Application of the Ordered-Subsets Transmission Reconstruction Algorithm to Contrast-Enhanced Dual-Energy Digital Breast Tomosynthesis

Contrast-enhanced dual-energy digital breast tomosynthesis (CE-DE-DBT) has a potential clinical impact for diagnostic breast imaging. Typically, CE-DE-DBT involves iodine contrast injection, image acquisitions at high energy (HE) and low energy (LE), weighted subtraction of the acquired data and finally reconstruction of the subtracted data. The resulting reconstruction displays iodinated structures against a noisy background that is the residual of imperfect subtraction of the anatomical breast structure. We hypothesize that CE-DE-DBT can be improved by using a reconstruction that yields better signal contrast and stronger background suppression as compared to that obtained using conventional reconstructions. We replace the commonly used FBP and SART reconstructions by a penalized likelihood reconstruction, OSTR (ordered-subsets transmission reconstruction). We used an ordinary quadratic smoothing regularizer as well as an edge-preserving Lange prior. OSTR is applied separately to the HE and LE data with weighted subtraction performed in the reconstructed domain. We obtained projection data using a Siemens CEDET DBT scanner and a customized CIRS020 phantom that reflected adipose and glandular spatial variability and included an array of iodine inserts of varying contrast and size. The data was scatter corrected and reconstructed using the three methods. We measured image quality by SDNR (signal-difference-to-noise-ratio) at each of the 16 iodine inserts. In all 16 cases, SDNR was best for OSTR and worst for FBP. Our iodine inserts had high contrast, but in a clinical setting, it may be important to visualize weaker iodine signals. OSTR, a form of penalized likelihood reconstruction, may find use in improving CE-DE-DBT.