Classification of fatty and dense breast parenchyma: comparison of automatic volumetric density measurement and radiologists’ classification and their inter-observer variation

Abstract

The optimal volumetric threshold of 10% using automatic assessment would classify breast parenchyma as fatty or dense with substantial accuracy and consistency compared to radiologists' BI-RADS categorization, which suffers from high inter-observer variation.

Evaluation of Quantra Hologic Volumetric Computerized Breast Density Software in Comparison With Manual Interpretation in a Diverse Population

“…The single radiologist read also poses a potential limitation of this study in that it decreases possible application to other clinics/radiology practices. However, it is reassuring that the findings were consistent with the published studies from Amsterdam and Oslo, 18 , 19 as well as with the unpublished United States–based study of Rafferty et al 21 It remains a conundrum why these 3 studies and our study are at variance with the studies from Turin, New Delhi, and Sydney. 16 , 17 , 20 , 23 …”

“…Because Q D increases in half-percent increments, this means that if a patient’s Q D was 13.5% or less, she belonged to the D1-2 category, but if her Q D was 14.0% or more, then she belonged to the D3 category. This finding of a 14% cutoff is similar to the 13.8% cutoff reported by the Amsterdam study 18 and the 10% cutoff reported by the Oslo study, 19 and also similar to the unpublished 13% cutoff attributed by Ciatto et al 20 to the United States–based study of Rafferty et al 21 In contrast, the independent studies done by Ciatto et al 20 and Regini et al 23 in an Italian population reported 22% and 21%, respectively, as the optimal Quantra density cutoff value. Also in contrast were the high Quantra density cutoffs of 19.5% and 20%, respectively, reported by the studies from India and Australia.…”

“… 16 , 17 A third study from New Delhi, India, 16 reported a Quantra cutoff of 19.5%, whereas a fourth study from Sydney, Australia, 17 reported Quantra cutoffs of 20.0% overall and 20.5% in a validation subset. In contrast, a fifth study from Amsterdam, the Netherlands, 18 reported a Quantra cutoff of 13.8%, whereas a sixth study from Oslo, Norway, 19 reported a Quantra cutoff of 10.0% versus the median BI-RADS classification from 5 radiologists, along with cutoffs ranging from 8% to 15% versus BI-RADS classifications from the radiologists individually. In addition, one of the Turin studies contrasted their 22% cutoff with a much lower cutoff of 13% apparently shown in a 2009 oral presentation by investigators in a United States–based study.…”

Positive predictive values by mammographic density and screening mode in the Norwegian Breast Cancer Screening Program

“…No studies have previously examined the three-point scale in relation to estimates of volumetric breast density obtained by the fully automated software. However, several studies have compared fully automated methods for breast density assessment to BI-RADS (85,96,102,103,105,(156)(157)(158)(159)(160)(161)(162), indicating a clear increase in volumetric breast density as the BI-RADS density category increased (103) or a strong relationship between volumetric breast density and the BI-RADS categories (96). Similar trends have been reported in other studies (102,(157)(158)(159)162).…”

“…This misclassification could have weakened the agreement between BI-RADS and VDG. A previous study from Norway using the dichotomized BI-RADS density classification (1 and 2 versus 3 and 4) showed that the agreement (measured by kappa with quadratic weights) between reader's median classification and an individual score of five radiologists ranged between 0.76 and 0.93 (176). Thus, even though strong, this agreement is not perfect, and we may therefore have underestimated the agreement between BI-RADS and VDG somewhat in Study IV.…”