Synthetic Melanoma Image Generation and Evaluation Using Generative Adversarial Networks

ArXi:2603.13497v1 Announce Type: cross Melanoma is the most lethal form of skin cancer, and early detection is critical for improving patient outcomes. Although dermoscopy combined with deep learning has advanced automated skin-lesion analysis, progress is hindered by limited access to large, well-annotated datasets and by severe class imbalance, where melanoma images are substantially underrepresented.