Robust nonlinear parameter estimation with physics-enforced spectral-spatial priors for multi-pool CEST analysis

Motivation: Asymmetric spectral analysis of multi-pool CEST MRI directly from largely incomplete measurements is challenging due to its nonlinearity and data scarcity in $$$k$$$-$$$w$$$ space. Conventional Lorentzian analysis is limited, considering only one-dimensional variation in spectral signals while not fully exploiting physics constraints. Goal(s): To develop a robust parameter estimation method for multi-pool CEST spectral analysis with highly incomplete measurements. Approach: We proposed a multi-pool CEST-specific, nonlinear separable model with physics-constrained spectral-spatial priors and performed patient studies with brain tumors. Results: The proposed method yields clear CEST parameter maps in brain tumor regions, demonstrating feasible whole-brain CEST MRI (1.8mm$$$^3$$$) in under 4 minutes. Impact: The proposed method yields robust multi-pool CEST spectral analysis even with highly incomplete measurements, enabling whole brain CEST MRI (1.8mm$$$^3$$$) in less than 4 minutes.