[1]翟凤文,高康霞,金静,等.多尺度编码器结合卷积Transformer对癫痫的分类[J].中国医学物理学杂志,2025,42(12):1621-1631.[doi:DOI:10.3969/j.issn.1005-202X.2025.12.012]
　ZHAI Fengwen,GAO Kangxia,JIN Jing,et al.Integration of multi-scale encoder and convolutional Transformer for epileptic classification[J].Chinese Journal of Medical Physics,2025,42(12):1621-1631.[doi:DOI:10.3969/j.issn.1005-202X.2025.12.012]
点击复制

多尺度编码器结合卷积Transformer对癫痫的分类()

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
42
期数:
2025年第12期
页码:
1621-1631
栏目:
医学信号处理与医学仪器
出版日期:
2025-12-29

文章信息/Info

Title:
Integration of multi-scale encoder and convolutional Transformer for epileptic classification
文章编号:
1005-202X（2025）12-1621-11
作者:
翟凤文1; 高康霞1; 金静1; 魏帮财2
1.兰州交通大学电子与信息工程学院， 甘肃 兰州 730070； 2.中电万维信息技术有限责任公司， 甘肃 兰州 730000
Author(s):
ZHAI Fengwen1;  GAO Kangxia1;  JIN Jing1;  WEI Bangcai2
1. School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China 2. China Telecom Wanwei Information Technology Co., Ltd., Lanzhou 730000, China
关键词:
脑电图信号; 癫痫; 深度学习; MTSDCformer; IP单元
Keywords:
Keywords: electroencephalogram signal epilepsy deep learning MTSDCformer information purification unit
分类号:
R318；TP391
DOI:
DOI:10.3969/j.issn.1005-202X.2025.12.012
文献标志码:
A
摘要:
针对多数结合Transformer的模型不能充分学习脑电信号中丰富的时空信息这一局限性，设计多尺度时空卷积编码器模块和密集卷积Transformer模块，并将这两种模块相结合提出MTSDCformer模型以实现对癫痫患者脑电图（EEG）信号的预测与分类。首先，该模型使用多尺度时空特征编码器对分段的EEG信号提取时间维度以及通道维度的特征信息；其次，将学习局部信息的卷积分支集成到Transformer编码器上形成密集卷积Transformer（DCT）模块，通过多层的DCT模块来学习EEG信号的全局依赖关系以及更细粒度的时间特征；然后，进一步通过信息纯化单元对EEG信号中的频率信息进行编码以提取中间层的详细特征；最后，根据提取的特征对输入的EEG信号进行分类，判断患者是否患有癫痫。实验部分在UPenn & Mayo诊所挑战数据集和麻省理工学院的波士顿儿童医院（CHB-MIT）公共数据集上讨论模型的有效性，实验结果表明MTSDCformer模型相较于LSTM、Vision Transformer、Compact Convolution Transformer、EEG-Conformer、EEGNet、R-Transformer和ARNN等基线方法表现出更好的性能。
Abstract:
To address the limitation that most Transformer-based models fail to fully learn the rich spatiotemporal information in electroencephalogram (EEG) signals, this study designs a multi-scale spatiotemporal convolutional encoder module and a dense convolutional Transformer (DCT) module, and integrates these two modules to propose MTSDCformer which enables the prediction and classification of EEG signals in epilepsy patients. The model uses a multi-scale spatiotemporal feature encoder to extract feature information from the segmented EEG signals across both the temporal and channel dimensions, and integrates convolutional branches dedicated to local information learning into the Transformer encoder, thereby forming the DCT module. Through the utilization of multi-layers of the DCT module, the model captures the global dependencies and finer-grained temporal features in EEG signals. Furthermore, an information purification unit is employed to encode the frequency information inherent in the EEG signals, facilitating the extraction of detailed features from the intermediate layers. Finally, based on the extracted features, the model classifies the input EEG signals to determine whether the patient has epilepsy. The effectiveness of the model is evaluated on the UPenn & Mayo Clinic Challenge dataset and the public Childrens Hospital Boston-Massachusetts Institute of Technology (CHB-MIT) dataset. Experimental results demonstrate that the MTSDCformer model outperforms baseline methods including LSTM, Vision Transformer, Compact Convolution Transformer, EEG-Conformer, EEGNet, R-Transformer, and ARNN.

