUSSAIN S A, BERNARDO D, et al. Scalp EEG Interictal High Frequency Oscillations as an Objective Biomarker of Infantile Spasms. Clinical Neurophysiology, 2020, 131(11): 2527-2536. [本文引用:2]
[54] ANDRZEJAK R G, LEHNERTZ K, MORMANN F, et al. Indications of Nonlinear Deterministic and Finite-Dimensional Structures in Time Series of Brain Electrical Activity: Dependence on Recor-ding Region and Brain State. Physical Review E, 2001, 64(6). DOI: DOI:10.1103/PhysRevE.64.061907. [本文引用:1]
[55] GOLDBERGER A L, AMARAL L A, GLASS L, et al. PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals. Circulation, 2000, 101(23): e215-e220. [本文引用:1]
[56] SWAMI P, PANIGRAHI B, NARA S, et al. EEG Epilepsy Datasets[DB/OL]. [2023-10-12]. https://DOI.org/10.13140/RG.2.2.14280.32006. [本文引用:1]
[57] SHAH V, VON WELTIN E, LOPEZ S, et al. The Temple University Hospital Seizure Detection Corpus. Frontiers in Neuroinforma-tics, 2018, 12. DOI: DOI:10.3389/fninf.2018.00083. [本文引用:1]
[58] DETTI P, VATTI G, ZABALO MANRIQUE DE LARA G. EEG Synchronization Analysis for Seizure Prediction: A Study on Data of Noninvasive Recordings. Processes, 2020, 8(7). DOI: DOI:10.3390/pr8070846. [本文引用:1]
[59] STAFSTROM C E, CARMANT L. Seizures and Epilepsy: An Overview for Neuroscientists. Cold Spring Harbor Perspectives in Medicine, 2015, 5(6). DOI: DOI:10.1101/cshperspect.a022426. [本文引用:1]
[60] AMIN U, NASCIMENTO F A, KARAKIS I, et al. Normal Variants and Artifacts: Importance in EEG Interpretation. Epileptic Disorders, 2023, 25(5): 591-648. [本文引用:1]
[61] DALY I, NICOLAOU N, NASUTO S J, et al. Automated Artifact Removal from the Electroencephalogram: A Comparative Study. Clinical EEG and Neuroscience, 2013, 44(4): 291-306. [本文引用:1]
[62] ISLAM M K, RASTEGARNIA A, YANG Z. Methods for Artifact Detection and Removal from Scalp EEG: A Review. Neurophysio-logie Clinique/Clinical Neurophysiology, 2016, 46(4/5): 287-305. [本文引用:1]
[63] MUMTAZ W, RASHEED S, IRFAN A. Review of Challenges Asso-ciated with the EEG Artifact Removal Methods. Biomedical Signal Processing and Control, 2021, 68. DOI: DOI:10.1016/j.bspc.2021.102741. [本文引用:1]
[64] DE VOS M, DEBURCHGRAEVE W, CHERIAN P J, et al. Automated Artifact Removal as Preprocessing Refines Neonatal Seizure Detection. Clinical Neurophysiology, 2011, 122(12): 2345-2354. [本文引用:1]
[65] PIPER D, STRUNGARU R, WITTE H. Artefact Removal Approa-ch for Epileptic EEG Data. UPB Scientific Bulletin(Series C), 2015, 77(4): 213-222. [本文引用:1]
[66] JIANG X, BIAN G B, TIAN Z. Removal of Artifacts from EEG Signals: A Review. Sensors, 2019, 19(5). DOI: DOI:10.3390/s19050987. [本文引用:1]
[67] PELTOLA M E, LEITINGER M, HALFORD J J, et al. Routine and Sleep EEG: Minimum Recording Stand ards of the International Federation of Clinical Neurophysiology and the International League Against Epilepsy. Clinical Neurophysiology, 2023, 147: 108-120. [本文引用:1]
[68] ZHENG R Z, FENG Y M, WANG T L, et al. Scalp EEG Functional Connection and Brain Network in Infants with West Syndrome. Neural Networks, 2022, 153: 76-86. [本文引用:1]
[69] FANG Z B, HU D H, ZHENG R Z, et al. Multiple Artifact Detection Based on Adaptive Scalp Region Selection and Classifier Fusion. IEEE Sensors Journal, 2024. DOI: DOI:10.1109/JSEN.2024.3358911. [本文引用:1]
