Fast Multiparametric
ASL MRI
via Multi‐Dimensional Acceleration and Subspace Reconstruction
Ershad Hassanpour Golagani, Yiran Li, Xiao Liang, Bo Li, Yulin Chang, Manuel Taso, John A. Detre, Ze Wang ABSTRACT
Purpose
Multi‐post‐labeling delays (PLDs) multi‐echo times (TEs) arterial spin labeling (ASL) MRI enables simultaneous quantification of cerebral blood flow (CBF), arterial transit time (ATT), and blood–brain barrier (BBB) water exchange, but requires long scan times due to the combination of multi‐TEs, PLDs, and repeated control–label acquisitions. A five‐dimensional (5D) acquisition and reconstruction framework was developed to enable accelerated multi‐parametric ASL.
Methods
A coordinated undersampling strategy was designed across in‐plane k‐space, through‐plane k‐space, TE, and control–label measurement dimensions, achieving a combined acceleration factor of approximately 28. Reconstruction was performed using a Kronecker‐structured subspace model that exploited low‐rank correlations along the TE and control–label dimensions. Sparsity regularization was applied in the subspace coefficient domain, and temporal bases were estimated from a single fully sampled middle slice. The framework was evaluated using in vivo echo‐time‐resolved ASL (TASL) datasets from eight subjects.
Results
The proposed reconstruction outperformed a matched compressed sensing (CS) reconstruction across all echo times, yielding higher structural similarity index measure (SSIM) and peak signal‐to‐noise ratio (PSNR), with the largest improvements observed under severe undersampling ( R ≈28) and at late echo times. Higher‐fidelity CBF, ATT, BBB, and intravascular transit time (ITT) maps were obtained, with the greatest gains observed for BBB and ITT. Test–retest reliability, assessed using intraclass correlation coefficient (ICC), remained comparable to fully sampled reference data.
Conclusion
Coordinated multi‐dimensional undersampling combined with Kronecker subspace reconstruction enables highly accelerated multi‐parametric ASL while preserving image quality, quantitative parameter estimation, and test–retest reliability, supporting clinically feasible acquisition times.