July 18, 2025 - Researchers at University College London and Heidelberg Engineering have developed Eye2Gene, an AI system that uses multimodal imaging to diagnose inherited retinal diseases (IRDs) with 83% accuracy. The model, trained on 58,030 retinal scans from 2,451 patients, outperforms existing phenotyping tools and could revolutionise genetic counselling for IRD patients.
Eye2Gene combines spectral-domain OCT, infrared reflectance, and fundus autofluorescence imaging through an ensemble of 15 convolutional neural networks. "This approach captures subtle patterns that human experts might miss," explained Dr Nikolas Pontikos, lead author of the Nature Machine Intelligence study. The system's ability to predict the most likely causative gene enables targeted genetic testing, reducing diagnostic delays.
The breakthrough highlights AI's potential in precision medicine, particularly for rare genetic disorders. By integrating multiple imaging modalities, Eye2Gene addresses the limitations of single-modality approaches, demonstrating how multimodal AI can enhance diagnostic accuracy. This aligns with global efforts to improve healthcare accessibility through AI-driven tools.
Our view: While Eye2Gene shows promise, its clinical adoption requires rigorous validation across diverse populations. The model's success underscores the importance of interdisciplinary collaboration between clinicians and AI researchers. Ethical considerations around data privacy and algorithmic bias must remain central as such systems transition from research to real-world use.
Be the first to comment!