Effect of Bias in Contrast Agent Concentration Measurement on Estimated Pharmacokinetic Parameters in Brain Dynamic Contrast-Enhanced Magnetic Resonance Imaging Studies

• Main Author: Azimeh Dehkordi
• Format: Article
• Language: English
• Published: Mashhad University of Medical Sciences 2020-05-01
• Series: Iranian Journal of Medical Physics
• Subjects: dynamic contrast-enhanced magnetic resonance imaging; kinetic analysis; contrast concentration bias; mpe of pk parameters
• Online Access: http://ijmp.mums.ac.ir/article_13664_ca4bc25a67a9b52d85a698628cbce7d4.pdf

<strong><em>Introduction:</em></strong> Pharmacokinetic (PK) modeling of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is widely applied in tumor diagnosis and treatment evaluation. Precision analysis of the estimated PK parameters is essential when they are used as a measure for therapy evaluation or treatment planning. In this study, the accuracy of PK parameters in brain DCE-MRI studies was quantified in relation to two major sources of error(including pre-contrast longitudinal-relaxation time, T_1,0 and flip angle, α). <strong><em>Material and Methods:</em></strong> 3470 dynamic contrast-enhanced-curves were simulated using a wide variation of the PK parameters. The bias of contrast concentration due to the systematic biases in α and T_1,0 was calculated and added to both contrast concentration and AIF profiles. Thereafter, the PK parameters were estimated for each simulated curve in the presence of different percentages of relative biases in α and T_1,0. The mean percentage error (MPE) of PK parameters was then calculated for all simulated curves. <strong><em>Results:</em></strong> The results indicated that plasma volume(v_p) was the most sensitive parameter to bias of contrast concentration, which may overestimate up to 700% in 10% coincidence relative bias in α and T₁,₀. The lowest MPE was related to the backward transfer constant (k_ep), which was ~2%-15% in 10% coincidence relative bias in each α and T₁,₀. <strong><em>Conclusion: </em></strong>Utilization of a nested model selection technique, along with an accurate estimator, such as maximum-likelihood estimation, created a unique approach for investigating the effect of the bias in the concentration measurement to the estimated PK parameters without the addition of any extra biases to the parameters during the estimation.