Motivation: Deep learning-based reconstructions could be used to further accelerate MR acquisitions by optimizing sequence parameters that reduce scan time, at the cost of reduced SNR, as an alternative to simply increasing undersampling factors. Goal(s): To compare a clinical MPRAGE protocol (CS x5, TA: 4min) at 7T with a x12 undersampled acquisition (Fast-Standard) and a time-matched protocol with x5 undersampling but higher turbo factor and optimized parameters to preserve contrast (Fast-Optimized). Approach: Reference and fast MPRAGE protocols were acquired in three subjects. Reconstructed images were compared qualitatively and quantitatively. Results: Fast-Optimized scans were more comparable to the reference than the Fast-Standard scans. Impact: This study demonstrates the effectiveness of using deep learning image reconstruction in combination with optimized sequence parameters for scan speed to further accelerate a clinical MPRAGE acquisition, in contrast to simply increasing undersampling factors.
Yu et al. (Tue,) studied this question.