基于同步EEG-fMRI采集的情绪认知重评数据特征融合分析研究

脑电(Electroencephalography, EEG)与功能磁共振成像(Functional magnetic resonance imaging, fMRI)为脑科学研究提供了互补的时空信息. 为研究大脑在对情绪图片采取认知重评策略时的神经活动, 基于同步采集的EEG-fMRI数据, 应用典型相关分析、经验模态分解及k-均值聚类等算法对融合情绪数据进行交叉关联和盲源分离, 得到空间上的fMRI图像和与之对应的EEG时间演变信号. 结果表明: 时域上, CCA分离出的脑电成分在认知重评状态下有明显的晚期正电位(Late positive potential, LPP) (潜伏期200ms～900ms)出现, 而且认知重评策略诱发下的LPP 波幅明显小于观看负性诱发的LPP波幅(F(1, 224)= 28.72, P<0.01), 而大于观看中性诱发的LPP波幅(F(1, 224)= 63.32, P<0.01); 与之对应的空域上, 可以明显地看出和情绪调节相关的扣带回, 额叶、颞叶等区域有明显激活区, 采用情绪认知重评策略时的脑区激活强度明显小于观看负性状态, 而大于观看中性, 且观看中性状态下被激活的与情绪相关的区域相对较少. 研究表明, 这种融合数据分析技术通过计算两种模态数据之间潜在的线性相关性, 可以有效地分离出大脑在时空上神经活动情况, 达到了同时描绘出大脑神经活动的时间信息与空间信息的效果.

Feature Fusion Analysis of Simultaneously Recorded EEG-fMRI in Emotion Cognitive Reappraisal

Electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) provide complementary spatio-temporal information for brain function study. In order to study the neural activity during the process of cognitive reappraisal strategy when emotional images are taken, canonical correlation analysis (CCA), empirical mode decomposition (EMD) and k-means clustering algorithms are used to finish the steps of cross-correlation and blinded source separate for simultaneously acquired EEG-fMRI emotion data. Spatial maps for fMRI data and the corresponding temporal evolution features of EEG signals are thus obtained. The results show that EEG components separated by CCA in cognitive reappraisal state have obviously late positive potential (LPP), with latency from 200ms～900ms, meanwhile the amplitudes of LPP evoked by taking cognitive reappraisal strategy are less than that evoked by watching negative images (F(1, 224)= 28.72, P<0.01), but more than that evoked by watching neutral images (F(1, 224)= 63.32, P<0.01). The corresponding spatial regions have apparent activation regions, such as cingulate, frontal, temporal lobe and other regions related to emotion regulation. What is more, the activation intensity evoked by taking cognitive reappraisal strategy is significantly less than that evoked by watching negative images, while it is more than that evoked by watching neutral images. The activated areas related to emotion are small when participants are watching neutral images. This fusion data processing technology computes the potential linear correlation between two modal data, which can effectively separate neural activities of the brain both in spatial and temporal domains, and can achieve the effects of obtaining temporal and spatial information of the brain activity at the same time.