Generative adversarial networks (GANs) show promise for synthesizing realistic medical images; however, the quantitative evaluation of accuracy remains difficult. This study utilizes an auxiliary GAN to generate synthetic chest X-rays and then use these images to train a VGG-16 convolutional neural network (CNN) classifier. The CNN's performance in classifying real X-rays was evaluated to assess the efficacy of GAN-generated training data. The auxiliary GAN was trained on real X-rays and then used to synthesize images modeled after the original data distribution. The VGG-16 model was trained on a synthetic dataset and tested on reserved real X-rays, which had not been seen during model development. Its performance was compared with classifiers trained solely on real data. The results analyze the VGG-16 testing accuracy between synthetic and real training data to quantify how effectively the auxiliary GAN captured the visual features critical for high CNN performance. Techniques for evaluating GAN-generated content as part of the clinical adoption of generative models are discussed. This study presents a methodology for assessing GANs in the production of synthetic medical training data while preserving vital information for analysis.

1. A. Madani, M. Moradi, A. Karargyris, and T. Syeda-Mahmood, “Chest x-ray generation and data augmentation for cardiovascular abnormality classification,” in 2018 IEEE 31st International Symposium on Computer-Based Medical Systems (CBMS), 2018, pp. 360–364.
2. L. Guibas, T. Virdi, and P. Li, “Synthetic Medical Images from Dual Generative Adversarial Networks,” arXiv:1709.01872 [cs], Sep. 2017, Accessed: Dec. 05, 2023. [Online]. Available: http://arxiv.org/abs/1709.01872
3. H. Salehinejad, S. Valaee, T. Dowdell, E. Colak, and J. Barfett, “Generalization of Deep Neural Networks for Chest Pathology Classification in X-Rays Using Generative Adversarial Networks,” in 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018, pp. 990–994.
4. A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” in Advances in Neural Information Processing Systems, vol. 25, F. Pereira, C. J. C. Burges, L. Bottou, and K. Q. Weinberger, Eds. Curran Associates, Inc., 2012, pp. 1097–1105.
5. X. Wang et al., “ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 3462–3471.
6. Paul Mooney, (2018, January). Chest X-Ray Images (Pneumonia), Version 2. Retrieved December 1, 2023 from https://www.kaggle.com/datasets/paultimothymooney/chest-xray-pneumonia

Auxiliary Generative Adversarial Networks Classification Deep Learning Neural Networks