AbstractPhysical experiments are described which simulated the main practical and technical problems encountered in estimating iron concentrations in liver by dual-energy computed tomography. In aqueous solutions located in an oval water phantom, the presence of Fe at concentrations of 1 and 2 mg/ml was detected and estimated with good accuracy and reproducibility. The main physical problem was that water showed a non-zero dual-energy difference reading whose magnitude varied with location, but this potential error was largely eliminated by measuring the location-specific water values using a technique readily adapted to patient studies. Experiments with pieces of bovine tissue in the same system suggested that the dual-energy response shown by liver tissue may include an Fe-independent contribution, reflecting its organic composition (including fat content): if attributed to Fe, this would correspond to an error<± 1 mg/g wet weight in the evaluation.