Fig. 4
From: Synthetic CT generation for pelvic cases based on deep learning in multi-center datasets
sCT images generated by different model. The first and third rows show real MR (a1, c1), real CT (a2, c2) and the sCT images generated by CycleGAN (a3, c3), CycleGAN with CL loss (a4, c4), RTGAN (a5, c5), 2.5D RTGAN with CL loss (a6, c6) and our model (a7, c7). The second and fourth rows highlight the ROIs outlined by the yellow box on each corresponding image