DeepMind, Alphabet Inc.’s U.K.-based artificial intelligence group, has developed a system that it claims can diagnose more than 50 different eye diseases with an accuracy of 94 percent.

The group detailed the project in a study published by scientific journal Nature Medicine today. The system is the fruit of a more than two-year collaboration between DeepMind and Moorfields Eye Hospital, an ophthalmic center based in the Alphabet subsidiary’s native U.K.

As part of the project, Moorfields supplied DeepMind with 16,000 anonymized, three-dimensional eye scans taken during patient visits. The group used the files to train the AI system on how to look for symptoms of different diseases. Its software performs analysis using two different neural networks. One processes eye scans to identify problems, while the other turns those findings into a diagnosis and decides if a referral recommendation is in order.

DeepMind tested the system’s accuracy by having it diagnose 1,000 scans that were not used in the training. According to the group, the software ended up making the correct call in 94 percent of the cases, with two false positions and no false negatives. If that wasn’t impressive on its own, DeepMind said the results put the AI ahead of the eight eye care professionals who also assessed the scans.

DeepMind’s goal is to reduce care times in an era when more and more people require eye treatment. Due to factors such as an aging population, ophthalmology referrals in the U.K. rose by 37 percent between 2007 and 2017, a trend that’s creating longer wait times, potentially putting some patients at risk.

“The time it takes to analyse these scans, combined with the sheer number of scans that healthcare professionals have to go through (over 1,000 a day at Moorfields alone), can lead to lengthy delays between scan and treatment – even when someone needs urgent care,” Mustafa Suleyman, DeepMind’s co-founder and dead of applied AI, wrote in a blog post. “The system we have developed seeks to address this challenge.”

The Alphabet subsidiary said that the software can not only diagnose diseases but also prioritize patients based on the severity of their conditions. That said, DeepMind isn’t looking to take human doctors out of the loop. The group said that the system was implemented as two separate neural networks specifically so the thought process behind a diagnosis can be made accessible to medical professionals for verification.

“This functionality is critically important, since eyecare professionals are always going to play a key role in deciding the type of care and treatment a patient receives,” Suleyman wrote. “Enabling them to scrutinise the technology’s recommendations is key to making the system usable in practice.”

DeepMind and Moorfields Eye Hospital plan to launch clinical trials of the system in 2019. As part of their agreement, DeepMind will make the technology available at no charge for five years to 30 of the organization’s U.K. hospitals and community clinics.

The Alphabet subsidiary is also exploring other ways to apply AI in the medical field. Earlier this year, DeepMind struck a partnership with the U.S. Department of Veterans Affairs to build a system for facilitating faster detection of acute kidney injury in patients. And not long before that, another Alphabet business called Verily introduced a method of using eye scans to estimate a person’s susceptibility to heart disease.

Image: DeepMind