A Comparative Analysis of Digital Mammography and Digital Breast Tomosynthesis in Japanese Women

Abstract

Introduction Digital breast tomosynthesis (DBT) addresses the limitations of digital mammography (DM). While DM+DBT is gradually replacing DM in the United States and other Western countries, its adoption in Japan has been limited. This study assessed and compared the diagnostic performance of DM and DM+DBT, with a novel focus on breast density, reader experience, and workflow efficiency in Japanese women who underwent examinations during our early experience shortly after the introduction of DBT. Methods A retrospective review of 48 patients (19 with breast cancer and 29 with benign lesions) who underwent both DM and DBT at Jyoban Hospital was conducted. Five readers (two physicians, one medical resident, and two radiologic technologists) assessed the images. Diagnostic accuracy was evaluated using receiver operating characteristic (ROC) curve analysis for breast density and reader experience in both screening and clinical settings. Inter-reader agreement and reading times were assessed, with paired t-tests used to analyze differences in reading times. Results DM+DBT achieved higher inter-reader agreement than DM (Kappa: 0.708, 95% CI: 0.567-0.849 vs. 0.661, 95% CI: 0.511-0.811), though this difference was not statistically significant (p = 0.653). In screening settings, DM+DBT significantly improved the area under the curve (AUC; 0.750, 95% CI: 0.717-0.784 vs. 0.709, 95% CI: 0.675-0.743; p = 0.005); in clinical settings, the increase was not significant (0.878, 95% CI: 0.834-0.921 vs. 0.844, 95% CI: 0.797-0.891; p = 0.153). For dense breasts, AUC improved notably with DBT (screening: 0.766, 95% CI: 0.699-0.833 vs. 0.684, 95% CI: 0.612-0.756; p = 0.007; clinical: 0.936, 95% CI: 0.883-0.990 vs. 0.827, 95% CI: 0.740-0.914; p = 0.023), but no significant differences were observed in non-dense breasts. All readers improved with DBT, except the medical resident. Reading times increased significantly, from 32.5-82.5 seconds (DM) to 71.3-113.8 seconds (DM+DBT) (p = 0.006). Conclusion DM+DBT improves diagnostic accuracy, particularly for dense breasts in Japanese women, but longer reading times and reader experience may limit its widespread adoption.

Keywords: breast cancer; digital breast tomosynthesis; digital mammography; roc analysis; screening.

Figure 1. Reader-specific performance analysis of digital mammography and digital breast tomosynthesis using receiver operating characteristic curves†.

† In the ROC curve for the medical resident, the dashed lines representing the DM+DBT clinical setting and the DM clinical setting are completely overlapping. DM: Digital Mammography; DBT: Digital Breast Tomosynthesis; ROC: Receiver operating characteristic.

Figure 2. Comparative diagnostic performance of digital mammography versus digital breast tomosynthesis using receiver operating characteristic curves.

ROC: Receiver operating characteristic; DM: Digital Mammography; DBT: Digital Breast Tomosynthesis.

Figure 3. Impact of breast density on diagnostic performance: receiver operating characteristic analysis of digital mammography versus digital breast tomosynthesis.

ROC: Receiver operating characteristic; DM: Digital Mammography; DBT: Digital Breast Tomosynthesis.