Measurements of rCBF by the Xe/CT method are based on the assumption of identity between the end-tidal xenon curve which is applied as input function, and the arterial xenon curve being the true input function to the brain. In this study corresponding end-tidal and arterial xenon curves were measured in an experimental animal model (part 1) and in 5 patients with traumatic brain injury (part 2) and used for rCBF calculation. In both studies rCBF was underestimated by using the end-tidal xenon concentration curve as brain input function. In part 1 rCBF underestimation was depended on pulmonary gas exchange; high or low levels of rCBF; tissue type; and xenon inhalation protocols. In part 2 the mean rCBF underestimation was 18.8 +/- 8.3%. In conclusion, non-invasive estimate of the input function should be considered as a source of error when defining quantitative blood flow values e.g. the flow thresholds of ischaemic infarction.