HARMONIZATION MITIGATES DIFFUSION MRI SCANNER EFFECTS IN INFANCY: INSIGHTS FROM THE HEALTHY BRAIN AND CHILD...

Area of Science:

• Neuroscience
• Developmental Neuroscience
• Medical Imaging

Background:

• Large-scale longitudinal studies like the HEALthy Brain and Childhood Development (HBCD) Study aim to understand brain maturation.
• Site-related variance in neuroimaging data can obscure biologically relevant signals.
• Diffusion MRI data, including diffusion tensor imaging (DTI) metrics, are crucial for studying white matter development.

Purpose of the Study:

• To characterize site-specific effects in diffusion MRI data within the HBCD Study.
• To evaluate the impact of scanner model variance on HBCD bundle metrics.
• To implement and assess a harmonization method for HBCD diffusion MRI data.

Main Methods:

• Investigated sensitivity of HBCD bundle metrics to scanner model variance.
• Applied ComBat-GAM harmonization to diffusion MRI data from HBCD data release 1.1.
• Analyzed data across six different scanner models.

Main Results:

• ComBat-GAM harmonization effectively reduced scanner model-related variance.
• Zero statistically significant differences were observed between scanner model distributions post-harmonization (FDR corrected).
• Reduced effect sizes (Cohen's f) across all diffusion MRI metrics.

Conclusions:

• Rigorous harmonization is essential for large-scale neuroimaging studies.
• ComBat-GAM is an effective method for mitigating site effects in HBCD diffusion MRI data.
• Future HBCD data analyses should account for these harmonization efforts to ensure reliable findings.