Generative Models in Medical Imaging
Aarti Paresh Pimpalkar, Vijaya S. Patil, Nilesh N. Thorat, Mayuresh Gulame, Jayashri D. Palkar, Poonam Suraj Gham · 2025
A set of artificial intelligence systems known as generative adversarial networks (GANs) consists of a generator and a discriminator that are simultaneously learned via adversarial training. GANs have been shown to be quite useful in a number of industries, including medical imaging. GANs help in picture segmentation, disease diagnosis, medical image synthesis, and data quality improvement. They also produce synthetic medical images. Their significance stems from their capacity to provide lifelike visuals, which enhances medical professional training, research, and diagnosis. It is essential to comprehend their uses, algorithms, recent developments, and difficulties if the field of medical imaging is to continue advancing. Nevertheless, no research has examined the most current advancements in GAN technology for medical imaging. In order to close this research gap, we started this large-scale study by examining the many uses of GANs in medical imaging and comparing them to other recent studies. The popular datasets and pre-processing methods for improving comprehension are then covered in detail. The GAN algorithms are then thoroughly discussed, with an emphasis on their individual advantages and disadvantages. Once that was done, we carefully examined the findings and experimental specifics of a few recent state-of-the-art studies to get a deeper grasp of how GANs are currently developing in medical imaging. In conclusion, we deliberated over the various obstacles faced and the potential avenues for future study to address these issues. This comprehensive review provides a thorough overview of GANs in medical imaging, covering their models, application domains, state-of-the-art results analysis, difficulties, and future directions. It is an invaluable tool for interdisciplinary research.