Generative Artificial Intelligence for Data Augmentation and Synthesis in Rare Disease Imaging

Rare diseases are characterized by extreme data scarcity, heterogeneity, and class imbalance, which severely limit the development of reliable artificial intelligence (AI) models for medical imaging. In recent years, data augmentation and synthetic data generation have emerged as key strategies to address these challenges. This scoping review systematically examines the state of the art in generative AI–based methods for creating synthetic training data in rare disease imaging. Following PRISMA guidelines, relevant studies published between ۲۰۱۰ and ۲۰۲۵ were identified and analyzed with respect to clinical objectives, disease domains, data modalities, and methodological approaches. The results reveal a rapid increase in research activity since ۲۰۱۸, with diagnostic applications dominating the field. While classical data augmentation remains widely used due to its simplicity and interpretability, deep generative models—such as generative adversarial networks and variational autoencoders—are increasingly adopted to capture complex pathological variability. Despite promising results, challenges related to validation, reproducibility, and clinical integration remain. This review highlights current trends, limitations, and future directions for leveraging generative AI to advance rare disease imaging research.

کلیدواژه ها:

Rare disease imaging ، generative artificial intelligence ، synthetic data generation