Clinical implementation of synthesized mammography with digital breast tomosynthesis in a routine clinical practice.

Abstract

Most published studies evaluating digital breast tomosynthesis (DBT) included a separate 2-dimensional full-field digital mammogram (FFDM) for DBT screening protocols, increasing radiation from screening mammography. Synthesized mammography (SM) creates a 2-dimensional image from the DBT source data, and if used in place of FFDM, it reduces radiation of DBT screening. This study evaluated the implementation of SM+DBT in routine screening practice in terms of recall rates, cancer detection rates (CDR),% of minimal cancers, % of node-positive cancers, and positive predictive values (PPV). A multivariate retrospective institutional analysis was performed on 31,979 women who obtained screening mammography (10/2013-12/2015) with cohorts divided by modality (SM+DBT, FFDM+DBT, and FFDM). We adjusted for comparison mammograms, age, breast density, and the interpreting radiologist. Recall type was analyzed for differences (focal asymmetry, asymmetry, masses, calcifications, architectural distortion). SM+DBT significantly decreased the recall rate compared to FFDM (5.52 vs. 7.83%, p<0.001) with no differences in overall CDR (p=0.66), invasive and/or in situ CDR, or percentages of minimal and node-negative cancers. PPV1 significantly increased with SM+DBT relative to FFDM (9.1 vs. 6.2%, p=0.02). SM+DBT did not differ significantly in recall rate or overall CDR compared to FFDM+DBT. There were statistically significant differences in certain findings recalled by screening modality (e.g., focal asymmetries). SM+DBT reduces false positives compared to FFDM, while maintaining the CDR and other desirable audit outcome data. SM+DBT is more accurate than FFDM alone, and is a desirable alternative to FFDM+DBT, given the added benefit of radiation reduction.