Rapid high-fidelity mapping using single-shot overlapping-echo acquisition and deep learning reconstruction.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Neuroimaging
• Quantitative MRI

Background:

• Quantitative B1 mapping is crucial for accurate functional MRI (fMRI) analysis.
• Conventional methods for B1 mapping are often time-consuming, limiting their application in dynamic studies.
• Rapid and accurate B1 quantification is needed to improve the reliability of fMRI.

Purpose of the Study:

• To develop and validate a novel, single-shot quantitative MRI technique, GRE-MOLED, for rapid B1 mapping.
• To assess the efficiency, accuracy, and robustness of GRE-MOLED in various imaging scenarios.

Main Methods:

• GRE-MOLED utilizes gradient-echo acquisition with multiple overlapping-echo detachment (MOLED) encoding and EPI readout in a single k-space shot.
• A deep neural network was employed for end-to-end parametric mapping from the captured overlapping-echo signals.
• Self-registered B0 information was used for distortion correction, and the technique was evaluated in phantoms and human volunteers.

Main Results:

• GRE-MOLED B1 mapping showed excellent agreement with multi-echo GRE methods (r > 0.97).
• High intrasubject repeatability (CV < 1.0%) was achieved in scan-rescan tests.
• GRE-MOLED demonstrated robustness to distortions and motion, offering higher temporal SNR and BOLD sensitivity compared to conventional fMRI.

Conclusions:

• GRE-MOLED is an efficient, high-quality, real-time technique for quantitative B1 mapping.
• This method has the potential to serve as an improved quantitative BOLD detection technique for fMRI.