Pharmacokinetic Modeling for Hyperpolarized MRI of 13C Urea
K Michel1*, C Walker1, S Lai1, M Merritt2, J Bankson1, (1) The University of Texas MD Anderson Cancer Center, Houston, TX, (2) The University of Florida, Gainesville, FL
Presentations
PO-GePV-I-30 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: To evaluate the effects of SNR in pharmacokinetic analyses of tumor perfusion using hyperpolarized (HP) ¹³C urea.
Methods: Simulation experiments were performed in MATLAB to assess the performance of three pharmacokinetic models for quantitative analysis of dynamic imaging of HP urea. Model A is similar to the extended Tofts model for DCE-MRI, and is defined by three parameters corresponding to capillary permeability (kᵥₑ) and volume fractions of blood (v(b)) and extravascular/extracellular space (EES, vₑ). Model B resembles the Tofts model, with parameters for permeability (kᵥₑ’) and EES volume fraction (vₑ). Model C assumes well-mixed vascular and extravascular spaces without an inaccessible cellular volume fraction (parameters kᵥₑ’’ and v(b)), and mirrors a model used in analyses of metabolic HP agents. Tissue signal curves were generated from model A using parameters estimated from prior imaging experiments. All models were fit to these synthetic data after the addition of zero-mean Gaussian noise, and fitting was repeated 100 times with fresh noise to evaluate error and bias in the parameter values derived from each model at different peak SNRs.
Results: For model A, accurate and reproducible estimation of vₑ requires greater peak SNR than both kᵥₑ and v(b). Mean errors in vₑ estimates of <5% required a peak SNR >15, whereas the mean errors in kᵥₑ and v(b) were both <3% for a SNR 10 in fitting model A. Compared to the ground truth values input to model A, fitting using model B underestimates vₑ and greatly overestimates permeability (kᵥₑ’>kᵥₑ). Model C overestimates v(b) and underestimates permeability (kᵥₑ’’
Conclusion: Selection of a pharmacokinetic model must balance physiological accuracy and model complexity. Our work demonstrates these trade-offs in the context of HP urea imaging.
Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by R01-CA211150, R01-DK105346, P30-CA016672 and CPRIT RP170366.
Keywords
Pharmacokinetic Modeling, Simulation, MRI
Taxonomy
IM- MRI : Hyperpolarized Imaging
Contact Email