Multi-shot diffusion-weighted MRI reconstruction with magnitude-based spatial-angular locally low-rank regularization...

Area of Science:

• Magnetic Resonance Imaging
• Medical Imaging
• Image Reconstruction

Background:

• Motion during scanning causes phase variations in diffusion-weighted imaging (DWI).
• These artifacts degrade image quality and limit resolution in multi-direction DWI.
• Accurate phase and magnitude estimation is crucial for robust DWI reconstruction.

Purpose of the Study:

• To resolve motion-induced phase variations in multi-shot multi-direction DWI.
• To improve image reconstruction by applying regularization to magnitude images.
• To achieve high-resolution DWI without sequence modification.

Main Methods:

• Developed a nonlinear model to estimate phase and magnitude images separately.
• Applied locally low-rank (LLR) regularization to magnitude images across diffusion directions.
• Exploited spatial and angular correlations for joint reconstruction.

Main Results:

• The proposed method significantly reduces noise compared to conventional reconstruction.
• Achieved submillimeter resolution (0.8mm, 0.9mm) at b=1,000 s/mm².
• Obtained 1-mm isotropic DWI at b=2,000 s/mm² without sequence changes.

Conclusions:

• Joint reconstruction with spatial-angular LLR regularization improves multi-direction DWI.
• The method effectively removes motion-induced phase artifacts.
• Simultaneously denoises images, enhancing overall DWI quality.