Multimodal Brain Parcellation Based on Functional and Anatomical Connectivity.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Mapping

Background:

• Brain parcellation is crucial for neuroimaging network analysis, but current unimodal methods (primarily fMRI) overlook diverse brain attributes.
• High-dimensional neuroimaging data presents statistical, computational, and interpretational challenges, necessitating refined parcellation strategies.

Purpose of the Study:

• To develop and validate a novel multimodal brain parcellation approach integrating functional MRI (fMRI) and diffusion MRI (dMRI) data.
• To enhance the reproducibility, functional homogeneity, and predictive accuracy of brain parcellations compared to unimodal methods.

Main Methods:

• Proposed a nonlinear mapping between fMRI and dMRI connectivity values, adaptively weighted by voxel-wise test-retest reliability.
• Introduced an efficient region-level extension incorporating gyral and sulcal structural information.
• Validated the multimodal parcellations against unimodal parcellations and existing atlases using Human Connectome Project data.

Main Results:

• Multimodal parcellations demonstrated superior reproducibility and comparable or higher functional homogeneity and left-out data likelihood.
• Parcellation boundaries aligned with cyto-architectural features, and extracted subnetworks corresponded well with known brain systems.
• The integration of multimodal data significantly improved the quality and reliability of brain parcellation.

Conclusions:

• Multimodal brain parcellation, integrating fMRI and dMRI, offers significant advantages over unimodal approaches.
• This method enhances the accuracy of brain network analysis and provides a more comprehensive understanding of brain organization.
• The findings support the use of multimodal data for robust and reliable brain mapping.