Navigator-free multi-shot diffusion MRI via non-local low-rank reconstruction

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Image Reconstruction
• Diffusion-Weighted Imaging (DWI)

Background:

• Single-shot EPI (ss-EPI) in DWI is limited by geometric distortions and T₂* blurring.
• Multi-shot EPI (ms-EPI) offers higher spatial resolution but suffers from shot-to-shot phase variations.
• Existing navigator-based methods for phase correction can prolong scan times.

Purpose of the Study:

• To develop a Non-Local Low-Rank (NLLR) reconstruction method for ms-EPI in DWI.
• To address phase inconsistencies and noise while maintaining high spatial resolution.
• To achieve clinically feasible scan times.

Main Methods:

• The NLLR method utilizes non-local patch matching to group similar image patches across spatially distant locations.
• This approach enhances the exploitation of non-local redundancy for improved phase estimation and correction.
• NLLR was validated through simulations and in vivo experiments, compared against denoising and navigator-free techniques.

Main Results:

• NLLR demonstrated superior noise suppression and structural preservation compared to post-processing denoising algorithms in simulations.
• In vivo experiments showed NLLR outperformed conventional navigator-free approaches, especially in noise reduction.
• Fractional anisotropy maps reconstructed with NLLR exhibited enhanced visualization of fine structures and improved signal-to-noise ratio (SNR).

Conclusions:

• The NLLR approach offers an efficient solution for high-resolution DWI reconstruction.
• It effectively mitigates phase variations and noise, leading to improved image quality.
• NLLR facilitates high-quality, high-resolution DWI within practical scan times.