Sub-second whole brain T₂mapping via multiband SENSE multiple overlapping-echo detachment imaging and deep learning.

Area of Science:

• Magnetic Resonance Imaging
• Quantitative Imaging
• Medical Physics

Background:

• Quantitative magnetic resonance imaging (qMRI) methods are often time-consuming.
• Multiple Overlapping-Echo Detachment (MOLED) offers rapid single-slice T2 mapping (around 100 ms).
• Whole-brain imaging with MOLED still requires several seconds, limiting its clinical utility.

Purpose of the Study:

• To accelerate whole-brain T2 mapping using simultaneous multi-slice imaging.
• To integrate multiband SENSE (MB-SENSE) technology with MOLED (MB-MOLED).
• To enhance image quality in accelerated MB-MOLED using a plug-and-play (PnP) algorithm with deep learning denoisers.

Main Methods:

• Developed and implemented a multiband MOLED (MB-MOLED) pulse sequence.
• Utilized deep learning (U-Net) for T2 map reconstruction.
• Applied a PnP algorithm with DRUNet denoiser to improve image quality at high multiband factors (MB).
• Validated the approach through numerical simulations, phantom experiments, and in vivo human brain imaging.

Main Results:

• Numerical simulations demonstrated PnP algorithm's effectiveness in improving T2 map quality at low signal-to-noise ratios.
• MB-MOLED maintained MOLED's accuracy, with results consistent with reference methods in phantom studies.
• In vivo experiments showed PnP algorithm significantly improved T2 map quality at MB=4.
• Achieved the first whole-brain T2 mapping within 600 ms using MB-MOLED with MB=4.

Conclusions:

• MB-SENSE and MOLED can be effectively combined for rapid quantitative imaging.
• The developed MB-MOLED technique enables sub-second whole-brain T2 mapping.
• The PnP algorithm is crucial for maintaining image quality during accelerated acquisition.
• This approach holds significant promise for dynamic and functional qMRI applications requiring high temporal resolution.