Topological Features of Electroencephalography are Robust to Re-referencing and Preprocessing

Area of Science:

• Neuroscience
• Data Science
• Signal Processing

Background:

• Electroencephalography (EEG) is a widely used, cost-effective neuroimaging technique.
• EEG data analysis is sensitive to reference site selection and preprocessing steps.
• Developing reference- and preprocessing-independent analytical methods is crucial.

Purpose of the Study:

• To investigate the subject-specificity and robustness of topological features derived from EEG data.
• To compare the stability of embedding spaces versus correlation spaces under different referencing and preprocessing conditions.
• To establish a topological analysis framework for individual brain dynamics in EEG.

Main Methods:

• Construction of embedding spaces from multi-channel EEG time-series and their temporal structures.
• Analysis of topological structures within these embedding spaces.
• Comparison of topological properties with correlation spaces derived from EEG data.
• Evaluation of robustness against re-referencing and preprocessing variations.

Main Results:

• Topological structures of EEG embedding spaces are subject-specific and robust to re-referencing and preprocessing.
• Correlation spaces derived from EEG data lack subject specificity and robustness to reference changes.
• The shape of EEG signal configuration spaces encodes individual brain dynamics.
• Temporal correlations significantly constrain the dynamics of resting-state EEG signals.

Conclusions:

• Topological analysis of EEG embedding spaces provides a robust method for characterizing individual brain dynamics.
• This approach offers a valuable addition to conventional topographic analyses in EEG.
• The findings suggest a roadmap for integrating topological methods into standard EEG analysis pipelines.