The Non-prewhitening and Hotelling Observers for Parameter Selection for Linear Iterative Image Reconstruction in Breast Tomosynthesis

Sean D. Rose,Adrian A. Sanchez,Ingrid Reiser,Emil Y. Sidky,Xiaochuan Pan
DOI: https://doi.org/10.1109/nssmic.2017.8533045
2017-01-01
Abstract:Digital breast tomosynthesis (DBT) has seen widespread clinical adoption for breast cancer screening over the past decade as either a supplement to, or replacement for, digital mammography. As in computed tomography, iterative image reconstruction methods for DBT involve specification of a large number of parameters that can significantly impact image quality. Efficiently computable task-based metrics are needed for characterizing the parameter dependencies of these reconstruction methods. Here we investigate two task-based image quality metrics for assessing the effect of regularization strength on microcalcification detectability for an iterative reconstruction method in DBT.
What problem does this paper attempt to address?