Preference Study of Reconstructed Image Quality in Mammographic Tomosynthesis

Digital tomosynthesis mammography is an advanced x-ray application that provides detailed 3D information about the imaged breast. The involved 3D reconstruction step is a difficult problem due to incomplete data and the limited projection angle. In this preference study we compared results obtained with advanced reconstruction techniques, in particular ART (Algebraic Reconstruction Technique), FBP (Filtered Backprojection), OSBP (order statistics based backprojection), and GFBP (generalized filtered backprojection) in terms of their image quality. The 3D datasets were displayed slice-by-slice in “cine loop” mode. As observers we included radiologists, physicists and image processing experts. Phantom and mastectomy specimen datasets were included in the study. The results obtained with our dataset indicate that ART was overall preferred for the phantom datasets while GFBP was preferred for mastectomy specimen reconstructions.