The objective is to evaluate whether an ex-vivo model can be used to test intracellular contrast agents for MR imaging of the liver. T1 weighted inversion recovery, proton density spin echo and T2* weighted gradient echo images of the liver were acquired at 0.5 T in 10 rats before and 30 min after intravenous injection of 0.075 mmol/kg Gadolinium benzyloxypropionictetraacetate (Gd-BOPTA, n = 5) or 0.015 mmol/kg dextran magnetite (DM, n = 5), Four additional animals served as controls. After exsanguination and perfusion with saline and formalin, specimens of the liver and brain were embedded in an agar gel and examined with MR imaging one to three weeks later using the same protocol. Zn-vivo, the mean liver signal enhancement caused by Gd-BOPTA in T1, proton density and T2* weighted images was +23%, +28% and -70%, respectively. The mean liver signal enhancement caused by DM was -71%, -76% and -94%. In-vitro, no signal change was seen in the brain of animals injected with Gd-BOPTA and DM as compared to controls. Liver signal was increased by Gd-BOPTA and decreased by DM. Mean liver enhancement rate induced by Gd-BOPTA was +22%, +5% and +27% for T1, proton density and T2* weighted images, respectively. Mean liver enhancement rate induced by DM was -27%, -19% and -31%. MR imaging signal changes induced by liver intracellular contrast agents are still appreciable in an ex-vivo model. The latter might be useful for for preliminary investigation of intracellular contrast agents for MR imaging of the liver. (C) 1997 Elsevier Science, Inc.