Abstract
Ophthalmic diseases such as diabetic retinopathy, glaucoma, age-related macular degeneration, and inherited retinal dystrophies remain leading causes of visual impairment worldwide, necessitating accurate diagnosis and longitudinal monitoring. Modern ophthalmology increasingly relies on facial and ocular imaging, yet the integration of artificial intelligence (AI) into these workflows raises significant privacy concerns. ROFI (Reversible Ophthalmic Face Image anonymizer) is a novel deep learning-based framework designed to anonymize patient facial features while preserving ophthalmic signs essential for diagnosis. By employing weakly supervised learning and neural identity translation, ROFI achieves high accuracy in retaining ocular disease markers while ensuring reversibility for authorized clinical review. Comparative studies demonstrate that anonymized images maintain diagnostic fidelity for retinal disease classification, while AI-based intraoperative guidance systems further enhance surgical precision and patient safety. Despite these advances, challenges remain regarding dataset diversity, external validation, bias, and cost-effectiveness, particularly in resource-limited settings. Future strategies should emphasize multicenter validation, integration into electronic health records, and awareness campaigns to promote adoption. ROFI represents a significant step toward balancing patient privacy with diagnostic accuracy, offering a scalable solution for secure, AI-driven ophthalmic care.