Effects of Detector Blur and Correlated Noise on Digital Breast Tomosynthesis Reconstruction

Digital breast tomosynthesis (DBT) was developed to alleviate the problem of overlapping tissue in 2-D mammography. To improve DBT image quality, we are developing model-based iterative reconstruction method based on models of the imaging system physics. Our preliminary implementation (SQS-DBCN) incorporated detector blur into the system model and correlation into the noise model under some simplifying assumptions, and used a separable quadratic surrogate (SQS) algorithm for minimizing the cost function. We have shown that SQS-DBCN improves DBT image quality, especially for microcalcification clusters (MCs) that are an early sign of breast cancer. This work further improves the regularization method by incorporating neighbors along the diagonal directions, whereas the previous regularization used only the horizontal and vertical neighbors. To further understand the role of the different components in the system model of the SQS-DBCN method, we reconstructed DBT images without modeling either the detector blur or noise correlation for comparison. Visual comparison of the reconstructed images showed that regularizing with diagonal directions reduced artifacts and the noise level. It effectively eliminated the square texture resulting from enhancing sharp edges along only the horizontal and vertical directions in the previous implementation. The SQS-DBCN reconstructed images had better image quality than reconstructions without models for detector blur or correlated noise, as indicated by the contrast-to-noise ratios (CNR) of MCs and textural artifacts. These results indicated that regularized DBT reconstruction with detector blur and correlated noise modeling, even with simplifying assumptions, can improve DBT image quality compared to that without system modeling. 100-Word Summary: We have previously developed a DBT reconstruction method with detector blur and correlated noise modeling (SQS-DBCN). In this work, we further improved the SQS-DBCN method by introducing regularization along diagonal directions, which was found to reduce the appearance of “square” artifacts and noise. We also studied the effects of the detector blur model or the correlated noise model on image quality. The results show degradation in contrast-to-noise ratio of subtle microcalcifications and increased artifacts when either model was suppressed, indicating that regularized reconstruction with detector blur and correlated noise modeling, even with simplifying assumptions, could improve DBT image quality compared to without system modeling. Purpose: We are developing a model-based iterative reconstruction method for DBT by incorporating models of the imaging system physics. Our preliminary implementation incorporated detector blur into the system model and correlation into the noise model under some simplifying assumptions, and used the separable quadratic surrogate (SQS) algorithm for minimizing the cost function [1]. We referred to this reconstruction method as SQS-DBCN. In the current study, we continue to refine the SQS-DBCN algorithm and to investigate the effects of the detector blur model and the correlated noise model on DBT image quality. We introduced regularization along the diagonal directions to reduce the textural artifacts in the reconstructed images. We performed reconstructions over a range of parameters with and without the detector blur model or the noise correlation model alternately. The contrast-to-noise ratios (CNR) of the reconstructed images and the visual quality of the images obtained under different reconstruction conditions were compared to evaluate the effectiveness of the modeling components in the SQS-DBCN. Method and materials: In the SQS-DBCN method for DBT [1], we consider the following reconstruction problem: