Fig. 2From: Independent attenuation correction of whole body [18F]FDG-PET using a deep learning approach with Generative Adversarial NetworksRepresentative data set showing non-attenuation-corrected [18F]FDG-PET (left), generated pseudo CT (middle), and acquired CT (right) in axial (top), sagittal (middle), and coronal (bottom) orientation. CTGAN data—while capturing the coarse distribution of CT anatomy—showed blurring and step formation in z-direction on coronal and sagittal reconstruction as well as irregular depiction of anatomical detailsBack to article page