Abstract Archives of the RSNA, 2009
Assessing Radiologist Performance in Dense versus Fatty Breasts Using Combined Full-Field Digital Mammography and Breast Tomosynthesis Compared to Full-Field Digital Mammography Alone
Presented on December 1, 2009
Presented as part of VB31: Breast Series: High-Risk Screening
Elizabeth Ann Rafferty MD, Presenter: Research grant, Hologic, Inc
Andrew Paul Smith PhD, Abstract Co-Author: Employee, Hologic, Inc
Loren Thomas Niklason PhD, Abstract Co-Author: Employee, Hologic, Inc
Patent holder, Hologic, Inc
Patent holder, Breast Tomosynthesis
The purpose of this study was to compare radiologists’ ROC area under the curve (AUC) using conventional full-field digital mammography (FFDM) plus breast tomosynthesis (Tomo) to the ROC area under the curve using FFDM alone as a function of breast density.
Fifteen radiologists evaluated 310 cases. These cases included cancer, benign, recalled screening and negative screening cases. The images were read in three modes; 1) FFDM alone, 2) FFDM plus MLO Tomo and 3) FFDM plus Tomo (both CC and MLO). For each case, the radiologists recorded a Probability of Malignancy (POM) score of 0-100. The POM scores were used for ROC analysis. Breast density based on the FFDM images was used to classify the cases as fatty (BIRADS density 1 and 2) or dense (BIRADS density 3 and 4).
52% of the cases were classified as fatty and 48% as dense. For fatty breasts the ROC AUC was 0.880 for FFDM, 0.898 for FFDM plus MLO Tomo and 0.915 for FFDM plus Tomo. For dense breasts the ROC AUC was 0.786 for FFDM, 0.832 for FFDM plus MLO Tomo and 0.877 for FFDM plus Tomo. FFDM plus Tomo was statistically superior to FFDM alone for both fatty (AUC difference 0.035, p-value 0.0008) and dense breasts (AUC difference 0.091, p-value 0.0001). FFDM plus MLO Tomo was superior to FFDM alone for dense breasts (AUC difference 0.046, p-values 0.041) but not for fatty breasts (AUC difference 0.018, p-value 0.077). FFDM plus Tomo was significantly better than FFDM plus MLO Tomo in dense breasts (AUC difference 0.045, p-value 0.043) but not for fatty breasts (AUC difference 0.17, p-value 0.104).
FFDM plus Tomo was significantly better than FFDM in ROC performance in both fatty and dense breasts. However, the gain in ROC AUC was more than twice as much in dense breasts compared to fatty breasts. The addition of one view Tomo provided about one half of the gain in ROC performance as obtained with two Tomo views. FFDM plus Tomo has ROC performance in dense breasts equivalent to FFDM alone in fatty breasts. Therefore the addition of tomosynthesis may provide a significant increase in breast cancer detection for women with dense breasts.
Tomosynthesis used in combination with FFDM will improve the sensitivity of breast cancer screening in women with dense breasts.
Assessing Radiologist Performance in Dense versus Fatty Breasts Using Combined Full-Field Digital Mammography and Breast Tomosynthesis Compared to Full-Field Digital Mammography Alone. Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8016152.html