Artificial Intelligence-Guided Glaucoma Screening Shows Promise

While artificial intelligence (AI) has already proven effective in screening for diabetic retinopathy, applying similar technology to glaucoma has been more complex. Unlike diabetic retinopathy, glaucoma is not a single disease but a group of conditions characterized by progressive optic nerve damage. Diagnosis relies on a constellation of findings—including structural and functional tests, clinical judgment, and long-term monitoring.

Despite these challenges, research into AI-assisted glaucoma detection continues to gain momentum, offering hope for improving early diagnosis of one of the world’s leading and most underdiagnosed causes of irreversible blindness. A new study presented today at the 129th annual meeting of the American Academy of Ophthalmology (AAO) adds compelling evidence to that promise, showing that a machine learning algorithm outperformed trained human graders in identifying patients at risk for glaucoma.

Researchers from the University College London (UCL) Institute of Ophthalmology and Moorfields Eye Hospital analyzed 6,304 fundus images collected from participants in the EPIC-Norfolk Eye Study, a large population-based cohort. The team compared the performance of their AI algorithm with that of trained human graders in estimating the vertical cup-to-disc ratio—a key metric for assessing glaucomatous optic nerve damage. A glaucoma specialist later confirmed diagnoses to establish ground truth.

The results were striking: the algorithm correctly identified glaucoma in 88 to 90 percent of cases, while human graders achieved an accuracy of 79 to 81 percent. Although the AI system did not distinguish between confirmed and suspected cases, its performance demonstrates significant potential as a front-line screening tool.

The authors note that this study stands apart from previous research because the algorithm was validated on a dataset reflective of typical screening populations, rather than one enriched with known glaucoma cases. Only 11 percent of the eyes included were glaucoma suspects, providing a realistic test of the algorithm’s clinical applicability.

Lead investigator Anthony Khawaja, PhD, FRCOphth, said he was surprised by how strongly the machine learning system outperformed human graders. He envisions a future in which AI serves as a cost-effective preliminary screening tool, especially when combined with other risk indicators such as intraocular pressure measurements or genetic profiling.

“Glaucoma remains one of the most common causes of vision loss that can’t be repaired globally,” Dr. Khawaja said. “To date, screening is too expensive for glaucoma, but I hope that artificial intelligence solutions, in combination with other approaches such as targeting by genetic risk, will be the solution.”

While additional validation and integration studies are needed, the findings point to a future in which AI-driven screening tools could help close the global gap in glaucoma detection—particularly in underserved regions where access to specialists is limited. By enabling earlier diagnosis, such tools could play a critical role in preventing irreversible vision loss for millions worldwide.