Researchers at the National Institutes of Health in the United States have applied artificial intelligence (AI) to a technique capable of generating high-resolution images of eye cells. The new technology increases the speed of retinal imaging by 100 times and enhances image contrast by 3.5 times. This advancement will provide researchers with better tools for assessing age-related macular degeneration and other retinal diseases. The related research was published in the latest issue of the journal "Communications Medicine."
This adaptive optics (AO) technology is used to improve optical coherence tomography (OCT) devices. Like ultrasound, OCT is non-invasive, fast, painless, and is a standard equipment in most ophthalmology clinics.
Retinal pigment epithelium (RPE) cells are a layer of tissue behind the light-sensitive retina, supporting the metabolically active retinal neurons. Many retinal diseases occur when the RPE is compromised.
When imaging RPE cells with AO-OCT, there is a phenomenon called speckle. Speckle interference in AO-OCT is similar to cloud cover interference in aerial photography. Certain parts of the image may be obscured at any given moment.
To address this, the research team developed a parallel Generative Adversarial Network (P-GAN) technology based on AI. By providing P-GAN with nearly 6000 human RPE images collected using AO-OCT, the team trained the AI to identify and restore cell features obscured by speckle.
During testing on new images, P-GAN successfully removed speckles from RPE images, restoring cell details. Researchers estimate that P-GAN reduces imaging acquisition and processing time to just 1% of the original. The new technology enables cellular-level resolution of 3D retinal structures, amplifying early signs of disease.
