Comparison of Supervised and Unsupervised Deep Learning Methods for Medical Image Synthesis between Computed Tomography and Magnetic Resonance Images

<div>(a–d) From left to right: the 4 columns are input images, ground truth images, synthetic images, and the difference maps. sCT results generated by U-Net (a) and CycleGAN (b), respectively; sMR results by U-Net (c) and CycleGAN (d), respectively.</div>

BioMed Research International

fig3

Figure 3

Figure 3: Comparison of Supervised and Unsupervised Deep Learning Methods for Medical Image Synthesis between Computed Tomography and Magnetic Resonance Images 

Figure 3 | Comparison of Supervised and Unsupervised Deep Learning Methods for Medical Image Synthesis between Computed Tomography and Magnetic Resonance Images 