PURPOSE: To investigate use of the macromolecular contrast agent feruglose for differentiating and grading of human benign and malignant breast tumors on the basis of their microvascular characteristics. MATERIALS AND METHODS: Sixty-three women with 63 primary breast lesions were examined with dynamic T1-weighted gradient-echo magnetic resonance (MR) imaging after intravenous injection of feruglose. A two-compartment unidirectional kinetic model applied to the dynamic data yielded estimates of the endothelial transfer coefficient, KPS, and the fractional plasma volume of the tumors. These MR imaging-derived parameters were correlated with the histologic tumor grade and quantified according to the Scarff-Bloom-Richardson (SBR) score by means of Pearson product moment correlation analyses. Differences between malignant and nonmalignant breast lesions with respect to KPS for feruglose were evaluated by means of the chi2 test and by calculating the sensitivity, specificity, and positive and negative predictive values. RESULTS: Histologic analysis revealed 26 benign and 37 malignant tumors. A moderate yet statistically significant correlation between KPS and SBR score was found (R = 0.496, P<.001). No significant correlation was observed between fractional plasma volume and SBR score (R = 0.085, P =.507). The KPS values were zero for 19 (73%) of the 26 benign tumors and were greater than zero for 27 (73%) of the 37 carcinomas. This distribution was significantly different (chi2 = 13.035, P =.001). With the criterion KPS> 0 in carcinomas, sensitivity was 0.73, specificity was 0.73, and the positive predictive value was 0.79. CONCLUSION: Quantitative measures of tumor microvascular permeability can be used for breast tumor characterization. The probability of breast tumor microvascular hyperpermeability to be associated with malignancy is 79%.