Source-Modeling Auditory Processes of EEG Data Using EEGLAB and Brainstorm.

Maren Stropahl,Anna-Katharina R Bauer,Martin G Bleichner,Stefan Debener

doi:10.3389/fnins.2018.00309

Maren Stropahl, Anna-Katharina R Bauer + Show 2 more

Open Access

https://doi.org/10.3389/fnins.2018.00309

Copy DOI

Abstract

Electroencephalography (EEG) source localization approaches are often used to disentangle the spatial patterns mixed up in scalp EEG recordings. However, approaches differ substantially between experiments, may be strongly parameter-dependent, and results are not necessarily meaningful. In this paper we provide a pipeline for EEG source estimation, from raw EEG data pre-processing using EEGLAB functions up to source-level analysis as implemented in Brainstorm. The pipeline is tested using a data set of 10 individuals performing an auditory attention task. The analysis approach estimates sources of 64-channel EEG data without the prerequisite of individual anatomies or individually digitized sensor positions. First, we show advanced EEG pre-processing using EEGLAB, which includes artifact attenuation using independent component analysis (ICA). ICA is a linear decomposition technique that aims to reveal the underlying statistical sources of mixed signals and is further a powerful tool to attenuate stereotypical artifacts (e.g., eye movements or heartbeat). Data submitted to ICA are pre-processed to facilitate good-quality decompositions. Aiming toward an objective approach on component identification, the semi-automatic CORRMAP algorithm is applied for the identification of components representing prominent and stereotypic artifacts. Second, we present a step-wise approach to estimate active sources of auditory cortex event-related processing, on a single subject level. The presented approach assumes that no individual anatomy is available and therefore the default anatomy ICBM152, as implemented in Brainstorm, is used for all individuals. Individual noise modeling in this dataset is based on the pre-stimulus baseline period. For EEG source modeling we use the OpenMEEG algorithm as the underlying forward model based on the symmetric Boundary Element Method (BEM). We then apply the method of dynamical statistical parametric mapping (dSPM) to obtain physiologically plausible EEG source estimates. Finally, we show how to perform group level analysis in the time domain on anatomically defined regions of interest (auditory scout). The proposed pipeline needs to be tailored to the specific datasets and paradigms. However, the straightforward combination of EEGLAB and Brainstorm analysis tools may be of interest to others performing EEG source localization.

Highlights

Despite strong competition from other imaging techniques, the scalp-recorded electroencephalogram (EEG) is still one of the key sources of information for scientists interested in the study of large-scale human brain function
Source modeling on the other hand allows to draw inferences about the timing and the location of brain processes of interest and may resolve to some degree the ambiguity we are faced with sensor level analysis (Michel et al, 2004; Lopes da Silva, 2013; Baillet, 2017)
Despite the fact that a source level analysis does not solve the inverse problem (Musha and Okamoto, 1999; Grech et al, 2008), high-density EEG in combination with source modeling is considered as an electrical brain imaging tool (Michel and Murray, 2012), which helps to confirm predictions about the likely spatial origin of EEG sensor level features

Summary

Introduction

Despite strong competition from other imaging techniques, the scalp-recorded electroencephalogram (EEG) is still one of the key sources of information for scientists interested in the study of large-scale human brain function. We have used source level analysis of 96-channel EEG recordings to study cross-modal processing in the auditory cortex of cochlear implant users (Stropahl et al, 2015; Stropahl and Debener, 2017), and showed that in these individuals, auditory cortex is recruited for the processing of visual stimuli This pattern has been repeatedly confirmed with EEG source analysis, as well as imaging modalities such as functional near infrared spectroscopy (fNIRS), but could not be obtained with functional magnetic resonance imaging (fMRI), which cannot be used for cochlear implant users (Sandmann et al, 2012; Stropahl et al, 2015, 2017; Chen et al, 2016). In other studies we used source level analysis to disentangle left and right auditory cortex activation patterns (Hine and Debener, 2007; Hine et al, 2008; Sandmann et al, 2015) and investigate the temporal evolution of auditory entrainment using 64-channel EEG recordings (Bauer et al, 2018)

Objectives

Methods

Results

Discussion

Conclusion