Improved ASL in Water-Unsuppressed MRSI Using Generalized Series Modeling and Statistical Learning

Motivation: Distortion-free ASL integrated with MRSI has been recently proposed. However, the ASL component requires several minutes of averaging to achieve sufficient SNR. Goal(s): To improve the SNR of ASL thereby reducing its scan time. Approach: In acquisition, a short-TE sampling of (k, t)-space was used to improve SNR efficiency. In processing, generalized series modeling and statistical learning were applied for reconstruction and denoising. Results: The proposed method significantly enhanced the SNR of ASL, enabling a fourfold reduction in scan time to less than one minute. This enabled the acquisition of ASL at 2×2×2 mm³ and MRSI at 2×3×3 mm³ resolution within 8 minutes. Impact: With the enhanced SNR and imaging speed, this integrated ASL and MRSI method offers a powerful multi-modal imaging tool for investigating cerebral blood flow and brain metabolism under healthy and disease conditions.