相似文献/References:

[1]李凯,姜永涛,于海波,等.癫痫治疗效果的复杂度评估分析[J].中国医学物理学杂志,2013,30(03):4169.[doi:10.3969/j.issn.1005-202X.2013.03.019]
[2]崔招焕,等.大鼠癫痫脑电信号采集[J].中国医学物理学杂志,2016,33(2):118.[doi:10.3969/j.issn.1005-202X.2016.02.003]
　[J].Chinese Journal of Medical Physics,2016,33(12):118.[doi:10.3969/j.issn.1005-202X.2016.02.003]
[3]李春,黄波,唐海宁,等. 癫痫伴发焦虑抑郁患者脑神经递质活动的脑电超慢涨落图表现分析[J].中国医学物理学杂志,2018,35(3):369.[doi:DOI:10.3969/j.issn.1005-202X.2018.03.022]
　LI Chun,HUANG Bo,TANG Haining,et al. Characteristics of intracerebral neurotransmitter activity on encephalofluctuograph in epileptic patients with anxiety and depression[J].Chinese Journal of Medical Physics,2018,35(12):369.[doi:DOI:10.3969/j.issn.1005-202X.2018.03.022]
[4]张宣,刘安康,张培玲.基于嵌入式AI的癫痫发作监测系统实现[J].中国医学物理学杂志,2022,39(9):1151.[doi:DOI:10.3969/j.issn.1005-202X.2022.09.016]
　ZHANG Xuan,LIU Ankang,ZHANG Peiling.Implementation of seizure monitoring system based on embedded AI[J].Chinese Journal of Medical Physics,2022,39(12):1151.[doi:DOI:10.3969/j.issn.1005-202X.2022.09.016]
[5]汤秣雄,李效龙.一种无线闭环迷走神经刺激器及系统[J].中国医学物理学杂志,2022,39(12):1530.[doi:DOI:10.3969/j.issn.1005-202X.2022.12.012]
　TANG Moxiong,LI Xiaolong.Wireless closed-loop vagus nerve stimulator and its system[J].Chinese Journal of Medical Physics,2022,39(12):1530.[doi:DOI:10.3969/j.issn.1005-202X.2022.12.012]
[6]陈明武,郑诗豪,杨波,等.CTP灌注参数联合脑电图对aSAH继发癫痫及预后的价值分析[J].中国医学物理学杂志,2023,40(2):182.[doi:DOI:10.3969/j.issn.1005-202X.2023.02.009]
　CHEN Mingwu,ZHENG Shihao,YANG Bo,et al.Predictive value of CTP parameters combined with electroencephalogram for epilepsy secondary to aSAH and the prognosis[J].Chinese Journal of Medical Physics,2023,40(12):182.[doi:DOI:10.3969/j.issn.1005-202X.2023.02.009]
[7]胡保华,朱宗俊,穆景颂,等.基于C-FFuzzyEn的神经电生理信号同步性分析[J].中国医学物理学杂志,2023,40(5):589.[doi:DOI:10.3969/j.issn.1005-202X.2023.05.011]
　HU Baohua,ZHU Zongjun,MU Jingsong,et al.Neural electrophysiological signal synchronization analysis using C-FFuzzyEn[J].Chinese Journal of Medical Physics,2023,40(12):589.[doi:DOI:10.3969/j.issn.1005-202X.2023.05.011]
[8]于永强,刘健,孙仁诚,等.基于多特征多关系图卷积神经网络的癫痫脑电分类[J].中国医学物理学杂志,2023,40(5):595.[doi:DOI:10.3969/j.issn.1005-202X.2023.05.012]
　YU Yongqiang,LIU Jian,SUN Rencheng,et al.Classification of epileptic electroencephalogram signal using graph convolutional neural network with multiple features and multiple relations[J].Chinese Journal of Medical Physics,2023,40(12):595.[doi:DOI:10.3969/j.issn.1005-202X.2023.05.012]
[9]高晨洋,吴凯,李文豪,等.EEG-fNIRS技术在神经精神疾病研究中的应用进展[J].中国医学物理学杂志,2024,41(3):348.[doi:DOI:10.3969/j.issn.1005-202X.2024.03.013]
　GAO Chenyang,WU Kai,,et al.Advances in application of EEG-fNIRS technology in researches on neuropsychiatric disorders[J].Chinese Journal of Medical Physics,2024,41(12):348.[doi:DOI:10.3969/j.issn.1005-202X.2024.03.013]
[10]张喜珍,张晓莉,吕洋,等.基于2D-CNN和Cox-Stuart早停机制的癫痫预测模型[J].中国医学物理学杂志,2025,42(1):82.[doi:DOI:10.3969/j.issn.1005-202X.2025.01.012]
　ZHANG Xizhen,ZHANG Xiaoli,et al.Epilepsy prediction model based on 2D-CNN and Cox-Stuart early stopping mechanism[J].Chinese Journal of Medical Physics,2025,42(12):82.[doi:DOI:10.3969/j.issn.1005-202X.2025.01.012]

备注/Memo

备注/Memo:
【收稿日期】2025-03-23 【基金项目】甘肃省高校创新基金（2022A-047）；甘肃省科技计划项目重点研发计划-工业类（23YFGA0047） 【作者简介】翟凤文，副教授，研究方向：数字图像处理，E-mail: zhaifw@mail.lzjtu.cn 【通信作者】高康霞，硕士研究生，研究方向：脑电信号分析，E-mail: 1664663203@qq.com
更新日期/Last Update: 2025-12-29