Improving the diagnosis of inherited eye diseases with AI

Artificial Intelligence (AI) technology could use retinal scans to aid clinicians in diagnosing rare inherited eye conditions.

Inherited retinal disorders are rare, genetic conditions that can cause vision loss and blindness. Identifying the genetic cause of these disorders from retinal imaging scans requires expert knowledge available in only a few centres in the UK. This lack of availability can lead to delays in testing and diagnosis. However, a newly developed AI technology (Eye2Gene) has been shown to perform as well or better than experts in the majority of cases.

'We all know that a picture is worth a thousand words, so we had some expectation that retinal scans interpreted by AI could out-perform HPO [Human Phenotype Ontology] terms only.' said Dr Nikolas Pontikos, who led the project. 'We hope that AI will help patients and their families by making specialist care more efficient, accessible, and equitable'.

Historically, diagnosis of these disorders has been aided by standardised descriptions of patient symptoms taken from the HPO, which was developed to help doctors and scientists communicate. However, these descriptions represent only an imperfect description of observable eye characteristics.

Researchers at the University College London Institute of Ophthalmology and Moorfields Eye Hospital in London, presented these findings at the European Society of Human Genetics conference earlier this month. The researchers were able to use data from Moorfields Hospital's historic inherited retinal disorders diagnoses, which spans 30 years of research. This included 4000 patients for whom the genetic cause of the disorder was known, the largest such cohort in the world.

When given whole exome/genome results, retinal scan and the HPO descriptions, the Eye2Gene algorithm returns a list of possible causative genes, ordered from most to least likely. On a set of 140 cases with already known genetic causes, the Eye2Gene algorithm provided lists where the correct causative gene was listed as high or higher than when only HPO characteristics were used in 70 percent of cases.

These are preliminary results and the results have not yet been peer-reviewed. The researchers emphasised that considerable further work, including clinical trials, will be needed before this software is available to aid clinical decision-making.

'We need further evaluation of Eye2Gene in order to assess its performance for different types of [inherited retinal disorders] patients from different ethnicities, different types of imaging devices, and in different types of settings, for example, primary vs. secondary care' said Dr Pontikos.

This application of AI aims to enhance clinical decision-making, which will in turn move clinicians towards a precision medicine approach for each patient. These tools are not designed to replace clinical expertise but are instead envisioned as a useful guidance tool.

'While real-life experts are essential, the use of AI will help in mitigating biases and will allow diagnoses for all in the future' said Professor Alexandre Reymond, president of the European Society of Human Genetics.