The long-term goal of this research is to optimize digital breast tomosynthesis (DBT). The principal advantage of DBT is that it can reduce the complexity of breast structure in the reconstructed slices;however, there is no valid 3D model of breast structure that can be used in the optimization. Previous work by Burgess et al showed that breast structure in a mammogram follows a power law (i.e., the log of the power spectrum of a mammogram is proportional to 2 times the log of the spatial frequency);that the detectability of a mammographic masses depends on 2;and this 2 dependence follows the same trend using either actual mammographic background or random noise filtered to have the same power-law relationship as a mammogram. The implication of their work is that if these findings can be applied to DBT images, then filtered noise can be used in lieu of a 3D breast model. The goal of this research is to examine whether Burgess'results for conventional mammography hold for DBT. Our hypothesis is that filtered noise can be used to determine the detectability of masses in tomosynthesis imaging. To prove this hypothesis, we will conduct observer studies to determine detection performance of human observers in clinical tomosynthesis images, as well as in simulated noise tomosynthesis images. We will further investigate the relationship of human detection performance in 2D reconstructed slices versus 3D viewing. The specific aims are: 1. Determine lesion detectability in real breast backgrounds and filtered noise both in 2D and 3D imaging. 2. Determine relationship between lesion detectability and acquisition geometry. 3. Compare human and ideal observer performance in filtered noise backgrounds. If we can prove our hypothesis, then we will have a tool that will allow us to optimize tomosynthesis imaging while accounting for anatomic background variations. This should significantly accelerate system optimization and shorten the time to make an optimized system clinically available. PUBLIC HEALTH RELEVANCE: The relevance of this research is that we will develop a method to optimize an emerging technology for detecting breast cancer that has the potential to have find cancers at an earlier and more treatable stage, while reducing the number of well-women recalled for more imaging exams. This should lead to a reduction in breast cancer mortality and morbidity.

R21EB008801

2009-04-01

2012-03-31