- An, L., Chen, J., Chen, P., Zhang, C., He, T., Chen, C., Zhou, J., Yeo, B.T., 2022. Goal-specific brain MRI harmonization, NeuroImage, 119570
There is significant interest in pooling magnetic resonance image (MRI) data from multiple datasets to enable mega-analysis. Harmonization is typically performed to reduce heterogeneity when pooling MRI data across datasets. Most MRI harmonization algorithms do not explicitly consider downstream application performance during harmonization. However, the choice of downstream application might influence what might be considered as study-specific confounds. Therefore, ignoring downstream applications during harmonization might potentially limit downstream performance. We propose a goal-specific harmonization framework that utilizes downstream application performance to regularize the harmonization procedure. Three datasets from three different continents with a total of 2787 participants and 10085 anatomical T1 scans were used for evaluation. Using three large-scale datasets, we demonstrated that our approach (gcVAE) compared favorably with existing approaches in terms of preserving relevant biological information, while removing site differences.
Since the whole Github repository is too big, we provide a stand-alone version of only this project and its dependencies. To download this stand-alone repository, visit this link: https://github.com/ThomasYeoLab/Standalone_An2022_gcVAE
If you want to use the code from our lab's other stable projects (other than An2022_gcVAE), you would need to download the whole CBIG repository.
-
To download the version of the code that was last tested, you can either
or
- run the following command, if you have Git installed
git checkout -b An2022_gcVAE v0.23.0-An2022_gcVAE
- Our code uses Python and R, here is the setup:
- The example of our code is detailed in
examples/README.md
- If you have access to ADNI, AIBL and MACC dataset, you can replicate our result using the instructions detailed in
replication/README.md.
- Release v0.23.0 (25/08/2022): Initial release of An2022_gcVAE project
Please contact Lijun An at [email protected] and Thomas Yeo at [email protected]