[Diagnostic value of a breast MRI score for the prediction of malignancy of breast lesions detected solely with MRI].

Purpose: To determine whether lesion scoring allows valid BI-RADS ((R)) classification and prediction of malignancy of breast lesions detectable solely with MRI.

Materials And Methods: Retrospective analysis of 86 patients who underwent MRI-guided localization and excisional biopsy of 100 breast lesions detectable only with MRI. Breast MRI was performed at 1.5 Tesla by means of a T 1w dynamic GE sequence. The positive predictive value (PPV) for malignancy was calculated for each score criterion separately, for the total score, and for corresponding BI-RADS classes. The PPV was correlated with histology.

Results: 31% (31/100) of breast lesions detectable only with MRI were malignant. Of the single score criteria only the lesion morphology criterion was significantly correlated with malignancy. The malignancy rate according to the breast MRI score was 0% for score 3, 18.2% for score 4, 32.1% for score 5 and each 50% for lesions of score 6 and 7. After translation into BI-RADS the malignancy rates were 0% for BI-RADS 3, 24.6% for BI-RADS 4 and 48.5% for BI-RADS 5 lesions. The thus defined BI-RADS classes were significantly correlated with malignancy.

Conclusion: The combination of different lesion criteria to form a total breast MRI score and its translation into BI-RADS is useful in case of lesions detectable only with MRI. It enables standardized BI-RADS classification with satisfying PPV of malignancy for each BI-RADS class. Because of their low malignancy rate (18.2%), we suggest classifying lesions with a breast MRI score of 4 points as BI-RADS 4a.