Controlling sharpness, SNR, and specific absorption rate for 3D fast-spin echo at 7T by end-to-end learning

Area of Science:

• Magnetic Resonance Imaging
• Biomedical Engineering

Background:

• 7T MRI offers higher signal-to-noise ratio (SNR) but faces challenges with long echo trains in 3D Fast Spin Echo (FSE) sequences.
• Optimizing pulse sequences is crucial for maximizing image quality and diagnostic utility at ultra-high field strengths.

Purpose of the Study:

• To develop non-heuristic variable flip angle (VFA) schemes for 3D FSE at 7T.
• To optimize VFA schemes for both point-spread function (PSF) and SNR across multiple tissues.
• To achieve this optimization under specific absorption rate (SAR) constraints and target contrast requirements.

Main Methods:

• An end-to-end learning framework was employed to integrate SAR constraints, contrast fidelity (SNR), and image blurring (PSF) into a cost function.
• Variable flip angle (VFA) schemes were optimized by adjusting the weights of PSF and SNR components in the cost function.
• The derived PSF- and SNR-optimized VFAs were tested in vivo using both open-source Pulseq and vendor protocols on a 7T MRI system.

Main Results:

• PSF-optimized VFAs significantly reduced image blurring in T2-weighted images compared to standard VFA, enhancing visibility of small structures and blood vessels.
• Quantitative analysis revealed optimized VFA yielded 50% less deviation from a reference PSF (sinc) than standard VFA.
• SNR-optimized VFA significantly improved SNR in white and gray matter regions (77.1 vs. 40.7) at the expense of increased blurring.

Conclusions:

• End-to-end learning frameworks can effectively optimize VFA schemes for 3D FSE at 7T, balancing PSF and SNR.
• This approach enables fast and flexible adjustments of the PSF/SNR trade-off for 3D FSE acquisitions.
• The optimized VFA schemes hold promise for improved image quality and diagnostic capabilities in ultra-high field MRI.