Hepatic arterial infusion of floxuridine has shown benefit in intrahepatic cholangiocarcinoma. Due to the complexity and expense of this treatment, a pre-treatment or early biomarker of efficacy would be advantageous. We employed non-model-based (NMB) analysis to maximize information from breath-motion degraded DCE-MRI in 24 patients. Model-based and NMB parameters were compared to RECIST response. The kurtosis value of pre-treatment time-to-half-maximum correlated with response, as did 1 month changes in perfusion parameters (Ktrans and signal at half maximum). NMB analysis resulted in fewer voxels being discarded due to motion and thus may be more representative of tumor physiology while not requiring modeling.