Removing scanner effects with a multivariate latent approach: A RELIEF for the ABCD imaging data?

Area of Science:

• Neuroimaging analysis
• Developmental neuroscience
• Data harmonization techniques

Background:

• Scan site harmonization is essential for multi-site neuroimaging studies.
• Established methods exist, but new techniques like RELIEF (REmoval of Latent Inter-scanner Effects through Factorization) aim to improve performance.
• The Adolescent Brain and Cognitive Development (ABCD) study is a valuable resource for developmental brain imaging.

Purpose of the Study:

• To evaluate the utility of the RELIEF harmonization method on data from the ABCD study.
• To compare RELIEF's performance against unharmonized, ComBat, and CovBat methods.
• To investigate the influence of manufacturer type, sample size, and age range on harmonization effectiveness.

Main Methods:

• Benchmarking RELIEF against unharmonized, ComBat, and CovBat data.
• Analysis of harmonization performance across different site sample sizes and manufacturer types.
• Utilizing data from the Adolescent Brain and Cognitive Development (ABCD) study.

Main Results:

• RELIEF outperformed other methods when harmonizing sites with sufficiently large samples.
• Substantial performance variation was observed with RELIEF when including sites with very small samples.
• Harmonization effectiveness was influenced by sample size and data imbalance.

Conclusions:

• RELIEF demonstrates effectiveness in neuroimaging data harmonization, particularly with larger sample sizes.
• Careful quality control is crucial when applying RELIEF to datasets with imbalanced sample sizes, such as the ABCD cohort.
• Shared scripts and insights can guide researchers in applying best practices for ABCD data harmonization.