Skip to main content
Fig. 3 | EJNMMI Research

Fig. 3

From: Independent attenuation correction of whole body [18F]FDG-PET using a deep learning approach with Generative Adversarial Networks

Fig. 3

Two example data sets of PET data reconstructed with the generated pseudo CT (PETGAN, left), reconstructed using the acquired CT (PETAC, middle) and their voxel-wise percent difference map (Δ%-map, right). The top example shows typical results with no visually appreciable differences between PETGAN and PETAC in most anatomic regions and with more pronounced deviations localized along the lung border (black arrows) and in areas of air-filled bowel (black circle). The bottom example depicts a so-called banana artifact in PETAC due to acquisition of PET and CT in different respiratory states resulting in relative overestimation of SUVs along the diaphragm and relative underesitmation of SUVs in the abdomen (blue arrows). These artifacts were not present on PETGAN

Back to article page