Multisite Harmonization of Structural DTI Networks in Children: An A-CAP Study

Area of Science:

• Neuroimaging
• Network Neuroscience
• Pediatric Traumatic Brain Injury

Background:

• Multisite neuroimaging studies require harmonization to manage scanner variability.
• ComBat is a promising tool for harmonizing structural neuroimaging data.
• Its application to graph theory metrics in structural brain connectomes is unexplored.

Purpose of the Study:

• Evaluate ComBat for multisite harmonization of structural brain network analysis.
• Assess ComBat's impact on graph theory metrics derived from diffusion-weighted scans in children.
• Compare matrix harmonization versus parameter harmonization.

Main Methods:

• Utilized diffusion-weighted scans from 484 children in the Advancing Concussion Assessment in Pediatrics (A-CAP) study.
• Constructed 90x90 weighted adjacency matrices using diffusion tensor tractography.
• Applied ComBat harmonization to adjacency matrices (matrix harmonization) or global network metrics (parameter harmonization).

Main Results:

• Scanner effects were significant in unharmonized and matrix-harmonized metrics.
• Parameter harmonization eliminated scanner differences in global network metrics.
• Age showed stronger correlations with harmonized network metrics.
• Both harmonization approaches maintained good within-scanner consistency.

Conclusions:

• Parameter harmonization effectively controls for scanner variability in structural network analysis.
• This method preserves network topology and connectivity weights.
• ComBat harmonization of global network parameters offers optimal results for multisite pediatric neuroimaging data.