Applications of Fairness Training in Artificial Intelligence to Ophthalmic Disease Diagnosis using External Eye Photographs
Authors
Affiliations (1)
Affiliations (1)
- Harvard Ophthalmology AI Lab, Schepens Eye Research Institute of Massachusetts Eye and Ear, Harvard Medical School, Boston, MA, USA
Abstract
Diabetic retinopathy (DR), age-related macular degeneration (AMD), and glaucoma are leading causes of preventable blindness, yet whether external eye photographs carry phenotypic information about posterior segment disease remains unclear, and methods to mitigate demographic disparities in such models are largely unexplored. We developed and evaluated deep learning and vision-language models for binary classification of vision-threatening DR (VTDR), advanced AMD, and glaucoma using 40,687 external eye photographs from 9,878 patients at Massachusetts Eye and Ear. Convolutional networks (WResNet, EfficientNet), a vision transformer (ViT), and a finetuned vision-language model (LLaVA) were assessed by area under the receiver operating characteristic curve (AUC), with fairness across sex, race, and ethnicity quantified by the equity-scaled AUC (ES-AUC). Five fairness-training strategies were applied to the best models. LLaVA achieved the highest baseline AUC for VTDR (85{middle dot}4%, comparable to ViT 82{middle dot}6%); advanced AMD was similar across architectures (EfficientNet 86{middle dot}3%; LLaVA 84{middle dot}8%; ViT 84{middle dot}7%), whereas glaucoma was substantially worse (67-74%). Fairness training improved VTDR and AMD AUCs to 86{middle dot}3% and 85{middle dot}9% while partially mitigating demographic disparities. In summary, although AI models that rely on external eye photography to detect posterior segment disease may not yet be suitable for clinical deployment, the fairness methods demonstrated in this proof-of-concept study could be applied to other AI systems to decrease performance disparities and potentially improve model performance.