Mammograms generated from DBT accurately represent density

Synthetic 2D mammograms built from digital breast tomosynthesis (DBT) imaging don’t change or mask the effects of breast density, soothing industry concerns about obfuscation according to a study published in the American Journal of Roentgenology.

Dense breasts present a problem for breast cancer screening: they both reduce the sensitivity of mammography and increase the risk of cancer, creating a sort of feedback loop. DBT is proliferating because it improves detection and recall rates in dense breasts, but the 3D images don’t always give radiologists the information they need, prompting DBT to be performed in conjunction with traditional mammography.

This approximately doubles the radiation dose received during the procedure, a major stumbling block for practices looking to adopt DBT. One way to reduce such exposure is to generate 2D mammography images from DBT’s 3D model by flattening dozens of layers and approximating the image. While studies have shown these synthetic 2D images have interpretation performance equal to traditional mammograms, the effects it has on dense breasts are less researched, according to the article.

A group of radiologists and a biostatistician from the University of Virginia School of Medicine reviewed more than 300 cases of women who were screened with a combination of traditional mammography and DBT, presenting the 2D and synthetic mammograms to a panel of three radiologists.

All 309 cases had a consensus reading, meaning at least two out of the three radiologists agreed on the density classification.

“Consistency in breast density interpretation has important implications in clinical practice, especially for identifying women who may benefit from supplemental screening,” wrote lead author and Visiting Assistant Professor in the Department of Breast Imaging Taghreed I. Alshafeiy, MD et al. “The results of our study suggest that the use of synthetic 2D images generated from tomosynthesis images does not significantly change the assignment of women to dense or nondense categories.”

This partially alleviates concerns of overtesting resulting from too many women being described as dense. The authors did note systematic variability in interreader agreement, but said it was caused by moderate differences in the appearance of the mammogram (film-screen, digital, or synthetic) and would have minimal clinical impact.

“Given the overall only moderate interreader agreement for both digital and synthetic 2D mammograms, systematic variability by some readers classifying more women with synthetic 2D mammograms as higher and some as lower density will have minimal clinical impact, wrote Alshafeiy et al. “Using synthetic 2D images will not result in a significant change in the number of women classified as having dense or nondense breast tissue; therefore, similar numbers of women would be considered for supplemental screening if synthetic 2D mammograms are used.”