EVALUATION OF STYLEGAN2 AND STYLEGAN3 FOR SYNTHETIC MEDICAL IMAGE GENERATION ON BUSI AND CBIS-DDSM DATASETS

Deep learning from medical images is typically hindered by limited access to images and severe imbalance of classes that reduces the effectiveness of typical machine learning algorithms. Generative adversarial networks can be employed to address such issues by creating natural-appearing synthetic images to complement training sets. In this study, we compare two advanced GAN architectures, StyleGAN2 and StyleGAN3, using two publicly available breast imaging datasets: BUSI (ultrasound, 210 malignant cases) and CBIS-DDSM (mammography, 509 malignant cases). Evaluation was based on Fréchet Inception Distance and Kernel Inception Distance. On BUSI, StyleGAN3 achieved FID = 140.7 and KID = 0.06 at 1000 epochs, whereas StyleGAN2 achieved FID = 259.7 and KID = 0.25. On CBIS-DDSM, StyleGAN3 achieved FID = 90.6 and KID = 0.06, and StyleGAN2 achieved FID = 124.8 and KID = 0.10. These results demonstrate that StyleGAN3 has a tendency to synthesize images that are more natural and diversified under limited dataset conditions, at the cost of increased training times, whereas StyleGAN2 provides similar quality at less expensive computational costs. The results indicate the potential of generating medical images and the tradeoff between image quality and efficiency in data augmentation for breast cancer image improvement.

1. Generative adversarial nets / I. Goodfellow et al // Advances in Neural Information Processing Systems. – 2014. – Vol. 27. – P. 2672-2680.

2. Arjovsky M. Wasserstein Generative Adversarial Networks / M. Arjovsky, S. Chintala, L. Bottou // Proceedings of the 34th International Conference on Machine Learning (ICML). – 2017. – Vol. 70. – P. 214-223.

3. Improved training of Wasserstein GANs / I. Gulrajani et al // Advances in Neural Information Processing Systems (NeurIPS). – 2017. – P. 5767-5777.

4. Karras T. A Style-Based Generator Architecture for Generative Adversarial Networks / T. Karras, S. Laine, T. Aila // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). – 2019. – P. 4401-4410.

5. Analyzing and Improving the Image Quality of StyleGAN / Т. Karras et al // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). – 2020. – P. 8110-8119.

6. Alias-Free Generative Adversarial Networks / Т. Karras et al // Advances in Neural Information Processing Systems (NeurIPS). – 35th Int. Conf. on Neural Information Processing Systems. – 2021. – P. 1190-1201.

7. Ryspayeva M. Multi-domain synthetic medical image generation and dataset balancing with DGAN-WP-TL / M. Ryspayeva, O. Salykova // Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization. – 2025. – Vol. 13, № 1. https://orcid.org/10.1080/21681163.2025.2556687.

8. Dataset of Breast Ultrasound Images / W. Al-Dhabyani et al // Data in Brief. – 2019. – Vol. 28. – P. 104863. https://orcid.org/10.1016/j.dib.2019.104863.

9. Curated Breast Imaging Subset of Digital Database for Screening Mammography (CBIS-DDSM) / R. Sawyer-Lee et al // The Cancer Imaging Archive. – 2016. https://orcid.org/10.7937/K9/TCIA.2016.7O02S9CY. – URL: https://www.cancerimagingarchive.net/collection/cbis-ddsm/.

10. GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium / М. Heusel et al // Advances in Neural Information Processing Systems (NeurIPS). – 2017. – Vol. 30. – P. 6626-6637.

11. Demystifying MMD GANs / М. Bińkowski et al // International Conference on Learning Representations (ICLR). – 2018.

12. Progressive Growing of GANs for Improved Quality, Stability, and Variation / Т. Karras et al // International Conference on Learning Representations (ICLR). – 2018. – Poster 2018.

13. Jolliffe I.T. Principal Component Analysis / I.T. Jolliffe. – 2nd ed. – New York: Springer, 2002. – 487 p. – (Springer Series in Statistics).

14. IGAN: Improved Generative Adversarial Network for Breast Ultrasound Image Synthesis / М. Alruily et al // Diagnostics. – 2023. – Vol. 13, № 15. – P. 2562. https://orcid.org/10.3390/diagnostics13152562.

15. 2S-BUSGAN: Two-Stage GAN for Breast Ultrasound Image Synthesis / W. Luo et al // Journal of Imaging. – 2023. – Vol. 9, № 3. – P. 54. https://orcid.org/10.3390/jimaging9030054.

16. Liu C. GSDA: Generative Semantic Data Augmentation for Breast Ultrasound Image Synthesis / C. Liu, S. Ding, W. Luo // Frontiers in Oncology. – 2023. – Vol. 13. – P. 1156789. https://orcid.org/10.3389/fonc.2023.1156789.

17. Cross-domain synthesis of mammograms using CycleGAN for data augmentation in breast cancer screening / L.Garrucho et al // Computers in Biology and Medicine. – 2023. – Vol. 165. – P. 107375. https://orcid.org/10.1016/j.compbiomed.2023.107375.

18. Fan J. BreastGAN: Adversarial Data Augmentation for Mammogram Classification / J. Fan, R. Huang, Z. Liu // Computers in Biology and Medicine. – 2021. – Vol. 135. – P. 104601. https://orcid.org/10.1016/j.compbiomed.2021.104601.

19. Prodan R. StyleGAN-XL for high-resolution mammogram synthesis with limited annotations / R. Prodan, P.F. Jäger, A. Kopp-Schneider // Medical Image Analysis. – 2023. – Vol. 87. – P. 102842. https://orcid.org/10.1016/j.media.2023.102842.