Table 3 Utility assessment. Average area under the ROC curve (AUC) and coefficient of variation (CV) computed for two ResNet-18 classifiers pretrained on imagenet and on PACGAN-generated images. Both models were fine-tuned using datasets of 100 and 200 images, respectively, across 100 bootstrap repetitions, and tested on a test set of 100 images.
From: High-resolution conditional MR image synthesis through the PACGAN framework
Training set size | Average AUC | CV | ||
|---|---|---|---|---|
ImageNet | PACGAN | ImageNet | PACGAN | |
100 images | 0.678 | 0.703 | 0.1137 | 0.0727 |
200 images | 0.713 | 0.760 | 0.0840 | 0.0533 |
