A Review on Medical Image Generation Generative Adversarial Networks (GANs)
Shrina Patel, Ashwin Makwana · 2025
This review focuses on the way GANs can be applied in the medical image generation and how they can be helpful in diagnostics, training or data augmentation. Through GANs, it is possible to create artificial images for synthetic medical image generation because the current data collected in the healthcare field lacks sufficient large amount of data amount and diverse image quality, however, there are apparent problems in the production of high quality and diverse, and strongly interpretable synthetic images. The work examines several structures based on GANs and compares them, pointing out the challenges that need to be addressed when generating realistic images of severe medical conditions. Some of these are to increase GAN dimensions to improve model complexity and expand the training data to include therapeutic relevance. Thus, this review contributes to the identification of the state of AI-based medical image synthesis.