ICA on sensor or source data: A comparison study in deriving resting state networks from EEG.

Area of Science:

• Neuroscience
• Brain Imaging
• Signal Processing

Background:

• Resting state networks (RSNs) are crucial for understanding brain function during rest.
• Functional magnetic resonance imaging (fMRI) is the primary tool for RSN study.
• Electroencephalography (EEG) and magnetoencephalography (MEG) offer alternative RSN analysis methods.

Purpose of the Study:

• To compare the performance of source-space ICA and sensor-space ICA in reconstructing RSNs.
• To evaluate the spatial, temporal, and spectral feature reconstruction accuracy of both ICA approaches.
• To validate findings using simulated and real EEG data against fMRI-derived RSN templates.

Main Methods:

• Independent Component Analysis (ICA) applied in both sensor-space and source-space (using inverse source imaging).
• Analysis of simulated EEG data to assess reconstruction fidelity.
• Analysis of resting-state EEG data from seven healthy participants.
• Comparison of EEG-derived RSNs with established fMRI RSN templates.

Main Results:

• Source-space ICA demonstrated superior performance in reconstructing spatial, temporal, and spectral features of RSNs from simulated data.
• Analysis of real EEG data revealed performance differences between the two ICA methods.
• Source-space ICA showed relatively better performance compared to sensor-space ICA when validated against fMRI RSN templates.

Conclusions:

• Source-space ICA is a more effective method for reconstructing RSNs from EEG data than sensor-space ICA.
• The findings suggest that source-space ICA provides more accurate representations of RSNs, aligning better with fMRI data.
• This improved accuracy has implications for future research utilizing EEG for RSN analysis.