2with flow values from work by Van Beers 3 where the model was fitted to acquired dynamic contrast-enhanced (DCE) CT data. A population average was used for the arterial input function 4 and the portal input function was constructed by recreating the relationship seen to the shape of the AIF in published work 3 . The tumour was modelled using a single-input dual-compartment model 5 (the extended Kety model) with parameter values based on those found from fitting the model to previously acquired DCE-MR data. Zero mean Gaussian noise was added to the images with a standard deviation equivalent to a signal to noise ratio of 10 in the pre-contrast images. PVEs were emulated by defining tissue masks at twice the in-plane resolution and applying a Gaussian kernel with a width of three voxels. Spoiled gradient echo T1 weighted images were generated with a 128x128x25 matrix and a voxel size of 3x3x8 mm. Methods Firstly, K trans values from the synthetic images were compared against an acquired data set. Mean K trans values were calculated for a tumour VOI in the acquired data set. The VOI was then dilated in 3-D by a voxel width three times and the mean values for the three dilated VOIs calculated. The parameter maps for the tumour were then embedded in a synthetic liver with flow values corresponding to a healthy volunteer and also corresponding to cirrhosis Child-Pugh classification C 3