Figure 3

Breast CT image reconstructions (11.8 cm chest-wall diameter) with AFN and/or FDK using incomplete data (columns 1–4) are compared to the FDK reference using complete data (column 5). A trained AFN can either directly output the image volume (column 1) or synthesize projection data for subsequent reconstructions (columns 2–4). When FDK uses the AFN synthesized complete data (column 2), it produces images visually similar to the “AFN attenuation coefficients”, showing a loss of resolution. When the acquired projection data are reused and spliced with AFN synthesized data (column 3), a subsequent FDK reconstruction recovers the lost resolution yet exhibits truncation-like artifacts (yellow arrows) due to the slight inconsistency between AFN synthesized projections and the acquired projections. We incorporated an offset-detector weight into the FDK algorithm, denoted as FDK-M. This weighted FDK reconstruction using AFN inpainted data (column 4) effectively eliminates the line artifacts. The display window is [0.15, 0.35] cm⁻¹.