Interactions between different EEG frequency bands and their effect on alpha–fMRI correlations

Abstract

In EEG/fMRI correlation studies it is common to consider the fMRI BOLD as filtered version of the EEG alpha power. Here the question is addressed whether other EEG frequency components may affect the correlation between alpha and BOLD. This was done comparing the statistical parametric maps (SPMs) of three different filter models wherein either the free or the standard hemodynamic response functions (HRF) were used in combination with the full spectral bandwidth of the EEG.

EEG and fMRI were co-registered in a 30 min resting state condition in 15 healthy young subjects. Power variations in the delta, theta, alpha, beta and gamma bands were extracted from the EEG and used as regressors in a general linear model. Statistical parametric maps (SPMs) were computed using three different filter models, wherein either the free or the standard hemodynamic response functions (HRF) were used in combination with the full spectral bandwidth of the EEG.

Results show that the SPMs of different EEG frequency bands, when significant, are very similar to that of the alpha rhythm. This is true in particular for the beta band, despite the fact that the alpha harmonics were discarded. It is shown that inclusion of EEG frequency bands as confounder in the fMRI–alpha correlation model has a large effect on the resulting SPM, in particular when for each frequency band the HRF is extracted from the data.

We conclude that power fluctuations of different EEG frequency bands are mutually highly correlated, and that a multi frequency model is required to extract the SPM of the frequency of interest from EEG/fMRI data. When no constraints are put on the shapes of the HRFs of the nuisance frequencies, the correlation model looses so much statistical power that no correlations can be detected.

Introduction

In many studies where neurocognitive processes are investigated using fMRI and BOLD signals (e.g. Greicius et al., 2003, Damoiseaux et al., 2006), the so called awake “resting state” is used as the state of reference, although in many of these studies this “resting state” is not objectively controlled. This shortcoming has been compensated using simultaneously recorded EEG signals, assuming that in the awake and resting subject the EEG displays prominent alpha activity. This issue was investigated previously (Goldman et al., 2002, Laufs et al., 2003b, Laufs et al., 2006, Moosmann et al., 2003, Gonçalves et al., 2006). Since these studies were aimed at the localisation of the generators of the alpha band, temporal variations in alpha power were convolved with the standard hemodynamic response function, and correlated with the fMRI BOLD signals to obtain statistical parametric maps (SPMs) in which the significant alpha generators were indicated. Results of the above mentioned studies showed that generally the alpha rhythm was negatively correlated to the BOLD signal in cortical regions, including e.g. the visual cortex or the pre- and post-central gyrus, whereas positive correlations were found in the thalamus. Large inter-subject variability was found in alpha-BOLD SPMs (Gonçalves et al., 2006, De Munck et al., 2007) which appeared to be mainly related to individual variations in the EEG signal (Gonçalves et al., 2008). The few studies that were performed on other frequency bands (Laufs et al., 2003a, Tyvaert et al., 2008) were based on group analysis and showed contradicting results.

In De Munck et al., 2007, De Munck et al., 2008), the EEG/fMRI correlation model was extended by extracting the hemodynamic response of the alpha from the data using the GLM, instead of assuming the standard HRF. The impulse response function so obtained was called alpha response function (ARF) to stress that it reflects the response to alpha rhythm observed with the surface EEG instead of a true hemodynamic response to the local electrical input. These ARFs were found to vary over the cortex (De Munck et al., 2007) and the peak time delay was shorter for the thalamus with respect to the cortex. These differences of ARFs might be due to differences in regional vasculature, which is a characteristic that has been noted by Harrison et al. (2002), but may also be due to differences in local electric activities caused by the ignorance of other frequency bands than the alpha band in the correlation model.

In the present paper, we consider the EEG/fMRI correlation models used so far as incomplete because they are based on the selection of a single EEG frequency band from which a regressor of interest is derived. Ignoring the other frequency bands, reflecting related or unrelated associated cognitive processes, and their interactions with the frequency of interest, may lead to spurious correlations or to the observed spatial variations of the ARF. Tyvaert et al. (2008) have demonstrated, using the EEG/fMRI resting state data of a few subjects with epilepsy and using the standard HRF model, that accounting for the other frequencies may indeed have an effect on the resulting SPM.

The main goal of the present study is to investigate the interaction of different EEG frequency bands and their effects on the SPM systematically by comparing different correlation models with various numbers of degrees of freedom. Contrary to Tyvaert et al. (2008), EEG/fMRI data of healthy subjects are used that were all measured with the same protocol. Basically two models were used: (a) one that we call the Single Frequency Model, where a single frequency of interest is selected and the other frequencies are ignored and (b) another one that we call the Multi Frequency Model and where one frequency band is taken as regressor of interest and all the others are added to the confounders. Furthermore in model (a) we used a Free Hemodynamic Response Function (HRF); in model (b) we used two variations of the model. In model (b1) the Free HRF was used for each frequency component and in model (b2) we used the standard or canonical HRF. Related secondary goals of the present study were to explore the interactions between EEG regressors derived from different frequency bands and electrode sides and to study the hemodynamic response functions and SPMs for other frequency bands than the alpha band.