As artificial intelligence (AI) rapidly advances, its integration into dermatology has been mainly through Convolutional Neural Networks for skin cancer classification and the implementation of explainable AI in those classifications.
Another network type of growing importance in dermatology are Cycle-Consistent Adversarial Networks (Cycle-GANs). A GAN functions with two neural networks: a generator creating images and a discriminator evaluating them, producing increasingly realistic results. A Cycle-GAN links two GANs for bidirectional data transformation between domains. GANs have been increasingly employed in dermatology for various applications. GANs have already found a significant application in dermatology, particularly for data augmentation in skin cancer classification models. The classification models are limited by imbalances in the training datasets.
The implementation of GANs for data augmentation to balance the training data enhanced the robustness and accuracy of these diagnostic models. GANs have also been used for color constancy in medical imaging, ensuring consistent appearance in dermatological images under different lighting conditions. Color variability can lead to bias of dermatologists and impact the diagnosis. Generative models are effective in generalizing dermoscopic image appearance. Furthermore, Cycle-GANs have been utilized to transform dermoscopic images of melanoma into art works as a form of art therapy for melanoma patients. Despite these advancements, the simulation of disease progression, such as the transformation of nevi into melanoma, remains underexplored. Our study represents the first attempt to apply Cycle-GANs to simulate skin lesion progression, offering a novel approach to visualize potential changes in lesions over time.