Simultaneous EEG–fMRI at ultra-high field: Artifact prevention and safety assessment

Abstract

The simultaneous recording of scalp electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) can provide unique insights into the dynamics of human brain function, and the increased functional sensitivity offered by ultra-high field fMRI opens exciting perspectives for the future of this multimodal approach. However, simultaneous recordings are susceptible to various types of artifacts, many of which scale with magnetic field strength and can seriously compromise both EEG and fMRI data quality in recordings above 3 T. The aim of the present study was to implement and characterize an optimized setup for simultaneous EEG-fMRI in humans at 7 T. The effects of EEG cable length and geometry for signal transmission between the cap and amplifiers were assessed in a phantom model, with specific attention to noise contributions from the MR scanner coldheads. Cable shortening (down to 12 cm from cap to amplifiers) and bundling effectively reduced environment noise by up to 84% in average power and 91% in inter-channel power variability. Subject safety was assessed and confirmed via numerical simulations of RF power distribution and temperature measurements on a phantom model, building on the limited existing literature at ultra-high field. MRI data degradation effects due to the EEG system were characterized via B 0 and B 1 + field mapping on a human volunteer, demonstrating important, although not prohibitive, B 1 disruption effects. With the optimized setup, simultaneous EEG-fMRI acquisitions were performed on 5 healthy volunteers undergoing two visual paradigms: an eyes-open/eyes-closed task, and a visual evoked potential (VEP) paradigm using reversing-checkerboard stimulation. EEG data exhibited clear occipital alpha modulation and average VEPs, respectively, with concomitant BOLD signal changes. On a single-trial level, alpha power variations could be observed with relative confidence on all trials; VEP detection was more limited, although statistically significant responses could be detected in more than 50% of trials for every subject. Overall, we conclude that the proposed setup is well suited for simultaneous EEG-fMRI at 7 T.

Towards high-quality simultaneous EEG-fMRI at 7 T: Detection and reduction of EEG artifacts due to head motion

“…and instant t. This is an approximated description where the current path between C i and C Ref is treated as a linear path along the scalp surface (a more rigorous approximation would involve modeling the volume conduction properties of the head (Yan et al, 2009)). Additionally, flux contributions from wire segments leaving the cap surface are assumed to be minimal, as can be achieved by adequate cable optimization (Jorge et al, 2015). If B !…”

“…Abralyte gel (EasyCap) was used to reduce electrode impedances. The cap connectors were linked to the EEG amplifiers via two 12 cm bundled cables, with the amplifiers placed just outside the head array (Jorge et al, 2015). After bandpass filtering (0.016-250 Hz) and digitization (0.5 μV resolution), the EEG signals were transmitted to the control room via fiber optic cables.…”

“…Part III: motion artifact correction and ICA ICA is a powerful exploratory technique that is often used in EEG data denoising (Arrubla et al, 2013;Eichele et al, 2005;Jorge et al, 2015). In this last section, we compared the performance of motion artifact correction based on motion sensors with an approach based on ICA denoising, and investigated the potential benefits of combining the two techniques.…”

“…Unfortunately, however, simultaneous EEG-fMRI acquisitions at ultra-high field suffer from various undesirable interactions that can degrade data quality and potentially compromise subject safety (Dempsey et al, 2001;Neuner et al, 2014). Safety concerns have been effectively moderated (Lemieux et al, 1997;Noth et al, 2012), and although EEG components can reduce the signal-to-noise ratio (SNR) of MR images, numerous studies have found that losses in temporal SNR remain acceptable for fMRI, even at ultra-high field strengths with fairly high electrode densities (Jorge et al, 2015;Luo and Glover, 2012;Mullinger et al, 2008b). In contrast, the artifacts induced in EEG recordings by the magnetic fields used in fMRI can surpass the signals of interest by several orders of magnitude, and severely compromise data quality (Allen et al, 2000;Debener et al, 2008).…”

“…Compared to gradient and pulse artifacts, considerably less work has been devoted to these contributions. Vibration-induced noise can be avoided by switching off the associated sources during acquisition, when possible (Mullinger et al, 2008a;Nierhaus et al, 2013), and certain EEG hardware modifications have been proposed to minimize its incidence (Jorge et al, 2015). The periodic structure of He coldhead contributions, in particular, has motivated a few novel data-based correction approaches as well (Kim et al, 2015;Rothlubbers et al, 2013).…”