Diagnosing eye diseases is a tedious task. Trained ophthalmologists sit in darkened rooms and examine retinal images on a screen. They look for abnormalities that may indicate diabetic retinopathy, one of the leading causes of blindness in Singapore. The work is very tiring, shifts usually last only 4 hours.
A startup wants to change that. Singapore-based health technology company Iris is using artificial intelligence (AI) to diagnose 3 types of eye diseases: diabetic retinopathy, suspected glaucoma, and age-related macular degeneration.
That’s done with Selena+ deep learning software, which can automatically detect these diseases from images of retinas. When clinicians upload a retinal image, the software compares it to a data set it’s been trained on, which includes nearly 500,000 images of retinas with all 3 diseases. In a few minutes, it will bring the result of the diagnosis.
Selena+ was developed by the Singapore Eye Research Institute and the School of Computing at the National University of Singapore. Dr. Daniel Ting, associate consultant at Singapore National Eye, was told that “professional class can reduce the workload of students by 75%, allowing them to focus their work on more complex and complex tasks”. CASIA.
Some clinics and hospitals in Malaysia, Vietnam and the Philippines are using the software, and its footprint is set to increase. Through an agreement with Japan’s Topcon Healthcare Solutions, it will soon be available in 18 Southeast and South Asian countries.
Other organizations are similarly working with deep learning tools to detect eye diseases. American researchers have developed a method to predict the risk of developing retinopathy in diabetic patients within 2 years, according to a study published in The Lancet.
Scientists at New York’s Department of Eye and Ear Disease at Mount Sinai have developed an AI algorithm to detect age-related macular degeneration. The system evaluates if the patient has the disease and its severity.
Some of the images that Selena+ trained on were shared with the Singapore Integrated Diabetic Retinopathy Program (SiDRP), a screening project for local polyclinics, said Dr Ting. Therefore, the software mainly processes the retina images of Singaporean Chinese, Malays and Indians. That doesn’t mean it can’t figure out the situation for non-Asians either.
When examined in datasets from Latino, African-Caribbean, and Caucasian populations, retinography could still identify cases of moderate to severe diabetes. The color of the fundus—the inner part of the eye—is important because it varies by breed. “The white Caucasian population has a lighter colored fundus compared to Asians and Africans,” Dr. Ting said.
The most important thing is to feed the AI system with a large amount of data. “It is important to ensure that the training data sets contain representative data and robust truth,” said Dr. Ting. “It is equally if not more important to test the AI system on multiple independent data sets.”
Dr. Ting noted that Selena+ has not yet been evaluated on portable retina cameras. It is still awaiting US Food and Drug Administration approval to confirm the safety and efficacy of the medical device.
EyRis is currently developing a classifier for severity levels of various eye diseases, Dr. Ting added. And in addition, AI can also be used to diagnose eye conditions other than diabetic retinopathy, suspected glaucoma, and age-related macular degeneration.
“That said, we might not call those new algorithms Selena+,” Dr. Ting said. A new iteration of the product may be in the works.