[70] ABD RAHMAN F, OTHMAN M F, SHAHARUDDIN N A. Analysis Methods of EEG Signals: A Review in EEG Application for Brain Disorder. Jurnal Teknologi(Sciences and Engineering), 2015, 72(2). DOI: DOI:10.11113/jt.v72.3886. [本文引用:1]
[71] NEVADO-HOLGADO A J, MARTEN F, RICHARDSON M P, et al. Characterising the Dynamics of EEG Waveforms as the Path through Parameter Space of a Neural Mass Model: Application to Epilepsy Seizure Evolution. NeuroImage, 2012, 59(3): 2374-2392. [本文引用:1]
[72] OUYANG C S, CHIANG C T, YANG R C, et al. Quantitative EEG Findings and Response to Treatment with Antiepileptic Medications in Children with Epilepsy. Brain and Development, 2018, 40(1): 26-35. [本文引用:1]
[73] ACHARYA U R, SREE S V, CHATTOPADHYAY S, et al. App-lication of Recurrence Quantification Analysis for the Automated Identification of Epileptic EEG Signals. International Journal of Neural Systems, 2011, 21(3): 199-211. [本文引用:1]
[74] VAN KLINK N, FRAUSCHER B, ZIJLMANS M, et al. Relationships between Interictal Epileptic Spikes and Ripples in Surface EEG. Clinical Neurophysiology, 2016, 127(1): 143-149. [本文引用:1]
[75] KHAN Y U, FAROOQ O, SHARMA P. Automatic Detection of Seizure Onset in Pediatric EEG. International Journal of Embedded Systems and Applications, 2012, 2(3): 81-89. [本文引用:1]
[76] XIANG J, MAUE E, FAN Y Y, et al. Kurtosis and Skewness of High-Frequency Brain Signals Are Altered in Pediatric Epilepsy. Brain Communications, 2020, 2(1). DOI: DOI:10.1093/braincomms/fcaa036. [本文引用:1]
[77] SIKDAR D, ROY R, MAHADEVAPPA M. Epilepsy and Seizure Characterisation by Multifractal Analysis of EEG Subband s. Biomedical Signal Processing and Control, 2018, 41: 264-270. [本文引用:1]
[78] QUINTERO-RINCON A, PEREYRA M, D'GIANO C, et al. Fast Statistical Model-Based Classification of Epileptic EEG Signals. Biocybernetics and Biomedical Engineering, 2018, 38(4): 877-889. [本文引用:1]
[79] HU W B, CAO J W, LAI X P, et al. Mean Amplitude Spectrum Based Epileptic State Classification for Seizure Prediction Using Convolutional Neural Networks. Journal of Ambient Intelligence and Humanized Computing, 2019, 14: 15485-15495. [本文引用:1]
[80] HU D H, CAO J W, LAI X P, et al. Epileptic State Classification Based on Intrinsic Mode Function and Wavelet Packet Decomposition // Proc of the 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society. Washington, USA: IEEE, 2019: 2382-2385. [本文引用:1]
[81] ALICKOVIC E, KEVRIC J, SUBASI A. Performance Evaluation of Empirical Mode Decomposition, Discrete Wavelet Transform, and Wavelet Packed Decomposition for Automated Epileptic Seizure Detection and Prediction. Biomedical Signal Processing and Control, 2018, 39: 94-102. [本文引用:1]
[82] LIU S, WANG J, LI S S, et al. Epileptic Seizure Detection and Prediction in EEGs Using Power Spectra Density Parameterization. IEEE Transactions on Neural Systems and Rehabilitation Enginee-ring, 2023, 31: 3884-3894. [本文引用:1]
[83] SAMIEE K, KOVÁCS P, GABBOUJ M. Epileptic Seizure Classification of EEG Time-Series Using Rational Discrete Short-Time Fourier Transform. IEEE Transactions on Biomedical Engineering, 2015, 62(2): 541-552. [本文引用:1]
[84] ADELI H, ZHOU Z Q, DADMEHR N. Analysis of EEG Records in an Epileptic Patient Using Wavelet Transform. Journal of Neuroscience Methods, 2003, 123(1): 69-87. [本文引用:1]
