Statistical detection of EEG synchrony using empirical bayesian inference.

Area of Science:

• Neuroscience
• Cognitive Science
• Signal Processing

Background:

• Synchronized brain oscillations integrate information across brain regions.
• Phase-locking value (PLV) quantifies neural synchrony but faces multiple testing issues.
• Standard methods like False Discovery Rate (FDR) lack power in high-dimensional neuroimaging data.

Purpose of the Study:

• To introduce and validate the local FDR (locFDR) method for analyzing PLV synchrony in EEG data.
• To compare the performance of locFDR against existing methods like hierarchical FDR and optimal discovery procedures.
• To enhance statistical power in detecting neural synchrony while controlling false positives.

Main Methods:

• Application of Empirical Bayes-based local FDR (locFDR) for PLV synchrony analysis.
• Validation using Monte Carlo simulations to assess specificity and sensitivity.
• Experimental validation on a real EEG dataset from a visual search task.

Main Results:

• locFDR effectively controls false positives without sacrificing statistical power in PLV inference.
• locFDR identified more significant synchrony discoveries compared to hierarchical FDR and optimal discovery procedures.
• Standard FDR methods failed to detect significant discoveries in the experimental dataset.

Conclusions:

• Empirical Bayes-based locFDR offers a powerful and sensitive approach for analyzing neural synchrony using PLV in EEG.
• locFDR overcomes limitations of standard multiple testing procedures in high-dimensional neuroimaging data.
• This method advances the understanding of brain information integration through synchronized oscillatory activity.