Low-rank magnetic resonance fingerprinting

Area of Science:

• Medical Imaging
• Quantitative MRI

Background:

• Magnetic resonance fingerprinting (MRF) enables quantitative MRI using randomized acquisition and offline parameter extraction.
• High undersampling ratios are necessary for MRF, leading to spatial artifacts that hinder accurate tissue parameter estimation.

Purpose of the Study:

• Introduce magnetic resonance fingerprinting with low rank (FLOR), a novel approach to enhance quantitative MRI.
• Address spatial artifacts caused by MRF undersampling to improve quantitative accuracy.

Main Methods:

• Exploit the low-rank property of concatenated temporal MRF contrasts.
• Utilize sparse signal representation in the dictionary domain.
• Implement an iterative recovery scheme combining gradient steps and singular value decomposition-based low-rank projection.

Main Results:

• Demonstrate improved parameter accuracy in both retrospective and prospective experiments compared to existing methods.
• Achieved improved accuracy at low sampling ratios (5% retrospective, 9% prospective).

Conclusions:

• FLOR effectively recovers undersampled MRF images by leveraging the signal's low-rank nature.
• The method provides more accurate quantitative parameter maps than previous iterative approaches.