Assessing directed information as a method for inferring functional connectivity in neural ensembles

Area of Science:

• Neuroscience
• Information Theory
• Computational Biology

Background:

• Neural networks are complex, with functional connectivity traditionally assessed by correlation, which lacks directionality.
• Effective connectivity models aim to capture the direction of influence between neurons.
• Directed information, based on Granger causality, is a proposed measure for neural directionality.

Purpose of the Study:

• To introduce a novel estimation procedure for directed information in neural spike train analysis.
• To evaluate the performance of directed information against traditional correlation methods.
• To demonstrate the capability of directed information in assessing directional neural relationships.

Main Methods:

• Modeling neural spike trains using point process generalized linear models.
• Developing and applying a new estimation procedure for directed information.
• Utilizing physiologically realistic simulations for comparative analysis.

Main Results:

• The proposed directed information estimation method successfully captures directional relationships between neurons.
• Directed information demonstrates superior performance compared to correlation in identifying neural connections.
• The study confirms directed information's ability to provide directionality, a feature absent in correlation.

Conclusions:

• Directed information provides a more comprehensive understanding of neural connectivity than correlation.
• The novel estimation procedure enhances the utility of directed information in neuroscience.
• This approach advances the modeling of effective connectivity in the brain.