Eye Care - Is AI The Future?

The future of eye care and eye health improves with the help of Artificial Intelligence.

At the Singapore National Eye Centre (SNEC), screening machines with artificial intelligence prove that they can think and decide like humans.

Artificial Intelligence (AI) for Eye Care and Screening of Eye Diseases

These machines can screen patients for diabetic retinopathy and other related eye diseases by referencing a database of almost half a million retinal images from multiethnic populations in Singapore and globally.

The images are stored in what is called an Artificial Intelligence System (AI System), which was jointly developed by SNEC, Singapore Eye Research Institute (SERI) and the National University of Singapore’s (NUS) School of Computing.

Claimed as the world’s first, the system has proved to be highly accurate in identifying images with and without eye disease. It can also detect glaucoma and age-related macular degeneration.

Results of a study on its use were published in the peer-reviewed Journal of the American Medical Association in December last year.

For the study, researchers worked with several leading global eye centres, including those in Australia, China, the United States, Mexico and Hong Kong. Many of the retinal images used in the system came from these countries.

Screening for Eye Diseases

The AI System can screen for eye diseases like a trained professional because it has, at its core, a Deep Learning System.

According to Professor Wong Tien Yin, Medical Director, SNEC, and Chairman, SERI, the Deep Learning System uses an innovative algorithmic approach to “train” technology to think and decide like humans.

“It can process large amounts of raw data and recognise intricate structures and patterns that may not be visible to the human eye,” said Prof Wong, the study’s senior author and Vice-Dean, Duke-NUS Medical School.

He said it would be useful in screening patients for diabetic retinopathy in Singapore and elsewhere.

“In countries that have these screening programmes, such as the United Kingdom and Singapore, it will increase efficiency and reduce costs by replacing a large proportion of what now requires human assessment.”

It will also make it easier to set up screening programmes in future communities, as it could largely be done by artificial intelligence. “It will also save costs and improve the efficiency of healthcare systems by allowing ophthalmologists and optometrists to concentrate only on diabetic retinopathy cases that require treatment.”

There are plans to develop more complex algorithms to train the Deep Learning System to do even more.

Predicting Eye Diseases

“The next step is to train the AI system to predict which patients will have eye diseases in the future by simply looking at their retinal images before they develop any disease,” said Prof Wong.

“The Deep Learning System can process large amounts of raw data, and recognise intricate structures and patterns that may not be visible to the human eye.” – Professor Wong Tien Yin, Medical Director SNEC

Stocking the Image Bank

Researchers hope to boost the number of retinal images they have to five million in the future.

“Ours is currently the largest data set in a Deep Learning System to screen for diabetic retinopathy and other sight-threatening conditions such as glaucoma and age-related macular degeneration,” said Dr Daniel Ting, Associate Consultant, Cataract and Comprehensive Ophthalmology Department, SNEC.

“Artificial intelligence is deemed to be the fourth industrial revolution in human history. In healthcare, we need to embrace this technology early to improve work efficiency while maintaining the high standards of clinical care,” said Dr Ting, who is also Assistant Professor, Duke-NUS Medical School, and lead author of this study.

The AI System is being further tested in the Singapore Integrated Diabetic Retinopathy Programme alongside screening by professional graders or optometrists.

Once tests are satisfactorily completed, they could be in use by the end of 2018. Other possible objectives are to predict other diabetes-related complications, such as stroke, coronary diseases and chronic kidney diseases in people with the illness.

Why is Artificial Intelligence Needed for Detecting Eye Diseases?

Diabetic retinopathy is a leading cause of preventable blindness among working Singaporeans. There are about 600,000 people with diabetes aged between 18 and 69 here.

Diabetic patients can suffer vision loss from damaged blood vessels in the retina. An estimated one in three people with diabetes has this condition, but many affected are unaware they have it.

The most effective way of preventing unnecessary vision loss from diabetic retinopathy is early screening and treatment. The Singapore Integrated Diabetic Retinopathy Programme was set up to do just that.

Currently, the programme relies on the grading of retinal photographs by trained professional graders or optometrists. However, with diabetes on the rise, screening challenges include the training and retention of professional graders and optometrists, the availability of and access to screening services, and the financial sustainability of these programmes in the long run.

These challenges indicate a clear need for a more accurate, efficient and cost-effective method of detecting diabetic retinopathy. This is where artificial intelligence comes into the picture.

Download the HealthHub app on Google Play or Apple Store to access more health and wellness advice at your fingertips.

Back to Top