Assessing the assessments: Radiologists vs. automated breast density software

Breast density software programs from different vendors can reach conflicting conclusions when categorizing breast tissues as either “dense” or “non-dense” under current BI-RADS guidelines and can differ from traditional visual assessments from radiologists, according to results of a study published online in the American Journal of Roentgenology.

There is growing evidence that breast density can be a reliable indicator of risk regarding the development of breast cancer, with some estimates assessing the risk for women with dense breasts at 4-6 times higher than those with typical fatty breast tissues.

Traditional methods for assessing breast density—such as visual assessments and quantitative area–based measurements—have distinct limitations, including variability, subjectivity and irreproducibility, said lead author Ji Hyun Youk, MD, and his colleagues from Yonsei University College of Medicine in Seoul, South Korea.

“To overcome those limitations, fully automated methods for volumetric breast density estimation from digital mammograms have been developed and shown to be highly reproducible,” they wrote. “In addition, it is feasible to obtain quantitative measurements of dense breast tissue volume with results that are in good agreement with the BI-RADS breast density categories.”

Youk and his team set out to compare automated volumetric measurements from two commercially available software programs with visual assessments of breast density using mammography according to the BI-RADS guidelines. To do so, they performed a retrospective analysis of 1,185 full-field digital mammography examinations using radiologists’ visual assessments as well as two  automated breast density assessment programs—Hologic’s Quantra software and Matakina Technology’s Volpara software.

They found that the agreement of density category ranged from “moderate to substantial” in Quantra and “fair to moderate” in Volpara when compared to standard visual assessment, with significant differences in the distribution of density categories among visual and volumetric measurements. The Quantra software classified breast tissue as “non-dense” more frequently than both radiologists and Volpara, while the latter was more likely to assess a “dense” rating to breast tissues.

“More mammographic examinations were classified as non-dense breast tissue using the Quantra software and as dense breast tissue using the Volpara software, as compared with visual assessments according to the BI-RADS,” the authors concluded, adding that “the difference in the distribution of breast density categories in comparison with visual assessment and in volumetric breast density data between the two software programs might be attributable to an intrinsic difference in the algorithms’ abilities to estimate fibroglandular tissue volume or total breast volume and the established thresholds for density categories.”