A1 Journal article – refereed
Patient-specific pharmacokinetic parameter estimation on dynamic contrast-enhanced MRI of prostate: Preliminary evaluation of a novel AIF-free estimation method

List of Authors: Ginsburg SB, Taimen P, Merisaari H, Vainio P, Boström PJ, Aronen HJ, Jambor I, Madabhushi A
Publication year: 2016
Journal acronym: J Magn Reson Imaging
Volume number: 44
Issue number: 6



To develop and evaluate a prostate-based method (PBM) for estimating pharmacokinetic parameters on dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) by leveraging inherent differences in pharmacokinetic characteristics between the peripheral zone (PZ) and transition zone (TZ).

Materials and Methods

This retrospective study, approved by the Institutional Review Board, included 40 patients who underwent a multiparametric 3T MRI examination and subsequent radical prostatectomy. A two-step PBM for estimating pharmacokinetic parameters exploited the inherent differences in pharmacokinetic characteristics associated with the TZ and PZ. First, the reference region model was implemented to estimate ratios of Ktrans between normal TZ and PZ. Subsequently, the reference region model was leveraged again to estimate values for Ktrans and ve for every prostate voxel. The parameters of PBM were compared with those estimated using an arterial input function (AIF) derived from the femoral arteries. The ability of the parameters to differentiate prostate cancer (PCa) from benign tissue was evaluated on a voxel and lesion level. Additionally, the effect of temporal downsampling of the DCE MRI data was assessed.


Significant differences (P < 0.05) in PBM Ktrans between PCa lesions and benign tissue were found in 26/27 patients with TZ lesions and in 33/38 patients with PZ lesions; significant differences in AIF-based Ktrans occurred in 26/27 and 30/38 patients, respectively. The 75th and 100th percentiles of Ktrans and ve estimated using PBM positively correlated with lesion size (P < 0.05).


Pharmacokinetic parameters estimated via PBM outperformed AIF-based parameters in PCa detection.

Last updated on 2019-29-01 at 18:56