AI-assisted technology makes retinal imaging 100 times faster, with greater contrast
Click Here to Manage Email Alerts
Key takeaways:
- Researchers at NIH have developed AI-based technology that de-speckles retinal pigment epithelial cell images obtained via OCT.
- The technology improved contrast by 3.5-fold and end-to-end time by 99-fold.
Researchers at NIH have developed an AI-assisted strategy that can improve cellular contrast of retinal images captured via OCT while reducing image acquisition and processing time by nearly 100-fold.
“Adaptive optics combined with optical coherence tomography is an emerging technology that can reveal the status of individual cells in patients’ eyes,” Johnny Tam, PhD, senior investigator in the clinical and translational imaging section at NIH’s National Eye Institute, told Healio. “By incorporating artificial intelligence to speed up the imaging, we hope to bring this technology one step closer to practicing optometrists.”
In a study published in Communications Medicine, Tam and colleagues sought to test a novel AI-based method called parallel discriminator generative adversarial network (P-GAN), a deep-learning algorithm developed to identify and recover speckle-obscured features captured via adaptive-optics OCT (AO-OCT), thereby reducing the volume of images needed for adequate visualization of retinal pigment epithelial (RPE) cells.
From 2019 to 2022, researchers recruited seven healthy participants (mean age, 29.1 years) with no history or signs of ocular disease and imaged a total of eight eyes. They obtained AO-OCT volumes at four retinal locations, with 120 speckled volumes taken at each site.
According to results, P-GAN effectively recovered the cellular structure from speckle-obscured images, improving RPE cell contrast by 3.5-fold and end-to-end time needed to visualize cells by 99-fold.
With this AI-assisted method, researchers noted that cellular contrast can be improved with just one speckled image, substantially reducing the time currently associated with AO-OCT and making it more accessible for routine clinical use.
“It was surprising that artificial intelligence was able to take just a single, noisy snapshot of the retinal pigment epithelial cells and transform that into something meaningful,” Tam said. “We should start to think about artificial intelligence as a part of the imaging strategy, and not just as something to be applied to images after they have been obtained.”
Reference:
- AI makes retinal imaging 100 times faster, compared to manual method. https://www.nei.nih.gov/about/news-and-events/news/ai-makes-retinal-imaging-100-times-faster-compared-manual-method. Published April 10, 2024. Accessed April 23, 2024.
Collapse
Read more about
Click Here to Manage Email Alerts