Kinetic Modeling Framework for Hyperpolarized 129Xe Chemical Exchange Saturation Transfer (HyperCEST) MRI
Abstract
Purpose
Hyperpolarized chemical exchange saturation transfer (HyperCEST) MRI using 129Xe gas enables detection of signal loss from perfluorooctyl bromide nanodroplets (PFOB NDs). We developed a kinetic model for 129Xe HyperCEST that enables quantitative analysis of these signals, including estimation of nanodroplet volume fraction (vB).
Methods
In vitro HyperCEST MR spectroscopy was performed by acquiring alternating saturation measurements (CEST OFF/ON) to produce exchange-mediated signal depletion in samples of PFOB NDs (Pool B) suspended in cell media (Pool A). A two-pool kinetic model was developed to generate and fit synthetic HyperCEST signal trains. Nanodroplet geometry was incorporated using per-dataset surface-area-to-volume (S/V) ratios to compute droplet size and total surface area. Independently measured vB values at multiple nanodroplet concentrations were used as ground-truth references for model validation. Model parameters were estimated using a nested optimization strategy: an outer nonlinear least-squares fitter solved for global permeability (P) and pool A and B relaxation rates (R1A, R2A, R1B, R2B), while inner fits estimated dataset-specific vB by minimizing error between measured and model-generated HyperCEST signals.
Results
The global optimization converged to a set of five kinetic parameters describing exchange permeability and relaxation in pools A and B. Using these calibrated parameters, dataset-specific inner fits recovered vB values that closely matched independently measured vB. Linear regression of fitted versus measured vB demonstrated agreement (R2=0.998), indicating accurate recovery of nanodroplet volume fraction from HyperCEST measurements.
Conclusion
We developed a kinetic modeling framework that enables accurate quantitative estimation of vB from HyperCEST measurements. This global fitting strategy calibrates a physically constrained HyperCEST kinetic model across heterogeneous datasets by separating shared exchange and relaxation parameters from dataset-specific vB. The agreement between fitted and independently measured vB values supports the use of this modeling framework for in vivo applications, where direct measurement of vB is not feasible.