by Jacob Gnieski, Emory University

(a), (c) Original Control Images; (b), (d) Feature map for 'median-Laws L3S3L3' feature for (a) and (c), respectively; (e), (g) Original Safety Images; (f), (h) Feature map for 'median-Laws L3S3L3' feature for (e) and (g), respectively. (i), (j) Zoomed-in version of (f) and (h), respectively. The PHF and vitreous debris are highly expressed as reflected by the warmer color tones of the heatmap compared to the rest of the vitreous compartment with lower feature expression (reflected by cooler color tones). Credit: Heliyon (2024). DOI: 10.1016/j.heliyon.2024.e32232

Age-related macular degeneration (AMD) is a leading cause of vision loss in the U.S., affecting 11 million people, particularly older adults. The more severe form, neovascular age-related macular degeneration (nAMD), is characterized by abnormal blood vessel growth under the retina. These vessels leak fluid or blood, leading to vision loss. Besides age, smoking, poor diet, and lack of physical activity also contribute to the risk.

The primary treatment for nAMD is anti-VEGF drugs. This treatment involves injecting a drug into the eye that blocks a protein called vascular endothelial growth factor (VEGF), which is responsible for the growth of abnormal blood vessels in the retina; however, it can cause eye inflammation as a serious side effect.

A team of researchers from Emory AI.Health and Cleveland Clinic aimed to predict which patients might develop this inflammatory response. By combining routine optical coherence tomography (OCT) scans with machine learning and precision medicine, they sought to identify patterns in eye scan images that could appear before or during inflammation caused by anti-VEGF drugs.

Identifying these patterns early could help doctors detect inflammation sooner and adjust treatment to prevent vision loss.

Detecting trouble early: The study

Published in Heliyon, the study analyzed images of 67 eyes from a retrospective clinical trial involving patients with nAMD. Researchers extracted specific texture-based features from OCT scans, focusing on the vitreous compartment—the clear gel in the eye. Using a machine learning model developed by Emory AI.Health, they identified patterns signaling inflammation before it was clinically visible.

The machine learning model accurately distinguished which patients would develop inflammation, achieving a 76% accuracy rate before anti-VEGF treatment and 81% accuracy at the time of injection. This data suggests its potential as a valuable tool for early detection.

"Macular degeneration is personal to me because my father suffers from it. As our population ages, more people will experience nAMD. Anti-VEGF agents can slow down macular degeneration but come with risks," said Anant Madabhushi, Ph.D., executive director of Emory AI.Health and principal investigator of the study.

"Our study provides valuable data for clinicians to make better treatment decisions, potentially reducing the dosage or combining these agents with anti-inflammatory drugs to prevent severe complications."

"This study validates our AI algorithms in a retrospective clinical trial and underscores the potential of precision medicine in ophthalmology," said Sudeshna Sil Kar, Ph.D., first author of the study and associate scientist at Emory AI.Health. "Next, we hope to embed our algorithms in prospective clinical trials to identify patients likely to develop these adverse events in real-time."

More information: Sudeshna Sil Kar et al, Optical coherence tomography-derived texture-based radiomics features identify eyes with intraocular inflammation in the HAWK clinical trial, Heliyon (2024). DOI: 10.1016/j.heliyon.2024.e32232

Journal information: Heliyon 

Provided by Emory University