Clinical diagnostic accuracy and non-inferiority of digital breast tomosynthesis by synthetic 2D mammography compared with full-field digital mammography for the diagnosis of breast cancer: a single-center retrospective study.

Background: Synthetic two-dimensional (2D) mammography (SM), which is reconstructed from digital breast tomosynthesis (DBT) data, has been developed as a possible substitute for full-field digital mammography (FFDM). Our purpose was to compare the diagnostic accuracy and non-inferiority of S-view SM with FFDM combined with DBT in the detection of breast cancers, in addition to evaluating inter-observer agreement for the visibility score of suspicious calcifications on S-view SM images. Methods: Between July 2019 and June 2021, a retrospective study was conducted, examining both screening and diagnostic mammography using Amulet Innovality, Fujifilm, with DBT, S-view SM, and FFDM. A total of 485 women with ages ranging from 31 to 90 years old, consisting of 177 breast malignancies from 175 patients, 70 biopsy-proven benign cases, and 240 non-biopsy benign cases, were included in the study. Two observers reviewed S-view SM plus DBT and FFDM plus DBT independently, with a 4-week washout period between assessments. Sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC) were compared using McNemar's test and DeLong test, respectively. The non-inferiority of S-view SM plus DBT to FFDM plus DBT was also assessed with a one-sided confidence interval (CI). Additionally, inter-observer agreement on the visibility of suspicious calcification on S-view SM was measured using Cohen's kappa. Results: Sensitivity, specificity, and AUC for S-view SM plus DBT were 98.9% (95% CI: 96.0-99.9%), 85.2% (95% CI: 80.7-88.9%), and 0.97 (95% CI: 0.96-0.98), respectively. When compared to FFDM plus DBT, the sensitivity was 99.4% (95% CI: 96.9-100%, P value >0.999), the specificity was 84.8% (95% CI: 80.0-88.6%, P value >0.999), and the AUC was 0.97 (95% CI: 0.96-0.98, P value =0.40). No statistically significant difference in clinical performance was observed between S-view SM plus DBT and FFDM plus DBT. The non-inferiority of the sensitivity and specificity of S-view SM plus DBT compared to FFDM plus DBT was proven, with lower CIs of 96.0% and 80.7%, respectively. Similarly, the non-inferiority analysis for AUC also demonstrated consistent results, with a lower CI of 0.87, confirming that S-view SM plus DBT is not inferior to FFDM plus DBT. Interrater agreement on the assessment of morphology classification and distribution of suspicious calcification was excellent (correlation coefficient =0.96). Conclusions: S-view SM has a non-inferior clinical performance to FFDM for the detection of breast cancers when used in combination with DBT. The utilization of SM alongside DBT emerges as a favorable option, offering the potential to mitigate radiation dose without compromising diagnostic precision.