The arterial input function measured for brain dynamic contrast-enhanced MRI is contaminated by the signal contribution of surrounding tissues. This work corrects these partial volume effects on signal level by using the surrounding gray matter enhancement to discriminate pure arterial signal. The method also accounts for the high contrast agent concentration reached in arteries and veins that leads to signal non-linearity, saturation, and concurrent unwanted $$$T_2^*$$$ effects. This partial volume correction method is compared to concentration scaling on a digital reference object and on eight subjects. Better recovery of the arterial first pass and recirculation are shown.