Multimodal subspace independent vector analysis effectively captures latent relationships between brain structure and...

Area of Science:

• Neuroscience
• Medical Imaging
• Biostatistics

Background:

• Inferring brain structure-function relationships from high-dimensional, multimodal neuroimaging data is challenging.
• Conventional methods often oversimplify statistical dependencies within and between modalities.
• Existing approaches typically assume shared components across subjects, limiting individual variability capture.

Purpose of the Study:

• Introduce Multimodal Subspace Independent Vector Analysis (MSIVA) for advanced neuroimaging data analysis.
• Capture complex, multi-dimensional statistical dependencies within and between modalities.
• Enable subject-level variability analysis at the voxel level within identified subspaces.

Main Methods:

• Developed MSIVA to define cross-modal and unimodal subspaces with variable dimensions.
• Enabled flexible estimation of independent subspaces and their cross-modal linkages.
• Validated MSIVA using synthetic datasets and large multimodal neuroimaging (sMRI, fMRI) datasets.

Main Results:

• MSIVA successfully recovered ground-truth subspace structures in synthetic data.
• MSIVA sources showed strong associations with phenotype variables: age, sex, schizophrenia, lifestyle, and cognition.
• Identified modality- and group-specific brain regions linked to age, sex, and schizophrenia.

Conclusions:

• MSIVA offers a flexible framework for analyzing multimodal neuroimaging data.
• The method effectively captures subject-level variability and complex inter-modal relationships.
• MSIVA aids in identifying linked phenotypic and neuropsychiatric biomarkers of brain structure and function.