Modeling brain connectivity dynamics in functional magnetic resonance imaging via particle filtering

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biophysics

Background:

• Functional brain connectivity alterations are linked to neurological diseases.
• Traditional methods often assume static brain networks, overlooking dynamic changes.
• Dynamic connectivity patterns offer valuable insights into brain physiology and pathology.

Purpose of the Study:

• To apply a Particle Filter (PF) computational methodology for analyzing non-stationarities in functional Magnetic Resonance Imaging (fMRI) brain connectivity.
• To investigate dynamic changes in brain network interactions.
• To explore the utility of PF in understanding brain function and disease markers.

Main Methods:

• Utilized a Particle Filter (PF) algorithm employing a Sequential Monte Carlo strategy.
• Estimated time-varying parameters of a first-order linear time-varying Vector Autoregressive (VAR) model.
• Applied the PF method to both simulated and real fMRI data.

Main Results:

• The PF approach successfully detected and tracked time-varying parameters in simulated data, capturing causal relationships.
• Analysis of real fMRI data, including periodic visual stimulation, showed PF estimates aligned with known brain functioning.
• Statistically significant modulations in cause-effect relationships between brain areas were detected, correlating with visual stimulation.

Conclusions:

• The Particle Filter (PF) is a robust method for studying dynamic brain connectivity in fMRI.
• This approach can reveal non-stationary interactions crucial for understanding brain function and neurological disorders.
• PF provides a powerful tool for detecting stimulus-induced changes in brain network causality.