Acquisition of typical EEG waveforms during fMRI: SSVEP, LRP, and frontal theta

Abstract

Recent work has demonstrated the feasibility of simultaneous electroencephalography (EEG) and functional magnetic resonance imaging (fMRI). Virtually no systematic comparisons between EEG recorded inside and outside the MR scanner have been conducted, and it is unknown if different kinds of frequency mix, topography, and domain-specific processing are uniformly recordable within the scanner environment. The aim of the study was to investigate several typical EEG waveforms in the same subjects inside the magnet during fMRI and outside the MR examination room. We examined whether uniform artifact subtraction allows the extraction of these different EEG waveforms inside the scanner during EPI scanning to the same extent as outside the scanner. Three well-established experiments were conducted, eliciting steady state visual evoked potentials (SSVEP), lateralized readiness potentials (LRP), and frontal theta enhancement induced by mental addition. All waveforms could be extracted from the EEG recorded during fMRI. Substantially no differences in these waveforms of interest were found between gradient-switching and intermediate epochs during fMRI (only the SSVEP-experiment was designed for a comparison of gradient—with intermediate epochs), or between waveforms recorded inside the scanner during EPI scanning and outside the MR examination room (all experiments). However, non-specific amplitude differences were found between inside and outside recorded EEG at lateral electrodes, which were not in any interaction with the effects of interest. The source of these differences requires further exploration. The high concordance of activation patterns with published results demonstrates that EPI-images could be acquired during EEG recording without significant distortion.

Introduction

The combination of the electroencephalogram (EEG) and functional magnetic resonance images (fMRI) is a potentially powerful approach in multi-modal brain research. Because the two methods are sensitive to different temporal and spatial properties of brain function, they have the potential to complement one another. The low time resolution of BOLD-sampling blurs interpretations of the functional specificity of a BOLD-activated brain structure. Peaks and latencies of event-related potentials (ERPs) provide additional functional information at high time resolution. They can be utilized to describe functional fMRI-activations more precisely. On the other hand, the spatial resolution of EEG is poor and the localization of sources for measured voltage distributions difficult. Estimation of cortical generators was shown to benefit from the use of spatial fMRI-constraints (Babiloni et al., 2003, Bonmassar et al., 2001). Recording EEG and fMRI simultaneously is of advantage if task repetition is supposed to induce additional processes, for example, memory, practice, subjective probability of stimulus occurrence, or other well known influences on repeated task performance. Only simultaneously recorded EEG and fMRI can ensure that the same composition of processes is represented in fMRI and EEG.

So far, only a few studies have been carried out on simultaneously acquired EEG and fMRI. Goldman et al. (2000) recorded EEG between periods of gradient-induced noise and showed increased power in the alpha band (8–12 Hz) when the subject's eyes were closed. After recording EEG with a specific sampling protocol (‘stepping stone sampling’) and the application of a gradient artifact subtraction method, Anami et al. (2003) retrieved alpha activity by inspection during eyes open/closed periods. Laufs et al. (2003) correlated the alpha power time course with the BOLD signals. In this study, the EEG measured during gradient switching was used for analysis as well. Using the MR gradient artifact subtraction technique and pulse artifact subtraction method which have been proposed by Allen et al., 1998, Allen et al., 2000, the authors concluded that artifact subtraction works for the spectral frequency analysis approach. Salek-Haddadi et al., 2003a, Salek-Haddadi et al., 2003b demonstrated the identification of epileptiform events in EEG continuously recorded during imaging. Bénar et al. (2003) compared the effectiveness of scanning artifact removal methods and for reducing the ballistocardiographic artifact in epileptic patients having interictal epileptiform discharges. They found that EPI-artifact subtraction (without using a triggering signal from the scanner, and sampling the EEG at a rate of 1 kHz) left less distortion than using a FFT filter, as it has been suggested by Hoffmann et al. (2000). Nevertheless, Bénar et al. (2003) reported about larger remaining imaging artifacts in the EEG, limiting the detection of epileptiform waves in some patients. In the same study, the ballistocardiographic artifact could be eliminated satisfactorily using both PCA and ICA, but also without any correction, the epileptiform discharges in the EEG were preserved. Because this is in distinction to most of the literature (cp. Allen et al., 1998, Bonmassar et al., 2002, Goldman et al., 2000), Bénar et al. attribute their results to the successful use of head restraints and electrode-wiring inside the scanner, but they limited this finding to epileptic EEG. Event-related potentials were recorded by Kruggel et al. (2000). They compared checkerboard visual evoked potentials (VEP) inside a 3T magnet with and without fMRI scanning to published reference data. Latencies of the P2 and N3 corresponded to these reference data outside a MR scanner. In a subsequent study (Kruggel et al., 2002), event-related potentials (ERPs) were computed on onsets of Kaniza figures and non-Kaniza figures (figures composed of the same elements as Kaniza-figures but they were arranged in such a way that prevented the illusion). Again ERPs were computed from MR-gradient-free epochs. MR-gradient artifact epochs were excluded from analysis. The ERPs were shown to vary with experimental conditions (Gestalt perception and target processing). Bonmassar et al. (2001) demonstrated the acquisition of 4 Hz checkerboard VEPs during fMRI scanning. Epochs for VEP averaging were scheduled between epochs of gradient switching to avoid gradient artifacts. Other studies showing the feasibility of simultaneous EEG and fMRI recordings were published by Goldman et al. (2002), Hoffmann et al. (2000), Lazeyras et al. (2002), without EEG-analysis, Lemieux et al. (2001), and Salek-Haddadi et al., 2002.

