Abstract C128: Protocol optimization for dynamic contrast enhanced-MRI in combination with kinetic modeling

Abstract

Abstract Introduction: Dynamic contrast enhanced-MRI (DCE-MRI) is increasingly used in combination with pharmacokinetic modelling (PKM) for tumor diagnosis, treatment planning and for the evaluation of novel antiangiogenic therapies. In vivo monitoring of drug response requires high reproducibility of the PKM-technique, especially for the Tofts-model prior parameter Ktrans, a measure for tumor vessel wall permeability. However, several clinical studies show high in-patient-variability and poor reproducibility of Ktrans in paired examinations, partially caused by improper experiment design. The aim of this study is to develop and to test a theoretical framework in which the uncertainty on Ktrans for a given DCE-MRI-protocol can be evaluated a priori, and the optimal values of imaging parameters, to achieve maximal reproducibility, can be determined. Methods: The DCE-MRI protocol under investigation consists of a series Turbo-Flash-images (Ti = 560 ms, flip angle = 12°, Te = 4.12 ms) with variable temporal resolution (Tr = Δt = 1–10 s) and total scan time (Tscan = 250 – 2000 s) after injection of Gd-DTPA (Dose = 0.1 mM kg−1). The a priori uncertainty on Ktrans is tested by means of the Cramer-Rao lower bounds (CRLB), a concept from system theory that predicts the minimal standard deviation σK of the fitted parameters for a given protocol design. In the first part of this study, these CRLB are used for testing the reproducibility of the protocol theoretically and selecting the optimal values for Δt and Tscan. In these simulations SNR of the MR-measurements varies as ∼√Δt. In the second part the protocols are compared experimentally. Two groups of two test animals (mice with induced ht29 - tumors) were imaged in paired examinations (36 hours in between) for 2000 s with Δt =1.1 s (group 1) and Δt = 3.3 s (group 2) respectively. For each animal and both examinations, 4 values of Ktrans are calculated (after 250 s , 500 s, 1000 s and 2000 s). The coefficient of within patient variation (CWV) is used as reproducibility measure. Results: Simulations show that for Gd-DTPA Ktrans-values between 0.02 and 0.5 min−1 can be measured with acceptable errors (K ∼ 10%, SNR =10 at t = 5 s). In this range, temporal resolution should be as high as possible (1–3 s), even if this implies reduced SNR. A minimal scan time of 1000 s is required. Low Ktrans (∼0.005 min−1) cannot be determined in a reproducible manner for Tscan < 5000 s, while high Ktrans (> 1 min−1) are inaccurate due to nyquist considerations (K > 50%). The in vivo experiments show similar trends: with all mean Ktrans in the range of 0.015 to 0.37 min−1, the CWV decreases with increasing scantime (group 1: 26% - 24% - 18% - 14%, group 2: 30% - 28 % - 20% -15% , Tscan = 250s - 500s -1000s -2500s). The reproducibility in group 1 is higher than in group 2 for all values of Tscan , as predicted by the Cramer-Rao theory. Conclusion: The CRLB offer an excellent framework for testing protocol design with respect to reproducibility demands. For Gd-DTPA, Ktrans-values in the range of 0.02 and 0.5 min−1 can be determined in a reproducible manner for protocols with high temporal resolution and prolonged scan time. In vivo experiments confirm the theoretical findings, showing that the CRLB can be applied for further optimization of the DCE-MRI protocol. A potential application is the assessment of optimal values of the inversion time Ti and the Gd-DTPA- dose, each within their respective bounds. Citation Information: Mol Cancer Ther 2009;8(12 Suppl):C128.