The aim of our study was to assess the accuracy of interstitial volume fraction measurement in low-perfused and low-vascularized tissue performed with dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI).The interstitial tissue volume fraction, v(e), was determined in the medial thigh muscles of twelve female pigs using a 3D gradient echo sequence with k-space-sharing and with a gadolinium-based contrast agent (gadoterate meglumine) ¹. Determination was performed using three pharmacokinetic models^2,3, the simple (TM) and the extended Tofts model (ETM) and the two-compartment exchange model (2CXM). We investigated the effect of varying acquisition durations (ADs) on the model parameter estimates of the three models and compared the v(e) values with the results of histologic examinations of muscle of the medial thigh muscle.Histology (illustrated in Fig. 1) yielded a median (25%-75% quartile) v(e) of 4.8(3.7-6.2)%. DCE-MRI measurement (illustrated by curve fits in Fig. 2) yielded similar interstitial tissue volume fractions depending on the model and AD. The lowest v(e) determined by DCE-MRI was 5.1(3.3-6.0)% and 5.2(3.3-6.1)% for the TM and ETM at 6 min AD, respectively. The highest v(e) was 7.7(4.4-8.9)% for the TM and 7.7(4.5-9.0)% for the ETM and 15 min AD. The variation of v(e) with AD was much smaller when the 2CXM was used: v(e)=6.2 (3.1-9.2)% for 6 min AD and v(e)=6.3 (4.3-9.8)% for 15 min. Fig. 3 summarizes median values of v(e), interstitium-to-plasma constant k(ep), and volume transfer constant K(trans) obtained with the three models and at the different ADs. The best fit was found for the 2CXM at 10 min AD (v(e)=6.6(3.7-8.2)%). With increasing AD, the values of K(trans) decreased while v(e) increased for both Tofts models. Figure 4 shows the behavior of k(ep) against AD for the three models used in our experiments. For the Tofts models, k(ep) decreased asymptotically with increasing AD and converged toward a value close to k(ep) determined with the 2CXM. The interplay of increasing v(e) values and declining values for K(trans) led to a decrease in the interstitium-to-plasma rate constant, k(ep)= K(trans)/v(e).The 2CXM uses more fitting parameters and thus yields better fits. As a result, v(e) is less dependent on AD; however, the uncertainty expressed by the 25%-75% quartile range is larger. The underfitting with the Tofts models results in an uncertainty in parameter determination, which has a major impact even for analysis of low-vascularized and low-perfused tissue, where the estimated v(e) values depend on AD. Artzi et al. already reported dependence of k(ep) (determined with the ETM) on AD in human glioblastoma⁴. In a simulation study⁵, Luypeart et al. found that, compared to the 2CXM, the ETM underestimated v(e) for short ADs and increasingly overestimated it for increasing AD, and that, conversely, K(trans) was overestimated for short ADs and increasingly underestimated for longer ADs using the ETM.Our results reveal good agreement between histologic estimates of interstitial volume and the estimates gained by DCE-MRI modeling. Due to its good fitting accuracy and independence of AD, the 2CXM appears to be suitable for use in clinical practice. In contrast to the 2CXM, estimation of v(e), K(trans), and k(ep) using both Tofts models significantly depend on AD. Thus, use of the Tofts models is more complex as AD must be taken into account when analyzing the fitting parameters.