Optimizing algorithm parameters based on a model observer detection task for image reconstruction in digital breast tomosynthesis

An iterative image reconstruction algorithm for digital breast tomosynthesis (DBT) is presented based on constrained, total-variation (TV) minimization. The algorithm is designed to yield the reconstructed volume in few iterations. As the algorithm does not solve the optimization problem upon which it is based, other stopping criteria are derived based on the imaging task of micro-calcification detection. We speculated that even though the algorithm optimization is based on this specific task, the resulting reconstructed volume may be near optimal for other tasks because of the sensitivity of micro-calcification contrast to image reconstruction algorithm parameters.