Sparsity and locally low rank regularization for MR fingerprinting

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Medical Physics
• Image Reconstruction

Background:

• Magnetic Resonance Fingerprinting (MRF) enables quantitative imaging but faces challenges with scan time and resolution.
• Existing low-rank reconstructions in MRF reduce artifacts but can still exhibit aliasing at high acceleration factors.
• Locally Low Rank (LLR) regularization is effective in other MR applications for exploiting temporal redundancy.

Purpose of the Study:

• To develop and evaluate a novel reconstruction method for accelerated MRF.
• To incorporate spatial sparsity and LLR regularization into MRF reconstruction, termed SLLR-MRF.
• To reduce aliasing artifacts and enable higher acceleration factors for improved MRF efficiency.

Main Methods:

• The SLLR-MRF approach combines spatial sparsity and LLR regularization within the MRF reconstruction framework.
• The method was validated using simulations, phantom T1/T2 mapping, and in vivo human brain imaging.
• Acquisition acceleration was explored in vivo to achieve higher in-plane spatial resolution within comparable scan times.

Main Results:

• SLLR-MRF demonstrated reduced aliasing artifacts compared to standard low-rank MRF reconstructions, particularly at high acceleration factors.
• Parametric maps derived from SLLR-MRF showed improved precision without compromising accuracy.
• In vivo results indicated potential for rapid MRF acquisitions (e.g., ~2.6s for 2x2mm resolution) with SLLR-MRF.

Conclusions:

• The SLLR-MRF reconstruction method enhances parametric map quality in accelerated MRF.
• This technique offers the potential for significantly shorter scan times or increased spatial resolution in MRF.
• SLLR-MRF represents a significant advancement for efficient and high-quality quantitative MRI using MRF.