Efficient mesoscale multiparametric quantitative MRI using 3D-QALAS at 7T with self-supervised learning

Motivation: Mesoscale quantitative MRI (qMRI) has the potential to provide unique insights into tissue composition, but technical challenges have so far precluded this application. Goal(s): To develop high-resolution multiparametric quantitative MRI using 3D-QALAS at 7T. Approach: We propose to: 1) modify the adiabatic T2-preparation module to account for shorter T2 values and increased B1+ inhomogeneity at 7T; 2) use multi-contrast/-slice zero-shot self-supervised-learning (ZS-SSL) for joint QALAS image reconstruction and 3) employ SSL parameter estimation algorithm that incorporates inversion efficiency estimation. Results: In vivo results demonstrate that high-fidelity whole-brain T1 and T2 maps at 500μm isotropic resolution can be achieved within 16min at 7T. Impact: We demonstrate high-fidelity mesoscale (500μm isotropic resolution) multiparametric qMRI with 3D-QALAS at 7T using self-supervised image reconstruction and parameter estimation.