A Bayesian Double Fusion Model for Resting-State Brain Connectivity Using Joint Functional and Structural Data.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biostatistics

Background:

• Current neuroimaging analyses often treat diffusion tensor imaging (DTI) and functional magnetic resonance imaging (fMRI) data separately.
• This separation limits the potential of DTI-derived structural connectivity (SC) to enhance the estimation of resting-state functional connectivity (FC).

Purpose of the Study:

• To develop a novel Bayesian hierarchical spatiotemporal model that integrates SC from DTI with FC estimation from fMRI.
• To improve the accuracy and robustness of functional connectivity analysis by leveraging multimodal neuroimaging data.

Main Methods:

• A Bayesian hierarchical spatiotemporal model was developed to incorporate structural connectivity (SC) into functional connectivity (FC) estimation.
• Structural connectivity (SC) from DTI data was used to create an informative prior for FC estimation from resting-state fMRI data via Cholesky decomposition.

Main Results:

• Simulation studies demonstrated that the integrated DTI-fMRI approach significantly reduced mean squared errors compared to analyzing data separately.
• Application to temporal lobe epilepsy data showed the model achieves lower false positive rates and greater robustness to data decimation.

Conclusions:

• The proposed model effectively integrates DTI and fMRI data for more accurate and robust resting-state functional connectivity analysis.
• This multimodal approach offers significant advantages over conventional methods, particularly in complex neurological conditions.