Although some studies have been performed on simultaneous EEG/fMRI recording, this area is still in its early stages. The main sources of artifacts in the EEG are identified: movements in the magnetic field, cardioballistic artifacts and HF-gradient artifacts. An influence of the magnetic field itself on the EEG, similar to the effect on the T-wave of the electrocardiogram (Wendt et al., 1988), has not been reported so far. Significant progress was made for the development of artifact-reduction methods. However, it is still unknown which effects remaining artifacts have on signal-extraction. Moreover, it is not known if different aspects of the EEG, for example, frequency-mix, topography, and domain-specific processing (e.g., motor, visual, cognitive) are uniformly recordable within the scanner environment. Accordingly, the aim of the study was to investigate several typical EEG waveforms in the same subject inside and outside the magnet. EEG was measured during fMRI scanning—including periods of gradient switching—and outside the MR examination room in the same subjects using identical technical EEG-equipment. We investigated if the application of artifact subtraction methods allows the extraction of typical waveforms to the same extent as outside the scanner.

Three well-established experiments were conducted in order to elicit a variety of EEG waveforms. In the first experiment, steady state visual-evoked potentials (SSVEP) were elicited by watching a rapid sequence of flashes. Müller (1997) showed that SSVEPs could be elicited with all stimulation frequencies. SSVEPs are sinusoidal in form and the waveform represents the stimulation frequency or its harmonic frequency (with slow stimulation rates). With constant stimulus intensity for all frequencies, the SSVEP amplitude decreases with higher stimulation frequencies. Estimation of equivalent current dipoles from magnetoencephalographic steady state visual-evoked fields (SSVEFs) were localized in the posterior occipital cortex near the calcarine fissure for 6.0 and 11.9 Hz SSVEF responses (Müller, 1997).

In the second experiment, readiness potentials (cf. Kornhuber and Deecke, 1965) were recorded. The readiness potential is part of the movement-related slow negativity, which occurs in the S1(warning stimulus)–S2(imperative stimulus)–R(Response) paradigm. It is mostly related to the S2-related negativity. The readiness potential (RP) consists of at least two components, an early (RP1) and a late (RP2) readiness potential. The RP1 principal generator is located in the mesial prefrontal cortex and supplementary motor area (SMA); the RP2 principal generator is the primary motor cortex (Deecke et al., 1998). The response-locked lateralized readiness potential (LRP) is a specific case of the RP2. It is computed by subtracting the ERP above the cortex ipsilateral to the side of motor response from the contralaterally recorded ERP. These difference waveforms are averaged across left hand and right hand motor-responses. The LRP is considered to represent response preparation or activation (Coles, 1989). Recently, Masaki et al. (2004) proposed that the LRP starts after the completion of response-hand selection and at the beginning of motor programming. Studying fMRI, Dehaene et al. (1998) used the time course of motor-cortex voxel activation to compute a lateralized bold response (LBR), which was higher with the certainty of the side of the required motor response.

In a third experiment, the EEG theta-frequency band was studied. There is numerous published work showing that theta band activity in the human EEG is associated with working memory engagement (e.g., Gevins et al., 1997), especially over the frontal midline (Inanga, 1998). Mental arithmetic calculations are an appropriate method to induce working memory processes in order to increase EEG theta power. Klimesch et al. (2001) proposed that theta activity reflects the encoding of new information into working memory. Imaging studies on mental arithmetic tasks have described a pattern of activation, including structures of the prefrontal and the parietal cortex (Burbaud et al., 1995, Dehaene et al., 1999). In our study, frontal theta in relation to mental arithmetic was compared to a baseline task, usually showing significantly less activation in the EEG-theta frequency range.

The EEG waveforms associated with the paradigms mentioned above represent a range of typical EEG characteristics. They differ in scalp-topography (occipital, central lateralized, frontal), characteristic frequency (stimulus-induced, slow potential, typical EEG-frequency band), and prominent components in the waveforms (sinusoidal oscillations, negative peak, power in frequency domain), and they represent activity in different cognitive or behavioral domains (visual perception, motor preparation, thinking).

Each subject performed the three experiments twice, inside the magnet during fMRI EPI scanning and outside the MR examination room. fMRI was acquired throughout the experiments. It has been shown that MR images can be acquired during EEG without significant distortion (Krakow et al., 2000). Most of the combined EEG/fMRI studies using simultaneous recordings were focused on the detection of epileptic EEG phenomena, showing that filtering methods enable to detect many of the epileptic events in the EEG (cp. Bénar et al., 2003). In this study, the attempt is made to detect several EEG standard waveforms, that is, event-related potentials and spectral compositions of the ongoing EEG, in the same subjects using subtraction-filtering techniques. Bénar et al. (2003) pointed to the fact that feasibility reports in laboratories where the methods were developed might not fully apply to other MR sites. Particularly, specific technical equipment could limit the replication of findings. Thus, feasibility studies in different laboratories contribute to the development of generally applicable simultaneous EEG-fMRI recording and analysis protocols.