Digital breast tomosynthesis improves the accuracy of the diagnosis of circumscribed lesions because of increase of margin visibility

Introduction

Circumscribed masses are a common cause of recall for assessment but most are benign. The aim of this study is to compare the accuracy of digital breast tomosynthesis (DBT) with digital mammography (DM) in assessing the margin of circumscribed lesions as a predictor of benign or malignant disease.

Methods

The study group consisted of women recalled from breast screening for further assessment of circumscribed masses. Clients underwent co-registered DM and DBT in both MLO and CC projections. Two experienced breast radiologists evaluated DM and DBT images and a consensus decision was reached on the percentage of the margin that was well defined on DM and DBT. The lesions were categorised 1 to 4 as follows: 1 = 0 to 25%, 2 = 26 to 50%, 3 = 51 to 75% and 4 = 76 to 100%.

Results

One hundred and twenty circumscribed lesions were evaluated. Data on 118 lesions seen on the MLO view are presented. There were 93 benign lesions and 25 cancers. There was a change in distribution of margin categories between DM and DBT. More lesions were categorised as 3 or 4 on DBT (59/118) compared with DM (18/118). Of the 93 benign lesions, 17 were categorised as 3 or 4 on DM and 57 on DBT. The difference between the two proportions was significant (P < 0.0002). There were more cancers categorised as 1 on DBT (21/39 = 54%) compared with DM (20/76 = 26%); Fisher's exact test (P < 0.004).

Conclusion

Increased margin visibility of circumscribed masses by DBT improves the accuracy of mammography interpretation and may decrease the recall rate in mammography screening.