An Information-Theoretic Framework to Measure the Dynamic Interaction Between Neural Spike Trains

Area of Science:

• Computational Neuroscience
• Information Theory
• Point Process Analysis

Background:

• Understanding neuronal interactions from spike train data is crucial in neuroscience.
• Existing methods often fail to account for the point-process nature of spike trains or rely on restrictive parametric assumptions.

Purpose of the Study:

• To develop a model-free, continuous-time information-theoretic framework for estimating interactions between spike trains.
• To quantify both undirected (symmetric) and directed (Granger-causal) interactions.

Main Methods:

• The framework computes the mutual information rate (MIR) and transfer entropy rate (TER) for point processes.
• Theoretical expressions for MIR and TER are derived.
• Efficient estimation strategies using nearest neighbor statistics are introduced.

Main Results:

• Simulations demonstrate the accuracy of MIR and TER in assessing interactions, even for weak coupling and short datasets.
• Continuous-time estimation shows superiority over standard discrete-time methods.
• Application to real-world neuronal culture data reveals the emergence of functional networks during maturation.

Conclusions:

• The proposed framework offers principled, efficient, and flexible measures for assessing spike train interactions.
• It surpasses previous discrete-time and parametric approaches.
• Opens new avenues for analyzing point-process data in neuroscience and beyond.