This site is intended for UK healthcare professionals
Medscape UK Univadis Logo
Medscape UK Univadis Logo

AI as Effective as Senior Ophthalmologists in Detecting Cause of Childhood Blindness

A UK-led research team has developed a deep learning AI model that can identify which at-risk infants have retinopathy of prematurity (ROP), which left untreated can lead to blindness.

ROP primarily affects premature babies, and whilst milder forms of ROP only require monitoring, more acute cases require prompt treatment. An estimated 50,000 children globally are blind because of the condition, highlighted the authors of a new study, published in The Lancet Digital Health.

"Retinopathy of prematurity is becoming increasingly common as survival rates of premature babies improve across the globe," explained lead author Dr Konstantinos Balaskas, director, Moorfields Ophthalmic Reading Centre & Clinical AI Lab at Moorfields Eye Hospital and associate professor at the UCL Institute of Ophthalmology. "It is now the leading cause of childhood blindness in middle-income countries and in the US," he pointed out.

The condition is usually diagnosed through interval screening by paediatric ophthalmologists. However, improved survival of premature neonates, coupled with a scarcity of available experts, has raised concerns about the sustainability of this approach, the authors pointed out. In addition, they said that without the proper infrastructure for comprehensive antenatal and postnatal care, the narrow window for screening and treatment could be missed, leading to preventable blindness.

Whilst treatments are now readily available, if not detected and treated quickly, ROP can cause blindness, which Dr Balaskas said was often due to a lack of eye care specialists.

AI as Effective as Senior Ophthalmologists

For the study, an international team of scientists and clinicians in the UK, Brazil, Egypt, and the US, set out to develop bespoke and code-free deep learning-based classifiers for plus disease – a hallmark of ROP – in an ethnically diverse population in London, and then externally validate them in ethnically, geographically, and socioeconomically diverse populations in four countries across three continents.

The team developed a deep learning AI model to screen for ROP. This was trained on a sample of 7414 images of the eyes of 1370 newborns who had been admitted to the Homerton Hospital in London between 2008 and 2018, and assessed for ROP by ophthalmologists. The tool was trained to work safely across different ethnic groups and its performance was then assessed on another 200 images and compared with the assessments of senior ophthalmologists - the majority vote of three senior paediatric ophthalmologists was used as the reference standard. The researchers further validated their tool by employing it on datasets sourced from the US, Brazil, and Egypt.

Individual-level ethnicity and socioeconomic deprivation data were not collected, the authors said. The aggregate ethnicity of the cohort was 44% White, 33% Black, 13% south Asian, 4% other Asian, 5% Chinese, and 1% other.

The AI tool was found to be as effective as senior paediatric ophthalmologists in discriminating normal retinal images from those with ROP that could lead to blindness.

Although the tool was optimised for a UK population, the researchers said it was promising that they found it to still be effective on other continents, and they added that it could still be further optimised for other environments.

Improve Access to Screening

AI can be a "game-changer" and open up access to sight-saving treatments, said Dr Siegfried Wagner, from the UCL Institute of Ophthalmology and Moorfields Eye Hospital, and first author.

The authors acknowledged that further validation and studies of effectiveness across different populations were needed before deployment, but they hoped that deep learning might provide a tool for mitigating the risk of lifelong sight impairment in young patients.

"We hope that the tool will enable a trained nurse to take images that could be assessed by the AI tool, in order for a referral for treatment to be made without the need for an ophthalmologist manually to review the scans," Dr Wagner said. In doing this the authors expressed hope that the new technique would improve access to screening in the many areas with limited neonatal services and few trained ophthalmologists.

"Given it is detectable and treatable, no child should be going blind from ROP," Dr Balaskas asserted.

Funding for the study was provided by the National Institute for Health Research Biomedical Research Centre based at Moorfields Eye Hospital NHS Foundation Trust and the University College London Institute of Ophthalmology.