|
|
|
Figure 4
Examples of our modified CycleGAN model training results: (a) direct training patch input/outputs (256 × 256 pixels); and (b) example WSI result generated through overlapping patch-based inference, from one WSI sample pair included in training (958 × 1000 pixels). The displayed µCT and histology images have been masked (replacing all background pixels with the mean value) with the included sample correspondence mask. |

journal menu![[Figure 4]](mo5315fig4.jpg)
access


