Tractography-Based Score for Learning Effective Connectivity From Multimodal Imaging Data Using Dynamic Bayesian...

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Effective connectivity (EC) describes causal influences between brain regions, constrained by anatomical connectivity (AC).
• Current methods often analyze functional MRI (fMRI) and diffusion tensor imaging (DTI) data separately.
• Integrating multimodal neuroimaging data offers a more comprehensive understanding of brain structure-function relationships.

Purpose of the Study:

• To propose a unified probabilistic framework for learning EC by combining fMRI and DTI data.
• To develop a novel, anatomically informed (AI) score for evaluating brain connectivity structures.
• To enhance the accuracy of EC inference through joint modeling of multimodal neuroimaging data.

Main Methods:

• Utilized dynamic Bayesian networks (DBNs) for exploratory learning of EC.
• Developed and implemented an AI score to simultaneously assess DTI and fMRI data fit for connectivity structures.
• Employed the AI score within the DBN structure learning process.

Main Results:

• Synthetic data experiments validated the AI score's effectiveness in structure learning compared to uninformed methods.
• Real-world fMRI-DTI data analysis demonstrated consistent EC inference aligned with existing literature.
• Cross-validation through classification experiments confirmed the practical utility of the proposed framework.

Conclusions:

• The proposed unified framework effectively integrates structural and functional brain data for EC inference.
• The AI score provides a robust method for evaluating connectivity models using multimodal neuroimaging.
• This multimodal approach offers superior reliability for distinguishing between normal and diseased brain states compared to single-modality analyses.