[85] KIYMIK M K, GULER I, DIZIBUYUK A, et al. Comparison of STFT and Wavelet Transform Methods in Determining Epileptic Seizure Activity in EEG Signals for Real-Time Application. Computers in Biology and Medicine, 2005, 35(7): 603-616. [本文引用:1]
[86] GUO S Y, ZHANG F. A SPCNN Model for Patient-Independent Prediction of Epilepsy Using MFCC Features // Proc of the 12th International Conference on Information Science and Technology. Washington, USA: IEEE, 2022: 68-73. [本文引用:1]
[87] DISSANAYAKE T, FERNANDO T, DENMAN S, et al. Deep Lear-ning for Patient-Independent Epileptic Seizure Prediction Using Scalp EEG Signals. IEEE Sensors Journal, 2021, 21(7): 9377-9388. [本文引用:1]
[88] LEE K, JEONG H, KIM S, et al. Real-Time Seizure Detection Using EEG: A Comprehensive Comparison of Recent Approaches under a Realistic Setting[C/OL]. [2023-10-12]. https://arxiv.org/pdf/2201.08780.pdf. [本文引用:1]
[89] RODRIGUEZ-BERMUDEZ G, GARCIA-LAENCINA P J. Analysis of EEG Signals Using Nonlinear Dynamics and Chaos: A Review. Applied Mathematics and Information Sciences, 2015, 9(5): 2309-2321. [本文引用:1]
[90] WILLIAMSON J R, BLISS D W, BROWNE D W, et al. Seizure Prediction Using EEG Spatiotemporal Correlation Structure. Epilepsy and Behavior, 2012, 25(2): 230-238. [本文引用:1]
[91] SARGOLZAEI S, CABRERIZO M, GORYAWALA M, et al. Scalp EEG Brain Functional Connectivity Networks in Pediatric Epilepsy. Computers in Biology and Medicine, 2015, 56: 158-166. [本文引用:1]
[92] VAN DIESSEN E, OTTE W M, STAM C J, et al. Electroencephalography Based Functional Networks in Newly Diagnosed Childhood Epilepsies. Clinical Neurophysiology, 2016, 127(6): 2325-2332. [本文引用:1]
[93] CUI X N, CAO J W, HU D H, et al. Regional Scalp EEGs Analysis and Classification on Typical Childhood Epilepsy Syndromes. IEEE Transactions on Cognitive and Developmental Systems, 2023, 15(2): 662-674. [本文引用:1]
[94] BRUZZO A A, GESIERICH B, SANTI M, et al. Permutation Entropy to Detect Vigilance Changes and Preictal States from Scalp EEG in Epileptic Patients. A Preliminary Study. Neurological Sciences, 2008, 29(1): 3-9. [本文引用:1]
[95] WU D P, LI J, DONG F, et al. Classification of Seizure Types Based on Multi-class Specific Band s Common Spatial Pattern and Penalized Ensemble Model. Biomedical Signal Processing and Control, 2023, 79. DOI: DOI:10.1016/j.bspc.2022.104118. [本文引用:1]
[96] HE F, BILLINGS S A, WEI H L, et al. A Nonlinear Causality Measure in the Frequency Domain: Nonlinear Partial Directed Coherence with Applications to EEG. Journal of Neuroscience Me-thods, 2014, 225: 71-80. [本文引用:1]
[97] YU H T, ZHU L, CAI L H, et al. Variation of Functional Brain Connectivity in Epileptic Seizures: An EEG Analysis with Cross-Frequency Phase Synchronization. Cognitive Neurodynamics, 2020, 14(1): 35-49. [本文引用:1]
[98] MA J Y, WANG Z Y, CHENG T Y, et al. A Prediction Model Integrating Synchronization Biomarkers and Clinical Features to Identify Responders to Vagus Nerve Stimulation Among Pediatric Patients with Drug-Resistant Epilepsy. CNS Neuroscience and Therapeutics, 2022, 28(11): 1838-1848. [本文引用:1]
[99] DAI X J, XU Q, HU J P, et al. BECTS Substate Classification by Granger Causality Density Based Support Vector Machine Model. Frontiers in Neurology, 2019, 10. DOI: DOI:10.3389/fneur.2019.01201. [本文引用:1]
