Medullary perfusion potentially is an early marker of renal damage, but cannot be measured using existing models in MR renography. In this study a novel 7-compartment, 10-parameter model is proposed and evaluated in synthetic and patient data. Robustness of an iterative fitting approach was assessed by the coefficient of variation over multiple fits, resulting in median values <2.5% both in synthetic and patient data. According to simulations, medullary perfusion was underestimated by 5.3%, but in diabetic patients medullary perfusion was relatively high (81 mL/100mL/min). Future studies will be needed to determine this model’s sensitivity to pathophysiological changes in medullary perfusion.