Artifacts denoising of EEG acquired during simultaneous EEG-FMRI

Abstract: Simultaneous electroencephalography and functional magnetic resonance imaging recordings (EEG-fMRI) have been widely used in neuroscientific and clinical research. Initially driven by the need to localize epileptogenic zones, the method evolved holding the promise to combine the high temporal resolution of EEG and high spatial resolution of fMRI. However, the MRI scanner’s influence on EEG recordings often hampers the analysis of EEG, limiting potential applications based on the fusion of these two modalities. To overcome these technical difficulties thus has been central along the development of EEG-fMRI.
EEG suffers primarily from induced voltages by fast changing gradients during MR imaging (gradient artifacts, GAs) and from physiological artifacts (mainly ballistocardiogram artifact, BCG). Due to their reproducible nature, they can often be eliminated by so-called average-artifact subtraction (AAS). However, motion during imaging and physiological variations will harm the effectiveness of the method. Therefore, the aim of this thesis is to improve the correction of these two main types of artifacts.
For GA removal, instead of adding hardware to model the motion-modulated GAs, a direct modelling of GA making use of sequence information is proposed. Apart from an improvement in artifact removal, regarding power spectral density (PSD) at the harmonics of 1/TR and the proportion of EEG epochs still affected by GA; complex motion information was also derived from this gradient modelling (GM) which showed high similarity with recorded motion tracking data.
BCG’s residual variance after AAS are usually further removed by PCA-based methods. However, the discrimination between the EEG signal and BCG artifact are usually ambiguous making the selection of number of principal components very tricky. To improve the effectiveness of the EEG-BCG discrimination, a moving average of the artifact across heart cycles was applied before the PCA (AA-PCA), thus the matrix to be PCA decomposed contains less EEG but purer BCG, resulting in an easier tuning between artifact removal and signal preservation. This was validated by the proportion of EEG epochs still affected by BCG and the signal-to-noise ratio of P3 event-related potentials.
An optimization of the PCA-based method to remove GA has been demonstrated as well. And the effectiveness of GA removal while preserving the signal of interest has been tested on the iEEG-fMRI data

Object Details

• Location: Deutsche Nationalbibliothek Frankfurt am Main
• Extent: Online-Ressource
• Language: Englisch
• Notes: Universität Freiburg, Dissertation, 2021

Contributors, Places and Time

• Event: Veröffentlichung
• (where): Freiburg
• (who): Universität
• (when): 2021
• Creator: Zhang, Shuoyue
• Contributor: Hennig, Jürgen ; Albert-Ludwigs-Universität Freiburg. Fakultät für Angewandte Wissenschaften

Further information

• DOI: 10.6094/UNIFR/193790
• URN: urn:nbn:de:bsz:25-freidok-1937909
• Rights: Der Zugriff auf das Objekt ist unbeschränkt möglich.
• Last update: 25.03.2025, 1:49 PM CET