Research has long implicated breast density as an important risk factor for breast cancer (BC). Compared to other risk factors, such as age, family history, hormone exposure, parity, etc., breast density shows an equivalent or greater association with BC risk, with odd ratios ranging from two through six or higher. Nevertheless, the breast density-BC risk association is complicated by several facts. First, dense mammographic tissue is present, to some degree, in the vast majority of women. Second, reported associations between breast density measures and BC risk show broad variability across studies, third, breast tissue is a biomarker both for BC risk and for past and present hormonal interactions. Moreover, currently there are no universally applied or accepted standards for measuring breast density. In this work we will build on our previous work in calibrated tissue metrics and the automated analysis of digitized mammograms and fully develop an approach for automatically assessing tissue-risk in mammograms acquired with the General Electric Senographe 2000D full field digital mammography (FFDM) system.
The aim i s to standardize the output images so that each pixel is calibrated to the amount of dense tissue that the x-ray beam passed through above the related detector location (above the pixel). The work includes comparing volumetric with calibrated two-dimensional measures. Our previous work in this area will be modified and built into a total tool for assessing density in the research environment. The resulting tissue metrics will be validated in terms of sensitivity and specificity with a case-control study. In order to better understand and quantify the breast tissue density-risk associations, large multi-center serial based studies will be necessary in the future. However, this will first require understanding the calibration stability over time, as well as designing the appropriate quality assurance procedures as proposed here. The intra-imager serial stability and the inter-imager concordance will be assessed by performing the same calibration on two similar FFDM systems located at the Pi's site and another site (Via Christi Regional Medical Center, Wichita, Kansas). The serial comparisons will be made with robust methods applied in industry for monitoring serial quality control. This work will produce a calibrated automated density assessment package that may be easily implemented across institutions without modifications for use in the FFDM systems that is tailored for tissue-related risk research.