Understanding retinal changes in rare inherited eye diseases 

We are funding a collaborative project between Dr Adam Dubis from the UCL Institute of Ophthalmology and Dr Watjana Lilaonitkul from the UCL Institute of Health Informatics. They will pool their expertise to analyse one of the world’s largest collections of retinal images of rare eye diseases from Moorfields Eye Hospital and other leading European Centres. They aim to use artificial intelligence to predict how the structure of the retina changes over time and how novel therapies, such as gene therapy, can affect these changes.

Revolutionary gene therapies are currently being developed - where the faulty gene causing a disease can be replaced with a working copy and in some cases prevent sight loss. 

Traditionally, patients are only treated in one eye to see if a new treatment works. The success of the treatment is then evaluated by comparing the vision between the treated and untreated eye. In order to fully harness the potential of new therapies, clinicians need to have better tools for predicting the progression of the disease and clinical outcomes of treatments.

Adam and Watjana will use artificial intelligence (AI) programmes to map, over time, changes at the back of the eye in four inherited retinal diseases (IRDs): Stargardt disease, recessive retinitis pigmentosa, USH2A-associated retinopathy and choroideremia. 

This project aims to create a platform where existing retinal images are used to train the AI to estimate retinal function and predict future degeneration. 

The ability to accurately predict the rate of retina degeneration is key for developing and implementing new therapies for IRDs. 

AI could analyse the images taken before a treatment to predict what would happen to the eye if left untreated. It could provide improved measures of therapeutic impact, inform the therapeutic window and enable better interpretation of clinical trial findings. 

The paramount ambition is to harness the power of AI to help inform personalised treatment options based on the predicted outcomes of both the treatment and disease progression for an individual patient.