[100] HU D H, CAO J W, LAI X P, et al. Epileptic State Classification by Fusing Hand -Crafted and Deep Learning EEG Features. IEEE Transactions on Circuits and Systems II(Express Briefs), 2021, 68(4): 1542-1546. [本文引用:1]
[101] SINGH K, MALHOTRA J. Two-Layer LSTM Network-Based Prediction of Epileptic Seizures Using EEG Spectral Features. Complex and Intelligent Systems, 2022, 8(3): 2405-2418. [本文引用:1]
[102] WANG Y M, CAO J W, WANG J Z, et al. Epileptic Signal Classification with Deep Transfer Learning Feature on Mean Amplitude Spectrum // Proc of the 41st Annual International Confe-rence of the IEEE Engineering in Medicine and Biology Society. Washington, USA: IEEE, 2019: 2392-2395. [本文引用:1]
[103] CAO J W, HU D H, WANG Y M, et al. Epileptic Classification with Deep-Transfer-Learning-Based Feature Fusion Algorithm. IEEE Transactions on Cognitive and Developmental Systems, 2022, 14(2): 684-695. [本文引用:1]
[104] JOSHI V, NANAVATI N. A Review of EEG Signal Analysis for Diagnosis of Neurological Disorders Using Machine Learning. Journal of Biomedical Photonics and Engineering, 2021, 7(4). DOI: DOI:10.18287/JBPE21.07.040201. [本文引用:1]
[105] SASIDHARAN A, DUTTA K K. Application of Machine-Lear-ning Techniques in Electroencephalography Signals // SAHU M, SINHA G R, eds. Brain and Behavior Computing. Totnes, UK: CRC Press, 2021: 61-84. [本文引用:1]
[106] SANEI S, CHAMBERS J A. EEG Signal Processing and Machine Learning. New York, USA: John Wiley and Sons, 2021. [本文引用:1]
[107] SAKKALIS V, CASSAR T, ZERVAKIS M, et al. Parametric and Nonparametric EEG Analysis for the Evaluation of EEG Activity in Young Children with Controlled Epilepsy. Computational Intelligence and Neuroscience, 2008. DOI: DOI:10.1155/2008/462593. [本文引用:1]
[108] JOSHI C N, PATRICK J. Eyelid Myoclonia with Absences: Routine EEG Is Sufficient to Make a Diagnosis. Seizure, 2007, 16(3): 254-260. [本文引用:1]
[109] SAKKALIS V, CASSAR T, ZERVAKIS M, et al. A Decision Su-pport Framework for the Discrimination of Children with Controlled Epilepsy Based on EEG Analysis. Journal of Neuroengineering and Rehabilitation, 2010, 7. DOI: DOI:10.1186/1743-0003-7-24. [本文引用:1]
[110] ADEBIMPE A, AARABI A, BOUREL-PONCHEL E, et al. EEG Resting State Functional Connectivity Analysis in Children with Benign Epilepsy with Centrotemporal Spikes. Frontiers in Neuroscience, 2016, 10. DOI: DOI:10.3389/fnins.2016.00143. [本文引用:1]
[111] ENGLOT D J, HAN S J, ROLSTON J D, et al. Epilepsy Surgery Failure in Children: A Quantitative and Qualitative Analysis. Journal of Neurosurgery: Pediatrics, 2014, 14(4): 386-395. [本文引用:1]
[112] CSERPAN D, ROSCH R, LO BIUNDO S P L, et al. Variation of Scalp EEG High Frequency Oscillation Rate with Sleep Stage and Time Spent in Sleep in Patients with Pediatric Epilepsy. Clinical Neurophysiology, 2022, 135: 117-125. [本文引用:1]
[113] GLABA P, LATKA M, KRAUSE M, et al. Changes in Interictal Pretreatment and Posttreatment EEG in Childhood Absence Epilepsy. Frontiers in Neuroscience, 2020, 14. DOI: DOI:10.3389/fnins.2020.00196. [本文引用:1]
[114] RIGHI M, BARCARO U, STARITA A, et al. Detection of Signs of Brain Dysfunction in Epileptic Children by Recognition of Transient Changes in the Correlation of Seizure-Free EEG. Brain Topography, 2008, 21(1): 43-51. [本文引用:1]
