Characterization of breast masses by dynamic enhanced MR imaging - A logistic regression analysis

Purpose. To identify features useful for differentiation between malignant and benign breast neoplasms using multivariate analysis of findings by MR i maging. Material and Methods: In a retrospective analysis, 61 patients with 64 brea st masses underwent MR imaging and the time-signal intensity curves for pre contrast dynamic postcontrast images were quantitatively analyzed. Statisti cal analysis was performed using a logistic regression model, which was pro spectively tested in another 34 patients with suspected breast masses. Results: Univariate analysis revealed that the reliable indicators for mali gnancy were first the appearance of the tumor border, followed by the washo ut ratio, internal architecture after contrast enhancement, and peak time. The factors significantly associated with malignancy were irregular tumor b order, followed by washout ratio, internal architecture, and peak time. For differentiation between benignity and malignancy, the maximum cut-off poin t was to be found between 0.47 and 0.51. In a prospective application of th is model. 91% of the lesions were accurately discriminated as benign or mal ignant lesions. Conclusion: Combination of contrast-enhanced dynamic and postcontrast-enhan ced MR imaging provided accurate data for the diagnosis of malignant neopla sms of the breast. The model had an accuracy of 91% (sensitivity 90%, speci ficity 93%).