Evaluation of the applicability of BI-RADS® MRI for the interpretation of contrast-enhanced digital mammography

Objective: To determine the usefulness of BI-RADS (R) MRI for the morphological description and categorization of images obtained with contrast-enhanced digital mammography in comparison with the final diagnosis. Material and methods: We included patients who had palpable breast lesions, those who needed clarification regarding abnormalities identified with another imaging technique, and those which needed a complementary evaluation of heterogeneously dense breast parenchyma. Three radiologists working independently used a template with the BI-RADS (R) MRI qualitative descriptors to evaluate the breast lesions studied with contrast-enhanced digital mammography. In a second phase, two other experienced radiologists reached a consensus about discrepant interpretations. Readers also classified each lesion (both benign and malignant lesions) on the BI-RADS (R) scale (1 - 5). All the results were compared with the real state of disease (determined by the appropriate gold standard for each type of lesion), and the statistical significance was assessed with the chi-square test. Results: A total of 218 benign lesions and 426 malignant lesions were included. The interobserver agreement among the three radiologists was high (Fleiss-Cohen kappa =0.805; 95% CI: 0.728-0.837). Similarly to what has been reported about breast MRI, on contrast-enhanced digital mammography, malignant lesions with mass effect tended to have an irregular shape with spiculated or ill-defined margins and a pattern of intense and heterogeneous enhancement (p <0.001). Nevertheless, unlike on breast MRI, ring enhancement was not an independent criterion of malignancy on contrast-enhanced digital mammography. For lesions without a mass effect, the only significant descriptor was the intensity of contrast material uptake (p < 0.05). Applying the BI-RADS (R) MRI material to contrast-enhanced digital mammography images enabled the correct classification of 85% of lesions in the benign categories (BI-RADS 1 and BI-RADS 2 and 3 lesions) and of 93% of the lesions in the malignant categories (BI-RADS 4-5); these values are similar to those reported for breast MRI. Conclusions: The morphological descriptors used in BI-RADS (R) MRI can be applied to the morphological analysis of breast lesions studied with contrast-enhanced digital mammography. The partial discrepancies in the interpretation did not influence the final BI-RADS (R) score, and the score enabled good differentiation between benign and malignant lesions. (C) 2019 SERAM. Published by Elsevier Espana, S.L.U. All rights reserved.