[115] THATCHER R W, WALKER R A, BIVER C J, et al. Quantitative EEG Normative Databases: Validation and Clinical Correlation. Journal of Neurotherapy, 2003, 7(3/4): 87-121. [本文引用:1]
[116] KEVRIC J, SUBASI A. The Effect of Multiscale PCA De-Noising in Epileptic Seizure Detection. Journal of Medical Systems, 2014, 38(10). DOI: DOI:10.1007/s10916-014-0131-0. [本文引用:1]
[117] SARDOUIE S H, ALBERA L, SHAMSOLLAHI M B, et al. An Efficient Jacobi-Like Deflationary ICA Algorithm: Application to EEG Denoising. IEEE Signal Processing Letters, 2015, 22(8): 1198-1202. [本文引用:1]
[118] KINNEY-LANG E, SPYROU L, EBIED A, et al. Tensor-Driven Extraction of Developmental Features from Varying Paediatric EEG Datasets. Journal of Neural Engineering, 2018, 15(4). DOI: DOI:10.1088/1741-2552/aac664. [本文引用:1]
[119] WEN T X, ZHANG Z N. Deep Convolution Neural Network and Autoencoders-Based Unsupervised Feature Learning of EEG Signals. IEEE Access, 2018, 6: 25399-25410. [本文引用:1]
[120] RAJAGURU H, PRABHAKAR S K. KNN Classifier and K-means Clustering for Robust Classification of Epilepsy from EEG Signals. Hamburg, Germany: Anchor Academic Publishing, 2017. [本文引用:1]
[121] LI G, JUNG J J. Seizure Detection from Multi-channel EEG Using Entropy-Based Dynamic Graph Embedding. Artificial Intelligence in Medicine, 2021, 122. DOI: DOI:10.1016/j.artmed.2021.102201. [本文引用:1]
[122] THOMAS E M, TEMKO A, LIGHTBODY G, et al. Gaussian Mix-ture Models for Classification of Neonatal Seizures Using EEG. Physiological Measurement, 2010, 31(7): 1047-1064. [本文引用:1]
[123] RIJO R, SILVA C, PEREIRA L, et al. Decision Support System to Diagnosis and Classification of Epilepsy in Children. Journal of Universal Computer Science, 2014, 20(6): 907-923. [本文引用:1]
[124] QARAQE M, ISMAIL M, SERPEDIN E, et al. Epileptic Seizure Onset Detection Based on EEG and ECG Data Fusion. Epilepsy and Behavior, 2016, 58: 48-60. [本文引用:1]
[125] LIU Y X, ZHOU W D, YUAN Q, et al. Automatic Seizure Detection Using Wavelet Transform and SVM in Long-Term Intracranial EEG. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2012, 20(6): 749-755. [本文引用:1]
[126] WANG Y X, CAO J W, LAI X P, et al. Epileptic State Classification for Seizure Prediction with Wavelet Packet Features and Rand om Forest // Proc of the Chinese Control And Decision Conference. Washington USA: IEEE, 2019: 3983-3987. [本文引用:1]
[127] YUAN S S, LIU J X, SHANG J L, et al. The Earth Mover's Distance and Bayesian Linear Discriminant Analysis for Epileptic Seizure Detection in Scalp EEG. Biomedical Engineering Letters, 2018, 8: 373-382. [本文引用:1]
[128] IBRAHIM S W, DJEMAL R, ALSUWAILEM A, et al. Electroencephalography(EEG)-Based Epileptic Seizure Prediction Using Entropy and K-nearest Neighbor(KNN). Communications in Science and Technology, 2017, 2(1): 6-10. [本文引用:1]
[129] JANA G C, SHARMA R, AGRAWAL A. A 1D-CNN-Spectrogram Based Approach for Seizure Detection from EEG Signal. Procedia Computer Science, 2020, 167: 403-412. [本文引用:1]
[130] AVCU M T, ZHANG Z, CHAN D W S. Seizure Detection Using Least EEG Channels by Deep Convolutional Neural Network // Proc of the IEEE International Conference on Acoustics, Speech and Signal Processing. Washington, USA: IEEE, 2019: 1120-1124. [本文引用:1]
