Breast density: Comparison of chest ct to mammography

BREAST DENSITY: COMPARISON OF CHEST CT TO MAMMOGRAPHY Breast density is an important factor to establish the optimal imaging strategy for breast cancer screening. Each year, many patients undergo chest CT imaging with various indications. This presents a unique chance to evaluate breast density before even first mammograms for screening are taken. In this study, we aimed to classify breast density from CT images and to determine the agreement of this classification with mammographic density findings as well as evaluate intra and interobserver variability of both methods. 209 patients with mammograms and thorax CTs taken in our department between November 2012-December 2014 had their CT and MMG images retrospectively evaluated for breast density by two radiologists. Breast density was categorized according to BI-RADS in mammographies and a classification with 4 categories that corresponds to BI-RADS categories was used for CT evaluation. A second classification scheme where grade 1 and 2 breasts were considered low density and grade 3 and 4 breasts high density was also used for CT evaluation. The agreement of mammography and CT findings was quite high (κ 0,639-0,691). With low density-high density classification, there was substantial agreement in MMG and CT evaluations of U.K and the first evaluation of O.K (κ 0,692 and 0,794 respectively) whereas O.K's second evaluation demonstrated excellent agreement (κ=0,853) of the two modalities. Mamographic evaluation had substantial inter-observer agreement (κ=0,680 for grade 1-4 classification and 0,686 for low-high density classification); however, CT had only a moderate inter-observer agreement (κ=0,480 for grade 1-4 classification and 0,491 for low-high density classification). Intra-observer agreement was substantial for CT (κ=0,616 for grade 1-4 classification and 0,765 for low-high density classification) and substantial with the first classification (κ=0,707) and excellent for the second classification scheme (κ=0,804)in MMG. With mammography considered the gold standard, CT had a sensitivity of 74,5-96,6;% specificity of 80,5-94,4%; positive predictive value of 88.5-92,7% and negative predictive value of 92,7-95,3% for differentiating between low density and high density breasts. The results of our study show that density evaluation in CT is compatible with mamographic findings. The agreement values of low denstiy-high density classification is higher than those of grade 1-4 classification. Our results also imply that the experience of the radiologist is quite important in tomographic evaluation. Cases where radiologists did not agree mostly had breast densities of grade 2 and 3. Since this corresponds to the distinction of low and high density breasts, more objective criteria to differentiate between them are required. Due to increased risk of breast cancer, patients with dense breasts on thorax CT images, with other risk factors also taken into consideration, can be considered for screening programs appropriate for their age. However; larger patient series on evaluation of breast density on CT, especially where the effect of the radiologist's experience on the results are evaluated are required.