Table 1 MAE, PSNR and SSIM obtained on the two independent test sets with various image quality. For each metric, we report the average and standard deviation across the corresponding test set. We compute the metrics for both real T1ce and synthetic T1nce in relation to the real T1nce, and so within the brain region
Test set | Compared images | Model | MAE (%) | PSNR (dB) | SSIM |
|---|---|---|---|---|---|
Testgood | T1nce / T1ce | - | 4.14 ± 1.59 | 23.03 ± 2.83 | 0.90 ± 0.05 |
T1nce / Synthetic T1nce | Res-U-Net | 3.06 ± 1.50 | 26.89 ± 4.30 | 0.95 ± 0.04 | |
Att-U-Net | 2.73 ± 1.69 | 29.07 ± 4.53 | 0.96 ± 0.05 | ||
Trans-U-Net | 2.80 ± 1.42 | 28.00 ± 4.13 | 0.96 ±0.04 | ||
cGAN Res-U-Net | 3.47 ± 1.59 | 23.89 ± 4.30 | 0.95 ± 0.04 | ||
cGAN Att-U-Net | 2.69 ± 1.68 | 28.89 ± 4.44 | 0.97 ± 0.05 | ||
cGAN Trans-U-Net | 2.86±1.59 | 28.00 ±4.32 | 0.96 ± 0.04 | ||
Testlow | T1nce / T1ce | - | 3.71 ± 1.99 | 24.20 ± 3.85 | 0.91 ± 0.06 |
T1nce / Synthetic T1nce | Res-U-Net | 2.93 ± 1.77 | 26.71 ± 4.32 | 0.95 ± 0.05 | |
Att-U-Net | 2.89 ± 1.85 | 27.15 ± 4.57 | 0.95 ± 0.05 | ||
Trans-U-Net | 2.98 ± 1.89 | 26.71 ± 4.38 | 0.94 ± 0.05 | ||
cGAN Res-U-Net | 3.20 ± 1.96 | 26.20 ± 4.42 | 0.93 ± 0.05 | ||
cGAN Att-U-Net | 2.86 ± 1.83 | 27.12 ± 4.50 | 0.95 ± 0.05 | ||
cGAN Trans-U-Net | 2.97 ± 1.83 | 26.68 ± 4.40 | 0.94 ± 0.05 |