Robust T1ρ , T2 , and T2 * mapping via spin-locked MOLED with synthetic data-driven deep learning reconstruction

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Quantitative Imaging
• Biomedical Engineering

Background:

• Quantitative MRI (qMRI) faces challenges in speed and robustness, particularly for T1ρ mapping.
• Existing methods are often sensitive to magnetic field inhomogeneities and spin-lock artifacts.
• Efficient multi-parametric mapping is crucial for clinical diagnostics.

Purpose of the Study:

• To develop and evaluate a novel MRI technique, SL-MOLED, for rapid and artifact-robust quantitative mapping.
• To enable simultaneous measurement of T1ρ, T2, T2*, proton density, and B0/B1 fields.
• To reduce sensitivity to B0/B1 inhomogeneities and spin-lock banding artifacts.

Main Methods:

• Integration of spin-lock preparation into the MOLED framework (SL-MOLED).
• Development of a deep learning reconstruction framework trained on Bloch-simulated data with artifact modeling.
• Validation through numerical experiments, phantom studies, healthy volunteers, and patient evaluation.

Main Results:

• Deep learning models with artifact modeling improved SSIM by 0.1-0.2 and reduced MAE by 2-5 ms for T1ρ, T2, T2*.
• SL-MOLED demonstrated excellent agreement with reference methods in phantoms (r > 0.997) and in vivo (r ≥ 0.976).
• High repeatability (CV < 3.3%) and detection of elevated T1ρ in MS lesions were observed.

Conclusions:

• SL-MOLED offers accurate, repeatable, and artifact-robust quantitative mapping in ~11 seconds per slice.
• The technique shows potential for reliable multi-parametric MRI in clinical settings.
• SL-MOLED facilitates the detection of subtle pathological changes, like elevated T1ρ in MS lesions.