Omni-QALAS: Optimized Multiparametric Imaging for Simultaneous $T_{1}$, $T_{2}$ and Myelin Water Mapping.

Abstract

3D-QALAS enables efficient multi-parametric $T_{\text{1}}$, $T_{\text{2}}$, and proton density (PD) mapping in a single scan; however, heuristic parameter selection limits image quality and myelin water fraction (MWF) quantification accuracy. While the subsequently proposed MWF-QALAS sequence improves MWF estimation, it requires an additional readout that prolongs scan time by approximately one minute and lacks precision for short $T_{\text{1}}$ and $T_{\text{2}}$ values. To address these limitations, we introduce Omni-QALAS (Optimized Myelin & Neurometric Imaging). The sequence flip angles, $T_{\text{2}}$ preparation durations, and $T_{\text{1}}$ recovery gaps are jointly optimized using a self-supervised multilayer perceptron (MLP). The MLP is trained to minimize a Cramér-Rao bound-based loss function subject to explicit total scan time constraints. Integrated with subspace reconstruction, the optimized Omni-QALAS reduces the required number of readouts from six to five compared to MWF-QALAS. This optimization shortens the overall scan time by 16% while yielding higher $T_{\text{1}}$ and $T_{\text{2}}$ accuracy alongside improved MWF maps. Simulation, phantom, and in vivo validations demonstrate that the Omni-QALAS protocol provides accurate whole-brain multiparametric quantification at 1 mm isotropic resolution within 5 minutes, with demonstrated feasibility for faster acquisition down to 3 minutes 30 seconds.