Our study was undertaken to develop diagnostic rules and to assess the reproducibility of dynamic and morphologic parameters for the characterization of suspicious breast lesions using dynamic high-spatial-resolution MR imaging. Fifty-seven patients with suspicious mammographic or palpable findings underwent preoperative contrast-enhanced MR imaging of the breast using a three-time-point method of acquisition. Each lesion was prospectively analyzed by two independent radiologists for morphologic and visual dynamic enhancement characteristics. A classification and regression tree was used to examine the optimal order, cutoff points, and combination of imaging parameters to build a diagnostic rule separating benign from malignant lesions using histopathology findings as the standard of reference. Kappa statistics were used to determine observer variability. Among 23 benign and 34 malignant lesions (12 invasive, three ductal carcinoma in situ, and 19 mixed cancer), margin morphology (p = 0.001) and enhancement pattern (p = 0.001) were the most significant MR imaging findings for lesion characterization. Focal mass lesions were classified as malignant when spiculated margins or both the washout enhancement pattern and "nonsmooth" margins were present. Interobserver agreement was almost perfect for washout pattern and substantial for margin assessment. In the limited population tested retrospectively, the diagnostic rule yielded a sensitivity and positive predictive value of 97% each and a specificity and negative predictive value of 96% each. The washout enhancement pattern combined with lesion margin assessment on dynamic contrast-enhanced high-resolution MR imaging of the breast allows reproducible lesion characterization and may be a highly specific diagnostic tool.