Trial-by-trial coupling between EEG and BOLD identifies networks related to alpha and theta EEG power increases during working memory maintenance

Abstract

PET and fMRI experiments have previously shown that several brain regions in the frontal and parietal lobe are involved in working memory maintenance. MEG and EEG experiments have shown parametric increases with load for oscillatory activity in posterior alpha and frontal theta power. In the current study we investigated whether the areas found with fMRI can be associated with these alpha and theta effects by measuring simultaneous EEG and fMRI during a modified Sternberg task This allowed us to correlate EEG at the single trial level with the fMRI BOLD signal by forming a regressor based on single trial alpha and theta power estimates. We observed a right posterior, parametric alpha power increase, which was functionally related to decreases in BOLD in the primary visual cortex and in the posterior part of the right middle temporal gyrus. We relate this finding to the inhibition of neuronal activity that may interfere with WM maintenance. An observed parametric increase in frontal theta power was correlated to a decrease in BOLD in regions that together form the default mode network. We did not observe correlations between oscillatory EEG phenomena and BOLD in the traditional WM areas. In conclusion, the study shows that simultaneous EEG–fMRI recordings can be successfully used to identify the emergence of functional networks in the brain during the execution of a cognitive task.

Introduction

Working memory (WM) has been one of the central themes in cognitive neuroscience research for the past decades. A considerable number of studies have either used hemodynamic (e.g. PET and fMRI) or electrophysiological recordings (e.g. MEG, EEG, intracranial recordings). PET and fMRI have been successful in linking different brain regions to task and modality specific WM processes (Cabeza and Nyberg, 2000, D'Esposito et al., 2000, Fletcher and Henson, 2001). Across modalities and tasks, dorso- and ventrolateral prefrontal and posterior parietal regions have been often linked to WM processes (Cabeza and Nyberg, 2000). In recent years, regions in the medial temporal lobe have also been implicated in WM, suggesting WM and long term memory share in part the same neural substrate (Cabeza et al., 2002, Petersson et al., 2006, Piekema et al., 2006, Ranganath et al., 2005, Ranganath and D'Esposito, 2001).

The high temporal resolution of EEG and MEG makes it possible to study WM-related processes at a millisecond time-scale. WM-related increases in oscillations observed in the EEG or MEG have been reported in the theta band (Gevins et al., 1997, Krause et al., 2000, Onton et al., 2005), alpha band (Klimesch et al., 1999); Jensen et al., 2002, Jokisch and Jensen, 2007, Schack and Klimesch, 2002, Tuladhar et al., 2007) and gamma band (Jokisch and Jensen, 2007, Kaiser et al., 2003, Lutzenberger et al., 2002, Tallon-Baudry et al., 1998). More specifically, parametric increases with WM load in the maintenance interval have been reported for posterior alpha (Jensen et al., 2002) and frontal theta power (Gevins et al., 1997, Jensen and Tesche, 2002, Onton et al., 2005). Different functional roles have been proposed for these different frequency bands. Theta and gamma have been hypothesized to be a direct neural correlate of WM maintenance, possibly in cooperation with medial temporal structures (Jensen, 2006, Jensen and Lisman, 1998). Alpha power increases have been linked to active inhibition of neuronal activity that could otherwise disturb the WM process (Jokisch and Jensen, 2007, Klimesch et al., 2007). Differential alpha power effects can be seen in upper and lower sub-bands (Klimesch, 1999). Decreases in power in the lower alpha band have been linked to higher task demands and attentional processing, whereas decreases in power in the upper alpha band reflects have been related to increased declarative memory performance. Klimesch et al. (1999) observed an increase during WM maintenance of character strings and they suggest that this is related to inhibition of memory activities that could otherwise disturb the WM process.

Based on these previous findings the question arises whether the regions observed with PET and fMRI are also functionally related to the oscillatory phenomena as measured with MEG and EEG. To address this question, we simultaneously measured EEG and fMRI while subjects performed a modified verbal Sternberg WM task. By relating single trial estimates of WM-related power increases to changes in the BOLD signal we hope to uncover which brain regions are functionally related to these WM-induced power increases. Successful single trial coupling between EEG measures (ERP's) and the BOLD signal has been demonstrated by Eichele et al., 2005, Debener et al., 2005.

The relation between alpha power on the one hand and the BOLD signal on the other hand has been investigated in previous studies. Posterior alpha is most prominent during eyes closed alert wakefulness, and usually decreases with increased visual processing (Klimesch, 1999). We therefore expected posterior alpha to correlate with the BOLD signal regions involved in visual processing. Posterior alpha has indeed been found to correlate with posterior visual areas in eyes closed resting state conditions (Feige et al., 2005, Goldman et al., 2002, Moosmann et al., 2003). More recently Laufs et al., 2006, Goncalves et al., 2006 showed that the observed networks that correlate with alpha power might depend on the relative strengths of other frequency bands in the entire power spectrum. Recently, Meltzer et al. (2007) showed in separate EEG and FMRI sessions, that load dependent increases in alpha power during WM maintenance correlate negatively with the midline parietal–occipital cortex across subjects.

Frontal theta is a prominent feature in the EEG that is reported to increase in tasks that require attention or WM. It is often studied during mental arithmetic (Inanaga, 1998, Inouye et al., 1994, Ishihara and Yoshii, 1972, Ishii et al., 1999, Lazarev, 1998, Mizuki et al., 1980, Sasaki et al., 1996, Smith et al., 1999), but more recently it has also been associated with WM maintenance (Gevins et al., 1997, Jensen, 2006, Jensen et al., 2002). Recently, two studies investigated the BOLD correlates of frontal theta activity during mental arithmetic (Mizuhara et al., 2004, Sammer et al., 2007). While Sammer et al. (2007) reported only positive correlations between BOLD and theta in the insula, medial temporal lobe, superior temporal cortex, cingulate cortex and various frontal regions. Mizuhara et al. (2004) reported predominantly negative correlations in medial frontal, posterior cingulate, temporal and inferior parietal regions. In a recent resting state experiment we only observed negative correlations with frontal theta power in a collection of regions that together form the default mode network (DMN; Scheeringa et al., 2008). This negative correlation between frontal theta and DMN activity was also observed by Meltzer et al. (2007) across subjects in a WM task.

Regions in which correlations between single trial EEG power measures and the BOLD signal are observed cannot be directly interpreted as being functionally related to the WM-induced parametric power increases. The reason for this is that the relation between BOLD signal in a region, and single trial estimates of power could also be related only to task independent fluctuations in power. These task independent power fluctuations could be related to EEG activity coming from other sources than the one that shows a WM-induced power change, that leak into the single trial estimates of power. Another potential source of task independent variation in power could lie in trial-by-trial coupling of task related regions with task unrelated regions, which has been observed between the motor cortices using fMRI (Fox et al., 2006). This task independent trial-by-trial coupling could possibly also be reflected in the trial-by-trial variation of EEG power components that do show an average effect of WM load. Therefore we argue here that regions functionally related to WM-induced alpha and theta EEG power increases should show a BOLD response that is in line with the observed WM effects in these bands and also show a relation with the single trial variation in power.