Inferring Neural Communication Dynamics from Field Potentials Using Graph Diffusion Autoregression

Area of Science:

• Neuroscience
• Computational Neuroscience
• Network Science

Background:

• Estimating dynamic network communication is crucial for understanding cognitive processes.
• Traditional methods for inferring neural communication have limitations, including lack of biological plausibility, neglect of spatial information, and static estimates.
• Advancements in multi-site neural recording technologies necessitate improved methods for dynamic network analysis.

Purpose of the Study:

• To introduce a novel graph diffusion autoregressive model for estimating dynamic network communication.
• To address limitations of traditional methods in modeling biologically plausible neural interactions and capturing rapid communication dynamics.
• To provide a high-resolution communication signal from distributed field potential recordings.

Main Methods:

• Developed a graph diffusion autoregressive model combining vector autoregression with a network communication process.
• Designed the model for distributed field potential recordings.
• Validated the model on simulated neural activity and in vivo recordings from macaque sensorimotor cortex.

Main Results:

• Successfully validated the model on simulated data and macaque neural recordings.
• Demonstrated the model's ability to describe rapid communication dynamics induced by optogenetic stimulation.
• Showcased the model's capacity to capture changes in resting state communication and trial-by-trial variability during a reach task.

Conclusions:

• The graph diffusion autoregressive model offers a powerful tool for estimating dynamic brain communication.
• This novel approach overcomes key limitations of traditional methods, providing high-resolution and biologically plausible network estimates.
• The model has significant potential for advancing our understanding of neural dynamics in various cognitive states and tasks.