Evaluation of EEG pre-processing and source localization in ecological research.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalography (EEG) source localization (SL) is crucial for understanding brain function in neurological and psychiatric disorders.
• A key limitation is the requirement for subject-specific anatomical data (MRI), often unavailable in real-world research.
• This study addresses the neurophysiological plausibility of SL using only EEG signals.

Purpose of the Study:

• To investigate if established EEG source localization methods can yield neurophysiologically plausible activation patterns without subject-specific anatomical information.
• To validate template-based SL techniques on naturalistic EEG datasets.
• To assess the feasibility of SL in research settings lacking individual MRI data.

Main Methods:

• An end-to-end pipeline integrating automatic pre-processing and eLORETA source estimation was developed.
• A shared forward model using the ICBM 2009c Nonlinear Symmetric template and CerebrA atlas was employed.
• Validation was performed on the Healthy Brain Network (HBN) and COGBCI datasets, comparing different cognitive states and workloads using permutation testing.

Main Results:

• Significant differences in brain activation were observed between resting and video-watching states, with increased posterior activity during visual tasks.
• Cognitive workload analysis demonstrated progressive activation increases in executive function regions with rising task difficulty.
• Permutation testing confirmed that reconstructed source activations align with expected neurophysiological patterns.

Conclusions:

• Established EEG source localization methods can generate neurophysiologically plausible results without subject-specific anatomical data.
• Template-based source analysis is a viable approach for research where individual MRI is impractical.
• The study highlights the strengths and limitations of applying SL to naturalistic EEG data.