Digital Breast Tomosynthesis Reconstruction with Detector Blur and Correlated Noise

This paper describes a new reconstruction method for digital breast tomosynthesis (DBT). The new method incorporates detector blur into the forward model. The detector blur introduces correlation in the measurement noise. We formulate it as a regularized quadratic optimization problem with data-fit term that accounts for the non-diagonal noise covariance matrix. By making a few assumptions based on the breast imaging process, we can model the detector blur in the optimization problem and solve it with a separable quadratic surrogate (SQS) algorithm. This method was applied to DBT reconstruction of breast phantoms and human subjects. The contrast-to-noise ratio and sharpness of microcalcifications and the visual quality of mass margins were analyzed and compared to those by the simultaneous algebraic reconstruction technique (SART). The results demonstrated the potential of the new method in improving the image quality of the reconstructed DBT images. This work is our preliminary step towards a model-based iterative reconstruction (MBIR) for DBT.