Robustly Coupling Motion Estimates in 4D Diffusion MRI

Develop a method for robustly coupling motion estimates across multiple diffusion encodings and slice groups in diffusion-weighted imaging, particularly under sparse gradient sampling and severe subject motion.

Background

Diffusion-weighted imaging requires multiple measurements acquired with different diffusion encodings, and subject motion can cause misalignment between slice groups and volumes. Because diffusion-direction-dependent contrast varies across measurements, motion estimation must be coordinated across both spatial and q-space dimensions. The paper frames this as a 4D registration problem.

Existing approaches either model diffusion contrasts without directly enforcing spatial consistency, estimate motion between volumes rather than between slice groups, or rely on sequential or cross-volume relationships that degrade under sparse gradient sampling and severe motion. The authors identify robustly coupling these motion estimates across the acquisition as an unresolved methodological problem.