Google's AI program: Building better algorithms for detecting eye disease

March 13, 2018

SAN FRANCISCO - March 13, 2017 - The ability of artificial intelligence (AI) to help screen patients for a common diabetic eye disease gains momentum with a new study published online today in Ophthalmology, the journal of the American Academy of Ophthalmology. Lily Peng, M.D., Ph.D., and her colleagues at Google AI research group, show that they could improve their disease detecting software by using a small subset of images adjudicated by ophthalmologists who specialize in retinal diseases. The specialists' input was then used to improve their computer's performance so that it was roughly equal to that of individual retinal specialists.

More than 29 million Americans have diabetes, and are at risk for diabetic retinopathy, a potentially blinding eye disease. People often don't notice changes in their vision in the disease's early stages. But as it progresses, diabetic retinopathy usually causes vision loss that in many cases cannot be reversed. That's why it's so important that people with diabetes have yearly screenings.

In earlier research, Dr. Peng and her team used neural networks--complex mathematical systems for identifying patterns in data--to recognize diabetic retinopathy. They fed thousands of retinal scans into these neural networks to teach them to "see" tiny hemorrhages and other lesions that are early warning signs of retinopathy. Dr. Peng showed the software worked roughly as well as human experts.

But Dr. Peng is interested in developing a system that would be good enough for her grandmother. So, to improve the accuracy of the software, she included the input of retina specialists, ophthalmologists who specialize in diseases of the retina.

"For my grandma, I would love to have a panel of subspecialists who actually treat the disease, to sit and debate her case, giving their opinion," Dr. Peng said. "But that is really expensive and it's hard to do. So how do you build an algorithm that gets close to this?"

To tease out how this could be done, Dr. Peng compared the performance of the original algorithm with manual image grading by either a majority decision of three general ophthalmologists, or a consensus grading by three retinal specialists.

The grading of diabetic retinopathy can be a complex process that requires the identification and quantification of fine features such as small aneurysms and hemorrhages. As a result, there can be a fair amount of variability among physicians examining images, looking for disease.

The retina specialists graded the images separately, then worked together to resolve any disagreements. Their review and subsequent consensus diagnosis offered considerable insight into the grading process, helping to correct errors such as artifacts caused by dust spots, distinguishing between different types of hemorrhages, and creating more precise definitions for "gray areas" that make it difficult to make a definitive diagnosis. At the end of the process, the retina specialists indicated that the precision used in the decision process was above that typically used in everyday clinical practice.

Using these specialist-graded images, Dr. Peng could then fine-tune the software, which improved their model's performance and improved detection of disease.

"We believe this work provides a basis for further research and raises the bar for reference standards in the field of applying machine learning to medicine," Dr. Peng said.
-end-
About the American Academy of Ophthalmology

The American Academy of Ophthalmology is the world's largest association of eye physicians and surgeons. A global community of 32,000 medical doctors, we protect sight and empower lives by setting the standards for ophthalmic education and advocating for our patients and the public. We innovate to advance our profession and to ensure the delivery of the highest-quality eye care. Our EyeSmart® program provides the public with the most trusted information about eye health. For more information, visit aao.org.

About Ophthalmology

Ophthalmology, the official journal of the American Academy of Ophthalmology, publishes original, peer-reviewed, clinically-applicable research. Topics include the results of clinical trials, new diagnostic and surgical techniques, treatment methods, technology assessments, translational science reviews and editorials. For more information, visit http://www.aaojournal.org.

American Academy of Ophthalmology

Related Retinopathy Articles from Brightsurf:

Promising discovery for patients with diabetic retinopathy
A study published in the journal Science has shed light on a cellular process that occurs in the retinas of people with diabetic retinopathy.

A vitamin A analog may help treat diabetic retinopathy
Diabetic retinopathy is a common complication of diabetes and a leading cause of blindness among the working-age population.

Scientists identify a potential treatment candidate for early type 2 diabetic retinopathy
Diabetic retinopathy is one of the main vascular complications of type 2 diabetes, and the most common cause of visual deterioration in adults.

No need to draw blood -- smart photonic contact lens for diabetic diagnosis and retinopathy treatment
Sei Kwang Hahn and his research team from POSTECH developed a smart LED contact lens.

Scientists identify a possible new treatment for diabetic retinopathy
About 1 in 3 diabetic patients develops diabetic retinopathy (DR), which can impair vision and lead to blindness.

Asthma medication inhibits changes in diabetic retinopathy in type 1 diabetes mouse
A new study found the asthma medication montelukast (brand name Singulair) can inhibit early changes in diabetic retinopathy, the eye disease which develops due to diabetes, in a mouse model of type 1 diabetes.

Earlier detection of diabetic retinopathy with smartphone AI
Equipping a smartphone to capture retinal images and utilizing artificial intelligence to interpret them may help overcome barriers to ophthalmic screening for people with diabetes, new Kellogg Eye Center research shows.

CHOP researchers develop easy-to-implement predictive screening tool for retinopathy
A multi-hospital collaboration led by researchers at Children's Hospital of Philadelphia (CHOP) has found a simple method of determining which premature infants should be screened for retinopathy of prematurity (ROP).

Researchers develop new tools in the fight against diabetic blindness
Estimates are that 600 million people will have some sort of diabetic retinopathy by 2040.

The microenvironment of diabetic retinopathy supports lymphatic neovascularization
'We asked whether proliferative diabetic retinopathy involves the growth of new lymphatic vessels in addition to blood vessels -- and, indeed, we found expression of lymphatic markers in the PDR tissues.' The new study, conducted at the University of Helsinki, Finland, was published in the Journal of Pathology.

Read More: Retinopathy News and Retinopathy Current Events
Brightsurf.com is a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to Amazon.com.