Pharmacokinetic modeling strategies for dynamic hyperpolarized urea imaging

To evaluate pharmacokinetic modeling methods for quantification of tissue perfusion/permeability with hyperpolarized ¹³C urea.

Three models for quantitative analysis of dynamic HP urea imaging data were proposed and evaluated in numerical simulations and a thyroid cancer mouse model. A multicompartment model resembling the extended Tofts model for DCE-MRI (Model I) and two simplified models were used. The simplified models each eliminate a volume parameter representing vascular (Model II) or impermeable cellular space (Model III). Signal curves were generated from Model I, and models were fit to these synthetic data to quantify the effects of acquisition settings, bias in simplified models, and noise.

For Model I, reproducible and accurate results from snapshot imaging occurred at excitation angles of roughly 10 to 40 degrees, with wider ranges of good performance at longer TRs. Models II and III exhibited bias in estimation of the trans-capillary transfer rate constant (k_ve), with high sensitivity in k_ve fitting to variations in the volume parameter not explicitly included. At a peak SNR of 25, k_ve coefficients of variation were 14.6%, 5.53%, and 4.9% for Models I-III, respectively. When vascular input function (VIF) amplitude was jointly estimated, these coefficients of variation increased to 26.9%, 8.86%, and 25.4%. Individual pharmacokinetic parameters exhibit added bias with VIF amplitude fitting, but the k_ve/v_e and k_ve/v_b ratios are independent of VIF scaling and provide accurate results for Models I and III.

Our results demonstrate the feasibility and relative performance of pharmacokinetic models for HP urea for quantification of tissue permeability and perfusion.