We analyzed magnetic resonance imaging (MRI) morphologic patterns of retroperitoneal fibrosis (RF) to identify those able to distinguish malignant RF (mRF) from idiopathic RF (iRF). This retrospective study concerned 50 consecutive patients with MRI-based RF diagnoses, 35 of whom also had histologically proven RF. Previous radiotherapy, abdominal or pelvic surgery or infection during the preceding 6 months, vascular aneurysm (aorta or iliac artery), presence of retroperitoneal multiple nodular masses, or enlarged lymph nodes with a diameter >15 mm constituted exclusion criteria. Patients with mRF differed from those with iRF by age, smoking habits, and follow-up duration but not by clinical manifestations, inflammatory syndrome, or renal insufficiency. MRI-documented mRF extension along the aorta, from above the renal arteries to below the aortic bifurcation, was more frequent than iRF (47% vs. 0%; p = 0.001) but less frequent between the renal arteries and the aortic bifurcation (18% vs. 50%; p = 0.04); mRF extension behind the aorta was wider than iRF (5.0 vs. 2.5 mm; p = 0.03). Neither urinary tract nor vessel involvement differed. Medial ureteral attraction was significantly less frequent in mRF than iRF (24% vs. 83%; p < 0.001), according to univariate and multivariate analyses. An algorithm based on the most discriminant criteria (RF extending from above the renal arteries to below the aortic bifurcation and the absence of medial ureteral attraction) for mRF diagnosis had 82% sensitivity and 83% specificity. When applied to the 15 iRF patients without histologic data, specificity was 73%. This mRF decision tree, consisting of the 2 most discriminant MRI criteria, could be used as a supplementary argument to support RF biopsy.