Accelerating 3D-T_1ρ mapping of cartilage using compressed sensing with different sparse and low rank models

Area of Science:

• Magnetic Resonance Imaging
• Biomedical Engineering
• Medical Physics

Background:

• 3D-T1ρ mapping is crucial for cartilage assessment.
• Reducing scan times in MRI is a persistent challenge.
• Compressed sensing (CS) offers potential for accelerating MRI acquisition.

Purpose of the Study:

• To assess the feasibility of using CS to speed up 3D-T1ρ cartilage mapping.
• To evaluate if CS can reduce scan times without compromising T1ρ estimation accuracy.

Main Methods:

• Retrospective undersampling of fully sampled 3D-T1ρ datasets (acceleration factors 2-10).
• Comparison of 12 sparsifying transforms for CS reconstruction, including finite differences, wavelets, and learned transforms.
• Evaluation on synthetic phantoms and in vivo human knee cartilage datasets.

Main Results:

• Most CS methods achieved satisfactory results at acceleration factor 2 (T1ρ error < 4.5%).
• Spatiotemporal finite difference (STFD), exponential dictionaries (EXP), and low rank plus sparse (L+S SFD) methods showed improved performance.
• These advanced methods achieved < 6.5% T1ρ error at acceleration factor 10 with spatial filtering, with STFD reaching 5.1% error.

Conclusions:

• Compressed sensing is feasible for accelerating 3D-T1ρ cartilage mapping up to acceleration factor 10.
• STFD, EXP, and L+S SFD regularizers are effective CS methods for this application.
• These methods maintain satisfactory accuracy for T1ρ estimation in cartilage.