Artificial intelligence can diagnose glaucoma with better accuracy than teleophthalmologists, Aravind Eye Hospital has found. The Pondicherry unit of the hospital was encouraged by the results to deploy AI technology to all its vision centres in the region.
The hospital conducted screenings, in six centres, on 303 persons and compared it with the results provided by teleophthalmologists who used the Fundus camera. It found that AI had generated diagnosis with better accuracy.
“A Bengaluru-based company has developed the technology, where you can take the image of the retina on a iPhone. The AI can be used in the offline mode as well. We check the eye pressure and the optic nerve changes. We trained the computer to identify the optic nerve changes and pick up glaucoma,” explained R. Venkatesh, chief medical officer at the hospital, adding: “The AI can refer the changes to a doctor.”
Much work has been done using AI to detect diabetic retinopathy, he acknowledged, but this was the first time that the technology was used to detect glaucoma. To study accuracy, the centres used Fundus cameras to take images and upload them to the electronic medical records at the hospital. The results were compared by reinvestigating the patients at the hospital. “We found that when a doctor referred a patient the chances of false positivity were high,” he said.
The AI presented 91% sensitivity in identifying glaucoma and 93% specificity to declare an image is normal. Sensitivity and specificity are measures of how well a test can correctly identify people with or without a disease. With AI, 91% of the time, the problem was accurately identified and 93% of the time, the AI was able to declare a person disease-free. The results have encouraged the hospital to expand the technology to all its 13 centres.
The hospital funded the hospital visit of the people with disease for investigation and visited people with normal vision at the centres to confirm the results. Swathi Upadhyaya is the chief author of the study, Diagnostic performance of the offline Medios Artificial Intelligence (AI) for glaucoma 2 detection in a rural tele-ophthalmology setting, published in Ophthalmology Glaucoma. Dr. Venkatesh is the mentoring author.
Hospital authorities said, “Operating offline and on-the-edge, the AI system provides a cost-effective solution for glaucoma screening, especially in areas with limited internet access. This technology has the potential to improve early detection, reduce healthcare costs, and enhance eye care in underserved communities”.
Published – September 19, 2024 10:04 pm IST
link