ComBat harmonization

ComBat harmonization reduces batch effects in multi-site diffusion-weighted magnetic resonance imaging (dMRI) data to enable aggregation and comparison of datasets while preserving biological variability.

Key Features:

• Batch effect mitigation: ComBat adjusts for non-biological variability introduced by differences in scanner models, acquisition protocols, and reconstruction settings.
• Multi-site dMRI harmonization: Tailored for harmonizing diffusion-weighted MRI data across multiple sites and scanners in collaborative studies.
• Preservation of biological variability: Designed to maintain inter-subject and between-group biological differences after harmonization, including effects relevant to neuropsychiatric disorders.
• Evaluation of harmonization performance: Performance has been assessed on a dMRI dataset of 542 individuals from five sites with focus on preservation of fractional anisotropy (FA) effect sizes and the influence of pre-processing variations.
• Site-specific challenges: Recognizes that non-linear scanner contributions can compromise harmonization at some sites.
• Pre-processing consistency: Efficacy can be affected by minor differences in pre-processing pipelines, underscoring the need for consistent processing across sites.

Scientific Applications:

• Neuropsychiatric research: Enables pooling and harmonization of dMRI data to study group- and individual-level differences relevant to neuropsychiatric disorders.
• Increased statistical power: Facilitates aggregation of multi-site datasets to increase sample size and detect FA-related effects with greater power.

Methodology:

ComBat employs a statistical approach to adjust for site-specific effects while preserving biological variability; evaluations include assessing preservation of between-group fractional anisotropy (FA) differences and analyzing the impact of minor pre-processing variations on harmonization effectiveness.