Computer Vision for Rare Diseases: A Scoping Review

Rare diseases, though individually infrequent, collectively affect over 300 million people worldwide. The clinical heterogeneity, delayed diagnosis, and limited expertise associated with rare diseases present substantial challenges for early detection and effective management. Computer Vision (CV) technologies, by analyzing medical images (such as X-rays and MRI), pathological slides, facial features, and other physical manifestations, demonstrate significant potential in various clinical aspects of rare diseases, including automated screening, assisted diagnosis, subtyping, monitoring, surgical support, and prognostic assessment. However, the landscape of CV research in rare diseases remains fragmented, with limited overviews of available datasets, state-of-the-art technologies, and clinical applications. In this scoping review, we provide a comprehensive overview of CV research in rare diseases, identifying and analyzing 772 relevant publications. We provide a detailed summary of publicly available datasets, discuss recent advances in CV methodologies—including data augmentation, interpretability, multimodal fusion, training strategies, and loss function design. Furthermore, we categorize and review clinical applications across major rare disease types, including ophthalmic, neurologic, developmental, skeletal, respiratory, genitourinary, cardiovascular, hematologic, rheumatologic, and endocrine/metabolic diseases. We also discuss key challenges hindering the widespread adoption of CV in clinical practice, such as data privacy, fairness, scarcity and imbalance, annotation quality, interpretability, and computing resource constraints, and propose possible future directions, including multi-disciplinary collaboration, privacy-preserving data sharing, domain adaptation, multimodal fusion, advanced interpretability, and efficient model deployment. This scoping review aims to provide a comprehensive reference for researchers and clinicians by systematically mapping the landscape of CV research in rare diseases, thereby facilitating the advancement of CV-driven diagnosis and management, and promoting the real-world clinical translation of these technologies in intelligent medicine for rare diseases.