IBM researchers use deep learning, neural networks to screen for diabetic retinopathy with 86 percent accuracy

By Heather Mack
Share

Out of the 422 million people around the world living with diabetes, one in three of them will develop diabetic retinopathy (DR), a common condition that can lead to permanent blindness if left untreated. While early detection and treatment can dramatically reduce that risk, a third of people with diabetes have never even been screened for DR, as many are living in low-income, medically underserved areas that make the much-needed clinical intervention impossible.

But new research from IBM suggests technology can rise up to fill those healthcare gaps. Using a mix of deep learning, convolutional neural networks and visual analytics technology based on 35,000 images accessed via EyePACs, the IBM technology learned to identify lesions and other markers of damage to the retina’s blood vessels, collectively assessing the presence and severity of disease. In just 20 seconds, the method was successful in classifying DR severity with 86 percent accuracy, suggesting doctors and clinicians could use the technology to have a better idea of how the disease progresses as well as identify effective treatment methods.

While it’s still early days for the technology, researchers are excited for the possibility of using IBM’s method to replace or augment traditional DR screening and treatment processes, which currently require a comprehensive in-person visit with a specialist.

“A comprehensive, dilated eye exam is required in order to identify DR and many other eye diseases, and that can't be done on a smart phone with the current camera technology in existence today. Where a smart phone could play a role is also in the cloud delivery of results to care teams and to patients - for example, you could use your smart phone to pull test results down from the cloud via any browser,” Joanna Batstone, the VP and lab director of the IBM Research center in Australia that published the study, told MobiHealthNews in an email.

The technology could also extend the research of eye specialists by connecting them with primary care providers or clinicians working in rural or underserved areas, Batstone wrote.

“For example, in Australia, legislation recently changed which allow general doctors to conduct these dilated eye exams and send these images away to specialists for remote analysis. While this is a big step forward for increasing access to screening, the risk of missed follow up appointments to get the results remains a challenge,” Batstone wrote. “Technology like this could potentially give patients an indication of their risk of the disease on the spot, referring them to specialists who would be able to focus their time and resources on those patients who've been identified with the disease, instead of screening every image.”

As the research continues, IBM hopes to work in partnerships with opthalmologists in a clinical setting, and Batstone said Watson Health Imaging is also working to bring cognitive imaging technologies into the eye health field. 

IBM isn’t the only company using deep learning to detect diabetic retinopathy. In November 2016, Google researchers published a paper in JAMA showing how their similar approach can the condition with better than 90 percent accuracy.

We’ve seen at least one real-world example that shows how scanning technology could improve access to DR screenings, albeit a bit more low-tech than what IBM or Google is working on. Using a mix of in-office visits, telemedicine and web-based screening software, the Los Angeles Department of Health Services was able to greatly expand the number of patients in its safety net hospital who got screenings and referrals.  In an article published in the journal JAMA Internal Medicine, researchers describe how the two-year collaboration using Safety Net Connect’s eConsult platform resulted in more screenings, shorter wait times and fewer in-person specialty care visits. By deploying Safety Net Connect’s eConsult system to a group of 21,222 patients, the wait times for screens decreased by almost 90 percent, and overall screening rates for diabetic retinopathy increased 16 percent. The digital program also eliminated the need for 14,000 visits to specialty care professionals.