Improving Between-Group Effect Size for Multi-Site Functional Connectivity Data via Site-Wise De-Meaning.

Area of Science:

• Neuroimaging
• Psychiatric Disorders
• Brain Connectivity

Background:

• Multi-site functional MRI (fMRI) databases are crucial for studying neurodevelopmental and psychiatric disorders.
• Site effects in multi-site fMRI data can confound functional connectivity (FC) measures.
• Existing methods to mitigate site effects often reduce the effect size of FC comparisons.

Purpose of the Study:

• To introduce and evaluate a novel site-wise de-meaning (SWD) strategy for multi-site FC analysis.
• To compare the performance of SWD against generalized linear model (GLM) and ComBat harmonization.
• To assess the preservation of effect sizes in FC alterations associated with psychiatric disorders.

Main Methods:

• Developed a site-wise de-meaning (SWD) strategy involving removal of the site-wise FC mean.
• Compared SWD with GLM and ComBat harmonization on two multi-site datasets: ABIDE and B-SNIP.
• Evaluated method performance by measuring the effect sizes (Hedge's g) of FC differences between patients and controls.

Main Results:

• SWD improved effect sizes in the B-SNIP dataset (schizophrenia) by 4.5-7.9%, while GLM and ComBat decreased them by 22.5-42.6%.
• SWD improved effect sizes in the ABIDE dataset (autism) by 2.9-5.3%, while GLM and ComBat decreased them by up to 11.4%.
• SWD demonstrated superior performance in preserving effect sizes compared to original data and other methods.

Conclusions:

• The site-wise de-meaning (SWD) method is effective in mitigating site effects in multi-site fMRI studies.
• SWD superiorly preserves effect sizes in functional connectivity alterations related to psychiatric disorders.
• This method offers a promising approach for robust analysis of multi-site neuroimaging data in clinical research.