OBJECTIVE. We performed a prospective clinical test of a high-spatial-resolution model-based parametric method for diagnosis of breast lesions detected on contrast-enhanced MR imaging. SUBJECTS AND METHODS. Fifty-seven women with 68 pathologically confirmed breast lesions were imaged (45 masses, 23 microcalcifications). Seven consecutive 2-min three-dimensional gradient-recalled echo acquisitions were performed after suitably timed gadopentetate dimeglumine injections. We derived a composite parametric image from three judiciously selected time points (three-time-point method), using a model-based kinetic algorithm. In this composite image, color brightness and hue signify contrast uptake and washout characteristics related to the product of microvessel surface area and permeability, as well as to the extracellular volume fraction. The reviewer was provided with the slice location of lesions, but with no other radiographic or clinical information. The reviewer then classified the lesions as benign or malignant using a 5-step receiver operating characteristic scale. RESULTS. Observers using the three-time-point method correctly diagnosed 27 of 31 malignant and 31 of 37 benign lesions (sensitivity, 87%; specificity, 84%). The area under the receiver operating characteristic curve was 0.911. False-negative results were found for three patients with low- to intermediate-grade ductal carcinoma in situ and one patient with 5-mm invasive ductal cancer. For the 45 solid lesions, sensitivity and specificity were 96% and 82%, respectively. CONCLUSION. Application of the three-time-point method permitted, in most cases, differentiation of malignant and benign lesions, even in the presence of complex breast enhancement patterns. Sensitivity for solid tumors was higher than for ductal carcinoma in situ.