Zeng, Song, Chen et al. present a StyleGAN2-based framework for generating class-conditional medical images to address data scarcity and imbalance in diagnostics. Lesion-specific augmentation is enabled without requiring per-class model training, demonstrating improved accuracy in computer-aided detection of maxillary sinus pathologies.
- Peisheng Zeng
- Rihui Song
- Zetao Chen
