The effect of cell-membrane water permeability on DWI has been discussed (1). To measure the effect quantitatively, we performed multi-b and multi-diffusion-time (MbMTd) DWI on aquaporin-4-expressing (AQ) and non-expressing (noAQ) cells, and demonstrated a clear difference between the signals from the two cell types. The data was interpreted with a two-compartment model including inter-compartmental exchange. It was also assumed that restricted diffusion of water molecules inside the cells leads to the intracellular diffusion coefficient being inversely proportional to the diffusion-time. Estimates of the water-exchange times with this model were comparable with those measured using an independent optical imaging technique (2), which suggests that this method might be used to characterize cell-membrane water permeability. As the technique can be applied in routine clinical examination, it has the potential to improve clinical diagnosis.A 7T animal MRI (Kobelco and Bruker, Japan) system was used for this experiment. Two types of cells (noAQ and AQ) suspended in PBS were set in the gantry. MbMTd DWI was obtained using a pulsed-gradient spin-echo (PGSE) sequence with multi-shot EPI acquisition (TR = 3s, TE = 115ms, matrix size = 128x128, spatial resolution = 0.2×0.2mm, and slice thickness = 2mm). The interval between the diffusion gradient lobes (Δ) were set at 40, 70, and 100ms to change the diffusion time (represented by Td_scan = Δ - δ/3) while keeping TE constant. The diffusion gradient duration (δ) was fixed at 7ms for all experiments. For each Td_scan, the b-value was increased from 0 to 8000 s/mm2 in 14 steps by increasing the gradient amplitude. Karger model (3-5), which is a simple two compartment model with inter-compartmental compound exchange, was used for the initial analysis of the data. Based on these results it was further assumed that Td_scan is sufficiently long that the diffusion coefficient in the extracellular space (D_ex) is approximately constant while that in the intracellular space (D_in) is inversely proportional to the diffusion time (6-7), and the data was analyzed again using a Td-independent D_ex and a/Td_scan^c instead of a constant D_in (Fig. 1). The parameter a has dimensions of length squared, while c is a parameter inserted to test the assumption that D_in is inversely proportional to Td_scan.Separate apparent diffusion coefficient (ADC) maps were calculated for the low- (b=0-1500 s/mm2) and high-b (b=4000-8000 s/mm2) ranges using single-exponential fitting to the images acquired with Δ=100ms (Fig. 2). The low-b range ADC map appears to depend on depth, while the high-b range ADC map is more sensitive to differences in AQP4 expression. The mean b-value dependent signal changes from ROIs drawn midway down the cell cultures are compared in Fig. 3. The fitted curves are consistent with the observed data across all values of Td_scan (Fig. 3). The exchange times (t_ie) for the noAQ and AQ cells were 88.9+-2.2ms and 62.7+-2.9 ms, respectively. These values are compatible with those obtained with optical imaging (2). The parameter c was estimated to be 0.985+-0.003 for noAQ and 1.01+-0.00 for AQ, respectively. The fact that the estimated values of c for both the noAQ and AQ cells were almost 1 is consistent with our assumption about the intracellular diffusion coefficient. All estimates are summarized in Table 2.