Dynamical differential covariance recovers directional network structure in multiscale neural systems.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Understanding neural interactions is crucial for deciphering the neural basis of behavior.
• Current statistical methods for analyzing neural activity struggle with accurately and efficiently estimating directional network interactions, especially under non-stationary conditions.

Purpose of the Study:

• To derive and validate a novel method, dynamical differential covariance (DDC), for detecting directional neural interactions.
• To assess DDC's performance in terms of bias, noise tolerance, scalability, and computational efficiency compared to existing methods.

Main Methods:

• Dynamical differential covariance (DDC) was derived based on dynamical network models.
• DDC was validated using simulated neural networks with known ground-truth connectivity, including those with false positive motifs.
• The method was applied to resting-state functional magnetic resonance imaging (rs-fMRI) data from over 1,000 subjects, with comparisons to diffusion MRI (dMRI) for structural connectivity.

Main Results:

• DDC demonstrated low bias and high noise tolerance, even under non-stationarity.
• The method scales well with the number of recording sites, requiring computation comparable to standard covariance methods.
• DDC successfully identified regional interactions with strong structural connectivity in rs-fMRI data.

Conclusions:

• Dynamical differential covariance (DDC) is a robust and efficient method for estimating directional neural connectivity.
• DDC shows promise for analyzing complex neural systems and can be generalized to various dynamical models and recording techniques.
• This method offers a valuable tool for system identification in neuroscience and beyond.