AAO: Machine Learning Outperforms Human Graders for Diagnosing Glaucoma

FRIDAY, Oct. 24, 2025 (HealthDay News) — Machine learning (ML) outperforms human graders for diagnosing glaucoma, according to a study presented at the annual meeting of the American Academy of Ophthalmology, held from Oct. 18 to 20 in Orlando, Florida.

Anthony P. Khawaja, M.B.B.S., from the University College London Institute of Ophthalmology and Moorfields Eye Hospital in London, and colleagues used data from fundus images in 6,304 participants to compare the accuracy of ML versus human grader-ascertained vertical cup-disc-ratio (VCDR) for glaucoma detection. VCDR was estimated from fundus images by human graders (H-VCDR) and a previously trained ML model (ML-VCDR).

Overall, 696 participants had glaucoma or suspect status in at least one eye. The researchers found that H-VCDR and ML-VCDR explained 17 and 31 percent of right-eye glaucoma status variance, with area under the receiver operating characteristic curves (AUROC) of 79 and 88 percent, respectively. H-VCDR and ML-VCDR explained 20 and 35 percent of the variance for left eyes, with AUROCs of 81 and 90 percent, respectively.

“Glaucoma remains one of the most common causes of vision loss that can”t be repaired globally. 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,” Khawaja said in a statement.

Several authors disclosed ties to the biopharmaceutical and technology industries.

Press Release

More Information