Artificial intelligence (AI) could represent a major breakthrough in ophthalmology. Researchers at Moorfields Eye Hospital NHS Foundation Trust and University College London have developed an AI model that predicts which patients with keratoconus require immediate treatment and which patients can be safely monitored.
Keratoconus is a progressive eye condition in which the cornea bulges outwards. This leads to distorted vision and can ultimately cause severe vision loss. The disease often manifests itself in the teenage years and young adulthood and affects up to 1 in 350 people. In mild cases, the condition can be treated with special contact lenses, but without timely intervention, a corneal transplant may be necessary. This is a major and risky procedure.
A commonly used treatment is corneal cross-linking, in which UV light and riboflavin (vitamin B2) drops strengthen the cornea. This procedure is effective in 95 per cent of cases, provided it is performed in a timely manner. The problem is that doctors often do not know which patients will experience a rapid worsening of the condition and therefore require immediate treatment. It is also difficult to estimate which patients will remain stable without intervention. As a result, patients are monitored intensively for years, which is not only burdensome but also requires costly healthcare resources.
AI as decision support
The British research team trained an AI algorithm with 36,673 OCT (optical coherence tomography) images from 6,684 patients. By combining these scans with patient data, the AI algorithm was able to predict which patients were at high risk of keratoconus progression.
The results are promising:
- Based on the initial scans, the algorithm was able to correctly classify two-thirds of patients as low risk and one-third as high risk.
- When a second check was also included, the accuracy rose to 90%.
This means that a significant proportion of patients can be treated earlier and in a more targeted manner, while others need to be checked less often. The research was presented at the 43rd Congress of the European Society of Cataract and Refractive Surgeons (ESCRS).
Added value for healthcare and patients
According to research leader Dr Shafi Balal, this development could have a major impact: ‘With AI, we can predict which patients need treatment and who can be safely monitored. This prevents unnecessary check-ups and saves patients from a late diagnosis resulting in severe vision loss.’
‘This research suggests that we can predict who is at risk from the first consultation. This allows us to intervene early and prevent complications,’ adds Dr. José Luis Güell of the Instituto de Microcirugía Ocular in Barcelona.
More powerful AI model
The researchers are now working on a more powerful AI model, trained with millions of eye scans, that can detect other eye diseases in addition to keratoconus. They are also looking at integrating the technology into clinical practice, including validation on different OCT devices.
If the results are confirmed, AI could not only contribute to better and more timely care for young patients, but also significantly relieve the pressure on ophthalmic care.
AI-driven screening for eye diseases
A few months ago, medical doctor and PhD candidate Sebastiaan van Meyel from Rotterdam Eye Hospital received the Bayer Ophthalmic Care Award 2025 for his research into AI-driven screening for glaucoma and age-related macular degeneration (AMD).
The project investigates how AI can analyse fundus photographs to identify at-risk patients at the GP's surgery. This prevents unnecessary referrals and alleviates the pressure on eye care. The research builds on an AI algorithm developed by Dr Hans Lemij, tested on large datasets such as Eyepacs
