AI-Powered Insights Could Prevent Vision Loss in Young Adults

Recent advancements in artificial intelligence (AI) have paved the way for predicting which young adults require treatment to stabilize their corneas and prevent vision loss, according to a study presented at the 43rd Congress of the European Society of Cataract and Refractive Surgeons (ESCRS).

The study primarily focused on keratoconus, a condition that typically manifests in teenagers and young adults, affecting approximately 1 in 350 individuals. While some patients can manage their symptoms with contact lenses, others may experience rapid deterioration of their condition, potentially necessitating a corneal transplant if left untreated. Traditionally, determining which patients need intervention involves long-term monitoring.

Researchers harnessed AI to analyze images of patients' eyes, integrating this data with other relevant information to accurately forecast which individuals required immediate treatment and which could continue to be monitored.

The research, conducted by a team from Moorfields Eye Hospital NHS Foundation Trust and University College London, revealed that keratoconus causes the cornea--the transparent layer at the front of the eye--to bulge. This condition is a leading cause of corneal transplants in Western countries. A specific treatment known as 'cross-linking' can effectively halt the progression of keratoconus when administered before permanent scarring occurs, thus reducing the likelihood of requiring a transplant. However, the medical community has struggled to predict which patients will experience progression.

In the study, a cohort of patients referred for keratoconus evaluation underwent extensive assessments, including optical coherence tomography (OCT) scans to analyze the shape of their corneas. The AI system evaluated a vast dataset comprising 36,673 OCT images from 6,684 patients, alongside additional patient information.

The results indicated that the AI algorithm could accurately identify whether a patient's condition was likely to worsen or remain stable based solely on data from their initial examination. The researchers were able to categorize approximately two-thirds of the patients into a low-risk group, indicating they did not require immediate treatment, while the remaining third fell into a high-risk category requiring prompt intervention. Furthermore, when incorporating data from a follow-up visit, the algorithm achieved an impressive success rate of 90% in accurately categorizing patients.

The cross-linking procedure utilizes ultraviolet light combined with riboflavin drops to reinforce the cornea, boasting a success rate exceeding 95%. The findings of this research highlight the potential of AI in distinguishing between patients who necessitate treatment and those who can safely continue their observation.

The implications of this study are significant, as it suggests that high-risk keratoconus patients could receive preventive treatment prior to the onset of severe complications. Additionally, patients deemed low-risk would benefit from reduced frequency of monitoring, optimizing healthcare resources.

The research team is now focused on developing a more sophisticated AI algorithm that can analyze millions of eye scans, which could aid in forecasting keratoconus progression as well as detecting other ocular conditions.

An expert in the field, who was not involved in the study, emphasized the challenges of managing keratoconus and the importance of timely treatment to avoid invasive procedures. The potential of AI to provide early predictions from initial consultations could significantly enhance patient outcomes and reduce unnecessary interventions.