Search

CET and skills guides

Study and gain CET points through OT’s online CET exams, and access archived CET, CPD articles and skills guides in our education library

Find out more

Science and vision

News and features about the latest scientific developments and advances in optometry, ophthalmology and eye medicine

Find out more

Professional support

News and features about the latest developments relating to professional support from across optics. This includes updates from optical organisations such as the AOP and the GOC

Find out more

Jobs

Explore the latest UK and global jobs in the optical sector for optometrists, dispensing opticians and more

Find out more

Moorfields researchers develop AI to track geographic atrophy

OT  speaks to consultant ophthalmologist, Konstantinos Balaskas, about AI technology that has been developed to grade geographic atrophy

Konstantinos

Understanding the state of a patient’s current geographic atrophy can be a difficult and time intensive task, Konstantinos Balaskas, a consultant ophthalmologist at Moorfields Eye Hospital, told OT.

As a result, Balaskas and a team of researchers from the Moorfields Clinical Centre and AI Hub have developed artificial intelligence (AI) technology that can quantify how much atrophy is present and how it is progressing.

Consultant ophthalmologist, Konstantinos Balaskas, talks to OT about developing AI to track geographic atrophy

The system, which uses optical coherence tomography scans of the retina, comprises four different AI models, each monitoring a different layer of the retina. The technology then recombines the models to determine if there are areas of atrophy.

“It was quite an elaborate approach,” Balaskas, who led the development team, told OT.

Balaskas highlighted that with more effective treatments for geographic atrophy being developed, the interobserver variability when determining the boundaries and surface area of the atrophic region that comes with a clinician grading the scans will become more relevant.

The AI system could also significantly save time, with Balaskas emphasising that it takes a human grader between 43 and 56 minutes to grade an image, whereas the AI system can complete the same task in 2.04 seconds.

While Balaskas acknowledges that there are still regulatory approval steps that the technology must go through, “the path to getting there is not as long as people think,” he concluded.