Improved microcalcification detection for breast tomosynthesis using a penalized-maximum-likelihood reconstruction method

Abstract

In this paper we explore the use of a penalized maximum likelihood (PML) based reconstruction method to improve the image quality and microcalcification detectability in digital breast tomosythesis (DBT). To evaluate performance, a human observer psychophysical study was performed with computer simulated images. The simulation used realistic structured breast models derived from CT scans of surgical mastectomy specimens giving sufficient statistical variability in terms of breast background structural noise. Sensitivity and specificity of microcalcification detectability measured with PML reconstruction was compared to that obtained with the filtered back projection (FBP) method for simulated breast tomosynthesis images. An observer study conducted using localized receiver operating characteristic (LROC) analysis showed significantly better sensitivity and specificity using the PML reconstruction method for simulated mean glandular dose levels of 1.0 mGy for a 5 cm compressed breast. This study suggests that MC detection accuracy is improved using PML reconstruction technique and that it might be feasible to reduce the imaging dose of DBT using this technique. © 2010 Springer-Verlag.