CT colonography (CTC) uses orally administered fecal-tagging agents to indicate residual materials that couldotherwise interfere with the interpretation of CTC images. To visualize the colon in virtual 3D endoluminalviews, electronic cleansing (EC) can be used to subtract the fecal-tagged materials from the CTC images. However,conventional EC methods produce subtraction artifacts that distract readers and computer-aided detectionsystems. In this study, we used generative adversarial learning to transform fecal-tagged CTC input image volumesto corresponding virtually cleansed image volumes. To overcome the need for paired training samples, weused a cycle-consistent 3D-generative adversarial network (3D EC-cycleGAN) scheme that can be trained withunpaired samples. The associated generator and discriminator networks were implemented as 3D-convolutionalnetworks, and the loss functions were adapted to the unique requirements of EC in CTC. To investigate thefeasibility of the approach, the 3D EC-cycleGAN was trained and tested with CTC image volumes of an anthropomorphicphantom filled partially with fecal tagging to recreate the attenuation ranges observed in clinicalCTC. Our preliminary results indicate that the proposed 3D EC-cycleGAN can potentially learn to perform ECwithout producing the kinds of subtraction artifacts that are observed with conventional EC methods.
展开▼
