Pharmacokinetic analysis of Dynamic Contrast-Enhanced MRI (DCE-MRI) data allows extraction and mapping of quantitative parameters of tissue biology in vivo, such as K^trans, v_e, and k_ep. However, the accuracy and precision of these parameters can be affected by many factors, with arterial input function (AIF) determination being a primary element. In this multicenter study, we sought to evaluate variations in DCE-MRI parameters estimated from shared prostate DCE-MRI data as a result of differences in AIFs.Prostate DCE-MRI data were collected on 11 subjects at one center and shared with eight other Quantitative Imaging Network (QIN) centers. The data acquisition details have been previously reported ¹. Each of the 9 centers determined individual AIFs with its site-specific method (fully automated, semi-automated, manual, etc.). All individual AIF files were submitted to a managing center (one of the 9 centers), which performed DCE-MRI pharmacokinetic modeling with all AIFs using the Tofts model (TM) ². A previously published population AIF ³ was also included in the analysis for comparison. In addition, a reference tissue (using obturator muscle in this study) approach ⁴ was also used for concentration adjustment of all individual AIFs. The same 11 data sets were then modeled with the reference-tissue-adjusted AIFs, resulting in a combined total of twenty sets of pharmacokinetic parameters (K^trans, v_e, and k_ep) for each subject, which were then classified into AIF-unadjusted (unadj.) and AIF-reference-tissue adjusted (adj.) groups. To insure variations in derived kinetic parameters were from AIF variations only, all other parameters for the model fitting including tumor ROI definition and pre-contrast T₁ were kept the same. Mean tumor K^trans, v_e, and k_ep (= K^trans/v_e) values for each patient were summarized. To assess agreements and variations of the DCE-MRI parameters obtained with different AIFs within a group (unadj. and adj.), intra-class correlation coefficient (ICC) and within-subject coefficient of variation (wCV) were computed, respectively.The top image of Figure 1 zoomed to the prostate area shows a DCE-MRI slice of one subject. The cyan-colored ROI demarks the lesion area for subsequent TM modeling and parameter comparisons. K^trans color maps generated by TM analysis using unadjusted AIFs from the 9 centers (numbered 1 to 9) are shown in the middle panels and those with reference-tissue-adjusted AIFs are shown at the bottom. Substantial variations, mostly in the magnitude of voxel K^trans but not in the distribution pattern, can be seen among the K^trans maps obtained with different unadjusted AIFs. These differences are lessened when the K^trans parameter was derived with reference-tissue-adjusted AIFs. Figure 2 shows the column graph of wCV values for the K^trans, v_e and k_ep parameters obtained with the unadjusted (shaded light grey) and adjusted (dark grey) AIFs. The respective 95% confidence intervals (CI) are shown as error bars. Figure 3 shows the column graph of ICC values with the same layout as Fig. 2. Results from this multicenter study clearly show that variations in AIF quantification result in variations in pharmacokinetic parameter values estimated from prostate DCE-MRI data, with K^trans and v_e having the largest and smallest variations, respectively. The agreement in the K^trans and v_e parameters obtained with muscle-reference-region-adjusted AIFs is improved compared to that with unadjusted AIFs (Figs. 1-3). It is important to note that DCE-MRI parameter variations caused by AIF variations are mostly systematic. As shown in Fig. 1, the differences among the prostate tumor K^trans maps obtained with different AIFs are mostly in voxel K^trans values. The pattern of voxel K^trans distribution largely remains similar. This suggests that assessment of tumor heterogeneity through texture analysis of DCE-MRI parametric maps may not be affected greatly by variations in AIF determination. Consequently, despite the significant challenge in accurate AIF determination, quantitative DCE-MRI remains a promising imaging tool for tumor characterization and therapeutic monitoring. This observation needs further validation in future studies. k_ep variation is shown to be less sensitive to AIF uncertainty than K^trans, suggesting it may be a more robust pharmacokinetic parameter for characterizing prostate microvasculature. In addition, parameters obtained using individually determined AIF demonstrates better consistency (higher ICC values, etc.) than those obtained with a generic population AIF (often derived from a different acquisition protocol) for substantial contrast agent extravasation situations like that in the prostate.