A tutorial on data-driven methods for statistically assessing ERP topographies.

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Computational Neuroscience

Background:

• Event-related potentials (ERPs) reveal cognitive processes through dynamic topographical changes.
• Microstates, brief periods of stable brain network activation, are key to analyzing these ERP dynamics.
• Current methods for microstate analysis suffer from reduced statistical power due to noise in individual data assignments.

Purpose of the Study:

• To develop a novel randomization-based procedure for analyzing microstate temporal features.
• To introduce a cross-validation criterion for selecting the optimal number of microstate prototypes.
• To improve the statistical power and sensitivity in assessing cognitive process timing.

Main Methods:

• A randomization-based procedure is proposed, avoiding the assignment of grand-mean microstate prototypes to individual data.
• A new cross-validation criterion is introduced for selecting the optimal number of microstate prototypes across subjects.
• The method was applied to N400 experiment data and simulated data with varying signal-to-noise ratios.

Main Results:

• The new method demonstrated increased robustness to noise compared to existing procedures.
• It showed higher sensitivity in detecting subtle effects of microstate timing.
• The approach proved effective even with simulated data exhibiting low signal-to-noise ratios.

Conclusions:

• The proposed method is well-suited for assessing timing differences in cognitive processes.
• Its increased statistical power aids in identifying subtle effects, crucial for small patient populations.
• This technique offers a more precise and sensitive tool for neuroscientific research.