A strategy of model space search for dynamic causal modeling in task fMRI data exploratory analysis

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Systems Neuroscience

Background:

• Dynamic Causal Modeling (DCM) is crucial for estimating effective connectivity (EC) in neuroimaging.
• Traditional DCM requires pre-defined network structures, limiting exploratory analysis.
• Existing methods face challenges in reducing computational cost and model space for broader application.

Purpose of the Study:

• To develop a novel model space exploration strategy for DCM.
• To enhance the exploratory capabilities of DCM in neuroimage analysis.
• To reduce the complexity and improve the accessibility of DCM for researchers.

Main Methods:

• Proposed a model space exploration strategy with two algorithms: GreedyEC (reducing EC from a full model) and GreedyROI (adding EC from a single node model).
• Applied GreedyEC and GreedyROI to task-based functional magnetic resonance imaging (fMRI) data from a visual object recognition study.
• Utilized Bayesian model comparison to select the optimal DCM model.

Main Results:

• Combining GreedyEC and GreedyROI algorithms improved the effectiveness of DCM exploratory analysis.
• The proposed strategy successfully identified the best DCM model from a reduced model space.
• The algorithms were implemented as MATLAB functions within the SPM framework for user-friendliness.

Conclusions:

• The developed model space exploration strategy enhances DCM's utility for exploratory brain network analysis.
• This approach facilitates the identification of optimal causal models and lowers the barrier to DCM implementation.
• The tool aids neuroscientists in analyzing brain causal information flow networks more efficiently.