ROFI: A Deep Learning-Based Ophthalmic Sign-Preserving and Reversible Patient Face Anonymizer

2025/10/13 by Tian, Yuan, Zhou, Min, Chen, Yitong +19
#Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2510.11073

Abstract

Patient face images provide a convenient mean for evaluating eye diseases, while also raising privacy concerns. Here, we introduce ROFI, a deep learning-based privacy protection framework for ophthalmology. Using weakly supervised learning and neural identity translation, ROFI anonymizes facial features while retaining disease features (over 98% accuracy, κ> 0.90). It achieves 100% diagnostic sensitivity and high agreement (κ> 0.90) across eleven eye diseases in three cohorts, anonymizing over 95% of images. ROFI works with AI systems, maintaining original diagnoses (κ> 0.80), and supports secure image reversal (over 98% similarity), enabling audits and long-term care. These results show ROFI's effectiveness of protecting patient privacy in the digital medicine era.

Citations

• MedOmni-45°: A Safety-Performance Benchmark for Reasoning-Oriented LLMs in Medicine
• Can Large Models Fool the Eye? A New Turing Test for Biological Animation
• Semantics versus Identity: A Divide-and-Conquer Approach towards Adjustable Medical Image De-Identification
• Image Quality Assessment: From Human to Machine Preference
• Information Density Principle for MLLM Benchmarks
• Towards All-in-One Medical Image Re-Identification
• Medical Manifestation-Aware De-Identification
• Free-VSC: Free Semantics from Visual Foundation Models for Unsupervised Video Semantic Compression
• G2Face: High-Fidelity Reversible Face Anonymization via Generative and Geometric Priors
• GAIA: Rethinking Action Quality Assessment for AI-Generated Videos
• SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis
• On the Challenges and Perspectives of Foundation Models for Medical Image Analysis
• RiDDLE: Reversible and Diversified De-identification with Latent Encryptor
• Region-Aware Face Swapping
• A Coding Framework and Benchmark towards Low-Bitrate Video Understanding
• High-Resolution Image Synthesis with Latent Diffusion Models
• High-Resolution Image Synthesis with Latent Diffusion Models
• DiffusionCLIP: Text-Guided Diffusion Models for Robust Image Manipulation
• Self-Conditioned Probabilistic Learning of Video Rescaling
• EAN: Event Adaptive Network for Enhanced Action Recognition
• One Shot Face Swapping on Megapixels
• DP-Image: Differential Privacy for Image Data in Feature Space
• Adversary Instantiation: Lower Bounds for Differentially Private Machine\n Learning
• GAN-Control: Explicitly Controllable GANs
• FACEGAN: Facial Attribute Controllable rEenactment GAN
• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• Data Efficient and Weakly Supervised Computational Pathology on Whole Slide Images
• The advantages of the Matthews correlation coefficient (MCC) over F1 score and accuracy in binary classification evaluation
• International evaluation of an AI system for breast cancer screening
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Password-conditioned Anonymization and Deanonymization with Face Identity Transformers
• Dataset Growth in Medical Image Analysis Research
• AnonymousNet: Natural Face De-Identification with Measurable Privacy
• Artificial Intelligence for Pediatric Ophthalmology
• Certified Adversarial Robustness via Randomized Smoothing
• Spectral Normalization for Generative Adversarial Networks
• CosFace: Large Margin Cosine Loss for Deep Face Recognition
• The Unreasonable Effectiveness of Deep Features as a Perceptual Metric
• Progressive Growing of GANs for Improved Quality, Stability, and Variation
• Attention Is All You Need
• SphereFace: Deep Hypersphere Embedding for Face Recognition
• Interspecies Knowledge Transfer for Facial Keypoint Detection
• Fast Face-swap Using Convolutional Neural Networks
• Image-to-Image Translation with Conditional Adversarial Networks
• Adversarial examples in the physical world
• Joint Face Detection and Alignment Using Multitask Cascaded Convolutional Networks
• Deep Residual Learning for Image Recognition
• Generative Adversarial Networks
• Peripheral Facial Palsy: Etiology, Diagnosis and Treatment

Cited by

• SDEval: Safety Dynamic Evaluation for Multimodal Large Language Models

Related

• VisionCLIP: An Med-AIGC based Ethical Language-Image Foundation Model for Generalizable Retina Image Analysis
• Artificial intelligence and deep learning in ophthalmology
• RetinaGuard: Obfuscating Retinal Age in Fundus Images for Biometric Privacy Preserving
• Privacy-Preserving Eye-tracking Using Deep Learning
• OpticalDR: A Deep Optical Imaging Model for Privacy-Protective Depression Recognition