Near-Real-Time Multi-Parametric Quantitative MRI using Parallel Non-Cartesian 6D Spatial-Temporal Dictionary Learning...

Area of Science:

• Medical Imaging
• Neuroimaging
• Magnetic Resonance Imaging

Background:

• Multi-parametric quantitative MRI (qMRI) is crucial for precise image-guided interventions.
• Current qMRI techniques for the brain lack the temporal resolution required for dynamic applications.

Purpose of the Study:

• To develop and validate an online pipeline for near-real-time multi-parametric qMRI of the brain.
• To improve temporal resolution for qMRI to support time-critical applications.

Main Methods:

• An alternating dual-flip-angle blipped multi-gradient-echo sequence with a stack-of-stars (SoS) trajectory was employed.
• K-space data were processed using the Gadgetron framework with a novel 6D spatial-temporal dictionary learning neural network (6D-stDLNN) for reconstruction.
• Quantitative calculations were performed efficiently on the GPU.

Main Results:

• Prospective volunteer studies demonstrated high consistency (ICC > 0.9) between accelerated and fully sampled 3D multi-parametric maps.
• Achieved a temporal resolution of 8.1 seconds per frame with a total algorithm runtime under 5 seconds per frame.
• Simultaneous 3D quantitative susceptibility mapping (QSM), R2* mapping, and T1 mapping were generated.

Conclusions:

• A near-real-time imaging pipeline for simultaneous multi-parametric brain mapping (QSM, R2*, T1) was successfully developed.
• This demonstrates the feasibility of immediate quantitative assessment, showing potential for time-critical interventions like deep brain stimulation.