1 Center for Evidence Based Imaging, Department of Radiology, Harvard Medical School, Brigham and Women's Hospital, 75 Francis St, Boston, MA 02115.
2 Present address: Department of Radiology, Yale University School of Medicine, New Haven, CT.
AJR Am J Roentgenol. 2017 Oct;209(4):929-934. doi: 10.2214/AJR.17.17983. Epub 2017 Jun 22.
The purpose of this study is to compare the diagnostic performance of screening digital breast tomosynthesis (DBT) to that of full-field digital mammography (FFDM) in a mixed DBT and FFDM imaging environment.
This retrospective observational study consisted of all female patients undergoing screening DBT or FFDM at an academic medical center and outpatient imaging facility between October 2012 and May 2015. Patient demographics and personal history of breast cancer were collected from the electronic medical record. A natural language processing algorithm extracted patients' breast density, current or prior imaging findings, and BI-RADS category from their most recent prior imaging examinations. To control for differential selection of FFDM versus DBT, we applied propensity score matching based on patient age, imaging site, and prior imaging findings. An institutional breast cancer registry identified cancer diagnoses. Primary outcomes of recall rate, cancer detection rate, and positive predictive value 1 (PPV1) were compared between matched FFDM and DBT groups.
Among 68,794 screening examinations, we matched 16,264 FFDM with 21,074 DBT examinations (total, 37,338 examinations) using nearest neighbor propensity score matching. Recall rates were 10.3% (1683/16,264) for FFDM and 10.7% (2254/21,074) for DBT (p = 0.26). Cancer detection rates (number of cancers/1000 examinations) were 1.8/1000 for FFDM and 3.8/1000 for DBT (p = 0.005). The PPV1 (number of cancers/number of recalls) was 1.8% (26/1478) for FFDM and 3.6% (37/1036) for DBT (p = 0.006).
FFDM and DBT recall rates were not significantly different in a mixed FFDM and DBT breast imaging practice. However, the PPV1 of recalled cases and the cancer detection rate (the primary screening objective) were significantly higher with DBT compared with FFDM.
本研究旨在比较混合数字乳腺断层摄影术（DBT）和全数字化乳腺摄影术（FFDM）在混合 DBT 和 FFDM 成像环境中的诊断性能。
这是一项回顾性观察研究，纳入了 2012 年 10 月至 2015 年 5 月期间在一家学术医疗中心和门诊影像机构接受筛查性 DBT 或 FFDM 的所有女性患者。患者的人口统计学数据和乳腺癌个人病史从电子病历中收集。自然语言处理算法从最近的影像学检查中提取患者的乳腺密度、当前或既往影像学表现以及 BI-RADS 分类。为了控制 FFDM 与 DBT 的选择差异，我们根据患者年龄、影像学部位和既往影像学表现应用倾向评分匹配。机构乳腺癌登记处确定了癌症诊断。比较匹配的 FFDM 和 DBT 组之间的召回率、癌症检出率和阳性预测值 1（PPV1）。
在 68794 次筛查检查中，我们使用最近邻倾向评分匹配将 16264 例 FFDM 与 21074 例 DBT 匹配（共 37338 例检查）。FFDM 的召回率为 10.3%（1683/16264），DBT 为 10.7%（2254/21074）（p=0.26）。FFDM 的癌症检出率（每 1000 例检查中的癌症例数）为 1.8/1000，DBT 为 3.8/1000（p=0.005）。FFDM 的 PPV1（每 1000 例召回病例中的癌症例数）为 1.8%（26/1478），DBT 为 3.6%（37/1036）（p=0.006）。
在混合 FFDM 和 DBT 乳腺成像实践中，FFDM 和 DBT 的召回率没有显著差异。然而，与 FFDM 相比，DBT 检查的病例召回的 PPV1 和癌症检出率（主要的筛查目标）显著更高。
