Model-based super-resolution reconstruction of T₂ maps

Area of Science:

• Magnetic Resonance Imaging
• Quantitative MRI
• Neuroimaging

Background:

• High-resolution isotropic T2 mapping of the brain using multi-echo spin-echo (MESE) is limited by 2D slice thickness or 3D specific absorption rate constraints.
• Current methods struggle to achieve both high resolution and efficient acquisition for quantitative T2 mapping.

Purpose of the Study:

• To develop and validate a method for reconstructing 1 mm³ isotropic T2 maps from undersampled 2D MESE acquisitions.
• To overcome the limitations of existing T2 mapping techniques in terms of resolution and scan time.

Main Methods:

• A super-resolution technique combined with an iterative model-based approach was used to reconstruct quantitative T2 maps from undersampled low-resolution data.
• The method involved multiple 2D MESE acquisitions with 10-fold undersampling.
• Validation was performed using numerical simulations, phantoms, and in vivo human brain data.

Main Results:

• Numerical simulations identified a 10-fold acceleration with 4 acquisitions (18 min scan time) as optimal.
• The proposed method demonstrated improved resolution compared to low-resolution images in phantoms and brain data.
• Phantom T2 values were comparable to fully sampled MESE, and in vivo brain T2 values showed consistency across subjects.

Conclusions:

• The developed method successfully reconstructs 1 mm³ isotropic T2 relaxation maps.
• This approach significantly accelerates acquisition time (10x faster) while addressing limitations in high-resolution T2 mapping.