[131] FENG Y M, ZHENG R Z, CUI X N, et al. 3D Residual-Attention-Deep-Network-Based Childhood Epilepsy Syndrome Classification. Knowledge-Based Systems, 2022, 248. DOI: DOI:10.1016/j.knosys.2022.108856V. [本文引用:1]
[132] XU Z D, WANG T L, CAO J W, et al. BECT Spike Detection Based on Novel EEG Sequence Features and LSTM Algorithms. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2021, 29: 1734-1743. [本文引用:1]
[133] TSIOURIS Κ Μ, PEZOULAS V C, ZERVAKIS M, et al. A Long Short-Term Memory Deep Learning Network for the Prediction of Epileptic Seizures Using EEG Signals. Computers in Biology and Medicine, 2018, 99: 24-37. [本文引用:1]
[134] RAGHU S, SRIRAAM N, TEMEL Y, et al. EEG Based Multi-class Seizure Type Classification Using Convolutional Neural Network and Transfer Learning. Neural Networks, 2020, 124: 202-212. [本文引用:1]
[135] TIAN X B, DENG Z H, YING W N, et al. Deep Multi-view Feature Learning for EEG-Based Epileptic Seizure Detection. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2019, 27(10): 1962-1972. [本文引用:1]
[136] JIANG T J, XU Z D, CAO J W, et al. BECT Spike Detection Based on Novel Multichannel Data Weighted Fusion Algorithm. IEEE Transactions on Circuits and Systems II(Express Briefs), 2022, 69(11): 4613-4617. [本文引用:1]
[137] MALDONADO R, HARABAGIU S M. Active Deep Learning for the Identification of Concepts and Relations in Electroencephalography Reports. Journal of Biomedical Informatics, 2019, 98. DOI: DOI:10.1016/j.jbi.2019.103265. [本文引用:1]
[138] CAO J W, CHEN L, HU D H, et al. Unsupervised Eye Blink Arti-fact Detection from EEG with Gaussian Mixture Model. IEEE Journal of Biomedical and Health Informatics, 2021, 25(8): 2895-2905. [本文引用:1]
[139] WANG M, WANG J H, CUI X N, et al. Multidimensional Feature Optimization Based Eye Blink Detection under Epileptiform Discharges. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2022, 30: 905-914. [本文引用:1]
[140] WANG M, CUI X N, WANG T L, et al. Eye Blink Artifact Detection Based on Multi-dimensional EEG Feature Fusion and Optimization. Biomedical Signal Processing and Control, 2023, 83. DOI: DOI:10.1016/j.bspc.2023.104657. [本文引用:1]
[141] WESIERSKI D, RUFUIE M R, MILCZAREK O, et al. Rating by Detection: An Artifact Detection Protocol for Rating EEG Quality with Average Event Duration. Journal of Neural Engineering, 2023, 20(2). DOI: DOI:10.1088/1741-2552/acbabe. [本文引用:1]
[142] WANG Z M, WU D P, DONG F, et al. A Novel Spike Detection Algorithm Based on Multi-channel of BECT EEG Signals. IEEE Transactions on Circuits and Systems II(Express Briefs), 2020, 67(12): 3592-3596. [本文引用:1]
[143] MONSOOR T, ZHANG Y P, DAIDA A, et al. Optimizing Detection and Deep Learning-Based Classification of Pathological High-Frequency Oscillations in Epilepsy. Clinical Neurophysiology, 2023, 154: 129-140. [本文引用:1]
[144] CAO J W, FENG Y M, ZHENG R Z, et al. Two-Stream Attention 3-D Deep Network-Based Childhood Epilepsy Syndrome Cla-ssification. IEEE Transactions on Instrumentation and Measurement, 2022, 72. DOI: DOI:10.1109/TIM.2022.3220287. [本文引用:1]
[145] MISIUNAS A V M, MEŠKAUSKAS T, SAMAITIENE R. Algorithm for Automatic EEG Classification According to the Epilepsy Type: Benign Focal Childhood Epilepsy and Structural Focal Epilepsy. Biomedical Signal Processing and Control, 2019, 48: 118-127. [本文引用:1]
