An artificial intelligence (AI) platform outperformed a group of ophthalmologists in correctly diagnosing glaucomatous optic neuropathy, a study in the Journal of Glaucoma reports.

The study tested a new deep learning system against six ophthalmologists when grading 110 color fundus photos randomly selected from the Singapore Malay Eye Study. The investigation compared the diagnoses between the two groups and then to the original clinical diagnosis, which was considered the standard.

The study found the AI system outperformed five of the six ophthalmologists in terms of diagnostic performance, and no statistically significant difference existed between the AI system and the best case consensus among the ophthalmologists (i.e., opinions that most closely matched the standard). Researchers noted the agreement between the AI and the standard was 0.715, while the highest agreement in the ophthalmology group was 0.613.

Additionally, the AI system took approximately 10% of the time the ophthalmologists did in making a classification.

“The high sensitivity […] makes it a valuable tool for screening patients with glaucomatous optic neuropathy,” the researchers wrote in their paper. Future work is planned to extend the study to a larger sample of patients.

Al-Aswad LA, Kapoor R, Chu CK, et al. Evaluation of a deep learning system for identifying glaucomatous optic neuropathy based on color fundus photographs. J Glaucoma. June 21, 2019. [Epub ahead of print].