A two-step low rank matrices approach for constrained MR image reconstruction.

Area of Science:

• Medical Imaging
• Signal Processing
• Applied Mathematics

Background:

• Low-rank structure is a key characteristic utilized in constrained magnetic resonance imaging (MRI).
• Exploiting this structure aids in improving image reconstruction quality and efficiency.

Purpose of the Study:

• To develop a novel, efficient, and SVD-free constrained MR image reconstruction method.
• To leverage low-rank matrix properties for enhanced image recovery from k-space data.

Main Methods:

• Constructed two low-rank matrices (T_V and T_H) from weighted k-space data.
• Proposed a two-step reconstruction: recovering difference images by enforcing low-rankness, followed by least squares reconstruction.
• Employed the alternating direction method of multipliers (ADMM) and replaced the nuclear norm with the minimum Frobenius norm for efficient computation.

Main Results:

• The proposed method successfully recovers vertical and horizontal difference images.
• Demonstrated fast image reconstruction due to the absence of Singular Value Decomposition (SVD).
• Experimental results show superior performance compared to existing low-rank based reconstruction methods.

Conclusions:

• The developed two-step method effectively reconstructs MR images by exploiting low-rank priors.
• The SVD-free ADMM approach offers a computationally efficient alternative for constrained MRI.
• This technique shows significant potential for improving MR image reconstruction in clinical and research settings.