AI moves cellular-level imaging of the eye one step closer to everyday clinic use

Using a novel AI method called residual in residual transformer generative adversarial network (RRTGAN), researchers were able to effectively restore pixel resolution from fewer measurements. Credit: Johnny Tam

By combining artificial intelligence with imaging, researchers at the National Institutes of Health (NIH) have reduced by three-fourths the data required to use an advanced ocular imaging technology capable of resolving individual cells in the eye, bringing it one step closer to being accessible for everyday use in eye clinics.

"Getting the most advanced ophthalmic imaging technologies into the hands of healthcare providers will vastly improve the ability to detect retinal diseases earlier, and guide treatments to prevent vision loss," said Johnny Tam, Ph.D., investigator at NIH's National Eye Institute (NEI) and senior author of the study report, which published in the Nature journal, npj Artificial Intelligence.

Many leading causes of blindness involve the retina, the light sensitive tissue at the back of the eye that breaks down in diseases such as age-related macular degeneration and diabetic retinopathy. In recent years, enormous advances in using an imaging technique known as adaptive optics have made it possible to capture detailed 3D images of retinal cells. Adaptive optics can be combined with readily available technologies such as optical coherence tomography (OCT) to capture cellular-level assessments of the retina that help determine if a therapy is working to guide treatment decisions.

But current methods of using adaptive optics imaging along with OCT are impractical for use in busy eye clinics. They require a person to hold their eye very still while hundreds of images of the retina are taken, which is challenging not only for the patients, but also for the clinics because of the huge amount of imaging data that is generated.

"When using adaptive optics, we face a tradeoff between taking fewer measurements to make the procedure more efficient and obtaining good image resolution, which ideally would be sufficient for visualizing microscopic structures of each cell," said Tam.

"Our solution circumvents this tradeoff by developing a novel AI method - called residual in residual transformer generative adversarial network (RRTGAN) - that effectively restores pixel resolution from fewer measurements," said the first author on the paper, Vineeta Das, Ph.D., postdoctoral fellow at the NEI.

"Our AI method requires only a fourth of the acquired data, which will enable rapid cellular-scale 3D imagine of the eye in a clinical setting," she said.

The work was supported by the Intramural Research Program at the NEI.

Das V; Bower AJ; Aguilera N; Li J; Tam J. "Artificial intelligence assisted retinal imaging enables dense pixel sampling from sparse measurements." Published December 9, 2025 in npj Artificial Intelligence. https://www.nature.com/articles/s44387-025-00038-2

Contact