Analysis of task-related MEG functional brain networks using dynamic mode decomposition

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Functional brain networks underpin diverse motor, cognitive, and sensory functions.
• Understanding the temporal dynamics of these networks is key to deciphering brain states during behavior.
• Extracting network configurations and their evolution is essential for advancing neuroscience.

Purpose of the Study:

• To introduce Dynamic Mode Decomposition (DMD) as a method for extracting brain network dynamics.
• To compare the efficacy of DMD with Principal Component Analysis (PCA) for analyzing brain network temporal evolution.
• To investigate the ability of these methods to capture network dynamics across various behavioral tasks.

Main Methods:

• Utilized Dynamic Mode Decomposition (DMD) to analyze brain network dynamics.
• Employed Principal Component Analysis (PCA) as a comparative method.
• Applied both methods to real magnetoencephalography (MEG) data acquired during motor and memory tasks.

Main Results:

• The proposed framework successfully generated dominant connectivity brain networks and their time dynamics for both simple and complex tasks.
• Both PCA-based and DMD-based approaches yielded comparable dominant connectivity networks.
• The temporal dynamics extracted by PCA and DMD showed strong similarity.

Conclusions:

• The methodology, utilizing both PCA and DMD, effectively deciphers spatiotemporal dynamics of electrophysiological brain network states.
• This approach holds significant potential for understanding brain function during various tasks.
• The findings support the utility of DMD and PCA in neuroscientific research for network analysis.