Table 3.
Pretraining method | Data augmentation | Classification accuracy [%] | |
---|---|---|---|
Benign | Malignant | ||
None | No | 25.9 | 78.8 |
ImageNet | No | 33.3 | 87.9 |
WGAN (60000) | No | 63.0 | 81.8 |
None | Yes | 51.9 | 84.9 |
ImageNet | Yes | 40.7 | 93.9 |
WGAN (60000) | Yes | 66.7 | 93.9 |