[1]郭舒雯,郝锋丽,喻良,等.基于EEG信号的卷积神经网络在癫痫检测中的应用价值研究[J].卒中与神经疾病杂志,2023,30(02):193-197.[doi:10.3969/j.issn.1007-0478.2023.02.013]
点击复制

基于EEG信号的卷积神经网络在癫痫检测中的应用价值研究()
分享到:

《卒中与神经疾病》杂志[ISSN:1007-0478/CN:42-1402/R]

卷:
第30卷
期数:
2023年02期
页码:
193-197
栏目:
论著
出版日期:
2023-04-20

文章信息/Info

文章编号:
1007-0478(2023)02-0193-05
作者:
郭舒雯郝锋丽喻良赵雄飞
620010 四川省眉山市心脑血管病医院神经内科癫痫门诊(郭舒雯),神经内科(赵雄飞); 延安大学咸阳医院神经内科癫痫中心[郝锋丽(通信作者)]; 四川省人民医院神经内科综合癫痫中心(喻良)
关键词:
脑电图 卷积神经网络 癫痫 功率谱密度能量图
分类号:
R742.1
DOI:
10.3969/j.issn.1007-0478.2023.02.013
文献标志码:
A
摘要:
目的 探讨基于脑电图(Electroencephalogram,EEG)信号的卷积神经网络在癫痫检测中的应用价值。方法 本研究使用了来自癫痫EEG信号数据CHB-MIT数据库中的8例患者的EEG信号; EEG信号分为3类:发作间期、发作前期(发作前持续时间至10 min)和癫痫发作期状态; 开发了一种基于迁移学习和功率谱密度能量图的深度卷积神经网络(Deep convolutional neural network,DCNN)的癫痫EEG信号分类方法(EEG signal classification method,EEGC),并对癫痫状态进行分类; 将在线硬示例挖掘(Online hard example mining,OHEM)损失函数集成到EEGC方法中以获得较高的分类准确率。结果 本研究提出的EEGC方法对癫痫状态分类的准确度较高,但发作前期没有像其他两种状态那样准确分类; 当将OHEM损失函数集成到EEGC方法中时发作前期的分类准确度提高了3%,并且它对3种癫痫状态(发作间期、发作前期和癫痫发作)的分类具有很高的敏感度(97.8%、93.6%和95.8%)和特异度(99.2%、97.1%和99.3%)。结论 本研究提出的EEGC方法具有较高的癫痫状态分类准确率,可辅助临床医生了解患者癫痫状态的类别,从而有效地预防和治疗癫痫。

参考文献/References:

[1] 顾丽,余年,狄晴.癫痫发作后状态及其临床意义[J].中华神经科杂志,2022,55(1):65-70.
[2] Jamy R, Kaur M, Pizarro D, et al. Practice trends and the outcome of neuromodulation therapies in epilepsy: a single-center study[J]. Epilepsia Open, 2019, 4(3): 493-497.
[3] 邓劼,方方,王晓慧,等.早发性癫痫脑病伴暴发-抑制脑电图患儿30例的病因与临床特点分析[J].中华实用儿科临床杂志,2020,35(24):1853-1857.
[4] Cao J, Zhu J, Hu W, et al. Epileptic signal classification with deep EEG features by stacked CNNs[J]. IEEE Trans Cogn Dev Syst, 2019, 12(4): 709-722.
[5] Ozcan AR, Erturk S. Seizure prediction in scalp EEG using 3D convolutional neural networks with an image-based approach[J]. IEEE Trans Neural Syst Rehabil Eng, 2019, 27(11): 2284-2293.
[6] Hu WB, Cao JW, Lai XP, et al. Mean amplitude spectrum based epileptic state classification for seizure prediction using convolutional neural networks[J]. J Ambient Intell Humaniz Comput, 2019, 32(56): 1-11.
[7] Peng P, Xie L, Wei H. A deep fourier neural network for seizure prediction using convolutional neural network and ratios of spectral power[J]. Int J Neural Syst, 2021, 31(8): 2150022.
[8] Mansour RF, Aljehane NO. An optimal segmentation with deep learning based inception network model for intracranial hemorrhage diagnosis[J]. Neural Computing and Applications, 2021, 33(20): 13831-13843.
[9] Zhang W, Li X, Ding Q. Deep residual learning-based fault diagnosis method for rotating machinery[J]. ISA Trans, 2019, 95(1): 295-305.
[10] Li J, Lu K, Huang Z, et al. Transfer independently together: a generalized framework for domain adaptation[J]. IEEE Trans Cybern, 2019, 49(6): 2144-2155.
[11] Deng J, Guo W, Zhao Y, et al. Identification of diatom taxonomy by a combination of region-based full convolutional network, online hard example mining, and shape priors of diatoms[J]. Int J Legal Med, 2021, 135(6): 2519-2530.
[12] Liu X, Wu S, Daif A, et al. Clinical implications of scalp ictal EEG pattern in patients with temporal lobe epilepsy[J]. Clin Neurophysiol, 2019, 130(9): 1604-1610.
[13] Jy Y, Jetté N, Kwon CS, et al. Brief potentially ictal rhythmic discharges and paroxysmal fast activity as scalp electroencephalographic biomarkers of seizure activity and seizure onset zone[J]. Epilepsia, 2021, 62(3): 742-751.
[14] Ananda A, Ngan KH, Karaba(ˇoverg)C, et al. Classification and visualisation of normal and abnormal radiographs; a comparison between eleven convolutional neural network architectures[J]. Sensors(Basel), 2021, 21(16): 5381.
[15] Lu Q, Li S, Yang T, et al. An adaptive hybrid X deep FM based deep Interest network model for click-through rate prediction system[J]. PeerJ Comput Sci, 2021, 7(1): e716.

备注/Memo

备注/Memo:
基金项目:四川省卫生健康科研基金项目(20210826)
更新日期/Last Update: 2023-04-20