基于连续小波变换和符号传递熵的脑功能网络构建方法

doi:10.12263/DZXB.20210298

电子学报 ›› 2022, Vol. 50 ›› Issue (7): 1600-1608.DOI: 10.12263/DZXB.20210298

基于连续小波变换和符号传递熵的脑功能网络构建方法

李明爱¹^,²^,³(), 张圆圆¹

^1.北京工业大学信息学部, 北京 100124
^2.计算智能与智能系统北京市重点实验室, 北京 100124
^3.教育部数字社区工程研究中心, 北京 100124

收稿日期:2021-03-02 修回日期:2022-01-24 出版日期:2022-07-25
- 通讯作者:
- 李明爱
- 作者简介:
- 李明爱女，1966年7月出生，河南鹤壁人.2006年于北京工业大学获得博士学位，现为北京工业大学教授、博士生导师，主要研究方向为脑机接口技术、人工智能与智能康复.E-mail: limingai@bjut.edu.cn
  张圆圆男，1996 年12 月出生，安徽安庆人. 现为北京工业大学信息学部硕士研究生，主要研究方向为脑机接口技术、信号处理与模式识别.E-mail: zyybjutemail@emails.bjut.edu.cn
- 基金资助:
- 国家自然科学基金 (62173010)

A Brain Functional Network Based on Continuous Wavelet Transform and Symbolic Transfer Entropy

LI Ming-ai¹^,²^,³(), ZHANG Yuan-yuan¹

^1.Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China
^2.Beijing Key Laboratory of Computational Intelligence and Intelligent System, Beijing 100124, China
^3.Engineering Research Center of Digital Community（Ministry of Education）, Beijing 100124, China

Received:2021-03-02 Revised:2022-01-24 Online:2022-07-25 Published:2022-07-30
- Corresponding author:
- LI Ming-ai
- Supported by:
- National Natural Science Foundation of China (62173010)

摘要/Abstract

摘要：

为有效利用运动想象脑电信号(Motor Imagery Electroencephalogram，MI-EEG)的频域信息并精确反映脑电极之间的非线性因果交互作用，本文提出一种基于连续小波变换和符号传递熵的脑功能网络构建方法.首先，对每导MI-EEG进行连续小波变换，求得其时-频-能量矩阵；然后，将与运动想象密切相关的频带内各频率所对应的时间-能量序列依次拼接，得到各导联的一维时频能量序列；最后，基于任意两电极时频能量序列间的符号传递熵计算连接矩阵，构建脑功能网络.实验结果表明，以电极时频能量序列间的符号传递熵构建的脑功能网络，能够有效反映MI-EEG的时频特征和非线性特征信息传递，相比于传统脑网络构建方法，更有利于增强不同运动想象任务的可分性.

关键词: 脑-机接口, 运动想象脑电信号, 连续小波变换, 符号传递熵, 脑功能网络

Abstract:

In order to utilize the frequency domain information of motor imagery electroencephalogram(MI-EEG) signals to effectively and accurately reflect the nonlinear causal interaction between different EEG electrodes, this paper presents a brain functional network based on continuous wavelet transform and symbolic transfer entropy. Firstly, the continuous wavelet transform is applied to each MI-EEG signal to compute the time-frequency-energy matrix. Then, the one-dimensional time-frequency energy sequence of each channel is obtained by joining serially spliced time-energy sequence in the frequency band closely related to motor imagery. Finally, the brain connectivity matrix is calculated based on the symbolic transfer entropy between the time-frequency energy sequences of any two channels, and the brain functional network is constructed.The experiment results show that the brain functional network constructed with the symbolic transfer entropy between time-frequency energy sequences can effectively reflect the time-frequency characteristics and nonlinear characteristic information transmission of MI-EEG. Compared with the traditional brain network construction method, it is beneficial to enhance the separability of different motor imagery tasks.

Key words: brain-computer interface, motor imagery electroencephalography, continuous wavelet transform, symbolic transfer entropy, brain functional network

中图分类号:

TP18

李明爱, 张圆圆. 基于连续小波变换和符号传递熵的脑功能网络构建方法[J]. 电子学报, 2022, 50(7): 1600-1608.

LI Ming-ai, ZHANG Yuan-yuan. A Brain Functional Network Based on Continuous Wavelet Transform and Symbolic Transfer Entropy[J]. Acta Electronica Sinica, 2022, 50(7): 1600-1608.

