Fast and ultra-high shot diffusion MRI image reconstruction with self-adaptive Hankel subspace.

Area of Science:

• Magnetic Resonance Imaging
• Image Reconstruction
• Diffusion Weighted Imaging

Background:

• Multi-shot interleaved echo planar imaging is crucial for high-resolution, low-distortion diffusion weighted images (DWI).
• Motion artifacts from inter-shot phase variations in DWI limit applications, especially with ultra-high shot counts.
• Current low-rank matrix reconstruction methods are effective for 4-8 shots but fail for 10-12 shots and are computationally intensive.

Purpose of the Study:

• To develop a faster and more effective method for reconstructing ultra-high shot diffusion weighted images.
• To address the limitations of existing low-rank matrix methods in handling variant low-rankness across k-space.
• To improve the quality and reduce reconstruction time for high-resolution DWI.

Main Methods:

• Proposed DONA (1D lOw-raNk HAnkel) matrices to decompose large 2D Hankel matrices for accelerated reconstruction.
• Introduced DONATE (DONA with self-adapTive subspacE estimation) to handle variant low-rankness by separating signal subspaces.
• Utilized initial image estimation for strong subspaces and adaptive estimation for uncertain subspaces to preserve image details.

Main Results:

• DONATE achieved 4-shot DWI reconstruction in 10 seconds.
• DONATE demonstrated a 10x faster computation for 12-shot DWI reconstruction compared to previous methods.
• The method showed superior performance in low-distortion spine DWI reconstruction and subjective image quality evaluations.

Conclusions:

• DONATE significantly accelerates DWI reconstruction, enabling ultra-high shot acquisition.
• The self-adaptive subspace estimation effectively preserves image details while reducing artifacts.
• DONATE offers a promising solution for high-resolution, high-quality diffusion weighted imaging.