TABLE 2.
Synthetic PET |
Synthetic MRI |
||||
PSNR | SSIM | PSNR | SSIM | ||
Pan et al., 2020 | HGAN | 30.24 | 0.6945 | 26.07 | 0.6683 |
Pan et al., 2019 | FGAN | 29.62 | 0.6817 | 25.10 | 0.6404 |
Hu et al., 2019 | Adversarial U-Net (2D slice) | 25.13 | − | − | − |
Pan et al., 2018 | 3D-cGAN | 24.49 | − | − | − |
The proposed method (ROI) | RevGAN | 29.34 | 0.8034 | 29.81 | 0.9389 |
The proposed method (Full Data) | RevGAN | 29.42 | 0.8176 | 24.97 | 0.6746 |
The quality of the reconstructed image. Results of image synthesis achieved by different methods. Just as a reference, because the selected data and data preprocessing are different. Bold values mean the best value in the comparative experiment.