Skin lesion analysis using generative adversarial networks: a review

Skin cancer is one of the primary causes of death in the world. Timely diagnosis of skin cancer can reduce the number of deaths. Skin cancer can be diagnosed early using deep learning-based systems. The performance of deep learning-based systems suffers from overfitting if we don’t have enough data to train them. Acquiring a large amount of skin lesion images for training a deep learning-based system is a difficult task. Overfitting can be avoided using data augmentation. Generative adversarial networks (GANs) are very popular in skin lesion tasks because of their ability to generate high-quality synthetic skin lesion images. GANs are used for the classification and segmentation of skin-lesion images. We review the most relevant papers discussing the use of GANs for augmenting skin lesion datasets in this work. We gave an overview of the most commonly used GAN architectures in skin lesion analysis.