Accelerating Quantitative MRI using Subspace Multiscale Energy Model (SS-MuSE)

Motivation: Multiple high-resolution contrast weighted scans are required for soft tissue characterization. While 3D schemes offer several benefits, the long scan times as well as high memory demand and computational complexity of the algorithms restrict their clinical applicability. Goal(s): We introduce a fast and memory efficient plug-and-play (PnP) algorithm for accelerated 3D quantitative MRI. Approach: We generalize the PnP multi-scale energy-based model (MuSE) to jointly regularize the 3D multi-contrast spatial factors in a subspace recovery formulation. Results: 384 3D images were recovered from a 4.5-minute MPnRAGE scan with improved image quality, evidenced by the lower variance and bias of T1 mapping. Impact: The proposed method enabled fast iterative reconstruction of the large-scale multi-contrast MRI data from the accelerated scan. The recovered source images can be used to differentiate the tissue types or quantitative mapping.