图/表 13

图1 序列符号化示意图

图2 电极位置与导联序号分布图

图3 C3导联MI-EEG的时频能量图

图4 C3导联β频带的一维时频能量序列

图5 受试者S1两类想象任务的大脑连接矩阵图

图6 受试者S1的分类准确率

表1 不同特征选择方法的平均准确率对比

	特征选择算法
	None	Wrapper	Fisher	Pearson
准确率	0.639 5	0.944 6	0.947 4	0.9475
特征数目	4 096	190	130	120

表2 不同小波函数的分类准确率

频带	准确率	小波函数
频带	准确率	None	Haar	Daubechies	Mexican Hat	Meyer	Morlet
$α$	平均准确率	0.833 6	0.841 3	0.833 6	0.872 6	0.862 7	0.9078
$α$	最高准确率	0.955 6	0.977 8	1.0000	1.0000	1.0000	1.0000
$β$	平均准确率	0.841 3	0.907 0	0.897 0	0.874 5	0.914 0	0.9475
$β$	最高准确率	0.977 8	1.0000	1.0000	0.977 8	1.0000	1.0000

表2 不同小波函数的分类准确率

频带	准确率	小波函数
频带	准确率	None	Haar	Daubechies	Mexican Hat	Meyer	Morlet
$α$	平均准确率	0.833 6	0.841 3	0.833 6	0.872 6	0.862 7	0.9078
$α$	最高准确率	0.955 6	0.977 8	1.0000	1.0000	1.0000	1.0000
$β$	平均准确率	0.841 3	0.907 0	0.897 0	0.874 5	0.914 0	0.9475
$β$	最高准确率	0.977 8	1.0000	1.0000	0.977 8	1.0000	1.0000

图7 不同符号阶次的平均识别准确率

图8 不同时间段的脑地形图

图9 不同时间段的分类准确率

图10 获得图9准确率所需的特征数目

表3 基于多种脑功能网络方法的109名受试者平均分类准确率和Kappa值对比

	平均分类准确率		平均Kappa值
方法	a_ij	Degree+BC	a_ij	Degree+BC
CC ［8］	0.701 0	0.609 6	0.660 8	0.587 7
CMI ［9］	0.622 8	0.457 0	0.598 2	0.365 6
GC ［11］	0.909 6	0.751 6	0.827 7	0.701 3
DTF ［12］	0.874 2	0.872 6	0.799 4	0.778 1
MS ［10］	0.831 6	0.809 7	0.775 3	0.767 7
Ours	0.9475	0.9238	0.8580	0.8090

