Information-geometric measures estimate neural interactions during oscillatory brain states

Area of Science:

• Neuroscience
• Computational Neuroscience
• Information Theory

Background:

• Understanding neural information processing requires characterizing functional network structures.
• Information geometry (IG) offers robust methods for estimating neural interactions from spike-train data.
• Real neural activity is often non-stationary and oscillatory, posing challenges for existing analysis methods.

Purpose of the Study:

• To investigate the influence of oscillatory neural activity on Information Geometry (IG) measures.
• To assess the applicability of IG methods to real-world, non-stationary neural data.
• To determine how externally driven and internally induced oscillations affect single- and pairwise-IG measures.

Main Methods:

• Utilized model networks of binary and spiking neurons.
• Simulated two types of oscillatory mechanisms: externally driven and internally induced oscillations.
• Analyzed the impact of oscillations on single- and pairwise-IG measures.

Main Results:

• The single-IG measure showed a linear relationship with the magnitude of external input.
• The pairwise-IG measure demonstrated a linear relationship with the sum of connection strengths between neurons.
• Pairwise-IG measures were found to be independent of oscillation frequency.

Conclusions:

• Information geometry (IG) provides valuable insights into neural interactions during oscillatory network states.
• The findings are consistent with previous results obtained under equilibrium conditions.
• IG is a robust method applicable to dynamic and non-stationary neural activity observed in the brain.