This repository stores code and analyses for our recent commentary in ImagingNeuroscience entitled

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

Read the paper here!

• This works builds upon Zhang et al (2023) and tests RELIEF´s performance in the ABCD study.

This code was performed in RStudio (R version 4.2.3) and python (version 3.9.16).

The following main packages were used

• neuroCombat version 1.0.13 in R see
• RELIEF version 0.1.0 in R see
• CovBat version 0.1.0 in R see
• skicit-learn version 1.3.2. in python

• ABCD_Harmonization_.R performs data loading, handling and harmonization procedure with ComBat and RELIEF - we perform the harmonization in a controlled and naturalistic setting
• ABCD_ROCAUC_Comparison_Fig1.py investigates scanner classification performance from (un)-harmonized data - comparisons are in controlled / naturalistic setting
• ABCD_SampleInflue_controlled.py investigates sample size influence on harmonization performance - only in controlled setting
• ABCD_BioML_Table1.py investigates the harmonization technique´s ability to retain signal related to covariates + provides demographics

Zhang, R., Oliver, L. D., Voineskos, A. N., & Park, J. Y. (2023). RELIEF: A structured multivariate approach for removal of latent inter-scanner effects. Imaging Neuroscience (Cambridge, Mass.), 1, 1–16. https://doi.org/10.1162/imag_a_00011