[99mTc]-d,l-HM-PAO (HM-PAO) was injected rapidly into the internal carotid artery and its retention in the brain was recorded by external scintillation cameras in eight human subjects. A model is described based on three compartments: the lipophilic tracer in the blood pool of the brain, the lipophilic tracer inside the brain, and the hydrophilic form retained in the brain. The retention curve initially drops abruptly, corresponding to the nonextracted fraction of the injectate leaving the brain; it then falls exponentially towards the asymptotic level of the fractional steady-state retention R. Cerebral blood flow (F) was measured using the xenon-133 intracarotid injection method. The first-pass extraction E of HM-PAO was calculated from F using an empiric regression equation. The residue curves for the whole brain after intracarotid HM-PAO injection were analyzed to yield a retention fraction (R') and the brain clearance backflux constant of lipophilic HM-PAO (k). From the kinetic model and the measured values of R', k, and F, the following parameter values could be calculated: the average retained fraction of all tracer supplied to the brain, R = 0.38 +/- 0.05 (mean +/- SD), the conversion rate constant (lipophilic to hydrophilic tracer) in the brain k3 = 0.80 +/- 0.12 min-1, the efflux rate constant (brain to blood) k2 = 0.69 +/- 0.11 min-1, the conversion/clearance ratio alpha = k3/k2 = 1.18 +/- 0.25, the influx (blood clearance) constant K1 = 0.45 +/- 0.11 ml/g/min, and the brain/blood partition ratio lambda = K1/k2 = 0.67 +/- 0.23 ml/g. Using the kinetic model and assuming constancy of alpha, an algorithm was developed that corrects for the blood flow dependent backflux of HM-PAO and results in a more linear relation between regional cerebral blood flow (rCBF) and HM-PAO distribution.