Researchers from Duke University have developed an artificial intelligence (AI) program that can accurately predict the worsening of a common eye condition, potentially leading to more timely and effective treatments for patients.
The program, which uses a deep learning algorithm, boasts a remarkable 94% accuracy rate in determining whether age-related macular degeneration (AMD) will progress within a year.
Detailed in JAMA Ophthalmology, the algorithm assesses eye scans to foresee the advance to geographic atrophy (GA)—a severe stage of AMD that significantly impairs central vision.
This breakthrough could greatly enhance the efficiency of clinical trials and speed up the delivery of new treatments aimed at preserving sight in older adults.
“Being able to predict the progression of AMD allows us to intervene sooner and help maintain our patients’ vision,” explains Eleonora Lad, MD, Ph.D., a senior study author, retina specialist, and vice-chair of clinical research at the Duke Eye Center.
DeepGAze: A Game-Changer for Screening
Developed in collaboration with experts from Duke AI Health and the Duke Department of Electrical and Computer Engineering, the tool known as DeepGAze could revolutionize screening processes, influencing the frequency of patient visits and timing of treatments.
Moving from intermediate AMD to GA means going from a condition that might not disturb daily life much to one that can severely impact vital vision functions, such as reading and driving.
With the FDA’s recent approval of drugs like pegcetacoplan and avacincaptad pegol, which slow GA progression, predicting the onset of GA has become even more critical.
Clinical Advancements and FDA Approvals
Lad played a key role in the clinical trials leading to the approval of pegcetacoplan, celebrated as the first FDA-approved drug for GA.
Published results from these trials have shown that regular injections can significantly slow the expansion of GA lesions.
In the quest to identify patients who will progress from intermediate AMD to GA, DeepGAze examined numerous spectral-domain optical coherence tomography scans.
It did so efficiently without needing the additional, labor-intensive steps of manual image annotations traditionally used by expert graders, which did not improve the algorithm’s performance.
Implications for Future Care
With its ability to predict disease progression to GA within a year at such a high success rate, this AI tool represents a significant step forward.
It could potentially reshape how doctors manage AMD, allowing for earlier intervention and better planning for treatments that could halt or slow down the progression of this life-altering eye condition.
If you care about eye health, please read studies about how vitamin B may help fight vision loss, and MIND diet may reduce risk of vision loss disease.
For more information about eye disease, please see recent studies about how to protect your eyes from glaucoma, and results showing this eye surgery may reduce dementia risk.
The research findings can be found in JAMA Ophthalmology.
Follow us on Twitter for more articles about this topic.
Copyright © 2023 Knowridge Science Report. All rights reserved.
