Date Published: November 10, 2014
Publisher: Public Library of Science
Author(s): Eleonora Maggioni, Jorge Arrubla, Tracy Warbrick, Jürgen Dammers, Anna M. Bianchi, Gianluigi Reni, Michela Tosetti, Irene Neuner, N. Jon Shah, Wang Zhan.
Simultaneous electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) allow for a non-invasive investigation of cerebral functions with high temporal and spatial resolution. The main challenge of such integration is the removal of the pulse artefact (PA) that affects EEG signals recorded in the magnetic resonance (MR) scanner. Often applied techniques for this purpose are Optimal Basis Set (OBS) and Independent Component Analysis (ICA). The combination of OBS and ICA is increasingly used, since it can potentially improve the correction performed by each technique separately. The present study is focused on the OBS-ICA combination and is aimed at providing the optimal ICA parameters for PA correction in resting-state EEG data, where the information of interest is not specified in latency and amplitude as in, for example, evoked potential. A comparison between two intervals for ICA calculation and four methods for marking artefactual components was performed. The performance of the methods was discussed in terms of their capability to 1) remove the artefact and 2) preserve the information of interest. The analysis included 12 subjects and two resting-state datasets for each of them. The results showed that none of the signal lengths for the ICA calculation was highly preferable to the other. Among the methods for the identification of PA-related components, the one based on the wavelets transform of each component emerged as the best compromise between the effectiveness in removing PA and the conservation of the physiological neuronal content.
The combination of electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) can provide a non-invasive comprehensive view of brain activity with high temporal (EEG) and spatial (fMRI) resolution. The EEG technique gives a measure of the synchronized electrical activity of large populations of neurons. Despite its high temporal resolution, which is in the order of tens of milliseconds, the EEG suffers from the spatial inverse problem, related to the difficulty in inferring the spatial location of neuronal sources in the brain from the potentials recorded at scalp level , .
Across datasets and ICA intervals, it emerged that the variance-based selection criterion removed less components compared to the others, with 9.2±2.6 (mean ± standard deviation) out of 63 ICs removed, against 19.6±7.3 of the correlation method, 21.4±4.9 of the wavelets method and 20±5.3 of the partial autocorrelation function method. The results of each validation method relative to both the datasets are shown below.
The main objective of the present study was to identify the optimal ICA parameters for the removal of PA from EEG data recorded in an MR environment, after OBS correction. Since our interest was the analysis of spontaneous brain activity with EEG and fMRI, we discussed the effects of different ICA parameter settings on resting-state EEG data recorded at 3T. We compared two intervals for the calculation of the ICA mixing matrix, 1) the entire signal and 2) the PA intervals, together with four methods for selecting the PA-related ICs, based on their 1) contribution to the artefact variance, 2) correlation with PA templates, 3) wavelets transform and 4) partial autocorrelation function. The quality of the EEG cleaning was assessed by looking at the changes occurring after ICA correction in the EEG signal around the R peaks (from −200 ms to 1 s after it). Three different criteria were considered, based on the EEG 1) peak to peak amplitude, 2) batch spectral content and 3) time-varying spectral content. The comparison was performed on two groups of datasets relative to the same 12 subjects: the general agreement between the outcomes of the two comparisons highlighted the reliability of each ICA correction, whose performances were usually reproducible across datasets. The selection of PA-related ICs based on their wavelets transform emerged as the best compromise between the amount of removed PA and the preservation of the neuronal alpha content.
A full exploitation of the potentials of EEG-fMRI integration is possible only if an optimal cleaning of the EEG signal from the MR related artefacts is performed. The cardiac-related artefact has variable characteristics over space and time that make it difficult to remove. This study focused on the PA correction based on the combination of OBS and ICA and compared eight different ICA corrections, i.e. two intervals for the ICA calculation and four methods for selecting the PA-related components. Different criteria for the assessment of the quality of PA removal were used, some sensitive to the artefact removal, others also to the preservation of the information of interest. The two intervals of ICA calculation led to similar results, whereas the selection of the artefactual components based on their wavelets transform emerged as preferable to the other selection methods, since it resulted in the ability to highlight the PA-related components, making them easily distinguishable from the neuronal ones. The results were usually in agreement across the two datasets, thus confirming the reproducibility of the performance of each ICA correction algorithm. Even though the quality of the PA removal largely depends on the performance of the ICA decomposition, the present work provides valuable information on the optimization of the selection of PA-related ICs and on the assessment of the effects that each PA correction has on the EEG signal.