Brain Network Analysis: A Review on Multivariate Analytical Methods

Area of Science:

• Neuroscience
• Complex Systems Analysis
• Data Science

Background:

• Neuroimaging studies face methodological gaps in analyzing the brain as a complex system.
• Current brain network analysis tools are often univariate and ill-suited for big, complex brain data.
• Existing graph-based methods have limitations, and principled multivariate models for brain network analysis are underdeveloped.

Purpose of the Study:

• To review and categorize important multivariate methods for brain network analysis.
• To address the challenge of selecting the most appropriate multivariate method for specific research questions.
• To aid investigators in choosing methods based on network type, data characteristics, and analytical goals (connectivity vs. topology).

Main Methods:

• Categorization of multivariate methods into data-driven and model-based approaches.
• Discussion of method suitability for analyzing connectivity (edge-level) and topology (system-level).
• Consideration of factors influencing method selection, including network size, number of subjects, and brain regions.

Main Results:

• Multivariate methods offer a promising avenue to overcome limitations of univariate and graph-based approaches.
• A framework is presented to help researchers navigate the selection of appropriate multivariate techniques.
• The review highlights the need for accessible guidelines in this multidisciplinary field.

Conclusions:

• Choosing the right multivariate method is crucial for advancing brain network analysis.
• This review provides a valuable resource for neuroimaging researchers to select appropriate analytical tools.
• Dissemination of advanced analytical tools is key to improving our understanding of human health through biomedical data.