Cross-scanner and cross-protocol diffusion MRI data harmonisation: A benchmark database and evaluation of algorithms.

Area of Science:

• Medical Imaging
• Neuroimaging
• Biophysics

Background:

• Diffusion MRI provides quantitative microstructural insights but suffers from inter-scanner and inter-protocol variability.
• This variability hinders the combination of datasets and reliable quantitative measurements in clinical studies.
• Multi-center research and increased statistical power are limited by data heterogeneity.

Purpose of the Study:

• To establish a benchmark diffusion MRI database for cross-scanner and cross-protocol harmonization method development.
• To evaluate the performance of different harmonization techniques using this database.
• To assess the effectiveness of harmonizing single-shell diffusion MRI data.

Main Methods:

• Acquired diffusion MRI data from the same subjects on three scanners with varying gradient strengths (40, 80, 300 mT/m).
• Utilized both standard and advanced protocols with higher spatial and angular resolution.
• Compared the performance of five distinct cross-scanner/cross-protocol mapping methods.

Main Results:

• Cross-scanner harmonization of single-shell diffusion MRI data can significantly reduce measurement variability.
• The benchmark database serves as a valuable resource for testing harmonization and quality enhancement methods.
• Current harmonization techniques show both promise and limitations in addressing data heterogeneity.

Conclusions:

• Harmonization is crucial for reliable quantitative diffusion MRI across different acquisition settings.
• The developed database facilitates the advancement of diffusion MRI data integration techniques.
• Further development is needed to fully overcome the challenges of inter-scanner and inter-protocol variability.