Rapid multi-parametric quantitative MRI via deep learning-based synthetic-to-real reconstruction and 3D SSFP-MOLED...

Area of Science:

• Magnetic Resonance Imaging
• Medical Physics
• Biomedical Engineering

Background:

• Multi-parametric quantitative MRI (mqMRI) offers clinical potential for tissue characterization.
• Current mqMRI is limited by long scan times and signal sensitivity, especially for high-resolution whole-brain imaging.

Purpose of the Study:

• To introduce a novel signal encoding method, 3D SSFP-MOLED, to overcome the limitations of current mqMRI.
• To enable rapid and accurate acquisition of whole-brain multi-parametric maps.

Main Methods:

• Developed a novel phase-modulated 3D SSFP with multiple overlapping-echo detachment (3D SSFP-MOLED) technique.
• Created a physics-constrained synthetic data pipeline for training deep learning models with realistic field variations.
• Achieved simultaneous encoding of six physiological parameters (M0, T1, T2, T2*, B1+, ΔB0) into k-space.

Main Results:

• Generated whole-brain parametric maps at 1x1x2 mm³ resolution in under 3 minutes with 2x parallel acceleration.
• Demonstrated high accuracy and reproducibility in phantom, volunteer, and clinical case studies (tumors, hemorrhage).
• Validated the method's ability to perform rapid multi-parametric quantitation.

Conclusions:

• The 3D SSFP-MOLED method establishes a new paradigm for efficient and reliable mqMRI.
• Synergistic integration of adaptive signal encoding, physics-informed training, and deep learning reconstruction enhances clinical applicability.
• This approach addresses key challenges in clinical signal processing for advanced MRI applications.