参考文献 22

1	WOLPAWJR, BIRBAUMERN, MCFARLANDDJ, et al. Brain-computer interface for communication and control[J]. Clinical Neurophysiology, 2002, 113(6): 767⁃791.
2	GOMEZ-PILARJ, CORRALEJOR, NICOLAS-ALONSOLF. Neurofeedback training with a motor imagery-based BCI: Neurocognitive improvements and EEG changes in the elderly[J]. Medical & Biological Engineering & Computing, 2016, 54(11): 1655⁃1666.
3	HAMEDIM, SALLEHS H, NOORA M. Electroencephalographic Motor imagery brain connectivity analysis for BCI: A review[J]. Neural Computation, 2016, 28(6):1⁃43.
4	RODRIGUESP G, FILHOCAS, ATTUXR, et al. Space-time recurrences for functional connectivity evaluation and feature extraction in motor imagery brain-computer interfaces[J]. Medical & Biological Engineering & Computing, 2019, 57: 1709⁃1725.
5	张娜, 孙炎珺, 李明爱. 一种个性化动态脑功能网络的构建与特征提取方法[J]. 北京生物医学工程, 2020, 39(6): 551⁃560.
	ZHANGN, SUNYJ, LIMA. A personalized dynamic brain functional network and feature extraction [J]. Beijing Biomedical Engineering, 2020, 39(6): 551⁃560. (in Chinese)
6	GAOZ, WANGZ, MAC, et al. A wavelet time-frequency representation based complex network method for characterizing brain activities underlying motor imagery signals[J]. IEEE Access, 2018, 6: 65796⁃65802.
7	彭丝雨, 周到, 张家琦, 王宇, 高军峰. 基于互信息的脑网络及测谎研究[J]. 电子学报, 2019, 47(7): 1551⁃1556.
	PENGS Y, ZHOUD, ZHANGJ Q, WANGY, GAOJ F. Research on mutual information-based brain network and lie detection[J]. Acta Electronica Sinica, 2019, 47(7): 1551⁃1556. (in Chinese)
8	CHUNGY G, KIMM K, KIMS P. Inter-channel connectivity of motor imagery EEG signals for a noninvasive BCI application[C]//IEEE International Workshop on Pattern Recognition in Neuroimaging. Cambridge: ACM, 2011: 49⁃52.
9	GONGA, LIUJ, CHENS, et al. Time-frequency cross mutual information analysis of the brain functional networks underlying multiclass motor imagery[J]. Journal of Motor Behavior, 2018, 50(3): 254⁃267.
10	FILHOC A S, ATTUXR, CASTELLANOG. EEG sensorimotor rhythms' variation and functional connectivity measures during motor imagery: Linear relations and classification approaches[J]. PeerJ, 2017, 5: 3983.
11	HUS, WANGH, ZHANGJ, et al. Causality from Cz to C3/C4 or between C3 and C4 revealed by granger causality and new causality during motor imagery[C]//International Joint Conference on Neural Networks. Beijing: IEEE, 2014: 3178⁃3185.
12	GHOSHP, MAZUMDERA, BHATTACHARYYAS, et al. Functional connectivity analysis of motor imagery EEG signal for brain-computer interfacing application[C]//2015 7th International IEEE/EMBS Conference on Neural Engineering. Montpellier, France: IEEE Press, 2015:210⁃213.
13	BAIGM Z, KAVAKLIM. Connectivity analysis using functional brain networks to evaluate cognitive activity during 3D modelling[J]. Brain Sciences, 2019, 9(2): 24.
14	STANIEKM, LEHNERTZK. Symbolic transfer entropy[J]. Physical Review Letters, 2008, 100(15): 158101.
15	PANCHEI D L P, ALVAREZ-MEZAAM, OROZCO-GUTIERREZA. A data-driven measure of effective connectivity based on renyi's α-entropy[J]. Frontiers in Neuroscience, 2019, 13: 1277.
16	GAOZ, WANGZ, MAC, et al. A wavelet time-frequency representation based complex network method for characterizing brain activities underlying motor imagery signals[J]. IEEE Access, 2018, PP(99): 1⁃1.
17	SCHALKG, MCFARLANDDJ, HINTERBERGERT, et al. BCI2000:A general-purpose brain-computer interface(BCI) System[J]. IEEE Transactions on Biomedical Engineering. 2004, 51(6): 1034⁃1043.
18	BRUNNERC, BILLINGERM, SEEBERM, et al. Volume conduction influences scalp-based connectivity estimates[J]. Frontiers in Computational Neuroscience. 2016,10: 121.
19	NEUPERP C. Motor imagery activates primary sensorimotor area in humans[J]. Neuroscience Letters, 1997,239(2): 65⁃68.
20	ZHUX, LIP, LIC, et al. Separated channel convolutional neural network to realize the training free motor imagery BCI systems[J]. Biomedical Signal Processing and Control, 2019, 49(3): 396⁃403.
21	ROMANB, GUISANDEN, GRANADOM. Characterization of visuomotor/imaginary movements in EEG: An information theory and complex network approach[J]. Frontiers in Physics, 2019, 7: 115.
22	FILHOC A S, ATTUXR, CASTELLANOG. Can graph metrics be used for EEG-BCIs based on hand motor imagery?[J]. Biomedical Signal Processing & Control, 2018, 40(2): 359⁃365.

[1]	常文文, 王宏, 化成诚. 基于听觉ERP功能脑网络特征和SVM的测谎方法研究[J]. 电子学报, 2016, 44(7): 1757-1762.
[2]	许敏鹏, 张力新, 明东, 綦宏志, 陈龙, 马岚, 万柏坤. 基于SSVEP阻断与P300特征的混合范式脑-机接口[J]. 电子学报, 2013, 41(11): 2247-2251.
[3]	綦宏志;孙长城;许敏鹏;明东;万柏坤;刘志朋;. P300 Speller中基于权值重采样的 ABSVM字符识别方法研究[J]. 电子学报, 2011, 39(11): 2534-2539.
[4]	杨立才, 李佰敏, 李光林, 贾磊. 脑-机接口技术综述[J]. 电子学报, 2005, 33(7): 1234-1241.
[5]	张晓冬;王桥;吴乐南. 利用脊的特征进行信号盲分离[J]. 电子学报, 2004, 32(7): 1156-1159.

基于连续小波变换和符号传递熵的脑功能网络构建方法

A Brain Functional Network Based on Continuous Wavelet Transform and Symbolic Transfer Entropy

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 22

相关文章 5

编辑推荐

Metrics