Research from the Centre for Eye Research Australia (CERA) has shown an artificial intelligence (AI) system can accurately detect and measure distinctive deposits in the eye critical in the progression of age-related macular degeneration (AMD).
The findings open the possibility to study the condition at a scale previously thought impossible.
These deposits - called reticular pseudodrusen (RPD) - have previously been associated with a higher risk of progressing to late-stage AMD.
The deposits are not fully understood, and learning more about them requires looking at eye scans of many people who have them. However, identifying and measuring them is a challenge.
"Because of how they present on scans, identifying these deposits can be difficult for many clinicians," says Associate Professor Zhichao Wu, one of the corresponding authors of the study.
"And accurately quantifying or measuring their extent would be too prohibitively time-consuming to do manually.
"We want to be able to do large studies of hundreds or even thousands of people with these deposits to learn more about what they mean for age-related macular degeneration, but there aren't the people or time to do it by hand."
Efficient scanning
Developed alongside Dr Himeesh Kumar and collaborators at the University of Washington, the AI model was trained on hundreds of scans to both automatically detect RPD and measure their amount in the eye.
They then compared the effectiveness of the model against eyecare professionals and found it performed just as well as expert human clinicians.
"As well as automating the process of detecting RPD, the model also gives us an objective way of measuring the amount of RPD present," says Associate Professor Wu.
"This is not research that would be feasible to do by hand and really unlocks the possibility for us to learn about these deposits much more quickly."
To facilitate this, the team have released their model publicly so other researchers can use in their own work.
"We want to make this research possible for people all round the world," says Associate Professor Wu.
"Being able to learn more from both scans and teams around the world moves us closer towards a greater understanding about the role of these deposits in AMD."
Read the research
H. Kumar, Y. Bagdasarova, S. Song, et al., " Deep Learning-Based Detection of Reticular Pseudodrusen in Age-Related Macular Degeneration," Clinical & Experimental Ophthalmology (2025): 1-8, https://doi.org/10.1111/ceo.14607.
A major collaboration is closer to discovering who is most at risk of vision loss from age-related macular degeneration.
CERA's Macular Research Unit is investigating links between sleep apnoea and age-related macular degeneration.
