Debener Stefan, Strobel Alexander, Sorger Bettina, Peters Judith, Kranczioch Cornelia, Engel Andreas K, Goebel Rainer
MRC Institute of Hearing Research Southampton, Royal South Hants Hospital, School of Medicine, University of Southampton, UK.
Neuroimage. 2007 Jan 15;34(2):587-97. doi: 10.1016/j.neuroimage.2006.09.031. Epub 2006 Nov 16.
EEG signals recorded simultaneously with fMRI are massively compromised by severe artefacts, among them the cardiac cycle-related ballistocardiogram (BCG) artefact. Different methods have been proposed to remove the BCG artefact focusing on channel-wise template subtraction procedures or spatial filtering approaches such as independent component analysis (ICA). Here we systematically compared the performance of the optimal basis set (OBS), a channel-wise correction approach, with ICA and a recently proposed combination of both (OBS-ICA). The three different procedures were applied to 60-channel EEG data from 12 subjects recorded during fMRI acquisition in a 3-T scanner. In addition to examination of the residual BCG artefact, the signal-to-noise ratio (SNR) and the topography of the resulting auditory evoked potential component N1 were compared. Whereas all three approaches led to a significant artefact reduction, the ICA procedure resulted in a significantly reduced N1 SNR and amplitude when compared to BCG-uncorrected data, indicating a rather poor performance. In contrast to ICA, OBS and OBS-ICA corrected data substantially improved the SNR of the N1. The quality of the auditory evoked potential N1 topography was investigated by means of equivalent current dipole modelling. On a descriptive level, all three correction procedures led to a reduced localization error when compared to BCG-uncorrected data. This improvement was significant for OBS-ICA. We conclude that OBS and OBS-ICA can efficiently remove BCG artefacts and substantially improve the quality of EEG signals recorded inside the scanner, a prerequisite for the successful integration of simultaneously recorded EEG and fMRI.
与功能磁共振成像（fMRI）同时记录的脑电图（EEG）信号受到严重伪迹的极大影响，其中包括与心动周期相关的心冲击图（BCG）伪迹。已经提出了不同的方法来去除BCG伪迹，主要集中在逐通道模板减法程序或空间滤波方法，如独立成分分析（ICA）。在这里，我们系统地比较了最优基集（OBS，一种逐通道校正方法）与ICA以及最近提出的两者结合方法（OBS-ICA）的性能。这三种不同的程序应用于在3-T扫描仪中进行fMRI采集期间记录的12名受试者的60通道EEG数据。除了检查残留的BCG伪迹外，还比较了所得听觉诱发电位成分N1的信噪比（SNR）和地形图。虽然所有三种方法都导致伪迹显著减少，但与未校正BCG的数据相比，ICA程序导致N1的SNR和幅度显著降低，表明其性能相当差。与ICA相反，OBS和OBS-ICA校正后的数据显著提高了N1的SNR。通过等效电流偶极子建模研究了听觉诱发电位N1地形图的质量。在描述层面上，与未校正BCG的数据相比，所有三种校正程序都导致定位误差减小。对于OBS-ICA，这种改善是显著的。我们得出结论，OBS和OBS-ICA可以有效地去除BCG伪迹，并显著提高在扫描仪内记录的EEG信号的质量，这是成功整合同时记录的EEG和fMRI的先决条件。
