Diffusion MRI sampling schemes bias diffusion metrics and tractography.

Area of Science:

• Neuroimaging
• Diffusion MRI
• White Matter Anatomy

Background:

• Diffusion MRI is vital for studying white matter architecture.
• Tractography and diffusion metrics can be influenced by varying sampling schemes.
• Understanding these differences is crucial for reliable data interpretation.

Purpose of the Study:

• To assess systematic differences in diffusion MRI metrics and tractography across four common protocols: single-shell HARDI, Siemens multi-shell (Sms), DSI, and HCP multi-shell (HCPms).
• To evaluate these differences in healthy adults and individuals with corpus callosum dysgenesis (CCD).

Main Methods:

• Acquired data on a 3T scanner and processed uniformly.
• Extracted fractional anisotropy (FA), mean diffusivity (MD), effective contrast-to-noise ratio (eCNR), and orientation dispersion.
• Measured tract volumes and streamline counts in specific white matter regions (CC, CR, CSO, AC, PC) and CCD-specific bundles (Probst, sigmoid).

Main Results:

• FA and MD showed moderate correlations across schemes, but matched means were infrequent.
• eCNR and dispersion had limited cross-scheme correlations, with some similarities between Sms/HCPms and HARDI/DSI/HCPms.
• Tract volumes correlated across Sms, DSI, and HCPms for CC (controls) and CCD bundles; DSI and HCPms showed consistent volumes.
• Sms volumes agreed with DSI/HCPms in CR but were lower in CC and CCD ROIs; HARDI produced higher volumes in CC, CR, and CCD ROIs.
• AC and PC tract metrics were consistent across schemes, but correlations varied.

Conclusions:

• Four diffusion-sampling schemes exhibit systematic differences in voxel-wise metrics and tractography outcomes.
• Findings inform efforts to consolidate or contrast data across schemes.
• Future research should explore harmonization methods to reduce bias and enable pooled analyses across diverse protocols.