Related Experiment Video
Updated: Sep 5, 2025
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
Cross-site harmonization of multi-shell diffusion MRI measures based on rotational invariant spherical harmonics
Alberto De Luca1, Suheyla Cetin Karayumak2, Alexander Leemans3
1Department of Neurology, UMC Utrecht Brain Center, University Medical Center Utrecht, Utrecht, the Netherlands; PROVIDI Lab, Image Sciences Institute, University Medical Center Utrecht, Utrecht, the Netherlands.
Rotation invariant spherical harmonics (RISH) harmonization effectively removes cross-site differences in multi-shell diffusion magnetic resonance imaging (dMRI) metrics like DKI and NODDI. This technique preserves longitudinal effects, enabling robust multicenter brain studies.
Area of Science:
- Neuroimaging
- Diffusion Magnetic Resonance Imaging (dMRI)
- Computational Neuroscience
Background:
- Multi-shell dMRI enables advanced models like Diffusion Kurtosis Imaging (DKI) and Neurite Orientation Dispersion and Density Imaging (NODDI) for in-vivo brain studies.
- Scanner and site variations in DKI and NODDI metrics hinder multicenter studies, limiting sample sizes and robust analysis.
- Existing harmonization methods require repeated application for each dMRI metric.
Purpose of the Study:
- To investigate the efficacy of Rotation Invariant Spherical Harmonics (RISH) features harmonization in mitigating cross-site differences for multi-shell dMRI data.
- To assess whether RISH harmonization preserves longitudinal effects in DKI and NODDI metrics.
- To evaluate the generalizability of RISH harmonization to advanced dMRI models beyond Diffusion Tensor Imaging (DTI).
Main Methods:
- Acquisition of two-shell dMRI data from 46 subjects across 3 imaging sites with a longitudinal design (up to 3 time points).
- Derivation of DKI and NODDI metrics before and after RISH features harmonization.
- Comparison of metrics at whole-brain and voxel levels to evaluate cross-site differences and longitudinal effects.
Main Results:
- RISH harmonization effectively removed statistically significant between-site differences in DKI metrics (FA, MD, MK) and NODDI metrics, particularly in white matter.
- The RISH method maintained the magnitude and variance of longitudinal changes across all considered metrics.
- Prominent between-site differences observed in frontal and posterior brain regions were successfully mitigated by RISH harmonization.
Conclusions:
- RISH feature-based harmonization is a robust method for removing cross-site variability in multi-shell dMRI data for DKI and NODDI analyses.
- This technique facilitates the implementation of large-scale, multicenter dMRI studies by ensuring data consistency across different scanners and sites.
- RISH harmonization preserves the integrity of longitudinal data, allowing for reliable tracking of brain changes over time.
