Growing evidence suggests that breast density is an independent risk factor for breast cancer. Currently, breast density is most commonly quantified from mammograms using semi-automated image thresholding techniques to segment the area of the dense tissue. Mammography, however, is a projection imaging technique that visualizes the addmixture of superimposed breast tissues. Therefore, mammograms do not allow estimating volumetric density but a rather rough area-based estimate measured from the projection image of the breast. Digital breast tomosynthesis (DBT) is an emerging 3D x-ray imaging modality in which tomographic breast images are reconstructed from multiple low-dose x-ray source projections. Knowing that the risk of breast cancer is associated with the amount of fibroglandular tissue in the breast (a.k.a. breast density), measures of volumetric breast density from DBT images could provide more accurate measures of breast density and ultimately result in more accurate measures of risk. This project will develop a new robust and fully-automated method for volumetric breast density estimation in DBT based on a novel algorithm that combines image texture analysis with scale-based fuzzy connectedness image segmentation. The main idea is to incorporate the notion of """"""""texture-affinity"""""""" in fuzzy-connectedness segmentation by performing texture analysis in the reconstructed DBT images as a first-level image analysis step for generating the corresponding """"""""texture-scene"""""""" of the parenchymal pattern. A scale-based fuzzy-connectedness algorithm will be applied to the obtained """"""""texture-scene"""""""" image to determine the size of homogeneous local breast tissue structures and segment the dense tissue voxels. A volumetric breast density measure will be derived by dividing the corresponding volume of dense tissue to that of the entire breast. Our preliminary data suggest that texture analysis in DBT can be used to distinguish the dense from the fatty breast tissue regions, indicating that the proposed segmentation approach is feasible. We propose to validate our algorithm using i) simulated DBT images, generated using our validated anthropomorphic breast software phantom, in which ground truth for breast density can be controlled, and ii) clinical DBT, MRI and digital mammography (DM) images collected retrospectively from clinical trials that have been completed in our department. This project will combine the unique expertise of Penn investigators in DBT image texture analysis and fuzzy-connectedness segmentation to develop a novel algorithm for volumetric breast density estimation in DBT. The rapidly evolving technology of DBT and the potential for superior clinical performance will determine the emerging role of DBT in clinical practice. A robust and fully-automated method for measuring volumetric breast density from DBT images could provide a non-invasive quantitative imaging biomarker for estimating breast cancer risk that could be used to guide clinical decision making for offering customized breast cancer screening recommendations and forming preventive strategies, especially for women at high risk of breast cancer.
We envision a unique setting in which breast cancer risk assessment and patient education can be combined to empower women with knowledge about their personal risk and provide a fully-automated risk assessment tool for referring physicians. The rapidly evolving technology of digital breast tomosynthesis (DBT) and the potential for superior clinical performance will determine the emerging role of DBT in clinical practice. A robust fully-automated method for estimating volumetric breast density from DBT images will provide a non-invasive quantitative imaging biomarker for estimating breast cancer risk that can be used to guide clinical decision making for offering customized screening recommendations and forming preventive strategies, especially for women at a high risk of breast cancer.
