REPRODUCIBILITY OF MAMMOGRAPHIC PARENCHYMAL PATTERNS AND BREAST DENSITY ON FULL-FIELD DIGITAL MAMMOGRAPHY (FFDM): COMPARISON OF TWO CLASSIFICATION SYSTEMS

Abstract

Objective: The aim of the study is to compare inter-observer agreement between general radiologists in the classification of mammographic breast density using TABAR’s pattern and BI-RADS classification from two-view Full-Field Digital Mammography (FFDM). Methods: A data set of 400 mammograms was evaluated by three general radiologists. The radiologists independently reviewed the images and classified the parenchymal pattern according to BI-RADS and TABAR classification systems. Inter-observer agreements were analyzed using kappa statistics. Results: Inter-observer agreement for the BI-RADS is slight to fair (Reviewer 1 versus Reviewer 2: k=0.19, Reviewer 1 versus Reviewer 3, k=0.07 and Reviewer 2 versus Reviewer 3, k=0.49) and for TABAR is fair to moderate (Reviewer 1 versus Reviewer 2: k=0.23, Reviewer 1 versus Reviewer 3, k=0.31 and Reviewer 2 versus Reviewer 3, k=0.50). Conclusion: Our study demonstrates a poor level of agreement in breast parenchymal pattern and density based on both BI-RADS and TABAR classifications. Thus, breast density in risk stratification of breast cancer should be used with caution in our local practice.