Revolutionizing Infant Eye Care
Detecting retinopathy of prematurity in at-risk newborns with a deep learning model
Karim Barake | | 2 min read | News
Although artificial intelligence (AI) has been a growing ophthalmic tool over the last couple of years, it has yet to scratch the surface on its pediatric implementation – until now. A group of vision scientists from University College London and Moorfields Eye Hospital have been working on the development of an AI mechanism to detect retinopathy of prematurity (ROP) in infants. The deep learning model aims to
identify infants at risk of ROP, potentially solving disparities in vision care accessibility.
ROP is a leading cause of blindness during childhood, and although it is readily identifiable at an early stage by pediatric ophthalmologists, it causes a larger problem in areas with few vision care experts. The new AI model works independently of eye care professional availability. Therefore, its potential use in low resource vision care settings is significant. By employing artificial intelligence as an early detection system, doctors can benefit from a greater impact per visiting patient, leaving the burden of screening to other healthcare workers.
The study, published in The Lancet in April 2023, is led by Konstantinos Balaskas, Director of Moorfields Eye Hospital’s Clinical AI Lab and Ophthalmic Reading Center, and Associate Professor at the University College of London’s Institute of Ophthalmology. The AI model was fed with internally and externally assessed retinal images from American, British, Egyptian, and Brazilian data sets, and graded by ophthalmologists based on severity.
After using the developed AI model to assess a dataset of 6141 diverse retinal images, the researchers found that the deep learning mechanism had a significantly similar identification success rate to senior ophthalmologists. Although the study does highlight limitations in the adaptability of the model, especially considering the proven difference in ROP based on ethnic and socio-economic status, it does serve as a major revelation in the employment of AI and deep learning to bridge the gap in treating life-harming sight impairment in infants.
- S Wagner et.al., “Development and international validation of custom-engineered and code-free deep-learning models for detection of plus disease in retinopathy of prematurity: a retrospective study,” The Lancet, [Online ahead of print] (2023). PMID: 37088692.
Trustee Scholar at Boston University and Research Intern at Schepens Eye Research Institute