Related Experiment Video
Updated: Jun 27, 2026
17:06
Co-analysis of Brain Structure and Function using fMRI and Diffusion-weighted Imaging
Published on: November 8, 2012
26.8K
Inter-site harmonization based on dual generative adversarial networks for diffusion tensor imaging: application to
Jie Zhong1,2, Ying Wang3, Jie Li2
1Department of Radiology, The First Affiliated Hospital of Xi'an Jiaotong University, Xi'an, 710061, China.
Biomedical Engineering Online
|January 17, 2020
Summary
This study introduces a dual generative adversarial network (GAN) method to harmonize diffusion tensor imaging (DTI) data across different research sites. This approach effectively reduces site-specific variations in neonatal brain imaging metrics, improving data pooling for multi-center studies.
Area of Science:
- Medical Imaging
- Neuroscience
- Data Science
Background:
- Site-specific variations in multi-center studies pose challenges for data pooling.
- Diffusion Tensor Imaging (DTI) derived metrics are sensitive to inter-site differences.
- Harmonization is crucial for reliable analysis of neonatal brain DTI data.
Purpose of the Study:
- To propose and evaluate an inter-site harmonization method for neonatal brain DTI metrics.
- To utilize dual generative adversarial networks (GANs) for addressing site-specific variations.
- To improve the feasibility of pooling analyses in multi-center neonatal neuroimaging studies.
Main Methods:
- Acquired DTI-derived metrics (FA, MD) from age-matched neonates across two sites.
- Applied a proposed dual GANs-based harmonization approach.
- Compared the dual GANs method with conventional scaling and ComBat harmonization techniques.
Main Results:
- The dual GANs method effectively removed inter-site differences in DTI metrics.
- The proposed method demonstrated lower median absolute and root mean square errors compared to conventional methods.
- Harmonization improved the correlation between FA and postmenstrual age and maintained effect sizes.
Conclusions:
- The dual GANs-based harmonization method is effective for neonatal DTI metrics from different sites.
- GANs-based harmonization is a feasible pre-processing step for multi-center DTI pooling analyses.
- This technique enhances the reliability and utility of multi-site neonatal brain imaging data.
