Statistical harmonization corrects site effects in functional connectivity measurements from multi-site fMRI data.

Area of Science:

• Neuroimaging
• Neuroscience
• Medical Imaging

Background:

• Multi-site resting-state functional MRI (fMRI) studies enhance statistical power but suffer from nonbiological variability due to scanner differences.
• This variability, or site effect, can obscure true findings and lead to erroneous conclusions in neuroimaging research.
• Existing methods have not adequately addressed the removal of these site effects in fMRI data.

Purpose of the Study:

• To investigate the impact of site effects on functional connectivity and network measures in multi-site fMRI data.
• To evaluate the effectiveness of the ComBat harmonization technique in removing site effects from fMRI connectivity measures.
• To assess whether ComBat harmonization can improve the detection of age-related associations in multi-site fMRI studies.

Main Methods:

• Utilized a large, multi-site (4 sites) resting-state fMRI dataset with homogenized acquisition protocols.
• Assessed site effect magnitudes across different functional connectivity metrics and brain atlases.
• Applied the ComBat harmonization technique to remove identified site effects.
• Examined the impact of ComBat on the power to detect age associations.

Main Results:

• Site effects significantly impacted functional connectivity and network measures, with variability depending on the chosen metric and brain atlas.
• ComBat harmonization successfully removed site-specific variability in connectivity and network measures.
• ComBat application, combined with optimal metric/atlas choices, enhanced the statistical power to detect age associations.

Conclusions:

• The ComBat technique is effective in harmonizing multi-site fMRI data by removing site effects from connectivity and network measures.
• This harmonization approach increases the reliability and efficiency of analyzing multi-site fMRI neuroimaging studies.
• ComBat facilitates more robust detection of neurobiological effects, such as age associations, in large-scale fMRI datasets.