Table 3 FID scores for cGAN and baseline models. All models were evaluated using 10-fold cross validation. FID scores were calculated against the matching real patient MRI. Bolded fold represents the fold that yielded the best FID score for all three models (fold 3)
| Â | cGAN | Pretrained autoencoder | Autoencoder |
|---|---|---|---|
CV Fold 0 | 1.51 ± 0.67 | 3.41 ± 0.82 | 4.89 ± 0.52 |
CV Fold 1 | 1.91 ± 0.57 | 3.52 ± 0.75 | 3.90 ± 0.85 |
CV Fold 2 | 1.79 ± 0.64 | 3.32 ± 0.94 | 3.51 ± 0.92 |
CV Fold 3 | 1.31 ± 0.57 | 3.80 ± 0.73 | 3.76 ± 0.75 |
CV Fold 4 | 3.42 ± 0.72 | 4.69 ± 0.82 | 5.21 ± 0.94 |
CV Fold 5 | 3.14 ± 0.49 | 4.57 ± 0.53 | 5.16 ± 0.53 |
CV Fold 6 | 2.97 ± 0.38 | 4.08 ± 0.63 | 4.98 ± 0.41 |
CV Fold 7 | 3.45 ± 0.59 | 4.59 ± 0.94 | 4.94 ± 0.36 |
CV Fold 8 | 3.24 ± 0.94 | 4.21 ± 0.82 | 4.60 ± 0.48 |
CV Fold 9 | 2.76 ± 0.76 | 3.87 ± 0.91 | 4.05 ± 0.83 |
All folds | 2.13 ± 0.33 | 3.79 ± 0.48 | 4.31 ± 0.53 |