Good practices in EEG-MRI: The utility of retrospective synchronization and PCA for the removal of MRI gradient artefacts

Abstract

The electroencephalogram (EEG) recorded during magnetic resonance imaging (MRI) inside the scanner is obstructed by the MRI gradient artefact (MGA) originating from the electromagnetic interference of the MRI with the sensitive measurement of electrical scalp potentials. Post-processing algorithms based on average artefact subtraction (AAS) have proven to be efficient in removing the MGA. However, the residual MGA after AAS still limits the quality and usable bandwidth of the EEG data despite further reduction through re-sampling, principal component analysis (PCA), and regressive filtering. We recently demonstrated that the residual MGA can largely be avoided by means of hardware synchronisation. Here we present a new software synchronisation method, which substitutes hardware synchronisation and facilitates the removal of motion artefacts by PCA. The effectiveness of the retrospective synchronisation algorithm (Resync) is demonstrated by comparison to the aforementioned techniques. For this purpose we also developed a method for simulating the MGA and we propose new concepts for quantifying and comparing the performance of post-processing algorithms for EEG-MRI data. Results indicate that the benefits of (retrospective) synchronisation and PCA depend largely on the relative contribution of timing errors and motion artefacts to the residual MGA as well as the frequency range of interest. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. ÔØ Å ÒÙ× Ö ÔØ A C C E P T E D M A N U S C R I P T ACCEPTED MANUSCRIPTGood practices in EEG-MRI: The utility of retrospective synchronization and PCA for the removal of MRI gradient artefacts. AbstractThe electroencephalogram (EEG) recorded during magnetic resonance imaging (MRI) inside the scanner is obstructed by the MRI gradient artefact (MGA) originating from the electromagnetic interference of the MRI with the sensitive measurement of electrical scalp potentials. Post-processing algorithms based on average artefact subtraction (AAS) have proven to be efficient in removing the MGA. However, the residual MGA after AAS still limits the quality and usable bandwidth of the EEG data despite further reduction through re-sampling, principal component analysis (PCA), and regressive filtering.We recently demonstrated that the residual MGA can largely be avoided by means of hardware synchronisation. Here we present a new software synchronisation method, which substitutes hardware synchronisation and facilitates the removal of motion artefacts by PCA. The effectiveness of the retrospective synchronisation algorithm (Resync) is demonstrated by comparison to the aforementi...

A personalized history of EEG–fMRI integration

“…Likewise, correction algorithms for the reduction of MR gradient and ECG (ballistocardiogram) artifact reduction have been refined. Not infrequently, when a group first engaged in EEG-fMRI research they explored different artifact correction methods and (fortunately) tried to make advances by developing their own by modifying and for their individual purposes improving mostly existing algorithms (Debener et al, 2007;Ellingson et al, 2004;Ford et al, 2004;Goncalves et al, 2007;Huiskamp, 2005;Koskinen and Vartiainen, 2009;Laufs et al, 2008;Mahadevan et al, 2008;Mandelkow et al, 2010;Masterton et al, 2007;Mullinger et al, 2011;Negishi et al, 2007;Otzenberger et al, 2007;Ritter et al, 2007;Ryali et al, 2009;Sijbersa et al, 2000;Srivastava et al, 2005;Vincent et al, 2007;Wan et al, 2006a,b). Some of these methods are freely available, e.g.…”

Statistical feature extraction for artifact removal from concurrent fMRI-EEG recordings

“…Since the same gradient waveform is applied every time when a new slice is acquired, the induced gradient artifacts in concurrent EEG recordings are highly repetitive and phase-locked to the onset time of each slice acquisition. For this reason most existing methods for removing gradient artifacts share a common strategy of subtracting (or regressing out) from the raw data a temporal template or a group of templates that collectively characterizes the reoccurring patterns at every acquisition, leaving out brain signals that are expected to be independent of the timing of fMRI scanning (Allen et al, 2000; Freyer et al, 2009; Mandelkow et al, 2010; Negishi et al, 2004; Niazy et al, 2005; Ryali et al, 2009; Sijbers et al, 1999; Wan et al, 2006). However, different ways to derive such artifact templates from recorded data can arrive at varying degrees of residual artifacts and a potential loss of neuroelectrical signals.…”

Pattern changes of EEG oscillations and BOLD signals associated with temporal lobe epilepsy as revealed by a working memory task

“…However, it is important to highlight that the simultaneous acquisition requires hardware compatibility and specific software for artifact removal [19-21]. …”