DeepMind, Google’s AI company, has created an AI with the ability to detect over 50 different types of eye illnesses by relying on 3D retinal scans.
The company revealed recently the results of its joint research operation with Moorfields Eye Hospital, which is a well-known London-based center that focuses on treating eye conditions.
According to the revelation from DeepMind, the AI used in the research had the accuracy level of professional clinicians, particularly in detecting diseases like macular degeneration and diabetic eye disease.
What’s more, the artificial intelligence technology could also recommend the ideal course of actions for those with eye conditions as well as suggest the one that called for urgent care.
According to Mustafa Suleyman, DeepMind’s co-founder, the significant part about the entire research is that the AI boasts of “explainability”, which could increase physicians’’ trust in its recommendations.
He also told Business Insider that clinicians can interpret what the AI algorithm is thinking, for instance, by focusing on the underlying segmentation.
Suleyman also revealed that the AI appears less like a mysterious black box that generates results. He went ahead to say that it labels pixels that appear on the eye scan, particularly those that match the signs of a given illness. Furthermore, the AI can compute its confidence in its findings in terms of percentage.
For a long time, British eye experts have issued warnings that patients are at the risk of losing their eyesight due to the overstretched NHS as well as the ageing population in the UK. Suleyman attributed the reason for DeepMind’s partnership with Moorfields to increased demand for eye scans, which has overwhelmed clinicians.
Dr. Pearse Keane, Moorfields Eye Hospital’s consultant ophthalmologist, said he envisions a person visiting a local street optician and receiving an OCT scan in the future. In turn, this algorithm would recognize those patients with sight-threatening conditions at an early stage.
DeepMind’s artificial intelligence (AI) relied on a database of nearly 15, 000 eye scans without any identifying information for training purposes.
It worked with clinicians to label areas of disease before running those labeled images through the system. In addition, Suleyman asserted that the project needed a considerable investment from DeepMind and involved not only 25 staffers but also researchers from Moorfields.
Google purchased DeepMind back in 2014 for a whopping £400million, which is approximately $509 million. Away from that, the British AI company is likely known for AlphaGo, one of its algorithm that is popular for beating the world champion at Go, a strategy game.
Although DeepMind maintains its headquarters in the UK and independence from Google, the relationship between the two has elicited scrutiny. The main point of concern is whether Google, which is a private US-based company, should be allowed to access sensitive medical data needed for DeepMind’s health subsidiary.
In 2016 when undertaking a project with Royal Free Hospital, DeepMind was a victim of criticism for its failure to divulge details regarding its access to past medical data.
Nevertheless, Suleyman said that the eye scans recently used by DeepMind were fully anonymized. He emphasized his statement by saying that it would be difficult to recognize the owner of the eye scans.