LAUSR

AIM To evaluate whether digital breast tomosynthesis (DBT) can predict if circumscribed masses are benign or malignant by assessing margin sharpness. MATERIALS AND METHODS Circumscribed masses were evaluated on co-registered two-dimensional digital mammography (2DDM) and DBT. Lesions were categorised as follows: category 1=visible sharp border 0-25% of the total margin; category 2 = 26-50% category 3= 51-75%, and category 4=76-100%. Changes in category between 2DDM and DBT were analysed; if the category was lower on DBT the change was negative, if higher the change was positive. RESULTS Of 759 lesions, 121 masses classified as circumscribed on DBT were included; 25 were malignant and 96 benign. Of the benign lesions, 8/96 were within category 3 or 4 on 2DDM compared with 48/96 benign lesions within category 3 or 4 on DBT (Fisher's exact test p<0.000527). Forty-eight of 51 (94.1%) lesions categorised as 3 or 4 on DBT were benign and 65/67 (97.01%) of the positive category change group were benign. Lesions in category 1 on DBT had 45.4% chance of being malignant (20/44) compared with 22.72% (20/88) on 2DDM (chi-squared test p<0.001). Sixty-five of 67 (97.01%) lesions in the positive category change group were benign and 23/54 (42.6%) lesions with either no or negative category change were malignant. CONCLUSION The present study demonstrates 97% accuracy in predicting circumscribed lesions as benign when using positive category change and 94% accuracy when >50% of the margin is sharply defined on DBT.