Quantitative analysis of dynamic MRI data obtained with a hollow fibre module.

This paper presents a model describing the diffusion of water and contrast agent (CA) in hollow fibre modules (HFM), which is able to calculate the contribution of water molecular exchange to the parameters resulting from dynamic R(1) relaxation data. To acquire these data, the intra-fibre volume of the HFM initially filled with pure water was perfused with a CA solution of constant concentration C(0). The relaxation rate R(1e)(t) time course was measured assuming fast water molecular exchange, with a saturation-recovery multi-gradient-echo snapshot sequence. These data led to uncorrected (uc) water exchange estimations of the constant k(CA_uc) characterizing the CA R(1e)(t) kinetics and of the intra-fibre volume fraction V(in_uc) of the HFM. The model, which assumes diffusion-controlled process and the absence of interaction of CA with the fibre membrane, led to the corresponding theoretical values for corrected water exchange k(CA) and V(in). k(CA_uc) was always slightly larger than k(CA) and independent of C(0). V(in_uc) was always much smaller than V(in) and depended strongly on C(0). Furthermore, V(in) was found to be very similar to the real intra-fibre volume fraction as determined directly from high-resolution MRI and specifications given by the supplier, thus confirming the validity of our model. It is shown that k(CA_uc) characterizing every CA diffusing without interaction with the membrane depends linearly on its diffusion coefficient in water D(s_CA). The experimental results suggest that interactions of Dotarem and P806 with the HFM membrane are not significant. This is also the case for the smaller and neutral molecule water, validating all the calculations involved in this work. For the largest CAs, experiments suggest that sterical effects dominate their permeability and that D(s_CA) rather than their molecular weight defines HFM membrane cut-off. However, if other interactions are also important, a multi-parameter function rather than a single value is expected to define this cut-off.