Harmonising MRI-based Metrics (QSM) for Motor Neurone Disease

Introduction

This project aims to harmonise multi-site MRI-derived quantitative susceptibility mapping (QSM) data for studying Motor Neurone Disease (MND). The focus is on eliminating batch effects using neuroCombat harmonisation method while preserving biological variation. Additionally, the study seeks to correlate QSM metrics with MND subtypes, phenotypes, and progression.

Requirements

Tools

• R: 4.2 or above
• RStudio: Version 2024.04.1+748 (2024.04.1+748) or above
• QSMxT: 6.4.4 or above
• R Packages:
  • neuroCombat
  • ggseg
  • dplyr
  • ggplot2
• Input Data: Multi-site QSM datasets with batch information and biological covariates.

Project Structure

├── R/                                # R scripts for data processing and harmonisation
│   ├── covariates.R                  # Handles covariate extraction and processing
│   ├── neuroComBat_harmonisation.R   # Implements neuroCombat harmonisation
│   └── subfeatures.R                 # Processes QSM features for ROI-based analysis
├── visualisation_results/            # Directory for storing visualisations and analysis outputs
│   ├── harmonisation_results/        # Visual outputs of harmonised QSM data
│   ├── qsm_features_outlier_detection/ # Plots for detecting outliers in QSM features
│   └── statistical_results/          # Statistical analysis results, such as p-values and test statistics
├── README.md                         # Project documentation

Instructions

1. Set up

• To set up the environment, make sure you have Conda. Use the code below to install with all necessary dependencies:

conda env create -f environment.yml
conda activate <environment_name>

• Install QSMxT framework
• Install neuroComBat R package

2. Preprocessing

Run subfeatures.R to clean and organise qsm data. Run covariates.R to preprocess covariates data including biology and batch information.

3. Harmonisation and Visualisation

Use neuroComBat_harmonisation.R to apply neuroCombat harmonisation and use brain maps and statistical summaries results for visualisation.

Issues and Improvements

Completed

• Implemented neuroCombat harmonisation for QSM data, which successfully eliminated batch effects while preserving biological information.
• Applied some visualisations for regional QSM metrics.

Limitation

The number of data might insufficient. Further research should include larger datasets and explore some more advanced approaches, such as machine learning- and deep learning-based harmonisation methods.

References

• Stewart, A. W., Robinson, S. D., O’Brien, K., Jin, J., Widhalm, G., Hangel, G., ... & Bollmann, S. (2022). QSMxT: Robust masking and artifact reduction for quantitative susceptibility mapping. Magnetic resonance in medicine, 87(3), 1289-1300. https://doi.org/10.1002/mrm.29048
• Fortin, J. P., Cullen, N., Sheline, Y. I., Taylor, W. D., Aselcioglu, I., Cook, P. A., Adams, P., Cooper, C., Fava, M., McGrath, P. J., McInnis, M., Phillips, M. L., Trivedi, M. H., Weissman, M. M., & Shinohara, R. T. (2018). Harmonization of cortical thickness measurements across scanners and sites. Neuroimage, 167, 104-120. https://doi.org/10.1016/j.neuroimage.2017.11.024
• Jfortin1. neuroCombat_Rpackage. GitHub https://github.com/raoyongming/GFNet
• ggseg. ggseg. GitHub https://github.com/ggseg/ggseg
• ggseg. Using geom_brain. GitHub https://ggseg.github.io/ggseg/articles/ggseg.html