Eye exams are becoming much more powerful than many people realize. Modern eye clinics now use advanced imaging technology that allows doctors to look deep inside the eye without surgery, needles, or discomfort.
These scans can reveal tiny changes in the retina, the light-sensitive tissue at the back of the eye that plays a key role in vision. Because many eye diseases develop slowly and often show few symptoms at first, these scans have become an important tool for protecting eyesight.
One of the most widely used imaging methods is called optical coherence tomography, often shortened to OCT. This technology creates detailed three-dimensional pictures of the retina and optic nerve.
A single scan can generate hundreds of images that show different layers of the eye in remarkable detail. Doctors use these images to diagnose conditions such as glaucoma, diabetic eye disease, and age-related macular degeneration, one of the leading causes of blindness in older adults.
Although these scans provide valuable information, they also create a challenge. Each scan produces a huge amount of data that doctors must carefully review.
Examining hundreds of images for every patient takes time and can sometimes make it difficult to spot subtle signs of disease. As the number of patients continues to grow, researchers have been searching for ways to help doctors process this information more efficiently.
A team of researchers from Washington University School of Medicine in St. Louis, working with scientists from the University of Washington and Genentech, has developed a new artificial intelligence system that may help solve this problem.
The technology, known as OCTCube-M, was designed to analyze three-dimensional eye scans and identify signs of disease more quickly and accurately than previous systems.
The research was published in Nature Biomedical Engineering.
Artificial intelligence has already shown promise in medical imaging. AI systems can examine large numbers of images much faster than humans and identify patterns that may be difficult to notice with the naked eye. Previous AI systems for eye disease mainly relied on two-dimensional images.
However, the retina is a three-dimensional structure, and many diseases spread across multiple layers of tissue. The researchers believed that training AI with full 3D scans would provide a more complete picture of what is happening inside the eye.
To build the new system, the researchers trained the AI using more than 26,000 eye scans containing over 1.6 million individual retinal images. This enormous dataset allowed the AI to learn how healthy and diseased retinas appear across many different patients and imaging devices.
The results were impressive. Compared with older AI systems trained on two-dimensional images, OCTCube-M was better at identifying six of eight major retinal diseases.
In practical terms, this improvement could help detect dozens of additional cases for every thousand patients examined. Earlier detection often means earlier treatment, which can help preserve vision and prevent permanent damage.
The system also showed particular strength in predicting the progression of geographic atrophy, an advanced form of age-related macular degeneration. This condition affects millions of people worldwide and currently has limited treatment options.
By estimating how quickly the disease is likely to worsen, doctors and researchers may be able to make better treatment decisions and design more effective clinical trials.
One of the most surprising findings was that the AI could identify health risks outside the eye. Researchers discovered that retinal images contain clues about the condition of blood vessels throughout the body. Because the blood vessels in the retina share similarities with those in the heart, brain, and kidneys, changes in the eye may reflect problems elsewhere.
Using only retinal images, the AI was able to estimate the risk of conditions such as heart attack, stroke, and kidney failure. This raises the possibility that a routine eye examination could someday help doctors detect serious diseases before symptoms appear.
The researchers later improved the system even further by combining OCT scans with two additional imaging methods.
By bringing together information from multiple types of eye images, the AI achieved an even more detailed understanding of eye health. In predicting geographic atrophy progression, the combined system outperformed existing methods by nearly 50 percent.
The researchers plan to continue improving OCTCube-M by training it with larger datasets and additional types of medical images. They hope the technology will eventually help doctors diagnose diseases earlier, personalize treatments, and accelerate the development of new therapies.
The findings are exciting because they show how eye scans may become much more than a tool for checking vision. In the future, a simple eye exam could provide important information about overall health.
If you care about kidney health, please read studies about how to protect your kidneys from diabetes, and drinking coffee could help reduce risk of kidney injury.
For more health information, please see recent studies about foods that may prevent recurrence of kidney stones, and eating nuts linked to lower risk of chronic kidney disease and death.
Source: Washington University School of Medicine in St. Louis.