[146] JIANG T J, ZHU J H, HU D H, et al. Early Seizure Detection in Childhood Focal Epilepsy with Electroencephalogram Feature Fusion on Deep Autoencoder Learning and Channel Correlations. Multidimensional Systems and Signal Processing, 2022, 33(4): 1273-1293. [本文引用:1]
[147] SHOJI T, YOSHIDA N, TANAKA T. Automated Detection of Abnormalities from an EEG Recording of Epilepsy Patients with a Compact Convolutional Neural Network. Biomedical Signal Processing and Control, 2021, 70. DOI: DOI:10.1016/j.bspc.2021.103013. [本文引用:1]
[148] CAO J W, CHEN Y H, ZHENG R Z, et al. DSMN-ESS: Dual-Stream Multi-task Network for Epilepsy Syndrome Classification and Seizure Detection. IEEE Transactions on Instrumentation and Measurement, 2023. DOI: DOI:10.1109/TIM.2023.3307724. [本文引用:1]
[149] CUI X N, WANG T L, LAI X P, et al. Cross-Subject Seizure Detection by Joint-Probability-Discrepancy-Based Domain Adaptation. IEEE Transactions on Instrumentation and Measurement, 2023, 72. DOI: DOI:10.1109/TIM.2023.3248101. [本文引用:1]
[150] CUI X N, CAO J W, LAI X P, et al. Cluster Embedding Joint-Probability-Discrepancy Transfer for Cross-Subject Seizure Detection. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2022, 31: 593-605. [本文引用:1]
[151] CHINAPPEN D M, XIAO G, JING J, et al. Spike Height Improves Prediction of Future Seizure Risk. Clinical Neurophysiology, 2023, 150: 49-55. [本文引用:1]
[152] KIRAL-KORNEK I, ROY S, NURSE E, et al. Epileptic Seizure Prediction Using Big Data and Deep Learning: Toward a Mobile System. EBioMedicine, 2018, 27: 103-111. [本文引用:1]
[153] 韩长明, 彭福来, 陈财, 等. 基于脑电信号的癫痫发作预测研究进展. 生物医学工程学杂志, 2021, 38(6): 1193-1202.
(HAN C M, PENG F L, CHEN C, et al. Research Progress of Epileptic Seizure Predictions Based on Electroencephalogram Signals. Journal of Biomedical Engineering, 2021, 38(6): 1193-1202. ) [本文引用:1]
[154] ZHAO X, WANG X H, CHEN C, et al. A Knowledge-Based Approach for Automatic Quantification of Epileptiform Activity in Children with Electrical Status Epilepticus During Sleep. Journal of Neural Engineering, 2020, 17(4). DOI: DOI:10.1088/1741-2552/aba6dd. [本文引用:1]
[155] ZHOU W, ZHAO X, WANG X H, et al. A Hybrid Expert System for Individualized Quantification of Electrical Status Epilepti-cus During Sleep Using Biogeography-Based Optimization. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2022, 30: 1920-1930. [本文引用:1]
[156] ABD EL-SAMIE F E, ALOTAIBY T N, KHALID M I, et al. A Review of EEG and MEG Epileptic Spike Detection Algorithms. IEEE Access, 2018, 6: 60673-60688. [本文引用:1]
[157] WANG J H, CAO J W, HU D H, et al. Eye Blink Artifact Detection with Novel Optimized Multi-dimensional Electroencephalogram Features. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2021, 29: 1494-1503. [本文引用:1]
[158] JIANG T J, WU D P, GAO F, et al. Improved Spike Detection Algorithm Based on Multi-template Matching and Feature Extraction. IEEE Transactions on Circuits and Systems II(Express Briefs), 2022, 69(1): 249-253. [本文引用:1]
[159] SHLOBIN N A, SINGH G, NEWTON C R, et al. Classifying Epilepsy Pragmatically: Past, Present, and Future. Journal of the Neurological Sciences, 2021, 427. DOI: DOI:10.1016/j.jns.2021.117515. [本文引用:1]
[160] ARIAS C R, DURÓN R M, DELGADO-ESCUETA A V. Identification of New Epilepsy Syndromes Using Machine Learning // Proc of the IEEE 39th Central America and Panama Convention. Washington, USA: IEEE, 2019. DOI: DOI:10.1109/CONCAPANXXXIX47272.2019.8977043. [本文引用:1]
