NBR

NBR extends Network-Based Statistics by integrating non-linear mixed-effects (LME) models and permutation testing to analyze longitudinal brain network data and detect subnetworks associated with covariates in resting-state fMRI connectivity matrices.

Key Features:

• Network-Based Statistics (NBS) foundation: Identifies clusters of connections in brain networks and evaluates their significance through permutation tests, following Zalesky et al. (2010).
• Non-linear mixed-effects (LME) models: Models within-subject variance and offers flexibility for handling missing data in longitudinal designs.
• Support for unbalanced longitudinal datasets: Accommodates varying numbers of observations per subject while preserving within-subject variance estimation.
• Whole-network connection testing: Examines every possible connection within a sample of networks using the NBS framework.
• Addresses GLHT limitations: Mitigates underestimation of within-subject variance associated with General Linear Hypothesis Testing (GLHT) in longitudinal analyses.
• Application to resting-state fMRI and psychometrics: Operates on connectivity matrices and relates them to psychometric measures such as state anxiety scores.
• Subnetwork detection across brain regions: Identifies subnetworks involving regions including the cingulum, frontal, parietal, occipital, and cerebellum.
• R implementation: Provided as an R package for application to neuroimaging connectivity data.

Scientific Applications:

• Longitudinal network neuroscience: Improved detection of connectivity changes over time in longitudinal neuroimaging studies with unbalanced samples.
• Psychometric–connectivity associations: Identification of subnetworks associated with state anxiety using resting-state fMRI connectivity matrices.
• Studies of treatment, development, and aging: Analysis of network dynamics across conditions such as treatment response, development, or aging where observations may be incomplete.
• Cross-lobar subnetwork mapping: Mapping subnetworks spanning cingulum, frontal, parietal, occipital, and cerebellum regions.

Methodology:

Extends the NBS framework (Zalesky et al., 2010) by replacing GLHT with non-linear mixed-effects (LME) models to represent within-subject variance and using permutation tests to assess significance of clusters of connections across connectivity matrices.