Comparison Between Digital and Synthetic 2D Mammograms in Breast Density Interpretation.

Abstract

The purpose of this study was to compare assessments of breast density on synthetic 2D images as compared with digital 2D mammograms. This retrospective study included consecutive women undergoing screening with digital 2D mammography and tomosynthesis during May 2015 with a negative or benign outcome. In separate reading sessions, three radiologists with 5-25 years of clinical experience and 1 year of experience with synthetic 2D mammography read digital 2D and synthetic 2D images and assigned breast density categories according to the 5th edition of BI-RADS. Inter- and intrareader agreement was assessed for each BI-RADS density assessment and combined dense and nondense categories using percent agreement and Cohen kappa coefficient for consensus and all reads. A total of 309 patients met study inclusion criteria. Agreement between consensus BI-RADS density categories assigned for digital and synthetic 2D mammography was 80.3% (95% CI, 75.4-84.5%) with κ = 0.73 (95% CI, 0.66-0.79). For combined dense and nondense categories, agreement reached 91.9% (95% CI, 88.2-94.7%). For consensus readings, similar numbers of patients were shifted between nondense and dense categories (11 and 14, respectively) with the synthetic 2D compared with digital 2D mammography. Interreader differences were apparent; assignment to dense categories was greater with digital 2D mammography for reader 1 (odds ratio [OR], 1.26; p = 0.002), the same for reader 2 (OR, 0.91; p = 0.262), and greater with synthetic 2D mammography for reader 3 (OR, 0.86; p = 0.033). Overall, synthetic 2D mammography is comparable with digital 2D mammography in assessment of breast density, though there is some variability by reader. Practices can readily adopt synthetic 2D mammography without concern that it will affect density assessment and subsequent recommendations for supplemental screening.