Effective connectivity: influence, causality and biophysical modeling.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Brain Imaging

Background:

• Functional magnetic resonance imaging (fMRI) is crucial for understanding brain networks.
• Identifying effective connectivity in the brain presents significant modeling challenges.
• Previous approaches to causal modeling in neuroscience have limitations.

Purpose of the Study:

• To discuss model selection, causality, and deconvolution in fMRI-based brain network identification.
• To advocate for state-space models with biophysically informed equations for effective connectivity.
• To compare Dynamic Causal Modeling (DCM) and Granger Causal Modeling (GCM) within a broader causal inference framework.

Main Methods:

• Utilizing state-space models with biophysically informed observation and state equations.
• Incorporating prior information for unknown model parameters.
• Performing identifiability checks for model validation.
• Comparing Dynamic Causal Modeling and Granger Causal Modeling.

Main Results:

• Biophysically informed state-space models are critical for discovering effective brain connectivity.
• Priors on parameters and identifiability checks are essential for robust modeling.
• Links are established between Dynamic Causal Modeling, Granger Causal Modeling, and broader statistical causal inference methods.
• Current challenges in the field have promising solutions.

Conclusions:

• Effective connectivity analysis using fMRI necessitates sophisticated, biophysically grounded models.
• The integration of Bayesian dependency graphs and influence measures enhances causal modeling.
• Future developments in computational neuroscience are expected to address remaining challenges in brain network analysis.