Cross-population coupling of neural activity based on Gaussian process current source densities.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Local field potentials (LFPs) are complex signals reflecting activity from multiple neural sources.
• Disentangling these sources is crucial for understanding coordinated neural population activity on a trial-by-trial basis.
• Existing methods struggle to resolve fine-grained, cross-population correlations.

Purpose of the Study:

• To introduce a novel framework for estimating current source densities (CSDs) to decouple neural signals.
• To improve the identification and analysis of trial-to-trial fluctuations in neural activity.
• To reveal cross-population correlations previously obscured in LFPs.

Main Methods:

• Developed a general framework for current source density (CSD) estimation incorporating biophysical forward models.
• Modeled CSD as a spatiotemporal Gaussian process with fast/slow components and trial-varying mean functions.
• Validated the approach using simulation studies and analyzing electrophysiological data from primate auditory cortex and mouse visual areas.

Main Results:

• The CSD estimation framework successfully improved source signal identification compared to existing methods.
• Discovered cortical layer-specific phase coupling between probes in primate auditory cortex, not evident in LFPs.
• Observed correlated task-evoked CSDs across probes at specific cortical depths.
• Found evidence for depth-specific phase coupling in mouse visual areas using Neuropixels data.

Conclusions:

• Decoupling neural sources via CSD estimation enhances the detection of coordinated neural activity.
• This method reveals layer- and depth-specific neural communication patterns.
• The framework offers a powerful tool for analyzing complex neural population dynamics across different species and brain regions.