[161] HANDA P, MATHUR M, GOEL N. EEG Datasets in Machine Learning Applications of Epilepsy Diagnosis and Seizure Detection. SN Computer Science, 2023, 4(5). DOI: DOI:10.1007/s42979-023-01958-z. [本文引用:1]
[162] MCCALLAN N, DAVIDSON S, NG K Y, et al. Epileptic Multi-seizure Type Classification Using Electroencephalogram Signals from the Temple University Hospital Seizure Corpus: A Review. Expert Systems with Applications, 2023, 234. DOI: DOI:10.1016/j.eswa.2023.121040. [本文引用:1]
[163] LARSEN P M, WÜSTENHAGEN S, TERNEY D, et al. Duration of Epileptic Seizure Types: A Data-Driven Approach. Epilepsia, 2023, 64(2): 469-478. [本文引用:1]
[164] IASEMIDIS L D. Epileptic Seizure Prediction and Control. IEEE Transactions on Biomedical Engineering, 2003, 50(5): 549-558. [本文引用:1]
[165] COOK M J, O'BRIEN T J, BERKOVIC S F, et al. Prediction of Seizure Likelihood with a Long-Term, Implanted Seizure Advisory System in Patients with Drug-Resistant Epilepsy: A First-in-Man Study. The Lancet Neurology, 2013, 12(6): 563-571. [本文引用:1]
[166] SCHLEGEL K, KLEYKO D, BRINKMANN B H, et al. Lessons from a Challenge on Forecasting Epileptic Seizures from Non-cerebral Signals. Nature Machine Intelligence, 2024, 6: 243-244. [本文引用:1]
[167] ZHENG R Z, CAO J W, FENG Y M, et al. Seizure Prediction Analysis of Infantile Spasms. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2022, 31: 366-376. [本文引用:1]
[168] NATU M, BACHUTE M, GITE S, et al. Review on Epileptic Seizure Prediction: Machine Learning and Deep Learning Approaches. Computational and Mathematical Methods in Medicine, 2022. DOI: DOI:10.1155/2022/7751263. [本文引用:1]
[169] CANTALUPO G, PAVLIDIS E, BENICZKY S, et al. Quantitative EEG Analysis in Encephalopathy Related to Status Epilepti-cus during Slow Sleep. Epileptic Disorders, 2019, 21(S1): S31-S40. [本文引用:1]
[170] BALARAM N, JOSE J, GAFOOR A V, et al. Classification of Electrical Status Epilepticus in Sleep Based on EEG Patterns and Spatiotemporal Mapping of Spikes. Epileptic Disorders, 2022, 24(6): 1060-1072. [本文引用:1]
[171] KRAMER U, SAGI L, GOLDBERG-STERN H, et al. Clinical Spectrum and Medical Treatment of Children with Electrical Status Epilepticus in Sleep(ESES). Epilepsia, 2009, 50(6): 1517-1524. [本文引用:1]
[172] CHAVAKULA V, FERNÁNDEZ I S, PETERS J M, et al. Automated Quantification of Spikes. Epilepsy and Behavior, 2013, 26(2): 143-152. [本文引用:1]
[173] JOSHI C N, CHAPMAN K E, BEAR J J, et al. Semiautomated Spike Detection Software Persyst 13 Is Noninferior to Human Readers When Calculating the Spike-Wave Index in Electrical Status Epilepticus in Sleep. Journal of Clinical Neurophysiology, 2018, 35(5): 370-374. [本文引用:1]
[174] NONCLERCQ A, FOULON M, VERHEULPEN D, et al. Spike Detection Algorithm Automatically Adapted to Individual Patients Applied to Spike and Wave Percentage Quantification. Neurophysiologie Clinique/Clinical Neurophysiology, 2009, 39(2): 123-131. [本文引用:1]
[175] WIWATTANADITTAKUL N, DEPOSITARIO-CABACAR D, ZE-LLEKE T G. Electrical Status Epilepticus in Sleep(ESES)-Treatment Pattern and EEG Outcome in Children with Very High Spike-Wave Index. Epilepsy and Behavior, 2020, 105. DOI: DOI:10.1016/j.yebeh.2020.106965. [本文引用:1]