Motivation: Analyzing diffusion MRI (dMRI) metrics across multiple sites is challenging due to variability between scanners. Goal(s): To investigate whether unified reconstruction algorithms have an impact on the inter-scanner variability in diffusion MRI. Approach: Vendor-provided dMRI sequence was used on two scanners (Siemens, GE), and two standard reconstruction algorithms were evaluated to assess their effects on the reproducibility of diffusion microstructure measures and tractography. Results: Harmonizing dMRI reconstruction across vendors did not significantly reduce inter-scanner variability in tractography measures, including fiber counts, and tract volumetric overlap. But, the Split-GRAPPA method enhanced the reproducibility of fractional anisotropy (FA). Impact: This study shows that inter-scanner variability in diffusion MRI cannot be mitigated using a unified reconstruction algorithm if the acquisition sequences themselves are different. Prospective harmonization of dMRI may benefit more from harmonizing the sequence itself.
Liu et al. (Tue,) studied this question.